A Review of Non-Invasive Sensory Feedback

This article has been accepted for publication in a future issue of this journal, but has not been
fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2018.2791583, IEEE Access
A Review of Non-Invasive Sensory Feedback

Methods for Transradial Prosthetic Hands
Benjamin Stephens-Fripp1,2,3, Gursel Alici1,2,3, Rahim Mutlu1,2,3
1
Intelligent Nano-Tera Systems Research Laboratory, University of Wollongong, Australia
2
School of Mechanical, Materials, Mechatronic and Biomedical Engineering, University of Wollongong, Australia
3
ARC Centre of Excellence for Electromaterials Science, University of Wollongong, Australia
Corresponding author: Gursel Alici (gursel@uow.edu.au). The order of the authors is determined as per their contributions.
This research has been conducted with the support of the Australian Government Research Training Program Scholarship, and with the support of the ARC
Centre of Excellence for Electromaterials Science (Grant No. CE140100012).
Abstract Any implant or prosthesis replacing a function or functions of an organ or group of organs should be biologically and sensorily integrated
with the human body in order to increase their acceptance with their user. If this replacement is for a human hand, which is an important interface
between humans and their environment, the acceptance issue and developing sensory-motor embodiment will be more challenging. Despite
progress in prosthesis technologies, 50-60% of hand amputees wear a prosthetic device. One primary reason for the rejection of the prosthetic
hands is that there is no or negligibly small feedback or tactile sensation from the hand to the user, making the hands less functional. In fact, the
loss of a hand means interrupting the closed-loop sensory feedback between the brain (motor control) and the hand (sensory feedback through
the nerves). The lack of feedback requires significant cognitive efforts from the user in order to do basic gestures and daily activities. To this
aim, recently, there has been significant development in the provision of sensory feedback from transradial prosthetic hands, to enable the user
take part in the control loop and improve user embodiment. Sensory feedback to the hand users can be provided via invasive and noninvasive
methods. The latter includes the use of temperature, vibration, mechanical pressure and skin stretching, electrotactile stimulation, phantom limb
stimulation, audio feedback, and augmented reality. This paper provides a comprehensive review of the non-invasive methods, performs their
critical evaluation, and presents challenges and opportunities associated with the non-invasive sensory feedback methods.
Index Terms Sensory Feedback, Prosthetics, Non-Invasive, Electrotactile stimulation, Mechanotactile stimulation, Vibrotactile Stimulation
potentials are then sent through our Peripheral Nervous System

I. INTRODUCTION to transmit this information to our Central Nervous System
(CNS) for decision making. However, as shown in Figure 2, the
T actile information is required for correction and control of
object grasps and manipulations as vision alone does not
provide enough of the information required [1]. Prosthetic users
feedback loop for a prosthetic device differs from our own
biological feedback system. A combination of sensors is
have also shown a strong desire to decrease the need for visual required in prosthetic devices to match the range of signals
attention to perform functions [2]. Prosthetic hand rejection detected by our mechanoreceptors in our skin. The signals from
rates are estimated to be as high as 40% [3], with some of user’s these sensors require signal processing to encode them into a
reasons for rejection and not wearing a prosthetic device being form that the user can understand. This encoded information is
that they believe it is more functional and easier to receive then sent to the CNS, either by direct stimulation of the PNS
sensory feedback through their stump without using the [11, 12] or CNS [13, 14] using electrode arrays as shown in
prosthetic hand [4]. Sensory feedback is also important for Figure 1, or via activation of the mechanoreceptors at a location
prosthetic devices as it can provide users with a sense of somewhere on the body.
embodiment in their prosthesis [5-7]. Sensory feedback for prosthetic devices can be provided by
Body-powered prosthetic limbs can transmit a limited applying a sensation to a different area of the body to represent
amount of sensory feedback through cable tension. However, the stimuli detected by the hand. This, however, requires the
with myoelectric prosthetic devices, this indirect feedback user to associate this sensation with the stimuli being detected.
pathway no longer exists [8]. This problem was identified early Having the feedback somatotopically and modality matched
on in the Boston Arm prosthetic [9] where the authors makes the feedback feel more natural and potentially easier to
introduced vibration feedback to give the user position understand. In modality matched feedback, the stimulus is
information on the elbow joint of an EMG controlled prosthetic perceived as the same method of stimulation. For example, a
resulting in a performance comparable to that of the cable pressing force on the finger is perceived by a feeling of pressure
driven prosthetic. Sensory feedback from the nerves within our [15, 16]. An example of a non-modality matched feedback is
hands provides feedback on our grasp, contact surface and its using vibration on the skin to represent the detected pressure on
roughness and shape, and grasp stability [1]. Biological skin a finger. Modality matching in non-invasive feedback can be
detects these features through four different types of achieved through mechanotactile feedback for grasping force,
mechanoreceptors in our skin [10], as shown in Figure 1. In a temperature feedback for temperature and vibrotactile feedback
simplified overview of a biological feedback system, action to communicate surface vibrations, as presented in Figure 3. In
2169-3536 (c) 2017 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2018.2791583, IEEE Access
addition, electrotactile feedback can be used to create modality is perceived to be in the same location that the stimulus is acting
matched sensations by varying the stimulation waveform on. For example, when the prosthetic pointer finger detects
properties to create the feeling of either vibration, tapping pressure, the communicated sensation is detected by the brain
and/or pressure/touch. In Somatotopical feedback, the stimulus
Figure 1 - Sensory feedback in biological skin vs artificial skin [10] © 2016 Springer. Reprinted, with Permission, from “Pursuing prosthetic
electronic skin” by Chortos et al. in Nature Materials 2016
Figure 2 - Sensory Feedback and Feed Forward Control Loops
at the pointer finger. Although the invasive methods of targeted 5) Feedback should be adjustable
reinnervation [17] and nerve electrode interfaces [11, 12] These user priorities are considered in this review when
communicate through somatotopical feedback, non-invasive assessing non-invasive sensory methods that have been
methods can also apply mechanotactile, electrotactile, presented within literature.
vibrotactile or temperature feedback to phantom hand maps Within literature, there are currently survey papers that have
[18-23] to produce somatotopical feedback. reviewed the methods deployed in sensory feedback, which
In a recent review conducted by Benz et al. [24], prosthetic have various degrees of invasiveness. A few surveys have
users felt a strong need for their prosthetic devices to be examined the role of implants into the CNS [13, 14]. These
lightweight, as the weight of their current prosthetic hand leads methods, however, require a high level of invasiveness as
to fatigue in the arm, shoulder and back. Users also raised subjects are required to undergo brain surgery to place the
concerns about their limited functionality and difficulty in appropriate implant. Recent developments have also been made
performing precise tasks. In addition to the requirement of low with direct nerve stimulation, which relies upon implants within
weight, Cipriani et al. [25] have also suggested that transradial the PNS. Normann and Fernandez’s review paper [27] focussed
prosthetic devices need to be low in their power consumption on the variety of nerve arrays available and their use within
so they can be used all day, and have a low cost. Peerdeman et control and feedback in prosthetic hands. Nghiem et al. [28]
al. [26] developed a survey, which examined the requirements also provided a comprehensive overview of current types of
for feedback (and control) from a combination of interviews feedback methods and prosthetic hands on market, with a large
with professionals who regularly interacted with users focus on direct nerve stimulation through the PNS. Although
(occupational therapists, physio etc.) and existing literature the work involving PNS electrodes has shown some good early
surveys. As a result, they produced the following feedback results [11, 12, 29, 30], it is still in an early stage of
priorities, in hierarchical importance development with limited numbers of test subjects in the
1) Continuous and proportional feedback on grasping force laboratory testing that has been undertaken, hence it is not ready
should be provided for real-life application. In addition, at present there remains a
2) Position feedback should be provided to user reluctance within prosthetic users to undergo surgery [24].
3) Interpretation of stimulation used for feedback should be The focus of this review is, therefore, on non-invasive
easy and intuitive methods (those not requiring surgery), and will therefore not
4) Feedback should be unobtrusive to user and others discuss recent advances in sensory feedback that require
surgery. Even though sensory perception can be communicated - Focus on full hand prosthetic devices, not partial
via non-invasive methods once a patient has undergone targeted amputees, with the emphasis being on transradial
reinnervation [17], these approaches will be not discussed as amputees (amputation through the forearm). One study
part of this review as patients are still required to undergo showed that out of 427 upper limb amputees interviewed,
surgery in preparation. 63% of them had either a forearm or wrist amputation [35].
Svensson et al. [31] provided a brief overview of sensors - Focus on feedback methods to the user, not the sensors
used to detect stimuli, as well as some of the invasive and non- used to detect information within the prosthetic hand.
invasive methods used to provide sensory feedback. Because of - Feedback to include the user as part of the feedback
the broad nature of this review, only an introduction to a few of loop. Will not include studies where the hand creates its
the stimulation techniques used in sensory feedback was own feedback loop without involving the user (such as
provided. In this paper, in addition to providing a deeper and camera to automatically recognise appropriate grip [36],
more comparative look at the techniques used, we shall also or automatically adjusting grip when slip occurs [37, 38]).
describe the use of temperature feedback, augmented reality - Focus on non-invasive methods, i.e. those not
and phantom hand map stimulation, which were not included in requiring surgery, such as [11, 12, 17, 29, 30].
their review.
Saudabayev and Varol [32] also provided an overview of the II. STIMULATION TECHNIQUES
development of sensors for feedback in prosthetic hands. They There are a variety of feedback methods that currently have
examined the development during the period of 2000-2015, but been deployed within literature including the use of temperature
had a focus on the sensors that provide the feedback, as opposed [39, 40], vibration [41-51], mechanical pressure and skin
to the feedback methods. Schofield et al. [33] presented the stretching [15, 16, 52-57], electrotactile stimulation[58-71],
most recent review that had a major focus on non-invasive audio feedback[72-74], and augmented reality [75, 76]. A mind
feedback methods for prosthetic hands in 2014, which map of the different feedback methods is shown in Figure 3.
discussed literature up until December 2013. Similarly, Antfolk Some of these stimulation techniques have been explored being
et al. [34] published a review on sensory feedback in upper limb deployed together [77-82]; whereas electrotactile, vibration and
prosthetics in 2013. However, there has been significant mechanical pressure have also been applied to phantom limb
progress in the field of research since then, presenting a need stimulation [18-23]. Each of these methods are discussed
for an updated comprehensive review. This paper is separately, with an assessment of the methodologies used and
differentiated from existing reviews by presenting a any challenges and opportunities that are involved in each
comprehensive overview of the different methods deployed in technique. Studies with limited subjects and/or a lack of
providing non-invasive sensory feedback for transradial performance metrics have still been included to give an insight
prosthetic hands and the recent developments that exist within into the different approaches currently being explored within
current literature, and challenges and opportunities associated
with the non-invasive sensory feedback methods.
When conducting a systematic search of the literature, the
following restrictions were, therefore, placed on studies to be
included in this review;
Figure 3 - Mind Map for Feedback Methods
this area.Vibrational FeedbackVibrational feedback typically compared to using visual feedback alone, however, this was not
uses small commercially available vibrators, which are applied consistent across all the subjects. Deploying vibrotactile
to the skin surface and activate the Pacinian corpuscle stimulation has also been shown to provide an amputee with a
mechanoreceptors in the skin. These are usually small and light higher sense of embodiment in their prosthetic [5] when
weight, as shown in Figure 4. The user learns to associate the undertaking simulations modelled after the rubber hand
vibration at that site with one of the senses from their prosthetic experiment. However, vibrational feedback requires users to
hand. Vibration has typically been used to communicate undergo training in order to develop the full benefit [88]. Ninu
grasping force, however, a few studies have examined its role et al. [42] examined the performance of vibrational feedback on
in communicating proprioceptive information [42, 50, 83], and the forearm to help improve grips for picking up objects. This
some hybrid systems have used vibration to provide modality study examined 13 subjects (11 able-bodied subjects and two
matched feedback on texture information [80, 84]. These amputees), using a commercially available myoelectrically
studies only contain preliminary testing and further controlled prosthetic arm. The authors used a constant
investigation into this form of modality matched feedback is frequency with varying amplitude to communicate velocity of
required. Using vibration as a source of force feedback has been the closing hand, and simultaneously modulated the amplitude
demonstrated to have improvements over using vision alone as and frequency of vibrations to the grasping force. The
a feedback tool [41, 44, 46], but some literature suggests that researchers demonstrated that using vibrotactile feedback to
this benefit is only visible during inadequate feedforward communicate hand velocity, point of contact and grasping force
control [85]. However, the drawbacks of using vibration without visual feedback was enough information for the
include: an extra delay of approx. 400ms to begin generating subjects to pick up objects. However, they also noted that the
vibration and a limited bandwidth being available [86]. In hand velocity was the most important feature and the addition
addition, it has also been suggested that perception of of grasping force feedback had a minimal effect. Other studies
vibrational frequency can be affected by how tightly a vibration have also demonstrated that the use of vibrotactile feedback
motor is attached [87], which raises difficulties in predictive results in an improvement in grasping objects [89-91]. Nabeel
and reliable sensory feedback. [43] developed a pressure sensor that could be applied to the
The use of three vibration feedback devices to communicate finger tip of any prosthesis and implemented a vibration
grasping force and grasping angle (separately) from a prosthetic feedback system to the forearm of the user. Their test was only
hand to its user was examined by Yamada et al. [41]. They conducted on one amputee, who, however, recognised the
concluded that by incorporating vibration feedback, there was improvements as a result. The authors also suggested that
a reduction in cognitive load required to pick up objects performance increases would require more training.
Table 1 – Comparison of Vibrotactile studies
Reference Type of Hand Location Number of No of feedback Range and number Performance
subjects/ channels and of feedback levels
Number of Sensor
Amputees
Yamada et al. Myoelectrically 3 on bicep – 5/0 1 - Single force PWM range matched 3 Subjects demonstrated 10%
[41] 2016 controlled 1-DOF one for each sensor for to strength of lower cognitive load from
robotic hand gripper level grasping force grasping force vibrotactile feedback on grasping
OR PWM range matched force
Potentiometer for to aperture angle 4 Subjects demonstrated a lower
aperture angle cognitive load (10-40%) from
vibrotactile feedback on aperture
angle
Ninu et al. [42] Myoelectrically 1 on forearm 13 / 2 2 - Single Force Varied Amplitude to Performance in achieving desired
2014 controlled Gripper Sensor and match closing grasping force for Low and High
Velocity sensor velocity Force levels:
Varied Frequency Visual Hand feedback – 76% &
and amplitude 52%
simultaneously Velocity and Contact Vibration
proportionally to feedback (No visual) – 74% & 33%
grasping force Velocity, Force &Contact
Vibration Feedback (No visual) –
84% & 53%
No visual or Vibration Feedback –
19% & 22%
Nabeel [43] Body powered 2 on the 7/1 1 - Single Force PWM range 94% of able bodied subjects could
2016 prosthetic hand forearm sensor corresponding to use feedback to determine whether
sensor values 0-255 bottle was half or completely full of
water
Rosenbau- Myoelectrically 2 on the 6/6 1 - Single force Varying Pulse Pulse Vibrational feedback improved
Chau et al. [44] controlled Robotic forearm below sensor on thumb rate and Frequency grip force accuracy by 129% for
2016 Hand (opens and the elbow to induce Light, light grip force, 21% for medium
closes) Medium and Strong. grip force. No statistical
improvement for strong grip force
Chaubey et al Myoelectrically 12 locations on 7/7 1 - Pressure Linearly mapped PW Vibrational feedback significantly
[45] 2014 controlled Robotic biceps (one sensor on target to pressure signal improved grasping force error at
Hand (opens and activated at a object input 60% maximum force but not at
closes) time 80% maximum force
Clemente et al. Myoelectrically 2 within a cuff 5/5 1 - Pressure 60ms length Less blocks were broken with
[46] 2016 controlled Robotic on bicep sensor on thumb vibration when hand vibrotactile feedback on compared
Hand (opens and and index finger made or broke to no vibrotactile feedback
closes) contact with object (p<0.001)
Hanif and N/A - Computer N/A - 0/0 1 - 1 piezoelectric Changed length of on N/A – No performance measures
Cranny [47] Simulation Computer sensor at fingertip and off pulses to listed
2016 simulation represent roughness
Li et al. [48] N/A – simulated 5 Vibrators, 5/0 5 (one each 3 Values for each N/A – No performance measures
2016 sensations for one on the back finger) – finger – Strong listed
perception test of each finger simulated Medium and Weak
of the opposite sensations for
hand mounted perception test of
in a sports forces on
glove individual fingers
Raveh et al. Myoelectrically 8 Vibrators 43 / 0 1 - 2 Force Full strength to No statistical difference in visual
[49] 2017 controlled artificial wrapped Sensors to indicate contact demand when using vibrotactile
hand around the determine force pressure above feedback to communicate contact
forearm predefined threshold, of object
otherwise off
Hasson and Virtual Arm, EMG 1 Vibrator on 9 / 0 (9 in 1 – Calculated Amplitude No significant improvement
Manczurowsky controlled angle Forearm each of the position of Arm modulated to resulting from velocity based
[50] 3 groups, 27 OR Calculated Velocity OR vibrotactile feedback or position
total) velocity of arm Amplitude based vibrotactile feedback in
modulated to achieving desired arm position
Position
Walker et al. Simulation of holding 1 vibrator on 23 / 0 2 – Force Vibration mapped to Recovery of slipping objects
2015 [51] an object, controlled bicep Feedback on objects acceleration - Visual feedback only 90%
by a stylus stylus and objects due to slip - No feedback 42%
slipping - Vibrotactile feedback 80%
acceleration
through vibration
Witteveen et Computer simulated An array of 8 10 / 10 1 – Hand Position of tactor No significant differences between
al. 2015[83] hand controlled vibrators for Aperture OR activated performance in grasping objects
through mouse aperture Grasping Force representing hand when using either Hand Aperture
scrolling opening / Feedback OR Grasping Force
1 Vibrator on 8 different levels of Feedback
forearm for intensity represent
force grasping forces
Rosenbau-Chau et al. [44] demonstrated that recognition of piezoelectric sensor at the fingertip and sent vibrational
grip force could be improved by using vibrotactile feedback, frequencies corresponding with each of the four surfaces. They
however, the impact was large for some users and not for others. only demonstrated the production of differing frequencies
The feedback system had three stages of force; low, medium visually, as the method was not tested on any subjects and their
and high; represented by differing pulse frequencies and perception of these varying vibrational frequencies.
strengths. They proposed that by incorporating more than three Li et al. [48] examined the use of vibrators on a sports glove
stages of feedback, the system could become more unreliable. on the remaining hand to provide force feedback from the
The effectiveness of sinusoidal, sawtooth and square prosthetic. This enabled the user to identify the level of force
vibrational waves on amputees with upper limb prosthetic on the back of the corresponding finger on the remaining hand
devices was examined and sinusoidal waveform performed the quickly. Each vibrator had three different intensities to
best. The bicep region was determined to be the most represent either a soft, medium or a hard level of force being
comfortable by the subjects and achieved the highest accuracy. applied to the prosthetic. Their results showed that users
Desensitisation occurred after 66 seconds and the authors quickly learnt how to interpret the vibrations and their
proposed to instead use a series of pulses, rather than performance in picking up objects improved as a result.
continuous vibrations, to achieve a higher success rate and However, it may be not as effective outside of the laboratory
reduce desensitisation. They also concluded that training when two hands are required to complete tasks.
increased the success of vibrotactile feedback. This research Raveh et al. [49] examined the effect of vibrotactile feedback
group also examined the effect of varying pulse frequency in on the visual attention required in performing tasks with a
vibrotactile feedback to communicate grasping force [45]. In prosthetic. Subjects drove a simulated car whilst performing
their home testing, the subjects’ batteries lasted a period basic tasks with their myoelectric controlled hand. Their data
ranging from one day to three days. The six subjects overall had showed no improvement in the required visual attention to
positive responses to the use of vibrational feedback, with one complete basic tasks. However, their subjects were new to
subject commenting that he enjoyed shaking his five-year-old myoelectric control, received minimal training on vibrotactile
granddaughter’s hand knowing that he was not squeezing too feedback and the system only used vibration feedback to
tight. communicate contact. The authors hypothesised that the
subjects may not have had enough time to begin to trust the
feedback and, therefore, still felt they needed to rely on visual
cues.
Hasson and Manczurowsky [50] examined the effect of
vibrational feedback on providing position and velocity
proprioception information. They only tested moving a virtual
Figure 4 - Examples of Vibrators used in Vibrotactile feedback arm to a target position, not in grasping objects. However, their
results showed no improvement from vibration feedback.
(a) Spatially Changed Vibrators [41] © 2016 IEEE. Reprinted, with Witteveen et al. [83] also compared using vibrotactile
permission, from “Investigation of a cognitive strain on hand grasping feedback to communicate grasping force to communicating
induced by sensory feedback for myoelectric hand” by Yamada et al,
in 20016 IEEE International Conference on Robotics and Automation grasping aperture. Both forms of vibration feedback improved
(ICRA) performance in grasping objects, however, there was no
significant difference between the two different approaches.
(b) Coin Vibration Motors [43] © 2016 IEEE. Reprinted, with Vibrational feedback offers a cheap and lightweight system
permission, from “Vibrotactile stimulation for 3D printed prosthetic of feedback that users prefer the sensation over electrotactile
hand” in 2nd International Conference on Robotics and Artificial
feedback [92]. One limitation, however, is the delay in
Intelligence (ICRAI)
stimulation and since the feedback delay can decrease
embodiment [93, 94], this may attribute towards some of the
Clemente et al. [46] also demonstrated a practical method of negative results.
using vibrational feedback to control grasping force. The
researchers placed pressure sensor thimbles on an existing A. Electrotactile Feedback
prosthetic and used a cuff on the upper arm to provide
Electrotactile stimulation contains no moving parts and has
vibrotactile feedback to the subjects for a period of 60ms when
an efficient power consumption. Multiple features can be easily
the hand either made or broke contact with an object. Their data
and reliably controlled including the intensity, pulse width,
showed that the subjects using vibrotactile feedback achieved a
frequency and location of stimulation (with multiple
higher success rate picking up blocks without breaking than
electrodes), which leads to a higher bandwidth being available
those only using visual feedback. The subjects maintained this
[95]. The electrodes are slim and lightweight, shown in Figure
performance whilst using this prosthetic hand with vibration
5, and electrotactile stimulation is safe and comfortable to use.
feedback at home over a period of 4 weeks. Hanif and Cranny
However, each person’s minimum sensation threshold and pain
[47] demonstrated the use of intermittent vibrational pulses as
threshold is different and the perception of electrotactile
a possible method to communicate different surface textures.
information changes with the placement of the electrodes [63],
The feedback system detected different surface textures using a
with movements as small as 1mm having an influence [96]. In

addition, skin conditions can also influence the comfort and
dynamic range of electrotactile stimulation [96].
Not only does this mean that re-calibration of thresholds are
required every time electrodes are placed on the user; but that
the pulse width, frequency and amplitude may need readjusting
to achieve the same perception each time. In addition, potential
Figure 5 - Examples of Electrotactile Electrodes
problems arise from interference between myoelectric sensors
for control and electrotactile stimulation, however, this has (a) Concentric Electrodes [58] “Virtual grasping: closed-loop force
begun to be addressed within literature [69-71]. control using electrotactile feedback” by Jorgovanovic et al. in
Electrotactile stimulation induces a sensation by directly Computational and Mathematical Methods in Medicine 2014, licensed
stimulating the primary myelinated afferent nerves in the under CC BY
dermis [97]. Concentric electrodes limit the current spread and (b) Four pairs of electrodes [62] © 2016 IEEE. Reprinted, with
can increase localisation and discernibility of the induced permission, from “Effects of different tactile feedback on myoelectric
sensation [95, 97] and can reduce the resulting noise on the closed-loop control for grasping based on electrotactile stimulation”
EMG used for myoelectric control [70]. Despite their by Xu et al. in IEEE transactions on Neural Systems and Rehabilitation
advantages, only approximately half of the electrotactile Engineering 2016
feedback systems examined use them [60, 64, 66, 68-71], which
The work by Xu et al. [62] compared communication of
may impact upon their performance.
A few studies have demonstrated the benefit of using pressure, slip, and pressure with slip information through
electrotactile stimulation, with visual feedback of lights
electrotactile feedback, such as [58]. The authors used a
representing the sensors information, and no feedback. They
constant 100Hz frequency and 3mA intensity sent to electrodes
tested 12 subjects, with six of them being amputees, using a
on the dorsal side of the forearm to communicate the force
applied to a joystick controlled robotic hand. The Pulse Width simulated environment gripping and picking up objects. Four
pairs of electrodes placed on the forearm, shown in Figure 5b,
(PW), however, was varied from 20% above their sensation
was used to deliver the electrotactile feedback. The frequency
threshold to 20% below their pain threshold to communicate the
was set to a constant value of 100Hz, and the PW was regulated
force level detected on the robotic hand by a pressure sensor.
from 0µs to 500µs to communicate any detected changes in
Their research indicated that training with electrotactile
grasping force. To communicate slip, the authors sent the
feedback helped improve the user’s recognition of grip strength
electrotactile stimulation through a sequence of the four
when picking up a variety of objects. Isakovic et al. [59] also
available pairs of electrodes (1-2-3-4-1 etc.), where the time
demonstrated that using electrotactile feedback helped users
interval between changing electrode pairs represented the
learn to regulate myoelectric control of grasping force quicker.
Schweisfurth et al. [60] showed that using electrotactile amount of measured slip in the hand grasp, ranging from 20ms
to 500ms. The data showed that pressure + slip feedback
stimulations to feedback the EMG control signals outperformed
through electrotactile feedback performed the best out of
force feedback in achieving a target initial grasping force. In
sensory feedback methods, however, visual feedback
EMG feedback, the processed myoelectric control signal was
sent to the subject via electrotactile stimulation from beginning outperformed all of them in grasping failure rate and ability to
keep the grasping force as constant as possible. The authors also
of trial to 0.35 seconds after contact with the object. In force
identified a performance difference between amputees and
feedback, the system detected the grasping force by a pressure
able-bodied test subjects, but they also recognised that their
sensor on the prosthetic finger, and then sent an electrotactile
able-bodied subjects used their dominant hand and were
signal corresponding to this level of pressure from contact until
younger than their amputee subjects.
0.35 seconds after contact. The range of pressures was matched
Although there has been success in incorporating one
to a varying amplitude and PW of the stimulation current, up to
feedback channel with electrotactile communication for one
90% of the pain value. The participants achieved closer to a
grasp, prosthetic devices often control more than one grasp.
target force when receiving electrotactile feedback based on
EMG control signals than electrotactile feedback based on Therefore, more than one feedback channel is beneficial when
closing the loop in feedback control with the user. Choi et al.
grasping force.
[63] demonstrated that subjects could distinguish two channels
Shi and Shen [61] demonstrated the effect of varying
of electrotactile feedback on their biceps. However, they did not
intensity, frequency, PW on electrical stimulation and the effect
on subject’s perception. The authors individually varied the connect the system to any sensors but instead showed that users
could distinguish between the two channels. They also
PW, frequency and amplitude, and applied these stimulation
demonstrated that better recognition was achieved when using
currents through 9mm diameter electrodes to the subject’s arm.
intermittent stimulation on both channels (switching between
The data showed that pulse width could be varied from 0.2-
20ms; intensity from 0.2mA-3mA; and frequency from 45- the two), rather than both channels being on at the same time,
resulted in better recognition.
70Hz. These ranges delivered an appropriate level of feeling in
Patel et al. [64] used four electrotactile feedback sensors to
the subject and proportionally increased grades of intensities
map the configurations of a 4-DOF prosthetic hand. They
felt by the subject.
maintained a constant PW and intensity but varied the by translating information from 64 pressure sensors into 32
frequency. Four channels of feedback were used on the subjects electrotactile electrodes on the subject’s arm. They only
to help them either control individual finger flexion, or different conducted tests on able-bodied subjects and the users could
hand grasps, with myoelectric control. However, tests were detect movement directions easily, but had trouble determining
only conducted on able-bodied patients, with feedback being on individual positions. Hartmann et al. [68] also demonstrated
the opposite arm to the myoelectric sensing. Patel et al. used that the recognition of simple movement patterns using
multiple electrotactile channels to communicate position electrotactile arrays could be learnt by able bodied subjects
whereas Pamungkas and Ward [65] demonstrated the potential through training. This opens future possibilities to be explored
of using six electrotactile feedback channels for force feedback. that could provide the prosthetic user with richer sensory
Six electrotactile locations were used to communicate feedback.
information from pressure sensors contained on a glove Surface electrodes are predominantly used for myoelectric
controlled robotic hand. Five of the locations were used to control of prosthetic devices. One problem that arises is the
communicate force acting on the prosthetic fingers, and the interaction of the electrotactile stimulation with the myoelectric
other location was used to communicate the force acting on the surface electrodes. In experiments, by using myoelectric control
palm. For each finger, three frequencies (100Hz, 60Hz and on the opposite arm to the one being stimulated, this effect is
30Hz) were used to represent the force on each phalange, and sometimes avoided, but in practical applications interference
20Hz was used for the palm. Only the highest pressure value needs to be addressed. One approach undertaken is time-
from each finger was sent to the fingers’ corresponding division multiplexing for myoelectric control and electrotactile
electrode to avoid confusion from multiple frequency signals. stimulation [69]. The system constantly switches between
Their data showed that the subject learnt how to use the myoelectric control and electrotactile stimulation so that the
feedback appropriately to pick up a range of objects, as they had two are never occurring at the same time, with a minimal
more success when alternating between picking up heavy and reduction in performance. Other studies have reduced noise
light objects. Their subject also stated that they preferred interference through redesigned electrodes. Jiang et al. [70]
electrotactile feedback to only using visual feedback when demonstrated a special designed electrode for electrotactile
operating the robotic hand. stimulation that, in combination with signal processing and
Strbac et al. [66] demonstrated a different electrode design optimisation of the stimulation waveform, limited the noise
that enabled users to distinguish up to 16 stimulation locations, interference from electrotactile stimulation feedback with the
with up to five different frequencies at once, to provide multiple myoelectric control. Xu et al. [71] produced a new flexible
levels of feedback. Test results from a small number of able- electrode design that incorporated stimulation and EMG
bodied and amputee subjects demonstrated that six electrodes recording at the one site simultaneously without interference.
with four different frequency signals could be identified with Their redesigned electrodes were used to control the robotic
more than 90% accuracy by the subjects after minimal training. hand and transmit electrotactile stimulation feedback. The
The highest number of channels recognised was from one able- electrotactile stimulations were proportional to grasping force
bodied subject identifying all 16 pads after two hours of and they resulted in a lower error rate when picking up a plastic
reinforced learning. Six amputees also recognised eight bottle. Xu et al. also demonstrated the use of tactile funnelling
different stimulation patterns that corresponded to different illusion in position feedback, whereby stimulation was
movements, with an average accuracy of 86%. The authors perceived at a location between two electrodes, depending upon
stated that their next development was to integrate this approach the intensity of each of the corresponding electrode. The higher
into the prosthetic socket connection with an automatic the ratio of intensity of one electrode in the pair, the closer the
calibration (minimum amplitude set at just above recognition perceived stimulation will be towards that electrode.
and maximum just below maximum pain threshold), but this is Electrotactile feedback shows potential for a quick and easily
yet to appear in any published literature. They also noticed that controllable method of feedback that users can identify multiple
there was a large difference between individual user’s sites of feedback at once. However, currently this sensation is
performances, indicating that this approach could work well for often referred to as a tingling feeling and occasional feeling of
some but not others. Although this study only used simulated touch. Further research is required to be undertaken on the
signal patterns instead of feedback from sensors, it particular waveform characteristics to improve the induced
demonstrated the potential of using a multichannel sensation to the subject to achieve a more natural feeling of
electrotactile feedback as a potential interface for prosthetic pressure, as has been demonstrated in direct nerve stimulation
hands. [98]. Additional care and analysis is also required to ensure that
A human hand does not contain pressure sensors, which minimal interference occurs with the EMG interface used for
communicate isolated forces back to the user, rather, nerves are myoelectric control, so it does not significantly impact the
embedded throughout the whole skin and each translates a control of the prosthetic device.
different feeling to the brain. Franceschi [67] investigated
possibilities of communicating information from artificial skin
10
Table 2 - Comparison of Electrotactile Feedback Studies
Reference Type of Hand Location Number No of feedback Range and number Performance
of channels & of feedback levels
subjects Sensor
Jorgovanovic Joystick controlled Two bipolar 10 / 0 1 - Simulated force PWM to correspond Success picking up objects:
et al. [58] 1-DOF gripping electrodes on to grasping force, - 72% wih feedback
2014 simulation dorsal side of increments of 50µ - 40% without feedback
forearm from 20% above
minimum sensation to
20% below pain
threshold
Isakovic et Myoelectrically Electrode 3/3 1 - Grasping force Six discrete Force 94% accuracy in recognition of 6
al. [59] 2016 controlled 2-DOF array – 16 levels represented by discrete force levels
prosthetic hand cathodes and different Reduction of error from 24.4% to
one anode combinations of 15.6% when using feedback
electrodes being
activated
Schweisfurth Myoelectrically Four 11 / 1 1 - Grasping force Eight, four electrodes EMG feedback resulted in 21% lower
et al. [60] controlled 2-DOF electrodes on each with two error than force feedback
2016 prosthetic hand, but forearm frequency options
only one movement
was used
Shi and Shen N/A – just rating One 1/0 1 - No sensor – Intensity ranging from No quantitative measurement -
[61] 2015 feelings from stimulation testing sensations 0-3mA, .1ma Increasing electrotactile sensation can
feedback method electrode on increment; be brought on by increasing amplitude,
wrist Frequency ranging frequency or pulse width
1Hz-100HZ, 5HZ
increment;
Pulse width 1ms –
50ms, 1ms increment
Xu et al. [62] Simulated Hand Four pairs of 12 / 6 2 - Slip sensor Pressure feedback: Pressure, Slip, Pressure +Slip feedback
2016 electrodes on and/or Grasping PWM 0µs to 500µs outperformed no feedback (p<0.05) in
bicep. One force Slip feedback: Time achieving desired grasping force and
pair used at a between switching grasping time. However, Visual
time pairs of electrodes Feedback outperformed all tactile
used 20ms-500ms feedback methods in achieving desired
grasping force and outperformed
pressure as well as slip feedback in
time of grasping force (P<0.05)
Choi et al. N/A – simulated Two pairs of 10 / 0 3 - simulated Four different levels Two channel stimulation resulted in
[63] 2016 sensations for electrodes, sensations for on two channels – recognition accuracy of 52.9% for
perception test one perception test resulting in 15 simultaneous stimulation and 73.8%
stimulating different stimuli for intermittent stimulation
electrode on across the two
either side of channels. An
upper arm additional on/off state
between was communicated
bicep and for the thumb through
tricep offset pulses
Patel et al. Myoelectrically Four 9 / 0 4 - Simulated Linearly mapped Finger flexion recognition – 94%
[64] 2016 controlled simulated concentric (Virtual thumb flexion, frequency from 3 to
prosthetic hand with electrodes on Finger) thumb opposition, 30Hz to represent Grasp pattern recognition – 79%
4-DOF and forearm 8 / 0 index flexion and flexion level
Myoelectrically (Virtual middle/ring/litter
controlled Robotic Grasp) finger flexion
hand with 4-DOF 11 / 0
(Robotic
Finger)
Pamungkas Data glove Six 1/0 6 - Pressure force Four intensity ranges No Measurements listed
and Ward controlled humanoid electrodes on on each of the five (Zero, light, light,
[65] 2015 robotic hand forearm fingers and on the medium, high)
palm corresponding to
change in intensity
Strbac et al. N/A – simulated 16 electrodes 16 / 6 8 different Tested 3,4,5,& 6 The concentric electrode pattern had a
[66] 2016 sensations for on flexible patteren used as different frequency recognition rate of 99%, 95%, 80%
perception test cuff placed channels - intervals and 74% for 3,4,5&6 different
on forearm simulated frequency levels respectively
sensations for
perception test
Franceschi N/A – simulated 32 channel 5/0 10 different 10 different Direction recognition ~ 90%
[67] 2015 sensations for electrode movement movement patterns – Orientation recognition ~70%
perception test array placed patterns on sensors on/off no in between Position recognition ~ 60%
on forearm
11
- Array of 60 (Measurements approximated from

pressure sensors graph)
Hartmann et N/A – simulated 8 electrodes 2/0 8 different Intensity of Subjects could recognise each of the
al. [68] 2014 sensations for placed on the locations - Array stimulation used to eight locations with 92% accuracy
perception test forearm of 60 pressure help provide location
sensors
Dosen et al. Myoelectrically 1 concentric 9/0 1 - Simulation Intensity proportional RMS tracking error increases from
[69] 2014 controlled electrode on error to error amplitude ~13% for normal feedback to ~21%
simulation the forearm with a 100ms delay. Overshoot
increased from ~13% for normal
feedback to ~27% feedback with
100ms delay. (Measurements
approximated from graph)
Jiang et al. 6 EMG electrodes to 1 stimulated 1/0 1 - Constant On and off value, Filtering increases Signal to Noise ratio
[70] 2014 detect noise electrode on Simulation current compared noise from from 15dB to 43dB
upper arm 6 different types of
EMG electrodes
Xu et al. [71] Myoelectrically 2 electrodes 1/0 1 - Position of 1 pressure sensor - No measurements given
2016 controlled virtual on bicep simulated elbow Intensity of
arm to move elbow joint stimulation
joint proportional to
gripping force
OR
Virtual Arm angle
mapped to varying
intensity of 2
electrodes
zones and up to two levels of force. The authors, however,

B. Mechanotactile Pressure
recognised that the ideal location for the bulbs was within the
Preliminary tests conducted by Aziziaghdam et al. [52] phantom digit zones and they had positive feedback from a pilot
showed that an object could be identified as either hard or soft test on one amputee with distinct phantom digit locations.
from the acceleration response obtained whilst tapping an
object. Pressure feedback on the clavicle bone could then be
used to communicate this acceleration profile to the user. Some
other studies have examined the role of wearable haptic devices
on feedback. Morita et al. [53] used a winding belt motor on the
upper arm to communicate grasping force feedback of a
myoelectric controlled prosthetic hand. The speed of winding
also gave the user an indication of the hardness of the object.
Casini et al. [54] demonstrated the application of distributed
haptic force to help a user determine an object as hard, medium Figure 6 - Pressure feedback cuff [54] © 2015 IEEE. Reprinted, with
or soft. A combination of pressure and skin stretch on the bicep permission, from “Design and realization of the CUFF- clenching
was used as the feedback mechanism for the subject. Godfrey upper-limb force feedback wearable device for distributed mechano-
tactile stimulation of normal and tangential skin forces”, by Casini et
et al. [15] also examined the use of a feedback band around the
al. in 2015 IEEE/RSJ International Conference on Intelligent Robots
arm to provide information to users on grasping force. and Systems (IROS)
However, although a trend was observed in grasping force
modulation, this was not statistically significant compared to
visual feedback. Also, as can be seen in Figure 6, all these
haptic feedback devices were quite large and provided
unnecessary bulk to prosthetic devices.
Antfolk et al. [16] demonstrated the use of five servo
controlled mechanical pressure devices, shown in Figure 7a.
This allowed the user to recognise touch within individual digits
and three levels of pressure feedback. The authors noticed, Figure 7 – (a) Mechanical Pressure Feedback device [18] © 2013
however, that it was not helpful for improving grip recognition, IEEE. E\Reprinted, with permission from “Artificial redirection of
but they suggested more training was necessary to overcome sensation from prosthetic fingers to the phantom hand map on
confusion between neighbouring areas. Antfolk et al. also transradial amputees: Vibrotactile versus mechanotactile sensory
suggested the use of improved actuators and placing them on feedback” by Antfolk et al. in IEEE Transactions on Neural Systems
the phantom hand map to further improve results. The use of and Rehabilitation Engineering 2013
silicon bulbs, shown in Figure 7b, has been shown as a novel
(b) Silicon Bulb Mechanical Feedback [55] from ”Sensory feedback
way to apply mechanotactile feedback [55] to communicate
from a prosthetic hand based on air-mediated pressure from the hand
touch and levels of grasping pressure. Three silicon bulbs were to the forearm skin” by Antfolk et al. in journal of rehabilitation
attached to the user’s forearm and they recognised three distinct medicine, licensed under CC BY-NC
12
Table 3 - Comparison of Mechanotactile feedback
subjects/Amputees channels & Sensor of feedback levels
Aziziaghdam N/A Simulated Mechanical 1/0 1 - Acceleration on 3 - one for hard, No performance
et al. [52] sensations from Actuator on tapping mechanism semi hard and soft measurement
2014 tapping mechanism Clavicle to simulate tapping
for perception test Bone finger on object
Morita et al. Myoelectrically Mechanically 5/0 1 - Pressure and Speed of winding Hardness sensitivity
[53] 2014 controlled prosthetic winding belt displacement of corresponds to of 0.59N/mm
hand with thumb and on bicep finger to calculate hardness
only one finger hardness
Casini et al. Robotic hand – Pressure and 1/0 1 - difference in 3 levels of hardness 100% accuracy in
[54] 2015 SoftHand Pro skin stretch current to close distinguishing
cuff worn on hand compared to between three levels
bicep look up table of hardness
Godfrey et Robotic Hand – Pressure and 6/0 1 - Estimation on 5 levels of tightness Measurements only
al. [15] 2016 SoftHand Pro skin stretch force based on mapped to grasping displayed in graphical
cuff worn on current drawn force form
bicep
Antfolk et al. N/A only tested 5 servo motor 10 / 5 Up to five - Up to 3 levels of (Amputee and Able
[16] 2013 recognition of controlled Pressure sensor pressure Bodied) Localisation:
sensations actuators on from prosthetic 75.2% & 89.6%;
forearm hand Pressure level: 91.7%
& 98.1%; Grip
recognition: 58.7 &
68.0%
Antfolk et al. N/A only tested Bulbs 32 / 12 Up to 3 - simulated 2 levels of pressure Pressure: 90% & 80%;
[55] 2012 recognition of attached to sensations for Localisation: 96%
sensations the forearm perception test
Akhtar et al. Myoelectrically Contact pads 5/0 3 - Driven by Range of 13mm of Single Finger
[56] 2014 controlled prosthetic on forearm motors that drive movement to identification error:
hand thumb, index and represent fingers NF – 17.75%, VT –
middle fingers range of motion 8.58%, Skin stretch –
9.79%
Wheeler et Myoelectrically Rotational 15 / 0 1 – rotational angle 60 of elbow range Error rate only
al. [57] 2010 controlled virtual Skin Stretch of elbow corresponds to 45 displayed in graphical
arm on back of skin rotation form; 23% reduction
tricep in visual demand
using skin stretch
device
Battaglia et Myoelecrically Rocker on 18/0 1 – aperture of hand Hand opening Discrimination of
al. [99] controlld SoftHand the bicep grip linearly mapped to different sized spheres
2017 0-60 rocker with an accuracy of
rotation 73.3%
Akhtar et al. [56] explored the use of linear skin stretch on proprioception information for a 1-DOF hand. Using this
the forearm to provide feedback on the position of fingers. As feedback, eighteen healthy subjects were able to discriminate
one of the three motors for thumb, index, remaining three between different spherical sizes with an average accuracy of
fingers, respectively, drives the tendon in the corresponding 73.3%.
finger, it pulls a contact pad attached to the forearm to increase
the skin tension. Participants described this as comfortable over
C. Temperature feedback
the whole experiment and the data indicated an improved grasp
recognition whilst using the feedback. However, testing was Temperature feedback has only been deployed to
only conducted on unimpaired subjects and the contacts pad communicate identify force of their grip and the position of
required tape or adhesive glue to attach to the skin. their fingers [26]. Temperature, however, provides users with
Bark et al. [100] examined the use of rotational skin stretch extra information about their environment, and potential
for position feedback. Although subjects had trouble with using dangers or warnings that involve heat. Producing heat on the
absolute position sensing, the authors concluded that rotational upper arm to correspond with temperature detected at the
skin stretch had some benefit for position feedback when prosthetic hand was the only method of temperature feedback
controlling movement, for an EMG controlled prosthetic. This found within literature. Cho et al. [39] used a disguised
would, however, only be for suitable for feedback for 1-DOF. temperature sensor in a prosthetic hand to sense temperature
Wheeler et al. [57] then investigated its application to position and wirelessly transmit the measured temperature range. The
feedback of an elbow of a myoelectric transhumeral prosthesis. corresponding temperature was then communicated to the
The authors found that the use of the rotation skin feedback subject via a Peltier element on their opposite hand. The
resulted in a lower target error and visual demand. subjects distinguished between high, warm and cold
Battaglia et al. [99] used skin stretch from a rotating temperature setting with reasonable accuracy, however, it drew
mechanical rocker on the bicep of the arm to communicate upon a large amount of power. Ueda and Ishii [40] also
13
Table 4 - Comparison of Temperature Feedback

subjects / channels & Sensor of feedback levels
Number of
Amputees
Cho et al. [39] Externally driven Peltier element 6/0 1 - Temperature 3 Temperature Temperature recognition
2007 prosthetic hand placed on users Sensor on prosthetic values – Hot, Mild of three temperature
(Myoelectric controls left hand finger and Cold ranges with an accuracy
bypassed) of 96.7%
Ueda and Ishii Myoelectrically Peltier element 10 / 0 1 - Temperature 5 Temperature Temperature recognition
[40] 2016 controlled prosthetic placed on users Sensor on prosthetic values - Hot, of five temperature
hand with thumb and bicep finger Lukewarm, not ranges with an accuracy
only one finger much, a little cold, of 88%
cold
examined the use of temperature feedback via a Peltier element. accurately completed the simulation and their success improved
However, they developed a prediction algorithm based upon with training, although they rated two frequencies playing
initial measurements to speed up their response times. This simultaneously as difficult to interpret. Gibson and Artemiadis
resulted in a quicker response time when providing temperature [73] showed that a subject could use auditory feedback alone to
information to the subject. Although these results are positive, pick up objects with a robotic hand. Within their study, the
with the desire for minimal weight and power consumption in variance in volume represented the level of grasping force and
prosthetic devices, and a higher need for other sensations sent the varying frequency corresponded with the location of two
to the user, this feedback method may not be deeply different regions of the hand. After training, subjects
investigated until further advances are made with force and incorporated feedback to pick up and identify objects. In
position feedback. A potential focus of research would be to
another approach, Gonzalez et al. [74] utilised triads to
incorporate temperature feedback with another feedback
communicate the movement of a robotic hand. The sound of
method so that they occur simultaneously, since it is not a
cello corresponded to the force on the thumb and a piano sound
priority to occur by itself.
represented the force on index finger. The subjects were also
D. Audio Feedback able to use the audio feedback to help improve their movements
Wilson and Diren [72] demonstrated the potential of and control when grasping objects. Each of these audio
deploying audio to communicate sensory feedback from a feedback experiments was conducted within the laboratory, and
prosthesis. They examined the test subject’s ability to interpret given their high cognitive load required, further investigation is
modulation of two audio channels to control a computer required to determine their effectiveness whilst background
simulation. Their data showed that the subject could interpret noise is occurring.
two channels, but there was a 602ms delay and the audio
feedback resulted in a high cognitive load. The subjects
Table 5 ‐ Comparison of Audio Feedback
Reference Type of Hand Location Number No of feedback Range and Performance

of channels & number of
subjects / Sensor feedback levels
Number
of
Amputees
Wilson and Diren N/A – Sensations Headphones 8/0 2 - Sensations Range of Frequency Frequency and beat modulation
[72] 2016 simulated for simulated for from 300-3400Hz, resulted in a mean squared error
perception test perception test Amplitude from 50- of 0.0406 and delay of 522ms for
65dB and beat frequency, and a mean squared
frequency 0-15Hz error of 0.0658 and a delay of
602ms for the beat frequency
channel
Gibson and 5 Fingered Headphones 12 / 0 2 - Pressure Amplitude Three groups of four subjects
Artemiadis [73] Myoelectrically Sensor on corresponded to with their own individual
2014 controlled prosthetic grasping force, 2 mappings of frequencies to hand
prosethesis fingers and different locations. They identified
position of frequencies used to objects with and accuracy of
robotic hands represent two 83%, 87% and 100%
different hand respectively.
locations
Gonzalez et al. [74] Tendon driven robot Headphones 8/0 3 - Pressure 8 different piano Subjects achieved a lower
2012 hand Sensor on triads to recognise duration completing tasks with
prosthetic different hand audio feedback (37.52s vs
fingers and configurations. 43.67s) and used a lower grip
position of Amplitude force (0.17V vs 0.25V)
robotic hands corresponded to
grasping force
14
E. Augmented Reality will also require individual customisation for those who possess
Markovic et al. [75] used Google glasses to communicate the them, however, prosthetic sockets are customised to each
aperture angle, contact time, grasping force and EMG strength individual and mapping stimulators to phantom digits could
for sensory feedback of a prosthetic hand to its user. Subjects potentially be part of this process. D'Alonzo et al. [5] were able
used the visual feedback to improve their task performance to demonstrate that by stimulating phantom digit locations
when moving objects that required various strengths without during a rubber hand experiment, they were able to promote a
breaking them. The subjects noted, however, that they typically sense of self attribution with the rubber hand.
only glanced at the information and did not use EMG strength
signals.
Clemente et al. [76] also examined the use of augmented
reality for sensory feedback for prosthetic devices. They
communicated information through an ellipse, with the axis
lengths corresponding to grasping force and angle of grasp
closure. The authors changed the proportions of the grip force
and grip closure feedback and examined if the users changed
their movements accordingly. The data indicated that the
subjects relied on the force feedback but not the closure
feedback, however, in the tasks they were constantly looking at
the objects so the grip closure information was redundant. The
grasp angle feedback may only become important when doing
tasks without looking at the hand as closely. Although there was
a lower variability in initial grip force using the feedback, there
was a significant increase in the duration of time required to
Figure 8 - Examples of Phantom Hand Maps and their corresponding
pick up the object. This suggests that although performance
Phantom digit locations [101] from "Sensory qualities of the phantom
repeatability can be increased with augmented feedback, it hand map in the residual forearm of amputees" by Bjorkman et al. in
increases the cognitive load required from the user. Journal of Rehabilitation Medicine 2016. Licensed under CC BY-NC
III. STIMULATION OF PHANTOM HAND Ehrsson et al. [6] examined 18 amputees, out of which 12 had
Amputees can not only experience phantom limb pain, but a phantom hand map. These 12 subjects underwent a human
also experience phantom limb sensations as explored in [101]. rubber hand illusion test whilst their phantom digit locations
Amputees can have locations known as phantom digits that, were stimulated. Their experimental data showed that
when touched, trigger a sensation that corresponds in their brain stimulating these sites induced a sense of ownership with the
to touching their missing finger. Phantom digits provide a prosthetic. In addition, another study [7] examined two
pathway for a natural and efficient communication for a variety amputees undergoing a functional MRI scan whilst completing
of sensations that would not require any training. However, the rubber hand illusion test. The MRI scans showed that
these phantom digit locations are not located in all amputees stimulating these phantom locations activated the
and their location and size can vary amongst individuals, as corresponding finger location within the brain.
shown in Figure 8. Wang et al. [102] suggested that the Antfolk et al. [18] examined multi-site stimulation through
distribution of phantom digits is located along the stump nerves. vibrotactile and mechanotactile feedback with amputees that
This approach, therefore, cannot be applied uniformly to all had complete phantom hand maps. They found that those with
patients, as it is unsuitable for those without phantom digits. It a complete phantom hand map recognised multiple sites of
Table 6 - Summary of Augmented Reality Feedback
subjects / channels & Sensor of feedback levels
Number of
Amputees
Markovic et al. Myoelectrically Graphical 20 / 0 4 - Aperture angle, Hand aperture on a The improvement in
[75] 2017 controlled prosthetic feedback Pressure sensor linear scale, Contact speed and accuracy of
hand displayed in (contact), Pressure with object grasping from using
Google glasses Sensor (Grasping displayed as on/off, augmented reality
Force), EMG Grasping force and feedback compared to no
sensors EMG force on a augmented reality was
linear scale statistically significant.
Clemente et al. Myoelectrically Feedback in 8/0 2 – Pressure sensor Horizontal axis of Smaller variability in
[76] 2017 controlled prosthetic ellipse form (force), ellipse representing initial grip force with
hand displayed in potentiometer grip closure feedback provided.
Google glasses (angle) Vertical axis of Significantly larger
ellipse representing duration in picking up the
grasping force object with feedback
provided
15
feedback with a higher success rate than those who had an who do not have complete mapping and/or have limited stump
incomplete or no phantom hand map. Zhang et al. [19] size to place the electrodes.
demonstrated that using Somatotopical (phantom digits) Transcutaneous Electrical Nerve Stimulation (TENS) can
Feedback (SF) outperformed Non-Somatotopical feedback induce sensations in these phantom digit locations for all fingers
(NF) on the upper arm in electrotactile stimulation feedback. [20]. This study demonstrated the effect of varying pulse width,
The SF was faster in response time (600ms), had a lower frequency and current density, and their corresponding
cognitive workload and achieved a higher recognition rate. One sensation induced. The feelings of pressure, pressure +
channel of feedback resulted in similar recognition rates for NF vibration, vibration, tingling and numbness in the
and SF; however, three channel SF performed as effectively as corresponding finger location were induced through TENS
one channel of NF. Five feedback channels in SF performed applied to the phantom digit location. Liu et al. further
marginally lower and was equivalent to the three channels of expanded on this work by showing that these signals could be
NF; although the authors suggested that interference and induced by pressing on a tactile sensor on each prosthetic finger
crossovers with the different electrodes due to their size may [21]. Chai et al. [22] went on to demonstrate that these
have affected the performance of the five channel SF feedback. sensations were stable for an 11-month period for nine
Zhang et al. also recommended to combine SF and NF for those amputees. Testing was only conducted using one electrode and
Table 7 - Comparison of Phantom Limb Stimulation
Reference Type of Location Number of No of Range and number of feedback levels Performance
Hand subjects / feedback
Number of channels &
amputees Sensor
Antfolk et al. N/A – On the forearm, up 8/8 5 – simulated Only on and off values were used Complete Phantom
[18] 2013 simulated to 5 vibrotactile or sensations for Map:
sensations 5 mechanotactile perception test Mechanotactile –
for 100%, Vibrotactile
perception – 91%
test Partial Hand Map:
Mechanotactile –
61%, Vibrotactile –
49%
Zhang et al. N/A – Up to 5 electrodes 7/7 1,3, and 5 Changed frequency from 1-75Hz Position: SF 97%,
[19] 2015 simulated On Phantom digits channels NF 90%
sensations for SF and Upper tested – Strength: SF 86%,
for arm for NF simulated NF 80%
perception stimulations sensations for
test perception test
Chai et al. N/A – 1 Stimulation 2/2 5 (only 1 Current: 0 to Upper limit (UL), .125mA Measurements
[20] 2013 simulated electrode on user’s tested at a increment, displayed in
sensations phantom digits time) – PW: 20µs to UL, 10µs increment graphical form
for simulated Frequency: 1Hz to UL, 10Hz increment
perception sensations for
Liu et al. N/A – 1 Stimulation 2/2 5 (only 1 Current varied proportional to pressure, from Measurements
[21] 2015 simulated electrode On tested at a 0mA to 25mA displayed in
sensations user’s phantom time) - graphical form
for digits Pressure
perception sensors to
test detect force
on prosthetic
finger
Chai et al. N/A – 1 stimulation 19 / 11 5 (only 1 Current: 0 to Upper limit (UL), .125mA Measurements
[22] 2015 simulated electrode On tested at a increment, displayed in
sensations user’s phantom time) – PW: 20µs to UL, 10µs increment graphical form
for digits simulated Frequency: 1Hz to UL, 10Hz increment
perception sensations for
Li et al. [23] N/A – 2 electrodes placed 6/6 2 – simulated On and off value (electrode size –
2015 simulated on PTP area sensations for discrimination
sensations perception test distance)
for Parallel electrode:
perception 12mm–39.0mm,
test 9mm-36.1, 7mm-
31.3mm, 5mm-
27.2mm
Perpendicular
electrode: 12mm-
36.1mm, 9mm-
33.5mm, 7mm-
29.1mm, 5mm-
26.5mm
16
further investigation is required on simultaneous stimulation of communicate two different stimuli simultaneously.
multiple electrodes. Furthermore, a thorough investigation into D’Alonzo et al. [77] demonstrated that subjects could
creating sensations that correspond to varying levels of identify nine levels of stimulation through a hybrid feedback of
grasping force has not yet been reported in published literature. electrotactile or vibrotactile stimulation, shown in Figure 9,
Although initial data suggests that variations in the TENS PW, compared with either mode in isolation. These same authors
amplitude and frequency ranges, could induce varying intensity also went on to show that subjects could identify patterns from
of sensations [20]. four stimulation devices, that used a combination of
Li et al. [23] examined the effect of electrode size and electrotactile and vibrotactile stimulation, with a higher
spacing on stimulating a phantom hand map with TENS. They accuracy than similar sized vibrotactile devices [78]. However,
demonstrated that the bigger electrode, the wider range of testing was only conducted on able-bodied subjects. D’Alonzo
sensations produced. However, a higher current is then required et al. suggested that their results were limited by the size of
and further space between electrodes is needed. They concluded electrodes and the performance may improve if their size was
that having an electrode sizing of 5-7mm was a good reduced. Combining mechanical pressure and vibration has also
compromise based on their preliminary investigations. been explored [79], but only an experimental prototype was
built, without any testing performed on subjects. The device
IV. COMBINING MODALITIES: HYBRID TACTILE FEEDBACK also appears very bulky.
METHODS Jimenez and Fishel [80] examined a prosthetic finger with a
The literature discussed thus far has only communicated one temperature, vibration and force sensor incorporated for
type of sensation at a time, this can often lead to an ability to sensory feedback. The weight of an object was translated into
only communicate one sensation at a time. A few studies have squeezing pressure on the arm, the temperature was produced
examined the potential of using multiple feedback methods on the bicep of the arm and surface textures were communicated
simulataneously. This may be to improve the recognition rates through vibration feedback. The subject accurately perceived
and/or range of one type of stimuli, or create the ability to the mass, temperature and roughness of the objects but each
Table 8 - Comparison of Hybrid Stimulation Techniques
Reference Type of Hand Location Number No of Range and Performance

of feedback number of
subjects / channels & feedback levels
Number Sensor
of
amputees
D’Alonzo et al. N/A – Sensations 1 Electrotactile/Vibrotactile 10 / 0 1 – 9 levels of Recognition of 9 levels using
[77] 2014 simulated for combination stimulator on the Sensations intensity hybrid setup – 56% & 72%,
perception test Forearm simulated for vibrotactile only – 29%,
perception electrotactile only 44%
test
D’alonzo et al. N/A – Sensations A combination of 3 10 / 0 1– Sensations 5 different single Single Finger: Hybrid – 98%,
[78] 2014 simulated for electrotactile stimulators and simulated for channels Electrotactile-94%,
perception test 2 vibrotactile stimulators perception (representing Vibrotactile 1 – 89%,
spread across three locations test each finger), 5 Vibrotactile 2 – 73%
on the forearm different grasp Pattern: Hybrid – 77%,
patterns Electrotactile-79%,
Vibrotactile 1 – 77%,
Vibrotactile 2 – 69%
Clemente et al. No Testing conducted – just prototype built 2 - Contact 5 levels of No performance
[79] 2014 made/break pressure measurement as no testing
& grasping Vibration undertaken
force frequency range
from 5Hz to
200Hz
Jimenez and Robotic Gripper Force Tactor, Vibration 1/1 3, only one Temperature Measurements only displayed
Fishel [80] 2014 Tactor and Temperature tested at a range +- 3C in graphical form
Tactor - all on bicep time: Vibration varied
Temperature, amplitude
Force and Pressure 0-
Vibration 200kpA of air
sensor muscle pressure
Li et al. [81] 2016 No Testing conducted – just prototype built 2 channels of Max Vibration No performance
15 actuators - 240Hz, Max measurement as no testing
No Testing, Pressure 4.4N undertaken
just prototype
built
Motamedi et al. N/A – Sensations Applied to forearm. Normal 14 / 0 1 channel of Three values of Measurements only displayed
[82] 2017 simulated for stress and Vibration applied at feedback normal stress, in graphical form
perception test same location OR 6cm away three values of
from each other vibration
feedback
17
modality was only tested one at a time. The subject also distractions that occurs within every day environments. For
suggested that the vibrational feedback mechanism was too example, will audio feedback be able to be heard as easily with
distracting. Li et al. [81] also presented a new design for a background noise, or will vibrational feedback be able to be felt
feedback mechanism that combined vibrational feedback with whilst undertaking everyday tasks?
mechanical pressure into a small, lightweight and power A large amount of testing was completed on the dominant
efficient module that can be used as part of arrays. However, at arm of able-bodied participants. However, when this same
the time of preparation of this review, there was no literature on feedback is fed to the forearm of an amputee, the perception,
the testing of this system on a person. sensitisation and response could be different.
Motamedi et al. [82] examined the perception of pressure and Both electrotactile stimulation and vibrotactile stimulation
vibration feedback at the same time. They found that pressure suffer from the disadvantage that perception can not only vary
by itself was perceived with the highest accuracy, followed by between people, but also by the location of applied contact. This
pressure and vibration at the same location, pressure and may affect the practicality of systems for use day after day.
vibration at different locations and lastly vibration by itself There has also been no examination on whether repeated
performed the weakest. application produces the same results. Vibrotactile feedback is
Hybrid tactile feedback systems are still in an early stage of dependent upon the pressure of the tactor against the skin, and
development, with half of the studies examined only displaying the tactor reapplication by the user therefore may not result in
a prototype without undertaking any experimentation. Further consistent sensations. In addition, when using multiple
testing is therefore, not only required to be undertaken to vibration tactors or electrotactile electrodes, electrode locations
determine a person's ability to recognise two different feedback may affect their repeatability. Recalibration may be required
systems simultaneously, but to also examine the effect on the each time the user places it on, and moving locations may
cognitive load. More experimental data on recognition rates and impact the cognitive load required in using the device. Further
cognitive load could help determine if hybrid tactile feedback research into these areas is required.
systems can be successfully incorporated into a feedback loop Another challenge that exists is communicating the location
to improve the user's control and embodiment with their of the feedback. Within current literature, most studies only
prosthetic hand. communicate the force that represents one location on the digit.
When grasping an object, however, subjects may want to feel
the difference between force on the fingertip and force on the
inside of the finger. Vibrotactile and electrotactile arrays appear
to be one potential solution to this problem.
There is a large amount of different approaches to test
sensory feedback methods. Some studies have only tested
simulations to ensure correct perception, whilst others have
incorporated a myoelectric controlled prosthetic hand. There
are also variances within the number of degrees of freedom
employed, the number of channels and levels of feedback, as
Figure 9 - Combination of Electrotactile and Vibrotactile feedback well as the type of sensation being communicated. These
[77] © 2014 IEEE. Reprinted, with permission, from “HyVe: hybrid
differences can make a performance comparison between
vibro-electrotactile stimulation for sensory feedback and substitution
in rehabilitation” IEEE Transactions on Neural Systems and studies difficult. However, in addition, it also appears that
Rehabilitaiton Engineering 2014 different approaches may be required for different prosthetic
users [66] and for different prosthetic hands. For example, if a
V. CONCLUSIONS AND FUTURE PERSPECTIVES prosthetic hand only contains a simple grasping motion, then
using a pressure cuff or single vibration motor could be well
Each of the different methods of sensory feedback have been
suited. Although current pressure cuffs are quite bulky, the
shown to be successful in providing extra information to the
winding belt mechanisms provide a simple and easy to learn
prosthetic user, often enabling them to make better decisions in
feedback device for single DOF devices. However, if feedback
the control and use of their prosthetic hand. Although some
is required for all five fingers, then an approach of using
studies included subjects’ reflections on their use of the
phantom digits or electrotactile stimulation could be better
prosthetic device with sensory feedback at home [46, 55], the
suited. Commercial prosthetic hands are further developing in
majority of testing, however, has been completed under
their dexterity and degrees of freedom [103] and will therefore
laboratory conditions, often involving an external computer.
require multiple channels of feedback. Initial results for
During simulated sensation testing, all concentration is on
vibrotactile and electrotactile arrays have shown some
perception of the sensation. However, during everyday tasks,
successes as users have been able to identify locations and
perception requires detection and understanding whilst
movements, however, more research should be undertaken to
undertaking other tasks, thus minimisation of cognitive load
connect them with a prosthetic hand through sensory feedback.
becomes more important. To use these feedback methods
Comparative testing is required to compare the effectiveness
within a real-life context, thorough home testing is required to
in improving control and user comfort when using the various
examine success rates with the normal background noise and
methods. This testing would be required to be specific for each
18
type of prosthetic hand. For example, one set of experiments on modalities running concurrently. There have been some
feedback mechanisms for a 1-DOF hand and then another series contradicting results on a person’s ability to understand
of tests for a 3-DOF hand, as they may not produce the same multiple sensory feedback cues. Ajoudani et al. [84]
result. These would need to not only incorporate grasping demonstrated multiple cues being used successfully, with
performance, but also measures from the subjects on areas such mechanical pressure cuff to communicate pressure forces and
as: comfort, ease of use and cognitive load. vibrational feedback to communicate texture information.
Electrotactile stimulation of the phantom hand [19-23] has However, in a study undertaken by Kim and Colgate [17], their
shown some potential for sensory feedback in a multiple DOF subject showed a lower performance picking up a virtual object
system. Current literature suggests that by stimulating the when receiving shear forces through vibrations at the same time
phantom digits, it can provide up to five separate as receiving pressure feedback on grasping force, although this
somatotopicaly matched feedback pathways that feel natural to experiment was only performed with one subject with five sets
the user. By using electrotactile stimulation, it provides a of trials. Other multimodal feedback systems [12, 19-23, 77-81]
lightweight, low-power, larger bandwidth mechanism that can have shown capability, with initial testing demonstrating that
be easily controlled. However, phantom hand maps are not users could distinguish multiple channels of information sent
located on every amputee, and their location and number of simultaneously. This could provide a method that allows for
digits appear to be unique to each person. Initial testing has only multiple channels of information to be provided back to the user
stimulated one site at a time, and no testing has been reported to make informed controlling decisions on their prosthetic hand.
on stimulating multiple phantom digits at once. Graczyk et al. Both electrotactile stimulation of the phantom hand and
[12] has reported a predictable linear relationship between multimodal sensory feedback are only at initial stages of
perceived intensity, amplitude, frequency and pulse rate in testing, with only simulated perception being examined. Further
intraneural stimulation. Further testing is required to determine testing is required to determine whether these feedback
if this same relationship exists within phantom digit mechanisms improve the user’s ability to take part in the control
stimulation. loop.
As previously discussed, the top two feedback priorities for Examination of effectiveness of sensory feedback techniques
prosthetic hand users are force and position feedback. Initial needs to progress away from being done in isolation from the
research on proprioceptive feedback has had mixed results. control system. In the case of electrotactile sensory feedback,
Hasson and Manczurowsky [50] concluded that providing interference may occur and compromises may need to be made
position information through vibrotactile feedback did not in the feedback or control system’s performance to enable then
result in any improvement. Blank et al. [104] concluded from to work together at the same time, as reported in [69]. In
their data that proprioceptive feedback alone improved the addition, as shown in Figure 9, it may be optimal for two
performance of a 1-DOF grasping task when no visual cues sensory feedback loops to exist, one to the controller and one to
were available. When visual cues were available, however, the the user. This is because currently there are limited pathways
feedback only improved tasks with a moderate level of to effectively transmit all stimulations back to the user. Too
difficulty. The authors suggested that for precise tasks, other much information may cognitively overload them or
tactile cues are required as well. Pistohl et al. [105] also incorporate too long of a delay. Instead when minor alterations
examined the role of proprioceptive feedback. Subjects are required, such as during an object slipping, a higher
controlled a cursor with EMG on one arm and fed performance may result from the prosthetic controller
proprioceptive information to the other user’s arm using a regulating the constant grasp rather than incorporating the user.
robotic manipulator. The proprioceptive information was However, further testing in this area is required to ensure the
beneficial to the user when no visual information was available, correct balance is achieved for improving grasping
but did not benefit the user when visual information was performance, user comfort, cognitive load and embodiment.
available. However, both Bark et al. [100] and Wheeler et al. Although there are a few longitudinal studies that examine
[57] concluded that rotational skin stretch had some benefit in the use of sensory feedback over a longer period [46, 88, 107],
providing position feedback, but only for 1-DOF actuator such these mainly repeat the testing regularly over a few days or
as an elbow joint. Similarly [99] also demonstrated success in weeks. However, further analysis should be done on whether
providing position information for a 1-DOF hand. Further performance is maintained when consistently using the sensory
research is therefore required to provide proprioceptive feedback throughout the day over a few weeks, similarly to the
information for hands with multiple degrees of actuation in the work done by Clemente et al. [46]. Potentially, over time, the
fingers. nervous system could become desensitised to the stimulation
At present, the majority of literature has focussed on using site, resulting in a higher cognitive load required to focus on the
feedback to send one sensation at a time. Using a single method stimulations. If such a problem exists, stimulation sites may
to communicate more than one sensation may be difficult for need to be moved up and down the arm to reduce the chance of
the user to understand or result in a high cognitive load for the desensitisation. Longitudinal studies are also required to
user. An effective approach could be to use multiple feedback examine the impact of the training and adaptation to using
methods to communicate combinations, with each feedback sensory feedback. Chai et al. [108] demonstrated that subjects
method communicating a different sensation, either were able to improve their recognition rate of electrotactile
simultaneously or by constantly switching between the two feedback on non-phantom digit sites over a three day period to
19
Figure 10 - Multiple Sensory Feedback loops (adapted from [106])
a performance comparable to phantom digit sites. Stepp[88] et timing delay between visual and tactile information can impact
al. showed that incorporating vibrational feedback, subjects the sense of body ownership in the prosthetic. A Rubber hand
continued to increase in performance over an eight day period illusion test performed by Shimadi et al. [94] and an FMRI
and they still saw a reduction in performance when the feedback study on body ownership by Bekrater-Bodmann et al. [109]
was removed on day eight. However, recently, Strbac et al. showed that 0-300ms delay occurred no loss in body ownership.
[107] demonstrated that sensory feedback was greatly This FMRI study also showed significant disconnect between
beneficial in the beginning of using the prosthetic and learning visual information and tactile information when there was a
to reliably manipulate the grasping force though their EMG separation of more than 600ms. However, a further refinement
control. However, overtime the user tended to rely more on study by Ismail and Shimadi [93] suggest that the feedback
feedforward control and their understanding of the relationship delay should be less than 200ms to maximise sense of body
between EMG commands and resulting grasping force. Further ownership. Therefore timing becomes very crucial when
investigation is therefore required to determine the role of considering the method of feedback. This gives an advantage to
sensory feedback long term and on its role in learning EMG using electrical stimulation and may limit the effectiveness of
control. mechanotactile systems. This effect of timing may also explain
In addition, studies currently examine how sensory feedback some of the conflicting results of techniques such as vibrotactile
assists a user in picking up objects, but no testing on holding feedback. Although it can be as low as 10ms to first detect
these objects for longer periods has been conducted to date. For vibration [5], it can be up to 400ms to reach the desired
example, how does the feedback mechanisms work in assisting vibration level and frequency [86]. However, although only
the user to hold a cup of coffee over the time it takes to drink mentioned in a vibrotactile study by Hasson and Manczurowky
it? The constant feedback over time, may be helpful, or it may [50], haptic drivers can be implemented to decrease start up
be distracting for the user and the feedback may need to be also times of vibration motors.
incorporated into the control mechanisms to successfully hold Although invasive methods show promise for providing a
objects. richer sensory feedback experience in the long term, non-
The speed in communicating sensations has not been widely invasive methods provide an opportunity to benefit users whilst
reported on when examining the performance of a sensory those more invasive methods are still being developed. In
feedback system. A healthy peripheral nervous system can take addition, not all users will be willing to undergo further surgery
approximately 14-28ms to deliver tactile information [1]. As a and may instead opt for the non-invasive feedback option.
result it was suggested by Antfolk et al. [34] that any surface Particularly within laboratory conditions, various approaches to
stimulation for sensory feedback should be communicated in providing sensory feedback through non-invasive methods
small percentage of that amount (3-5ms) in order to have a show promise. A focus, therefore, for the immediate future
minimal impact on the overall travel time. Additionally the should therefore be placed on implementing a simple feedback
20
strategy that can be practically used at home every day so that "Intraneural stimulation elicits discrimination of
prosthetic users can begin to take advantage of the benefits that textural features by artificial fingertip in intact
sensory feedback could provide them. and amputee humans," Elife, vol. 5, p. e09148,
Mar 08 2016.
REFERENCES
[12] E. L. Graczyk, M. A. Schiefer, H. P. Saal, B. P.
[1] R. S. Johansson and J. R. Flanagan, "Coding and Delhaye, S. J. Bensmaia, and D. J. Tyler, "The
use of tactile signals from the fingertips in neural basis of perceived intensity in natural
object manipulation tasks," Nat Rev Neurosci, and artificial touch," Science Translational
vol. 10, pp. 345‐59, May 2009. Medicine, vol. 8, pp. 362ra142‐362ra142, 2016.
[2] D. J. Atkins, D. C. Y. Heard, and W. H. Donovan, [13] G. W. Vidal, M. L. Rynes, Z. Kelliher, and S. J.
"Epidemiologic Overview of Individuals with Goodwin, "Review of Brain‐Machine Interfaces
Upper‐Limb Loss and Their Reported Research Used in Neural Prosthetics with New
Priorities," JPO: Journal of Prosthetics and Perspective on Somatosensory Feedback
Orthotics, vol. 8, pp. 2‐11, 1996. through Method of Signal Breakdown,"
[3] E. A. Biddiss and T. T. Chau, "Upper limb Scientifica (Cairo), vol. 2016, p. 8956432, 2016.
prosthesis use and abandonment: A survey of [14] D. Perruchoud, I. Pisotta, S. Carda, M. M.
the last 25 years," Prosthetics and Orthotics Murray, and S. Ionta, "Biomimetic rehabilitation
International, vol. 31, pp. 236‐257, 2007. engineering: the importance of somatosensory
[4] E. Biddiss and T. Chau, "Upper‐limb prosthetics: feedback for brain‐machine interfaces," J Neural
critical factors in device abandonment," Eng, vol. 13, p. 041001, Aug 2016.
American journal of physical medicine & [15] S. B. Godfrey, M. Bianchi, A. Bicchi, and M.
rehabilitation, vol. 86, pp. 977‐987, 2007. Santello, "Influence of force feedback on grasp
[5] M. D. Alonzo, F. Clemente, and C. Cipriani, force modulation in prosthetic applications: A
"Vibrotactile Stimulation Promotes preliminary study," in 2016 38th Annual
Embodiment of an Alien Hand in Amputees International Conference of the IEEE
With Phantom Sensations," IEEE Transactions Engineering in Medicine and Biology Society
on Neural Systems and Rehabilitation (EMBC), 2016, pp. 5439‐5442.
Engineering, vol. 23, pp. 450‐457, 2015. [16] C. Antfolk, C. Cipriani, M. C. Carrozza, C.
[6] H. H. Ehrsson, B. Rosen, A. Stockselius, C. Balkenius, A. Björkman, G. Lundborg, et al.,
Ragno, P. Kohler, and G. Lundborg, "Upper limb "Transfer of tactile input from an artificial hand
amputees can be induced to experience a to the forearm: experiments in amputees and
rubber hand as their own," Brain, vol. 131, pp. able‐bodied volunteers," Disability and
3443‐52, Dec 2008. Rehabilitation: Assistive Technology, vol. 8, pp.
[7] L. Schmalzl, A. Kalckert, C. Ragno, and H. H. 249‐254, 2013/05/01 2013.
Ehrsson, "Neural correlates of the rubber hand [17] K. Kim and J. E. Colgate, "Haptic feedback
illusion in amputees: a report of two cases," enhances grip force control of sEMG‐controlled
Neurocase, vol. 20, pp. 407‐20, Aug 2014. prosthetic hands in targeted reinnervation
[8] J. D. Brown, T. S. Kunz, D. Gardner, M. K. amputees," IEEE Trans Neural Syst Rehabil Eng,
Shelley, A. J. Davis, and R. B. Gillespie, "An vol. 20, pp. 798‐805, Nov 2012.
Empirical Evaluation of Force Feedback in Body‐ [18] C. Antfolk, M. D. Alonzo, M. Controzzi, G.
Powered Prostheses," IEEE Trans Neural Syst Lundborg, B. Rosen, F. Sebelius, et al., "Artificial
Rehabil Eng, vol. 25, pp. 215‐226, Mar 2017. Redirection of Sensation From Prosthetic
[9] R. W. Mann and S. D. Reimers, "Kinesthetic Fingers to the Phantom Hand Map on
Sensing for the EMG Controlled "Boston Arm"," Transradial Amputees: Vibrotactile Versus
IEEE Transactions on Man‐Machine Systems, Mechanotactile Sensory Feedback," IEEE
vol. 11, pp. 110‐115, 1970. Transactions on Neural Systems and
[10] A. Chortos, J. Liu, and Z. Bao, "Pursuing Rehabilitation Engineering, vol. 21, pp. 112‐120,
prosthetic electronic skin," Nat Mater, vol. 15, 2013.
pp. 937‐50, Sep 2016. [19] D. Zhang, H. Xu, P. B. Shull, J. Liu, and X. Zhu,
[11] C. M. Oddo, S. Raspopovic, F. Artoni, A. "Somatotopical feedback versus non‐
Mazzoni, G. Spigler, F. Petrini, et al., somatotopical feedback for phantom digit
21
sensation on amputees using electrotactile [29] M. Ortiz‐Catalan, B. Hakansson, and R.

stimulation," J Neuroeng Rehabil, vol. 12, p. 44, Branemark, "An osseointegrated human‐
May 02 2015. machine gateway for long‐term sensory
[20] G. H. Chai, S. Li, X. H. Sui, Z. Mei, L. W. He, C. L. feedback and motor control of artificial limbs,"
Zhong, et al., "Phantom finger perception Sci Transl Med, vol. 6, p. 257re6, Oct 08 2014.
evoked with transcutaneous electrical [30] T. S. Davis, H. A. Wark, D. T. Hutchinson, D. J.
stimulation for sensory feedback of prosthetic Warren, K. O'Neill, T. Scheinblum, et al.,
hand," in 2013 6th International IEEE/EMBS "Restoring motor control and sensory feedback
Conference on Neural Engineering (NER), 2013, in people with upper extremity amputations
pp. 271‐274. using arrays of 96 microelectrodes implanted in
[21] X. X. Liu, G. H. Chai, H. E. Qu, and N. Lan, "A the median and ulnar nerves," J Neural Eng, vol.
sensory feedback system for prosthetic hand 13, p. 036001, Jun 2016.
based on evoked tactile sensation," in 2015 [31] P. Svensson, U. Wijk, A. Björkman, and C.
37th Annual International Conference of the Antfolk, "A review of invasive and non‐invasive
IEEE Engineering in Medicine and Biology sensory feedback in upper limb prostheses,"
Society (EMBC), 2015, pp. 2493‐2496. Expert Review of Medical Devices, vol. 14, pp.
[22] G. Chai, X. Sui, S. Li, L. He, and N. Lan, 439‐447, 2017/06/03 2017.
"Characterization of evoked tactile sensation in [32] A. Saudabayev and H. A. Varol, "Sensors for
forearm amputees with transcutaneous Robotic Hands: A Survey of State of the Art,"
electrical nerve stimulation," J Neural Eng, vol. IEEE Access, vol. 3, pp. 1765‐1782, 2015.
12, p. 066002, Dec 2015. [33] J. S. Schofield, K. R. Evans, J. P. Carey, and J. S.
[23] P. Li, G. H. Chai, K. H. Zhu, N. Lan, and X. H. Sui, Hebert, "Applications of sensory feedback in
"Effects of electrode size and spacing on motorized upper extremity prosthesis: a
sensory modalities in the phantom thumb review," Expert Rev Med Devices, vol. 11, pp.
perception area for the forearm amputees," in 499‐511, Sep 2014.
Proceedings of the Annual International [34] C. Antfolk, M. D'Alonzo, B. Rosén, G. Lundborg,
Conference of the IEEE Engineering in Medicine F. Sebelius, and C. Cipriani, "Sensory feedback in
and Biology Society, EMBS, 2015, pp. 3383‐ upper limb prosthetics," Expert Review of
3386. Medical Devices, vol. 10, pp. 45‐54, Jan 2013
[24] H. L. Benz, "Upper extremity prosthesis user 2013.
perspectives on unmet needs and innovative [35] J. Kawamura, N. Fukui, M. Nakagawa, T.
technology," 2016, pp. 287‐290. Fujishita, T. Aoyama, and H. Furukawa, "The
[25] C. Cipriani, M. Controzzi, and M. C. Carrozza, Upper‐limb Amputees ‐ A Survey and Trends in
"Objectives, criteria and methods for the design Kinki Area of Japan," The Japanese Journal of
of the SmartHand transradial prosthesis," Rehabilitation Medicine, vol. 36, pp. 384‐389,
Robotica, vol. 28, pp. 919‐927, 2010. 1999.
[26] B. Peerdeman, D. Boere, H. Witteveen, H. [36] M. Markovic, "Stereovision and augmented
Hermens, S. Stramigioli, J. Rietman, et al., reality for closed‐loop control of grasping in
"Myoelectric forearm prostheses: State of the hand prostheses," Journal of neural
art from a user‐centered perspective," Journal engineering, vol. 11, p. 046001, 2014 2014.
of Rehabilitation Research and Development, [37] G. Sriram, A. N. Jensen, and S. C. Chiu, "Slippage
vol. 48, pp. 719‐738, 2011. control for a smart prosthetic hand prototype
[27] R. A. Normann and E. Fernandez, "Clinical via modified tactile sensory feedback," in IEEE
applications of penetrating neural interfaces International Conference on Electro/Information
and Utah Electrode Array technologies," J Technology, 2014, pp. 225‐230.
Neural Eng, vol. 13, p. 061003, Dec 2016. [38] W. Shaw‐Cortez, D. Oetomo, C. Manzie, and P.
[28] B. T. Nghiem, I. C. Sando, R. B. Gillespie, B. L. Choong, "Towards dynamic object manipulation
McLaughlin, G. J. Gerling, N. B. Langhals, et al., with tactile sensing for prosthetic hands," in
"Providing a sense of touch to prosthetic 2016 IEEE/RSJ International Conference on
hands," Plast Reconstr Surg, vol. 135, pp. 1652‐ Intelligent Robots and Systems (IROS), 2016, pp.
63, Jun 2015. 1164‐1169.
22
[39] Y. Cho, K. Liang, F. Folowosele, B. Miller, and N. prosthetic hand users," presented at the 2016
V. Thakor, "Wireless Temperature Sensing IEEE EMBS Conference on Biomedical
Cosmesis for Prosthesis," in 2007 IEEE 10th Engineering and Sciences (IECBES), 2016.
International Conference on Rehabilitation [48] T. Li, H. Huang, C. Antfolk, J. Justiz, and V. M.
Robotics, 2007, pp. 672‐677. Koch, "Tactile display on the remaining hand for
[40] Y. Ueda and C. Ishii, "Development of a unilateral hand amputees," Current Directions in
feedback device of temperature sensation for a Biomedical Engineering, vol. 2, 2016.
myoelectric prosthetic hand by using Peltier [49] E. Raveh, J. Friedman, and S. Portnoy,
element," in 2016 International Conference on "Visuomotor behaviors and performance in a
Advanced Mechatronic Systems (ICAMechS), dual‐task paradigm with and without
2016, pp. 488‐493. vibrotactile feedback when using a myoelectric
[41] H. Yamada, Y. Yamanoi, K. Wakita, and R. Kato, controlled hand," Assist Technol, pp. 1‐7, Jun 19
"Investigation of a cognitive strain on hand 2017.
grasping induced by sensory feedback for [50] C. J. Hasson and J. Manczurowsky, "Effects of
myoelectric hand," in 2016 IEEE International kinematic vibrotactile feedback on learning to
Conference on Robotics and Automation (ICRA), control a virtual prosthetic arm," J Neuroeng
2016, pp. 3549‐3554. Rehabil, vol. 12, p. 31, Mar 24 2015.
[42] A. Ninu, S. Dosen, S. Muceli, F. Rattay, H. Dietl, [51] J. M. Walker, A. A. Blank, P. A. Shewokis, and M.
and D. Farina, "Closed‐Loop Control of Grasping K. O’Malley, "Tactile feedback of object slip
With a Myoelectric Hand Prosthesis: Which Are facilitates virtual object manipulation," IEEE
the Relevant Feedback Variables for Force transactions on haptics, vol. 8, pp. 454‐466,
Control?," IEEE Transactions on Neural Systems 2015.
and Rehabilitation Engineering, vol. 22, pp. [52] M. Aziziaghdam and E. Samur, "Providing
1041‐1052, 2014. contact sensory feedback for upper limb robotic
[43] M. Nabeel, "Vibrotactile stimulation for 3D prosthesis," in 2014 IEEE Haptics Symposium
printed prosthetic hand," in Robotics and (HAPTICS), 2014, pp. 575‐579.
Artificial Intelligence (ICRAI), 2016 2nd [53] T. Morita, T. Kikuchi, and C. Ishii, "Development
International Conference on., 2016, pp. 202‐ of Sensory Feedback Device for Myoelectric
207. Prosthetic Hand to Provide Hardness of Objects
[44] T. Rosenbaum‐Chou, W. Daly, R. Austin, P. to Users," Journal of robotics and mechatronics,
Chaubey, and D. A. Boone, "Development and vol. s28 no. 3, 2016.
Real World Use of a Vibratory Haptic Feedback [54] S. Casini, M. Morvidoni, M. Bianchi, M.
System for Upper‐Limb Prosthetic Users," JPO: Catalano, G. Grioli, and A. Bicchi, "Design and
Journal of Prosthetics and Orthotics, vol. 28, pp. realization of the CUFF ‐ clenching upper‐limb
136‐144, 2016. force feedback wearable device for distributed
[45] P. Chaubey, T. Rosenbaum‐Chou, W. Daly, and mechano‐tactile stimulation of normal and
D. Boone, "Closed‐Loop Vibratory Haptic tangential skin forces," in 2015 IEEE/RSJ
Feedback in Upper‐Limb Prosthetic Users," JPO: International Conference on Intelligent Robots
Journal of Prosthetics and Orthotics, vol. 26, pp. and Systems (IROS), 2015, pp. 1186‐1193.
120‐127, 2014. [55] C. Antfolk, S.‐O. Bjorkman A Fau ‐ Frank, F.
[46] F. Clemente, M. D. Alonzo, M. Controzzi, B. B. Frank So Fau ‐ Sebelius, G. Sebelius F Fau ‐
Edin, and C. Cipriani, "Non‐Invasive, Temporally Lundborg, B. Lundborg G Fau ‐ Rosen, and B.
Discrete Feedback of Object Contact and Rosen, "Sensory feedback from a prosthetic
Release Improves Grasp Control of Closed‐Loop hand based on air‐mediated pressure from the
Myoelectric Transradial Prostheses," IEEE hand to the forearm skin," Journal of
Transactions on Neural Systems and Rehabilitation Medicine, pp. 702‐707, 2012.
Rehabilitation Engineering, vol. 24, pp. 1314‐ [56] A. Akhtar, M. Nguyen, L. Wan, B. Boyce, P.
1322, 2016. Slade, and T. Bretl, "Passive mechanical skin
[47] N. Mohamad Hanif, P. Chappell, N. White, A. stretch for multiple degree‐of‐freedom
Cranny, and N. N. Nik Hashim, "Tactile to proprioception in a hand prosthesis," in
vibrotactile sensory feedback interface for International Conference on Human Haptic
23
Sensing and Touch Enabled Computer [66] M. Strbac, M. Belic, M. Isakovic, V. Kojic, G.
Applications, 2014, pp. 120‐128. Bijelic, I. Popovic, et al., "Integrated and flexible
[57] J. Wheeler, K. Bark, J. Savall, and M. Cutkosky, multichannel interface for electrotactile
"Investigation of Rotational Skin Stretch for stimulation," J Neural Eng, vol. 13, p. 046014,
Proprioceptive Feedback With Application to Aug 2016.
Myoelectric Systems," IEEE Transactions on [67] M. Franceschi, L. Seminara, L. Pinna, S. Dosen,
Neural Systems and Rehabilitation Engineering, D. Farina, and M. Valle, "Preliminary evaluation
vol. 18, pp. 58‐66, 2010. of the tactile feedback system based on artificial
[58] N. Jorgovanovic, S. Dosen, D. J. Djozic, G. skin and electrotactile stimulation," in 2015
Krajoski, and D. Farina, "Virtual grasping: 37th Annual International Conference of the
closed‐loop force control using electrotactile IEEE Engineering in Medicine and Biology
feedback," Comput Math Methods Med, vol. Society (EMBC), 2015, pp. 4554‐4557.
2014, p. 120357, 2014. [68] C. Hartmann, J. Linde, S. Dosen, D. Farina, L.
[59] M. Isakovic, M. Belic, M. Strbac, I. Popovic, S. Seminara, L. Pinna, et al., "Towards prosthetic
Dosen, D. Farina, et al., "Electrotactile Feedback systems providing comprehensive tactile
Improves Performance and Facilitates Learning feedback for utility and embodiment," in 2014
in the Routine Grasping Task," Eur J Transl Myol, IEEE Biomedical Circuits and Systems
vol. 26, p. 6069, Jun 13 2016. Conference (BioCAS) Proceedings, 2014, pp.
[60] M. A. Schweisfurth, M. Markovic, S. Dosen, F. 620‐623.
Teich, B. Graimann, and D. Farina, [69] S. Dosen, M.‐C. Schaeffer, and D. Farina, "Time‐
"Electrotactile EMG feedback improves the division multiplexing for myoelectric closed‐
control of prosthesis grasping force," J Neural loop control using electrotactile feedback,"
Eng, vol. 13, p. 056010, Oct 2016. Journal of NeuroEngineering and Rehabilitation,
[61] P. Shi and X. Shen, "Sensation Feedback and vol. 11, p. 138, 09/15 2014.
Muscle Response of Electrical Stimulation on [70] L. Jiang, Q. Huang, J. Zhao, D. Yang, S. Fan, and
the Upper Limb Skin: A Case Study," pp. 969‐ H. Liu, "Noise cancellation for electrotactile
972, 2015. sensory feedback of myoelectric forearm
[62] H. Xu, D. Zhang, J. C. Huegel, W. Xu, and X. Zhu, prostheses," in 2014 IEEE International
"Effects of Different Tactile Feedback on Conference on Information and Automation
Myoelectric Closed‐Loop Control for Grasping (ICIA), 2014, pp. 1066‐1071.
Based on Electrotactile Stimulation," IEEE [71] B. Xu, A. Akhtar, Y. Liu, H. Chen, W. H. Yeo, S. I.
Transactions on Neural Systems and Park, et al., "An Epidermal Stimulation and
Rehabilitation Engineering, vol. 24, pp. 827‐836, Sensing Platform for Sensorimotor Prosthetic
2016. Control, Management of Lower Back Exertion,
[63] K. Choi, P. Kim, K. S. Kim, and S. Kim, "Two‐ and Electrical Muscle Activation," Adv Mater,
channel electrotactile stimulation for sensory vol. 28, pp. 4462‐71, Jun 2016.
feedback of fingers of prosthesis," in 2016 [72] S. Wilson and S. Dirven, "Audio sensory
IEEE/RSJ International Conference on Intelligent substitution for human‐in‐the‐loop force
Robots and Systems (IROS), 2016, pp. 1133‐ feedback of upper limb prosthetics," in 2016
1138. 23rd International Conference on Mechatronics
[64] G. K. Patel, S. Dosen, C. Castellini, and D. Farina, and Machine Vision in Practice (M2VIP), 2016,
"Multichannel electrotactile feedback for pp. 1‐6.
simultaneous and proportional myoelectric [73] A. Gibson and P. Artemiadis, "Object
control," J Neural Eng, vol. 13, p. 056015, Oct discrimination using optimized multi‐frequency
2016. auditory cross‐modal haptic feedback," in 2014
[65] D. Pamungkas and K. Ward, "Electro‐tactile 36th Annual International Conference of the
Feedback System for a Prosthetic Hand," IEEE Engineering in Medicine and Biology
presented at the International Conference on Society, 2014, pp. 6505‐6508.
Machine Vision and Mechatronics in Practice, [74] J. Gonzalez, H. Suzuki, N. Natsumi, M. Sekine,
Twoowoomba, Australia, 2015. and W. Yu, "Auditory display as a prosthetic
hand sensory feedback for reaching and
24
grasping tasks," in 2012 Annual International for intuitive control of the Pisa/IIT SoftHand,"
Conference of the IEEE Engineering in Medicine IEEE Trans Haptics, vol. 7, pp. 203‐15, Apr‐Jun
and Biology Society, 2012, pp. 1789‐1792. 2014.
[75] M. Markovic, H. Karnal, B. Graimann, D. Farina, [85] I. Saunders and S. Vijayakumar, "The role of
and S. Dosen, "GLIMPSE: Google Glass interface feed‐forward and feedback processes for
for sensory feedback in myoelectric hand closed‐loop prosthesis control," Journal of
prostheses," J Neural Eng, vol. 14, p. 036007, NeuroEngineering and Rehabilitation, vol. 8, p.
Jun 2017. 60, October 27 2011.
[76] F. Clemente, S. Dosen, L. Lonini, M. Markovic, D. [86] C. Cipriani, M. D’Alonzo, and M. C. Carrozza, "A
Farina, and C. Cipriani, "Humans Can Integrate Miniature Vibrotactile Sensory Substitution
Augmented Reality Feedback in Their Device for Multifingered Hand Prosthetics,"
Sensorimotor Control of a Robotic Hand," IEEE IEEE Transactions on Biomedical Engineering,
Transactions on Human‐Machine Systems, vol. vol. 59, pp. 400‐408, 2012.
47, pp. 583‐589, 2017. [87] J. Cohen, M. Niwa, R. W. Lindeman, H. Noma, Y.
[77] M. D'Alonzo, S. Dosen, C. Cipriani, and D. Farina, Yanagida, and K. Hosaka, "A closed‐loop tactor
"HyVE: hybrid vibro‐electrotactile stimulation frequency control system for vibrotactile
for sensory feedback and substitution in feedback," in CHI'05 Extended Abstracts on
rehabilitation," IEEE Trans Neural Syst Rehabil Human Factors in Computing Systems, 2005, pp.
Eng, vol. 22, pp. 290‐301, Mar 2014. 1296‐1299.
[78] M. D'Alonzo, S. Dosen, C. Cipriani, and D. Farina, [88] C. E. Stepp, Y. An Q Fau ‐ Matsuoka, and Y.
"HyVE‐hybrid vibro‐electrotactile stimulation‐is Matsuoka, "Repeated training with
an efficient approach to multi‐channel sensory augmentative vibrotactile feedback increases
feedback," IEEE Trans Haptics, vol. 7, pp. 181‐ object manipulation performance," 20120302
90, Apr‐Jun 2014. DCOM‐ 20120827 2012.
[79] F. Clemente and C. Cipriani, "A novel device for [89] C. E. Stepp and Y. Matsuoka, "Vibrotactile
multi‐modal sensory feedback in hand Sensory Substitution for Object Manipulation:
prosthetics: Design and preliminary prototype," Amplitude Versus Pulse Train Frequency
in 2014 IEEE Haptics Symposium (HAPTICS), Modulation," IEEE Transactions on Neural
2014, pp. 569‐573. Systems and Rehabilitation Engineering, vol. 20,
[80] M. C. Jimenez and J. A. Fishel, "Evaluation of pp. 31‐37, 2012.
force, vibration and thermal tactile feedback in [90] C. Pylatiuk, A. Kargov, and S. Schulz, "Design
prosthetic limbs," in 2014 IEEE Haptics and Evaluation of a Low‐Cost Force Feedback
Symposium (HAPTICS), 2014, pp. 437‐441. System for Myoelectric Prosthetic Hands," JPO:
[81] T. Li, H. Huang, J. Justiz, and V. M. Koch, "A Journal of Prosthetics and Orthotics, vol. 18, pp.
Miniature Multimodal Actuator for Effective 57‐61, 2006.
Tactile Feedback: Design and Characterization," [91] A. Chatterjee, P. Chaubey, J. Martin, and N.
Procedia Engineering, vol. 168, pp. 1547‐1550, Thakor, "Testing a Prosthetic Haptic Feedback
2016. Simulator With an Interactive Force Matching
[82] M. Reza Motamedi, M. Otis, and V. Duchaine, Task," JPO: Journal of Prosthetics and Orthotics,
"The Impact of Simultaneously Applying Normal vol. 20, pp. 27‐34, 2008.
Stress and Vibrotactile Stimulation for Feedback [92] H. J. Witteveen, E. A. Droog, J. S. Rietman, and
of Exteroceptive Information," J Biomech Eng, P. H. Veltink, "Vibro‐ and electrotactile user
vol. 139, Jun 01 2017. feedback on hand opening for myoelectric
[83] H. J. Witteveen, H. S. Rietman, and P. H. Veltink, forearm prostheses," IEEE Trans Biomed Eng,
"Vibrotactile grasping force and hand aperture vol. 59, pp. 2219‐26, Aug 2012.
feedback for myoelectric forearm prosthesis [93] M. A. F. Ismail and S. Shimada, "‘Robot’ Hand
users," Prosthet Orthot Int, vol. 39, pp. 204‐12, Illusion under Delayed Visual Feedback:
Jun 2015. Relationship between the Senses of Ownership
[84] A. Ajoudani, S. B. Godfrey, M. Bianchi, M. G. and Agency," PLOS ONE, vol. 11, p. e0159619,
Catalano, G. Grioli, N. Tsagarakis, et al., 2016.
"Exploring teleimpedance and tactile feedback
25
[94] S. Shimada, K. Fukuda, and K. Hiraki, "Rubber [104] A. Blank, A. M. Okamura, and K. J.
Hand Illusion under Delayed Visual Feedback," Kuchenbecker, "Identifying the role of
PLOS ONE, vol. 4, p. e6185, 2009. proprioception in upper‐limb prosthesis
[95] A. Y. J. Szeto and F. A. Saunders, control," ACM Transactions on Applied
"Electrocutaneous Stimulation for Sensory Perception, vol. 7, pp. 1‐23, 2010.
Communication in Rehabilitation Engineering," [105] T. Pistohl, D. Joshi, G. Ganesh, A. Jackson, and K.
IEEE Transactions on Biomedical Engineering, Nazarpour, "Artificial proprioceptive feedback
vol. BME‐29, pp. 300‐308, 1982. for myoelectric control," IEEE Trans Neural Syst
[96] K. A. Kaczmarek, J. G. Webster, P. Bach‐y‐Rita, Rehabil Eng, vol. 23, pp. 498‐507, May 2015.
and W. J. Tompkins, "Electrotactile and [106] F. Cordella, A. L. Ciancio, R. Sacchetti, A. Davalli,
vibrotactile displays for sensory substitution A. G. Cutti, E. Guglielmelli, et al., "Literature
systems," IEEE Transactions on Biomedical Review on Needs of Upper Limb Prosthesis
Engineering, vol. 38, pp. 1‐16, 1991. Users," Frontiers in Neuroscience, vol. 10, 2016‐
[97] A. Y. Szeto and R. R. Riso, "Sensory Feedback May‐12 2016.
Using Electrical Stimulation of the Tactile [107] M. Strbac, M. Isakovic, M. Belic, I. Popovic, I.
Sense," in Rehabilitation Engineering, J. H. L. Simanic, D. Farina, et al., "Short‐ and Long‐Term
R.V. Smith, Ed., ed: CRC Press, 1990, pp. 29‐78. Learning of Feedforward Control of a
[98] D. W. Tan, M. A. Schiefer, M. W. Keith, J. R. Myoelectric Prosthesis with Sensory Feedback
Anderson, J. Tyler, and D. J. Tyler, "A neural by Amputees," IEEE Trans Neural Syst Rehabil
interface provides long‐term stable natural Eng, Jun 06 2017.
touch perception," Sci Transl Med, vol. 6, p. [108] G. Chai, D. Zhang, and X. Zhu, "Developing Non‐
257ra138, Oct 08 2014. Somatotopic Phantom Finger Sensation to
[99] E. Battaglia, J. P. Clark, M. Bianchi, M. G. Comparable Levels of Somatotopic Sensation
Catalano, A. Bicchi, and M. K. O'Malley, "The through User Training With Electrotactile
Rice Haptic Rocker: skin stretch haptic feedback Stimulation," IEEE Transactions on Neural
with the Pisa/IIT SoftHand," in World Haptics Systems and Rehabilitation Engineering, vol. 25,
Conference (WHC), 2017 IEEE, 2017, pp. 7‐12. pp. 469‐480, 2017.
[100] K. Bark, J. Wheeler, P. Shull, J. Savall, and M. [109] R. Bekrater‐Bodmann, J. Foell, M. Diers, S.
Cutkosky, "Rotational Skin Stretch Feedback: A Kamping, M. Rance, P. Kirsch, et al., "The
Wearable Haptic Display for Motion," IEEE Trans Importance of Synchrony and Temporal Order
Haptics, vol. 3, pp. 166‐176, Jul‐Sep 2010. of Visual and Tactile Input for Illusory Limb
[101] A. Bjorkman, U. Wijk, C. Antfolk, I. Bjorkman‐ Ownership Experiences – An fMRI Study
Burtscher, and B. Rosen, "Sensory Qualities of Applying Virtual Reality," PLOS ONE, vol. 9, p.
the Phantom Hand Map in the Residual e87013, 2014.
Forearm of Amputees," Journal of
Rehabilitation Medicine, vol. 48, pp. 365‐370,
Apr 2016.
[102] H. Wang, P. Fang, L. Tian, Y. Zheng, H. Zhou, G.
Li, et al., "Towards determining the afferent
sites of perception feedback on residual arms of
amputees with transcutaneous electrical
stimulation," in 2015 37th Annual International
Conference of the IEEE Engineering in Medicine
and Biology Society (EMBC), 2015, pp. 3367‐
3370.
[103] J. T. M. S. B. S. Belter, J. L. Segil, A. M. P. S. M. B.
S. Dollar, and R. F. P. Weir, "Mechanical design
and performance specifications of
anthropomorphic prosthetic hands: A review,"
Journal of Rehabilitation Research and
Development, vol. 50, pp. 599‐618, 2013 2013.

A Review of Non-Invasive Sensory Feedback

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

A Review of Non-Invasive Sensory Feedback

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Review of Non-Invasive Sensory Feedback

Uploaded by

Copyright:

Available Formats

This article has been accepted for publication in a future issue of this journal, but has not been

A Review of Non-Invasive Sensory Feedback

potentials are then sent through our Peripheral Nervous System

Figure 2 - Sensory Feedback and Feed Forward Control Loops

Figure 3 - Mind Map for Feedback Methods

Table 1 – Comparison of Vibrotactile studies

with movements as small as 1mm having an influence [96]. In

Table 2 - Comparison of Electrotactile Feedback Studies

- Array of 60 (Measurements approximated from

zones and up to two levels of force. The authors, however,

Table 3 - Comparison of Mechanotactile feedback

Table 4 - Comparison of Temperature Feedback

Table 5 ‐ Comparison of Audio Feedback

Reference Type of Hand Location Number No of feedback Range and Performance

Table 7 - Comparison of Phantom Limb Stimulation

Table 8 - Comparison of Hybrid Stimulation Techniques

Reference Type of Hand Location Number No of Range and Performance

Figure 10 - Multiple Sensory Feedback loops (adapted from [106])

sensation on amputees using electrotactile [29] M. Ortiz‐Catalan, B. Hakansson, and R.

You might also like