See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/267491054

Control of Hand Prostheses: A Literature Review

Conference Paper · August 2013

DOI: 10.1115/DETC2013-13349

CITATIONS READS
8 3,642

2 authors:

Aimee Cloutier Jingzhou (James) Yang

Rose Hulman Institute of Technology Texas Tech University
26 PUBLICATIONS   106 CITATIONS    283 PUBLICATIONS   2,930 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Effects of Volitional Spine Stabilization on Asymmetric Lifting Task in Recurrent Low Back Pain Population View project

mechanical engineering View project

All content following this page was uploaded by Jingzhou (James) Yang on 12 March 2015.

The user has requested enhancement of the downloaded file.

 Proceedings of the ASME 2013 International Design Engineering Technical Conferences and Computers and
Information in Engineering Conference
IDETC/CIE 2013
August 4-7, 2013, Portland, Oregon, USA

DETC2013-13349

CONTROL OF HAND PROSTHESES- A LITERATURE REVIEW

Aimee Cloutier James Yang

Human-Centric Design Research Laboratory Human-Centric Design Research Laboratory
Department of Mechanical Engineering Department of Mechanical Engineering
Texas Tech University Texas Tech University
Lubbock, Texas, USA 79409 Lubbock, Texas, USA 79409
Email: aimee.cloutier@ttu.edu Email: james.yang@ttu.edu

ABSTRACT functional products. Within the last fifteen years, however,

In recent years, there has been a steep rise in the quality of there has been a steep rise in interest for creating new types of
prostheses for patients with upper limb amputations. One prosthetic limbs due to advancements in technology. Some
common control method, using electromyographic (EMG) improvements have occurred because of novel design
signals generated by muscle contractions, has allowed for an strategies, such as the use of underactuated mechanisms [1-3]
increase in the degrees of freedom (DOFs) of hand designs and and shape memory alloy (SMA) actuators [4-8] which allow for
a larger number of available grip patterns with little added more degrees of freedom (DOFs) or help reduce the weight of
complexity for the wearer. However, it provides little sensory the prosthesis. However, the increase in functionality is due in
feedback and requires non-natural control which must be large part to advancements in control strategies and sensory
learned by the user. Another recent improvement in prosthetic feedback for prosthetic hands. The introduction of control
hand design instead employs electroneurographic (ENG) using electromyographic (EMG) and electroneurographic
signals, requiring an interface directly with the peripheral (ENG) signals has created an avenue for the development of
nervous system (PNS) or the central nervous system (CNS) to much more highly advanced, anthropomorphic artificial limbs.
control a prosthetic hand. While ENG methods are more Despite major technological advancements, though,
invasive than using surface EMG for control, an interface with commercially available prosthetic hands are still far more
the PNS has the potential to provide more natural control and limited than the human hand in terms of their degrees of
creates an avenue for both efferent and afferent sensory freedom, amount of sensory feedback provided, and methods of
feedback. Despite the recent progress in design and control classifying different grip patterns and motions. Many amputees,
strategies, however, prosthetic hands are still far more limited using a body-powered hook, a simple 1 DOF gripper, or a 2
than the actual human hand. This review outlines the recent DOF hand, have expressed a desire for increased functionality,
progress in the development of EMG and ENG controlled natural movement and interaction with objects, reduced weight,
prosthetic hands, discussing advancements in the areas of higher grasping speeds and forces, and improved cosmetic
sensory feedback and control. The potential benefits and appearance [9]. There is a need for more natural designs and
limitations of both control strategies, in terms of signal more flexible control schemes. The high degree of complexity
classification, invasiveness, and sensory feedback, are and dexterity of the hand makes it difficult to replicate, and
examined. A brief overview of interfaces with the CNS is there is still vast room for improvement in upper limb
provided, and potential future developments for these control prostheses.
methods are discussed. Design methods have certainly improved, with many
researchers attempting to create reliable prostheses with more
DOFs. As more DOFs and more potential grip patterns are
1. INTRODUCTION added, the need for more advanced control schemes grows. A
For several decades, the development of new types of much larger body of research has been dedicated to developing
upper limb prostheses remained relatively stagnant due to a efficient, reliable, and diverse control schemes and
lack of technology advanced enough to create realistic and accompanying sensory feedback for prosthetic hands. Much

1 Copyright © 2013 by ASME

attention has been devoted to electrode-based methods of as books or briefcases), spherical grasp (thumb and fingers
control, including surface EMG control and implantable grasp a spherical object), and cylindrical grasp (thumb and
strategies, which place electrodes within the muscles fingers grasp a cylindrical object) [10]. Three additional
(implantable myoelectric), within or surrounding the stump’s important grasps were later identified: the flattened hand (for
nerve fascicles (for control via the peripheral nervous system), holding large surface objects), the centralized grip (used to hold
or within the brain or spinal cord (for control via the central items such as eating utensils), and wrist flexion [11-14]. The
nervous system). eight most important grip patterns for performing activities of
The purpose of this review is to outline the recent daily living (ADLs) are shown in Fig. 1.
developments in electrode-based control schemes for prosthetic
hands, discussing both surface EMG and implantable ENG
methods. The paper will be organized as follows: Section 2
examines advancements in the sensory feedback and control
schemes for prostheses controlled with EMG signals and
discusses the advantages and disadvantages of the control
method. Section 3 covers recent developments of interfaces
with the peripheral nervous system. Section 4 briefly presents
several control methods employing electrodes linked to the
central nervous system (CNS). Section 5 provides a
comparison of the EMG and ENG based control methods,
discusses the potential future of control schemes, and gives
concluding remarks.

2. MYOELECTRIC PROSTHESES
For many years, a prosthetic hand could either be
functional or cosmetically appealing, but not both. Body-
powered hooks, for example, did not mimic the size or shape of Figure 1. EIGHT CANONICAL HAND POSTURES
the human hand but were functionally capable, unlike passive RERODUCED FROM [14]
prostheses. In the early days of myoelectric hand design,
amputees would often choose a body-powered hook over a 2.1 Sensory Feedback in Myoelectric Prostheses
more realistic looking myoelectic hand because it significantly The primary disadvantage of EMG control is the lack of
lacked functionality. Several major factors contributed to the sensory feedback provided [15, 16]. When operating a
functional shortcomings of the early myoelectric hands. First, myoelectric prosthetic hand, the user is forced to visually judge
myoelectric hands provided a severely limited number of whether an object will slip or break within a grasp [17,
DOFs. Initial designs of myoelectric hands had only one DOF 18].Various methods of providing feedback have been
and performed a simple opening and closing motion. These employed to improve the hand’s grasp. One of the earliest
hands operated on a switch; later, proportional control of the methods was to use electrocutaneous stimulation for feedback.
opening and closing became available. Prior and Lyman [19] developed a 4 degree of freedom (DOF)
Several factors delayed progress toward creating a multi prosthetic arm with electrocutaneous sensory feedback to help
DOF hand. The main impeding factor was a lack of technology control grasp force and opening. Using pulse width modulation
advanced enough to allow for more than one DOF. One (PWM), they improved patients’ ability to duplicate a pinch
common requirement for the creation of a hand prosthesis is force but encountered problems with signal interference. A
that it must be roughly the same size and weight as the human study by Paciga et al. [20] reveals that signal interference with
hand. Ideally, all components would be contained inside the the prosthesis control system will inevitably occur, but
palm. The size of the motors, sensors, and additional electrical electrocutaneous feedback is still a viable option if the
components would exceed size and weight restrictions if more feedback system is placed at least 60 mm away from the pick-
than one DOF were allowed. Currently, there are prosthetic up electrodes. Additionally, this type of feedback can cause the
hands controlled with EMG capable of moving with up to 20 user to feel slight pain. In a similar study, electrocutaneous
DOFs and a thumb with 4 DOFs. The trade off is a large feedback was used in conjunction with pressure sensors to
increase in the number of motors required (in this case, 11) and, provide feedback for control. The inclusion of sensory
consequently, the weight of the prosthesis [3]. feedback received positive responses from test subjects and
There are five grasps that were initially identified as the resulted in ten times less grasping error than was present with
most important to the functional capabilities of a prosthetic no sensory feedback. Shannon [21] used strain gauges to
hand: pinch grasp (used to hold small objects), lateral grasp provide sensory feedback. The gauges provoked an electrical
(used to hold a key), hook grasp (used for carrying items such stimulus (a series of pulses); as pinch force increased, the pulse

2 Copyright © 2013 by ASME

repetition rate also increased. Other electrocutaneous feedback provides cutaneous sensory feedback to the amputee by
systems produced similar results—overall, patients felt relocating residual nerves to nonfunctional muscles. Then, the
significantly more comfortable using a prosthesis with reinnervated muscles function as amplifiers of motor
electrocutaneous feedback than without it [22-24]. commands to the prosthetic limb, and the generated EMG
Despite its advantages, electrocutaneous feedback was not signals can be used to improve control [38]. Kuiken et al. [39]
ideal because of the pain it could cause the wearer. implemented the TMR method by transferring the ulnar,
Electrocutaneous feedback also fails to provide sensory median, musculocutaneous, and distal radial nerves to the
patterns that mimic the natural hand. Patterson and Katz [25, pectoral and serratus muscles. Results showed significant
26] proposed an alternative to the electrocutaneous method in improvement in control; additionally, whenever the
which pressure sensors were used to provide feedback. reinnervated chest skin was touched, the amputee experienced
Experiments were performed involving pressure feedback, sensation in the amputated hand [40].
visual feedback, vibrotactile feedback, and combinations of the
three. The results showed that setting the pressure feedback 2.2 Signal Classification Techniques
equal to the grasping pressure of the hand reduced grasping For the past decade, the problem of signal classification
pressure errors. Mingrino et al. [27] attempted to approach the has been addressed by a large number of engineering
most natural feedback method for a prosthetic hand. A human researchers, undoubtedly because a robust combination of
hand prevents an object from slipping by detecting signal processing and classification algorithms is vital to a
microvibrations through receptors in the skin. Therefore, a functional, dexterous, and anthropomorphic electrode-based
vibrotactile method of sensory feedback was used to control prosthetic hand. Improved classifiers allow for more functions
slip. In contrast to sensing displacement, vibrotactile sensors of the hand, a quicker response time, and a more reliably
can provide an early warning for object slippage before and accurate system. While the problem of obtaining an adequate
during sliding. The sensory feedback system included a force number of degrees of freedom (DOF) is largely a design
sensing resistor, a dynamic sensor (piezoelectric), and a concern, signal classification is equally important. Without an
vibrotactile sensor. The resulting vibration was proportional to advanced signal classifier, an electrode-based prosthetic hand
the strain produced on the PVDF [28, 29]. would be incapable of reliably or accurately performing grasp
An additional method of sensory feedback is to mimic the patterns. The components of a signal classifier can be broken
neuromuscular system using compliance control [30]. into the basic stages of feature extraction (including
Compliance control is based on the characteristic of increased dimensionality reduction) and pattern recognition, along with
compliance around the joint with decreasing activity of the online and offline learning. An example of a typical classifier is
muscle. One research group used compliance control by shown in Fig. 2 [41]. Much research has been dedicated to
implementing an interferential current synthesizing low optimizing feature extraction and pattern recognition, and
frequencies in the body. Two waves were used—one fixed at various techniques for each of those classifier steps will be
4000 Hz, and another changing from 3700 Hz to 4000 Hz. discussed in the following sections.
Subjects successfully detected changes in the stimulus
frequency [31, 32].
One of the main concerns when developing methods of
providing sensory feedback is the prevention of object
slippage. The role of a slip detector is to increase the grip force
and prevent slippage from occurring. Cotton et al. [33]
developed a thick-film piezoelectric sensor to be placed on the
fingertip for slippage detection. Engeberg et al. [34] developed
an adaptive controller for slip prevention. Four methods were
tested: force control, proportional and proportional derivative
(PD) prevention of slip, and the adaptive slip prevention
controller, designed to amplify the vibrations when slippage
occurred. While all methods outperformed force control, the
adaptive slip prevention controller produced the best results. Figure 2. TYPICAL CONTROL CLASSIFIER
Similarly, other studies compared several alternate methods of REPRODUCED FROM [38]
slip control (sliding mode slip prevention controller, integral
sliding mode slip prevention controller, and PD controller) and 2.3 Feature Extraction
concluded that the integral sliding mode controller best The purpose of the feature extraction stage is to identify
detected object slippage while producing the least amount of and select the most useful components of the processed signal.
object deformation [35-37]. An intelligent choice of features is a highly important step in
One final technique of improving sensory feedback is to the signal classification process; by choosing features which
use the method of targeted muscle reinnervation (TMR). TMR separate patterns of movement most distinctly, a higher level of

3 Copyright © 2013 by ASME

accuracy can be obtained. Choosing the optimal number of PCA is used for dimensionality reduction and compared to the
extracted features is also important for creating a balance performance when the feature vector is self-organizing feature
between the reliability and computational burden of the system. map (SOFM). The use of PCA and SOFM are also considered
The extracted portion of the signal is contained in the together, and a multilayer ANN was used for pattern
transient part of the EMG signal. There are numerous features recognition. Results indicated a trade off when each method
that have been chosen and employed for extraction, and they was used alone; the SOFM obtained much higher accuracy but
are separated into several main categories. The two most also required a much larger computation time. A combination
commonly extracted features are time domain features and of the two results in an optimal compromise, achieving both
time-frequency domain features. Among the most popular time low error rates and low processing time. All processes (9
domain features are mean absolute value (MAV), zero crossing motions) could be performed in less than 125 msec.
(ZC), waveform length (WL), root mean square (RMS) and
slope sign change (SSC) [13]. Liu et al. [42] tested 2.4 Pattern Recognition Techniques
combinations of the MAV, ZC, WL, and SSC time domain The pattern recognition stage of signal classification uses
features for the classification of eleven different hand the extracted features to predict the intended motion and, given
movements and concluded that three time-domain features this information, choose the resulting movement of the
(regardless of the feature) and a sixth order autoregressive prosthesis. There are several common types of pattern
(AR) model provide optimal feature extraction. Using linear recognition techniques used: Bayesian classifiers, linear
discriminant analysis (LDA), a pattern classifier was created classifiers such as linear discriminant functions, the least
which achieved accuracy similar to an artificial neural network squares method, and support vector machines, nonlinear
(ANN) [43]. Huang and Chiang [44] used a combination of classifiers, commonly employing back propagation and neural
time domain features, including integral of EMG (IEMG), networks (as well as nonlinear support vector machines), and
variance, ZC, SSC, an AR model, Cepstrum analysis, WL, and the fuzzy optimization-based clustering algorithm, among
Willison Amplitude (WAMP) to determine which time domain others. As with feature extraction, it is important to strike a
features would produce the highest classification rate. Those balance among computational burden, accuracy, and stability.
features were extracted and used as input for the classifier. Oftentimes, different classifiers can be combined to produce
Time-frequency domain features are more accurate than time the optimal result or the classifier will be combined with an
domain features but are also more computationally expensive. optimization technique such as a genetic algorithm to improve
The most commonly used time-frequency domain features its performance.
are wavelet transform (WT) and wavelet packet transform
(WPT) [45]. Sometimes, it is useful to combine time and time- 2.4.1 Bayesian Classifiers. Many pattern recognition
frequency domain features. Arvetti et al. [46] combined the methods in signal classification systems are based on the
MAV, ZC, WL, and SSC time domain features with the probability of an intended motion given the feature vector
extraction of the time-frequency wavelet features as the driving input. Therefore, the Bayes rule is useful for determining the
features for an ANN. A list of common time domain features conditional probabilities (and thus, the intended motion). There
and time-frequency domain features, along with their are numerous types of Bayesian classifiers. One interesting use
governing equations, can be found in [47]. of a Bayes classifier was produced by Kurzynski and
Another technique associated with the feature extraction Wolczowski [50], who developed a recursive method which
stage is dimensionality reduction. Dimensionality reduction is would feed new parameters into a genetic algorithm after each
used to reduce the dimension of the feature vector, thereby iteration for better learning performance. A Bayes-optimal
alleviating some of the computational burden. One commonly algorithm was used for feature extraction and compared to the
used dimensionality reduction technique is principal performance of the RMS algorithm, and a Bayes classifier was
component analysis (PCA). Using PCA, the original data is compared to two different Markov models. The key difference
linearly mapped to a new, reduced set of uncorrelated features. between a Markov model and a Bayes classifier is that a Bayes
The mean square error between the original and projected classifier treats features as independent objects, while Markov
feature sets is minimized. [48] employed PCA for models treat the algorithm as a sequential decision task. Both
dimensionality reduction, using MAV, Wilson Amplitude, and Markov models significantly outperformed the Bayes classifier
waveform length features. With a feed forward artificial neural while the four Bayes-optimal feature extraction algorithms
network (ANN) and 32 surface electrodes, they were able to outperformed the RMS approach. The Markov model’s
decode individual finger movements with a discrimination consistently better performance shows that treating algorithms
accuracy of 98%. While the system achieves a higher accuracy as decision tasks always produces a better result. In another
than most classification systems, the large number of EMG study, a virtual model of a hand is created using a combination
channels may increase the amount of concentration and effort fuzzy-Bayesian classifier (called the Fuzzy Bayesian Inference
required for the patient to control it. PCA is also used in Chu et System) [51]. Kondo et al. [52] proposed an anticipatory
al. [49] for feature reduction. Extracting the time-frequency pattern recognition technique with the k-nearest neighbor
domain WPT feature from the transient portion of the signal, Bayesian classifier. Among input transient signals, the

4 Copyright © 2013 by ASME

anticipatory pattern recognition identifies possible grasp signals, and the coefficients were imported into a BP ANN
patterns given prior knowledge about the probability of which identified finger motions. Herle et al. [74] used an AR
transitioning from the current grip to another grip. Across all model with a feed forward NN and a higher level hierarchical
subjects and trials, anticipatory pattern recognition combined controller based on Finite State Machine (FSM) for
with k-NN outperformed k-NN alone by an average of 6.7%. classification. A 91% classification rate was achieved with low
computational cost. Nishikawa et al. [75] recognized the
2.4.2 Fuzzy Clustering Algorithms. Clustering importance of online learning and proposed an ANN that
methods, in general, are unsupervised and aim to identify adapts to an individual’s characteristics in real time. Their
groups (referred to as clusters) with similar patterns and to gain system was able to discriminate 6 forearm motions with a rate
useful information from these clusters. The advantage of the of 89.9% for online learning (compared to a rate of 80.3% for
optimization-based fuzzy clustering algorithm over other traditional offline methods). Tsuji et al. [60] used an ANN
clustering methods is that it does not require an assumed during two stages: one to discriminate the motion and one for
probability density function (PDF) [53]. Fuzzy inference impedance control by measuring the flexor and extensor
systems (FIS) are effective tools for information processing. In contractions from the other ANN. Ito et al. [76, 77] improved
a FIS, a set of membership functions are produced as input, and the speed of ANN-based pattern recognition using
a collection of if-then statements determine the output [54]. considerations of entropy to determine movement vs. non-
Weir and Ajiboye [55] used a fuzzy logic classifier to control a movement and extracting features from an initial rise zone. A
4 DOF prosthetic hand and produce discrimination rates discrimination accuracy of greater than 90% was attained, and
ranging from 86-100% across all subjects. Momen et al. [56] intended movements could be executed in approximately 70
recognized that pattern recognition approaches to control often ms, whereas most systems exceed 200 ms. Karimi et al. [78]
result in unnatural movement. They attempted to correct this extracted signals from 16 different forearm locations to
flaw by creating an EMG classifier which chooses from a set of discriminate 10 hand motions. The standard deviation of a
intended movements selected by the user. The RMS feature WPT was used as the feature vector, a two-layer ANN
extraction and fuzzy C-means classifier combination performed the pattern recognition, and a GA optimized the
discriminated movements with an accuracy of 92.7%. The algorithm. Discrimination rates exceeded 98%.
approach by Khezri et al. [57] uses a combination of back
propagation and least mean square for training a neuro-fuzzy
classifier. The combined approach classified six hand motions
with an accuracy of 96%.

2.4.3 Neural Networks. As a pattern classifier and a

learning tool, neural networks have been widely used for
controlling prosthetic hands [58-64]. A neural network is meant
to mimic the mapping patterns of the brain; based on the
behavior of neurons, an artificial neural network (ANN) is a
collection of nodes (to represent neurons) and the relationships
among those nodes. Neural networks also typically involve a
learning phase which allows the classifier to improve over
time. A neural network is composed of an input layer, a series
of hidden layers, and an output layer, as shown in Fig. 3, and it
is one of the most popular pattern recognition techniques for
prosthetic hand control applications [65-67]. Figure 3. ANN WITH EXAMPLE INPUT AND OUTPUT
Neural networks effectively model nonlinear data; the REPRODUCED FROM Karimi et al. (2011)
drawback is that they tend to be unstable. Usually, neural
networks are combined with other classifying strategies to 2.5 Hierarchical Control
improve stability [68]. One common combination is the use of One alternative to the traditional pattern recognition
a NN with an AR model [69]. Zhao et al. [70-72] used a approach is to implement hierarchical control. Hierarchical
combination of a NN, an AR model, and WT feature extraction control is an event-driven process which uses sensory
to control a five-fingered, underactuated prosthetic hand information and causes the limb to switch among states and
intended to perform power grasp, centralized grasp, fingertip which generally simplifies the task of posture selection. The
grasp, and cylindrical grasp. The NN was based on the concept of hierarchical control was first implemented in the
Levenberg-Marquardt or variable learning rate (VLR) methods. design of the MARCUS prosthetic hand. Initially, hierarchical
The AR and ANN combination was also used by Wang et al. control was used to transition among states such as touch, hold,
[73] to distinguish among various finger movements. The AR squeeze, and release [79]. Later, it was combined with pattern
model performed dimensionality reduction on the EMG recognition, where the pattern classifier would select a

5 Copyright © 2013 by ASME

prehensile posture, and the hierarchical controller performed amputee to set his own grip force and joint positions without
the transition among grasps [80]. the aid of visual feedback, as demonstrated in the study.
The most notable applications of interfaces with the PNS
arose from a series of European research projects, such as the
3. ELECTRONEUROGRAPHIC CONTROL WITH PNS CYBERHAND, SMARTHAND, and Revolutionizing
A prosthetic hand controlled by ENG is operated by Prosthetics projects. The CYBERHAND aimed to create a link
electrodes interfaced directly with the peripheral nervous between the CNS and the artificial limb by implanting
system or the central nervous system. Interfaces with the PNS electrodes within the peripheral nerves. The CYBERHAND
have been used for two major applications: functional electrical project exploits the combined advantage of more natural motor
stimulation (FES) and the creation and control of artificial commands and sensory communication back to the patient
limbs [81]. FES electrically stimulates the nerves of a patient through afferent nerves. The unique sensory system of the
with a spinal cord injury or neurological disorder to help CYBERHAND has two functions: providing input for a
restore natural sensor-motor function. While interfaces with the controller to enable grasp without requiring user attention, and
PNS are widely used for FES applications [82-84], the focus of supplying information about slippage, texture, and contact to
this section is the application to prosthetic limbs. the wearer [89]. The CYBERHAND was initially tested using
ENG control takes advantage of the remaining neural EMG control but was intended for ENG control [90].
pathways after amputation. Despite its invasiveness, control The purpose of the CYBERHAND project was extended
through the PNS is advantageous because it provides more and fulfilled with the SMARTHAND project. The
natural control of the prosthesis, as opposed to requiring SMARTHAND is a five-fingered, underactuated prosthetic
learned movement by the wearer. Also, unlike surface EMG, hand which was interfaced with the PNS of an amputated
ENG control allows for both the execution of motor commands patient. The hand is controlled by 4 motors and has 16 DOF
and the delivery of sensory feedback to the wearer by directly [91]. The prosthesis successfully communicated motor
stimulating the afferent nerves. commands and delivered sensory feedback [3].
An electrode acting as an interface with the PNS can be
extraneural or intraneural. The most common type of electrode
used in this context is the extraneural cuff electrode, which 4. ELECTRONEUROGRAPHIC CONTROL WITH CNS
surrounds the nerve fascicle and acquires signals from its The most invasive method of ENG control is to implant
exterior. Intrafascicular or intraneural electrodes are placed electrodes into the motor cortex of an amputated patient. As
directly within the tissue of the nerve and result in a higher with peripheral control, brain-machine interfaces within the
signal noise ratio (SNR) than extraneural electrodes. Edell et al. motor cortex or the spinal cord are also used for restoring
[85] created an intraneural electrode intended to be interfaced functionality in patients who do not have an amputated limb
with the PNS. The electrode was implanted in rabbits for a 15 but have a neurological disorder or paralysis. Numerous studies
month trial period, and an eye blink test was used to determine have been performed on primates with the intention of
the electrode’s effectiveness. The results indicated that the extending the research to human subjects. Moritz et al. [92]
electrode could effectively stimulate a small number of nerve used Macana Nemestrina monkeys to show that activity of the
fibers, and the interface would remain stable for at least a 15 neurons in the motor cortex can be used to stimulate muscles,
month period. However, most interest in intrafascicular indicating that goal-directed movements could be restored in a
electrodes has been given to longitudinal intrafascicular paralyzed arm. Similarly, Ethier et al. [93] developed a FES
electrodes (LIFEs) or thin-film longitudinal intrafasicular system in primates which is controlled by microelectrodes
electrodes (tfLIFEs), which provide a good tradeoff between implanted within the brain. Using the recordings from the
invasiveness and functionality; they deliver precise information neurons, predictions for the intended movement of paralyzed
from the CNS and are capable of neural stimulation on a muscles were made and used to control stimulation in those
diverse set of nerves [86]. One type of LIFE, a self-opening paralyzed muscles. This method allows the primate to bypass
intrafascicular electrode (SELINE), was developed by the injured spinal cord and regain control of the paralyzed
Carpaneto et al. [87]. The SELINE is inserted into an existing muscles. The most recent, successful application of this type of
hole made by a needle and is partially extracted so that the control was done by Collinger et al. [94], who implanted
wings of the electrode can open through the nerve tissue. electrodes into the motor cortex of a 52-year-old woman with
Results of simulations showed no failure of the electrode in the tetraplegia. The woman was trained for 13 weeks to operate a 7
main body or the wings. This electrode provides both high DOF prosthetic arm. She achieved a mean success rate of
selectivity and an efficient anchoring system. 91.6% on reach tasks, and over the course of the 13 weeks, her
The first successful demonstration of ENG to control an time to complete tasks dropped from 148 seconds to 112
artificial arm was done by Dhillon and Horch [88].Using LIFEs seconds, and her path efficiency increased. The results indicate
allows for a connection with phantom limb sensations. When that paralyzed or amputated patients may be able to regain
these sensations are used for motor control, it is possible for an some functionality in their limbs and will be able to perform

6 Copyright © 2013 by ASME

ADLs, including reaching, orientation, and hand placement algorithms are a critical point of development for ENG control.
tasks. There is a larger computational burden because the PNS signal
spikes are more difficult to observe than EMG signals.
5. CONCLUSIONS Decoding algorithms must also be capable of providing reliable
Despite their popularity, myoelectric hands are not optimal efferent and afferent communication. Finally, the
for creating a realistic, natural, and functional prosthetic hand. implementation of ENG control currently lacks long-term
The most distinct advantage of surface myoelectric control is clinical studies, which will be a critical hurdle in their success
that it does not require an invasive procedure. The trade off is as a useful method of prosthetic limb control [95].
that signals from the nervous system must be obtained through
muscle contractions, resulting in unnatural movement. A patient ACKNOWLEDGMENTS
with a myoelectric prosthesis must learn a set of unnatural This work was supported by the Graduate Fellowship and
movements to operate it successfully. If only two or three EMG CH Foundation Fellowship, Texas Tech University, Lubbock,
channels are used for control, excess concentration is not a TX, United States.
large concern, but when more channels are included for
increasing the reliability of the classification system, the REFERENCES
prosthesis can become difficult to use. Consequently, [1] Massa, B., Roccella, S., Carrozzza, M.C., Dario, P, 2002,
amputated patients will often choose a simpler and less “Design and Development of an Underactuated Prosthetic
functional prosthesis. However, the main drawback of Hand,” 2002 IEEE International Conference on Robotics
myoelectric hands is their inability to provide sensory feedback
and Automation, Washington, DC, 4, pp. 3374-3379. 
to the user. Sensory feedback is highly important for effective
control, and in its absence, the wearer must rely heavily on [2] Ozawa, R., Hashirii, K., Kobayashi, H., 2009, “Design and
visual cues to determine whether an object will slip or break Control of Underactuated Tendon-Driven Mechanisms,”
within a grasp. This is problematic because it requires 2009 IEEE International Conference on Robotics and
concentration from the user. Automation, Kobe, Japan, pp. 1522-1527.  
Rosahl [95] presents several points regarding the ethics of [3] Cipriani, C., Controzzi, M., Vecchi, F., Carrozza, M. C.,
implantable electrodes. Interfaces with the PNS provide several 2009, “Progress towards the Development of the
distinct advantages which cannot be gained using EMG
SmartHand Transradial Prosthesis,” 2009 11th
control. These interfaces have the potential to provide an
artificial limb which functions continuously without attention International Conference on Rehabilitation Robotics,
from the user and which does not require the user to alter Kyoto, Japan, pp. 682-687. 
normal behavior. Additionally, ENG control has the potential to [4] O’Toole, K. T., McGrath, M. M., Hatchett, D. W., 2007,
restore the natural physical functionality of disabled people; “Transient Characterisation and Analysis of Shape
these interfaces have succeeded in restoring neural function Memory Alloy Wire Bundles for the Actuation of Finger
where all other methods have failed [96]. There is some
Joints in Prosthesis Design,” Mechanika, 6(68), pp. 65-69. 
concern that interfaces with the nervous system (primarily the
CNS) could eventually be used to enhance human capabilities [5] Andianesis, K. and Tzes, A., 2008, “Design on an
beyond normal levels, but the main opposition to intrafascicular Anthropomorphic Prosthetic Hand Driven by Shape
electrodes is that they require an invasive surgery and there are Memory Alloy Actuators,” 2nd Biennial IEEE/RAS-EMBS
not sufficient clinical studies to determine how the implanted International Conference on Robotics and
electrodes will affect the body over time [97]. Additionally, if Biomechatronics, Scottsdale, AZ, pp. 517-522. 
the electrodes no longer worked after several years, another [6] O’Toole, K. T., McGrath, M. M., Coyle, E., 2009,
invasive surgery would be required to replace them. However,
“Analysis and Evaluation of the Dynamic Performance of
the ethical concerns associated with ENG interfaces with the
PNS are not significantly different than those related to medical SMA Actuators for Prosthetic Hand Design,” Journal of
procedures already in practice [95]. Materials Engineering and Performance, 18, pp. 781-786. 
Cortical implants are advantageous because they are [7] El Kady, A. M., Mahfouz, A. E., Taher, M. F., 2010,
connected directly to the brain, but they are generally too “Mechanical Design of an Anthropomorphic Prosthetic
invasive to be practical for amputated patients. Less invasive Hand for Shape Memory Alloy Actuation,” 2010 5th Cairo
cortical procedures do exist, but they do not provide natural International Biomedical Engineering Conference, Cairo,
control and are still limited in terms of their DOF [47].
Egypt, pp. 86-89. 
The use of peripheral interfaces seems to be a logical
evolution for prosthetic hand control, since it provides the best [8] El Kady, A. and Taher, M. F., 2011, “A Shape Memory
compromise between invasiveness and natural control [47]. Alloy Actuated Anthropomorphic Prosthetic Hand: Initial
However, there are many challenges to be overcome before this Experiments, 2011 1st Middle East Conference on
control method can be effectively implemented. First, decoding

7 Copyright © 2013 by ASME

 Biomedical Engineering, Sharjah, United Arab Emirates, [20] Paciga, J. E., Gibson, D. A., Gillepsie, R., Scott, R. N.,
pp. 41-44.  1980, “Clinical Evaluation of UNB 3-State Myoelectric
[9] Wiste, T. E., Dalley, S. A., Withrow, T. J., Goldfarb, M., Control for Arm Prostheses,” Bulletin of Prosthetics
2009, “Design of a Multifunctional Anthropomorphic Research, 17(2), pp. 21-33. 
Prosthetic Hand with Extrinsic Actuation,” 2009 IEEE 11th [21] Shannon, G. F., 1979, “A Myoelectrically-Controlled
International Conference on Rehabilitation Robotics, Prosthesis with Sensory Feedback,” Medical and
Kyoto, Japan, 14(6), pp. 699-706.  Biological Engineering and Computing, pp. 73-80. 
[10] Farry, K. A., Walker, I. D., Baraniuk, R. G., 1996, [22] Lewis, E. A., Sheredos, C. R., Sowell, T. T., Houston, V.
“Myoelectric Teleoperation of a Complex Robotic Hand,” L., 1975, “Clinical Application Study of Externally
IEEE Transactions on Robotics and Automation, 12(5), pp. Powered Upper-Limb Prosthetics Systems: the VA Elbow,
775-788.  the VA Hand, and the VA/NU Myoelectric Hand
[11] Huang, H., and Chen, C., 1999, “Development of a Systems,” Bulletin of Prosthetics Research, 10(24), pp. 51-
Myoelectric Discrimination System for a Multi-Degree 136. 
Prosthetic Hand,” 1999 IEEE International Conference on [23] Prior, R. E., Lyman, J., Case, P. A., Scott, C. M., 1976,
Robotics and Automation, Detroit, MI, 3, pp. 2392-2397.  “Supplemental Sensory Feedback for the VA/NU
[12] Huang, H., Liu, Y., Wong, C., 2003, “Automatic EMG Myoelectric Hand Background and Preliminary Designs,”
Feature Evaluation for Controlling a Prosthetic Hand Bulletin of Prosthetics Research, pp. 170-191. 
Using a Supervised Feature Mining Method: An Intelligent [24] Scott, R. N., Brittain, R. H., Caldwell, R. R., Cameron, A.
Approach,” 2003 IEEE International Conference on B., Dunfield, V. A., 1980, “Sensory-Feedback System
Robotics and Automation, Taipei, Taiwan, 1, pp. 220-225.  Compatible with Myoelectric Control,” Medical and
[13] Geethanjali, P., Ray, K. K., Shanmuganathan, P. V., 2009, Biological Engineering and Computing, 18, pp. 65-69. 
“Actuation of Prosthetic Drive Using EMG Signal,” 2009 [25] Patterson, P. E. and Katz, J. A., 1988, “A Sensory
IEEE Region 10 Conference, Singapore, Singapore, pp. 1- Feedback System for Grasping Pressure in a Myoelectric
5.  Hand,” IEEE Engineering in Medicine and Biology Society
[14] Dalley, S. A., Wiste, T. E., Varol, H. A., Goldfarb, M., 10th Annual International Conference, New Orleans,
2010, “A Multigrasp Prosthesis for Transradial Louisiana, 4, pp. 1564-1565. 
Amputees,” 32nd Annual International Conference of the [26] Patterson, P. E. and Katz, J. A., 1992, “Design and
IEEE EMBS, Buenos Aires, Argentina, pp. 5062-5065.  Evaluation of a Sensory Feedback System That Provides
[15] Cranny, A., Cotton, D. P. J., Chappell, P. H., Beeby, S. P., Grasping Pressure in a Myoelectric Hand,” Journal of
White, N. M., 2005, “Thick-film Force and Slip Sensors Rehabilitation Research and Development, 29(1), pp. 1-8. 
for a Prosthetic Hand,” Sensors and Actuators, 123, pp. [27] Mingrino, A., Bucci, A., Magni, R., Dario, P., 1994,
162-171.  “Slippage Control in Hand Prostheses by Sensing Grasping
[16] Cranny, A., Cotton, D. P. J., Chappell, P. H., Beeby, S. P., Forces and Sliding Motion,” Proceedings of the
White, N. M., 2005, “Thick-film Force, Slip and IEEE/RSJ/GI International Conference on Intelligent
Temperature Sensors for a Prosthetic Hand,” Measurement Robotics and Systems, 3, pp. 1803-1809. 
Science and Technology, 16, pp. 931-941.  [28] Mingrino, A., Dario, P., Sabatini, A., 1992, “A Hand
[17] Muridan, N., Chappell, P. H., Cranny, A., White, N. M., Prosthesis with Slippage Control by Tactile Sensors,”
2010, “Texture Sensor for a Prosthetic Hand,” Procedia Proceedings of the 15th Annual International Conference
Engineering, 5, pp. 605-608.  of the IEEE Engineering in Medicine and Biology Society,
[18] Anugolu, M., Potluri, C., Ilyas, A., Kumar, P., Chiu, S., pp. 1276-1277. 
Devine, N., Urfer, A., Schoen, M. P., 2011, “A Review on [29] Dario, P., Lazzarini, R., Magni, R.,Oh, S. R., 1996, “An
Sensory Feedback for SEMG Based Prosthetic Hands,” Integrated Miniature Fingertip Sensor,” Seventh
2011 International Conference on Artificial Intelligence, International Symposium on Micro Machine and Human
Las Vegas, Nevada.  Science, pp. 91-97. 
[19] Prior, R. E., Lyman, J., 1975, “Electrocutaneous Feedback [30] Ito, K., Pecson, M., Luo, Z., Yamakita, M., Kato, A.,
for Artificial Limbs Summary Progress Report February 1, Aoyama, T., Ito, M., 1994, “Compliance Control of an
1974, through July 31, 1975.”  EMG-controlled Prosthetic Forearm Using Ultrasonic
Motors,” Proceedings of the IEEE/RSJ/GI International

8 Copyright © 2013 by ASME

 Conference on Intelligent Robots and Systems, 3, pp. [41] Kato, R., Yokoi, H., Arieta, A., H, Yu, W., Arai, T., 2009,
1816-1823.  “Mutual Adaptation Among Man and Machine by Using f-
[31] Yoshida, M. and Sasaki, Y., 2001, “Sensory Feedback MRI Analysis,” Robotics and Autonomous Systems, 57,
System for Prosthetic Hand by Using Interferential pp. 161-166. 
Current,” Proceedings of the 23rd Annual International [42] Liu, X., Zhou, R., Yang, L., Li, g., 2009, “Performance of
Conference of the IEEE Engineering in Medicine and Various EMG Features in Identifying Arm Movements for
Biology Society, 3, pp. 1431-1432.  Control of Multifunctional Prostheses,” IEEE Youth
[32] Sasaki, Y., Nakayama, Y., Yoshida, M., 2002, “Sensory Conference on Information, Computing, and
Feedback System Using Interferential Current for EMG Telecommunication, Beijing, China, pp. 287-290. 
Prosthetic Hand,” Second Joint EMBS/BMES Conference, [43] Li, G. and Kuiken, T., 2009, “EMG Pattern Recognition
Houston, Texas, 3, pp. 2402-2403.  Control of Multifunctional Prostheses by Transradial
[33] Cotton, D. P. J., Chappell, P. H., Cranny, A., White, N. Amputees,” 31st Annual International Conference of the
M., Beeby, S. P., 2007, “A Novel Thick-Film Piezoelectric IEEE EMBS, Minneapolis, Minnesota, pp. 6914-6917. 
Slip Sensory for a Prosthetic Hand,” IEEE Sensors [44] Huang, H. and Chiang, C., 2000, “DSP-based Controller
Journal, 7(5), pp. 752-761.  for a Multi-Degree Prosthetic Hand,” 2000 IEEE
[34] Engeberg, E. D. and Meek, S. G., 2008, “Adaptive Object International Conference on Robotics and Automation,
Slip Prevention for Prosthetic Hands through Proportional- San Francisco, California, 2, pp. 1378-1383. 
Derivative Shear Force Feedback,” 2008 IEEE/RSJ [45] Zecca, M., Micera, S., Carrozza, M. C., Dario, P., 2002,
International Conference on Intelligent Robots and “Control of Multifunctional Prosthetic Hands by
Systems, Nice, France, pp. 1940-1945.  Processing the Electromyographic Signal,” Critical
[35] Engeberg, E. D. and Meek, S. G., 2009, “Backstepping Reviews in Biomedical Engineering, 30(4-6), pp. 459-485. 
and Sliding Mode Control Hybridized for a Prosthetic [46] Arvetti, M., Gini, G., Folgheraiter, M., 2007,
Hand,” IEEE Transactions on Neural Systems and “Classification of EMG Signals Through Wavelet Analysis
Rehabilitation Engineering, 17(1), pp. 70-79.  and Neural Networks for Controlling an Active Hand
[36] Engeberg, E. D., 2010, “Biomimetic Sliding Mode Control Prosthesis,” 2007 IEEE 10th International Conference on
of a Prosthetic Hand,” Proceedings of the 2010 IEEE RAS Rehabilitation Robotics, Noordwijk, The Netherlands, pp.
& EMBS International Conference on Biomedical Robotics 531-536. 
and Biomechatronics, pp. 343-348.  [47] Micera, S., Carpaneto, J., Raspopovic, S., 2010, “Control
[37] Engeberg, E. D. and Meek, S. G., 2011, “Adaptive Sliding of Hand Prostheses Using Peripheral Information,” IEEE
Mode Control of Grasped Object Slip for Prosthetic Reviews in Biomedical Engineering, 3, pp. 48-68. 
Hands,” 2011 IEEE/RSJ International Conference on [48] Tenore, F., Ramos, A., Fahmy, A., Acharya, S., Etienne-
Intelligent Robots and Systems, San Francisco, California, Cummings, R., Thakor, N. V., 2007, “Towards the Control
pp. 4174-4179.  of Individual Fingers of a Prosthetic Hand Using Surface
[38] Zhou, P. , Lowery, M. L., Englehart, K. B., Huang, H., EMG Signals,” 2009 Annual International Conference of
Li, G., Hargrove, L., Dewald, J. P. and Kuiken, T. A., the IEEE EMBS, Lyon, France, pp. 6146-6149. 
2007, Decoding a new neural machine interface for control [49] Chu, J., Moon, I, Mun, M., 2005, “A Real-Time EMG
of artificial limbs. Journal of Neurophysiology, 98(5), pp. Pattern Recognition Based on Linear-Nonlinear Feature
2974-2982. Projection for Multifunctional Myoelectric Hand,” IEEE
[39] Kuiken, T. A., Miller, L. A., Lipschutz, R. D., Lock, B., Transactions on Biomedical Engineering, 53(11), pp.
Stubblefield, K. , Marasco, P. , Zhou, P., Dumanian, G., 2232-2239. 
2007, Targeted Reinnervation for Enhanced Prosthetic Arm [50] Kurzynski, M., Woloszynski, T., Wolczowski, A., 2010,
Function in Woman with a Proximal Amputation: a case “A Multiclassifier System with Dynamic Ensemble
study. Lancet, 369(9558), pp. 371-380. Selection Applied to Recognition of EMG Signals for the
[40] Kuiken, T. A., Marasco, P. D., Lock, B. A., Harden, R. N., Control of Bio-prosthetic Hand,” 2011 11th International
Dewald, J. P., 2007, Redirection of Cutaneous Sensation Conference on Control, Automation and Systems, pp. 382-
from the Hand to the Chest Skin of Human Amputees with 386. 
Targeted Reinnervation. Proceedings of the National [51] Khandan, H. and Lucasa, C., 2008, “Tactile Based Fuzzy-
Academy of Science, USA, 104(50), pp. 20061-20066. Bayesian Controller for Prosthetic Hand,” 5th International

9 Copyright © 2013 by ASME

 Symposium on Mechatronics and Its Applications, [63] Zajdlik, J., 2006, “The Preliminary Design and Motion
Amman, Jordan, pp. 1-4.  Control of a Five-fingered Prosthetic Hand,” Proceedings
[52] Kondo, T., Amagi, O., Nozawa, T., 2008, “Proposal of of the International Conference on Intelligent Engineering
Systems, pp. 202-206.
Anticipatory Pattern Recognition for EMG Prosthetic
[64] Wu, Z., Niu, H., Li, D., Pu, F., Fan, Y., 2010, “A virtual
Hand Control,” International Conference on Systems, sEMG Prosthetic Hand Development System Based on
Man, and Cybernetics, 2008, pp. 897-902.  LabVIEW and PCI-1710HG A/D Card,” 2010 3rd
[53] Theodoris, S. and Koutroumbas, K., 2009, “Pattern International Conference on Biomedical Engineering and
Recognition,” Elsevier, Inc. Burlington, MA.  Informatics, 4, pp. 1743-1745.
[54] Potluri, C., Kumar, P., Anugolu, M., Chiu, S., Urfer, A., [65] Morita, S., Shibata, K., Zheng, X. Z., Ito, K., 2000,
Schoen, M. P., 2010, “sEMG Based Fuzzy Control “Prosthetic Hand Control Based on Torque Estimation
from EMG Signals,” 2000 IEEE/RSJ International
Strategy with ANFIS Path Planning for Prosthetic Hand,”
Conference on Intelligent Robots and Systems, 1, pp. 389-
Proceedings if the 2010 3rd IEEE RAS & EMBS 394.
International Conference on Biomedical Robotics and [66] Morita, S., Kondo, T., Ito, K., 2001, “Estimation of
Mechatronics, Tokyo, Japan.  Forearm Movement from EMG Signal and Application to
[55] Weir, R. F. and Ajiboye, A. B., 2003, “A Multifunction Prosthetic Hand Control,” 2001 IEEE International
Prosthesis Controller Based on Fuzzy-Logic Techniques,” Conference on Robotics and Automation. Seoul, Korea, 4,
25th Annual International Conference of the IEEE EMBS, pp. 3692-3697.
[67] Ma, S. and Moussa, M., 2008, “An Intelligent Grasping
Cancun, Mexico, 2, pp. 1678-1681. 
System for Applications in Prosthetic Hands,” Cairo
[56] Momen, K., Krishnan, S., Chau, T., 2007, “Real-time International Biomedical Engineering Conference, 2008,
Classification of Forearm Electromyographic Signals pp. 1-5.
Corresponding to User-Selected Intentional Movements [68] Wang, R. C., Li, F., Wu, M., Wang, J. Z., Jiang, L., Liu, H.,
for Multifunction Prosthesis Control,” IEEE Transactions 2006, “Multi-degree Prosthetic Hand Control Using a New
on Neural Systems and Rehabilitation Engineering, 15(4), BP Neural Network,” Proceedings of the 13th International
pp. 535-542.  Conference on Neural Information Processing, 3, pp. 589-
595.
[57] Khezri, M., Jahed, M., Sadati, N., 2007, “Neuro-Fuzzy
[69] Asres, A. Dou, H., Zhou, Z., Zhang, Y., Zhu, S., 1996, “A
Surface EMG Pattern Recognition for Multifunctional Combination of AR and Neural Network Technique for
Hand Prosthesis Control,” IEEE International Symposium EMG Pattern Identification,” 18th Annual International
on Industrial Electronics, 2007, pp. 269-274.  conference of the IEEE Engineering in Medicine and
[58] Saridis, G. N., and Gootee, T. P., 1982, “EMG Pattern Biology Society, Amsterdam, 4, pp. 1464-1465.
Analysis and Classification for a Prosthetic Arm,” IEEE [70] Zhao, J., Xie, Z., Jiang, L, Cai, H., Liu, H., Hirzinger, G.,
2005, “Levenberg-Marquardt Based Neural Network
Transactions on Biomedical Engineering, 29(6), pp. 403-
Control for a Five-fingered Prosthetic Hand,” 2005 IEEE
412.  International Conference on Robotics and Automation,
[59] Crowder, R. M., 1991, “Local Actuation of Multijointed Barcelona, Spain, pp. 4482-4487.
Robotic Fingers,” International Conference on Control, [71] Zhao, J., Jiang, L, Shi, S., Cai, H., Liu, H., Hirzinger, G.,
1991,   2006, “A Five-fingered Underactuated Prosthetic Hand
[60] Tsuji, T., Fukuda, O., Shigeyoshi, H., Kaneko, M., 2000, System,” 2006 IEEE International Conference on
“Bio-mimetic Impedance Control of an EMG-controlled Mechatronics and Automation, Luoyang, China, pp. 1453-
Prosthetic Hand,” 2000 IEEE/RSJ International 1458.
Conference on Intelligent Robots and Systems, 1, pp. 377- [72] Zhao, J., Xie, Z., Jiang, L., Cai, H., Liu, H., Hirzinger, G.,
382. 2006, “EMG Control for a Five-fingered Prosthetic Hand
[61] Secco, E. L. and Magenes, G., 2002, “A Feedforward based on Wavelet Transform and Autoregressive Model,”
Neural Network Controlling the Movement of a 3-DOF 2006 IEEE International Conference on Mechatronics and
Finger,” IEEE Transactions on Systems, Man, and Automation, Luoyang, China, pp. 1097-1102.
Cybernetics--Part A: Systems and Humans, 32(3), pp. 437- [73] Wang, J. Z., Wang, R. C., Li, F., Jiang, M. W., Jin, D. W.,
445. 2005, “EMG Signal Classification for Myoelectric
[62] Tanaka, Y., Noda, S., Tsuji, T., Maruishi, M., Fukuda, O., Teleoperating a Dexterous Robot Hand,” 2005 IEEE
2004, “A Virtual Prosthetic Hand Using EMG Signals for Engineering in Medicine and Biology 27th Annual
fMRI Measurements,” 2004 IEEE/RSJ International Conference, Shanghai, China, 6, pp. 5931-5933.
Conference on Intelligent Robots and Systems. Sendai, [74] Herle, S., Man, S., Lazea, Gh., Robotin, R., Marcu, C.,
Japan, 3, pp. 2954-2959. 2010, “Myoelectrical Signal Classification for the

10 Copyright © 2013 by ASME

 Hierarchical Control of a Human Hand Prosthesis,” [86] Szekely, I., Stoianovici, V. C., Machedon, M., Nedelcu, A.,
Proceedings of the 2010 IEEE International Conference on 2009, “Wireless Neural Signal Transmission in Biomedical
Automation, Quality and Testing, Robotics, 2, pp. 1-6. Prosthetic Systems,” 15th International Symposium for
[75] Nishikawa, D., Yu, W., Yokoi, H., Kakazu, Y., 2001, “On- Design and Technology of Electronics Packages, pp. 63-
line Learning Method for EMG Prosthetic Hand Control,” 68.
Electronics and Communications in Japan, Part 3, 84(10), [87] Carpaneto, J., Cutrone, A., Bossi, S, Sergi, P., Citi, L.,
pp. 1510-1519. Rigosa, J., Rossini, P. M., Micera, S., 2011, “Activities on
[76] Ito, K., Tsukamoto, M., Kondo, T., 2008, “Discrimination PNS Neural Interfaces for the Control of Hand
of Intended Movements Based on Nonstationary EMG for Prostheses,” 33rd Annual International Conference of the
a Prosthetic Hand Control,” 3rd International Symposium IEEE EMBS, Boston, Massachusetts, pp. 4637-4640.
on Communications, Control and Signal Processing, 2008, [88] Dhillon, G. S. and Horch, K. W., 2005, “Direct Neural
pp. 14-19. Sensory Feedback and Control of a Prosthetic Arm,” IEEE
[77] Ito, K., Ibe, A., Gouko, M., 2009, “Discrimination of Transactions on Neural Systems and Rehabilitation
Intended Motions for Prosthetic Hands Using Engineering, 13(4), pp. 468-472.
Nonstationary EMG,” 35th Annual Conference of IEEE [89] Dario, P., Carrozza, M. C., Guglielmelli, E., Laschi, C.,
Industrial Electronics, 2009, pp. 2050-2055. Menciassi, A., Micera, S., Vecchi, F., 2005, “Robotics as a
[78] Karimi, M., Pourghassem, H., Shahgholian, G., 2011, “A Future and Emerging Technology,” IEEE Robotics and
Novel Prosthetic Hand Control Approach Based on Automation Magazine, pp. 29-45.
Genetic Algorithm and Wavelet Transform Features,” 2011 [90] Carrozza, M. C., Micera, S., Massa, B., Zecca, M.,
IEEE 7th International Colloquium on Signal Processing Lazzarini, R., Canelli, N., Dario, P., 2004, “The
and its Applications, pp. 287-292. Development of a Novel Biomechatronic Hand--Ongoing
[79] Kyberd, P. J., Holland, O. E., Chappell, P. H., Smith, S., Research and Preliminary Results,” IEEE/ASME
Tregidgo, R., Bagwell, P. J., Snaith, M., 1995, MARCUS: Transactions on Mechatronics, 7(2), pp. 108-114.
a two degree of freedom hand prosthesis with hierarchical [91] Controzzi, M., Cipriani, C., Carrozza, M. C., 2008,
grip control. IEEE Transactions on Rehabilitation “Mechatronic Design of a Transradial Cybernetic Hand,”
Engineering, 3, pp. 70–76. 2008 IEEE/RSJ International Conference on Intelligent
[80] Cotton, D. P. J., Cranny, A., Chappell, P. H., White, N. M., Robots and Systems, Nice, France.
Beeby, S. P., 2007, Control Strategies for a Multiple [92] Moritz, C. T., Perlmutter, S. I., Fetz, E. E., 2008, “Direct
Degree of Freedom Prosthetic Hand. Measurement & Control of Paralysed Muscles by Cortical Neurons,”
Control, 40, pp. 24–27. Nature, 456, pp. 639-642.
[81] Navarro, X., Krueger, T. B., Lago, N., Micera, S., Stieglitz, [93] Ethier, C., Oby, E. R., Bauman, M. J., Miller, L. E., 2012,
T., Dario, P., 2005, “A Critical Review of Interfaces with “Restoration of Grasp Following Paralysis through Brain-
the Peripheral Nervous System for the Control of Controlled Stimulation in Muscles,” Nature, 485, pp. 368-
Neuroprostheses and Hybrid Bionic Systems,” Journal of 371.
the Peripheral Nervous System, 10, pp. 229-258. [94] Collinger, J. L., Wodlinger, B., Downey, J. E., Wang, W.,
[82] Crago, P. E., Chizeck, H. J., Neuman, M. R., Hambrecht, F. Tyler-Kabara, E. C., Weber, D. J., McMorland, A. J. C.,
T., 1986, “Sensors for Use with Functional Neuromuscular Velliste, M., Boninger, M. L., Schwartz, A. B., 2012,
Stimulation,” IEEE Transactions on Biomedical “High-performance Neuroprosthetic Control by an
Engineering, BME-33(1), pp. 256-268. Individual with Tetraplegia,”
[83] Hincapie, J. G. and Kirsch, R. F., 2009, “Feasibility of http://dx.doi.org/10.1016/S0140-6736(12)61816-9.
EMG-based Neural Network Controller for an Upper [95] Rosahl, S. K., 2009, “Neuroelectronic Interfaces with the
Extremity Neuroprosthesis,” IEEE Transactions on Neural Central Nervous System—Ethical Issues.”
Systems and Rehabilitation Engineering, 17(1), pp. 80-90. [96] Tsoli, A. and Jenkins, O. C., 2010, “Robot Grasping for
[84] Rossini, L. and Rossini, P. M., 2010, “Combining ENG Prosthetic Applications,” Robotics Research, 66, 1-12.
and EEG Integrated Analysis for Better Sensitivity and [97] Micera, S., Citi, L., Rigosa, J., Carpaneto, J., Raspopovic,
Specificity of Neuroprothesis Operations,” 32nd Annual S., Di Pino, G., Rossini, L., Yoshida, K., Denaro, L., Dario,
International Conference of the IEEE EMBS, Buenos Pl, Rossini, P. M., 2010, “Decoding Information from
Aires, Argentina, pp. 134-137. Neural Signals Recorded Using Intraneural Electrodes:
[85] Edell, D. J., Riso, R., Devaney, L., Larsen, B., Koris, M., Toward the Development of a Neurocontrolled Hand
DeLorenzo, D., 1996, “Intraneural Microstimulation for Prosthesis,” Proceedings of the IEEE, 98(3), 407-417.
Enhanced Prosthetic Control Using a Peripheral Nerve
Interface,” 18th Annual International Conference of the
IEEE Engineering in Medicine and Biology Society,
Amsterdam, 1, pp. 337-338.

11 Copyright © 2013 by ASME

View publication stats