DSP Coursework Instructions

Download as pdf or txt
Download as pdf or txt
You are on page 1of 3

Academic Year 2020/21

U21497 Digital Signal Processing

Coursework

Deadline For Submission: 2300 hours (UK time), 17th May 2021

Hand-In Instructions: The report is to be submitted as a pdf through the report


Moodle upload facility. Matlab code to be submitted as a
single text file (.txt) in the Matlab code submission box. Ex-
perimental results to be submitted as a single Matlab mat
file and uploaded to Moodle submission box for the course-
work. This work is to be carried out by individuals. It is not
group work.
Please note that all work will be checked for plagiarism.
Any work that is found to be copied from any source is likely
to be classed as plagiarised and there are strict penalties if
this is the case.

Instructions for completing


assessment: See inside for details.

Examiner: Dr John Chiverton


Mel Frequency Cepstral Coefficients
Mel Frequency Cepstral Coefficients (MFCC)s are widely used for the recognition of speech. These
MFCCs help to highlight some of the important qualities that are present in speech that can help
differentiate speakers and individual words. The computation of MFCCs can be typically split into the
following steps:

1. Windowing

2. Discrete Fourier Transform (DFT)

3. Filter Bank and Logarithmic Computations

4. Discrete Cosine Transform (DCT)

The aim of this coursework is for the these stages to be investigated, described, implemented, applied
and comparisons performed between different settings.
You will need to research this well known topic and understand how Matlab can be used to per-
form the various steps outlined above. You will be given a source of speech samples via a moodle
provided link to demonstrate the results at different stages but also the overall result. Please follow
the instructions that you will be provided to download the speech samples. These speech samples
may differ for each person. So please pay careful attention to this.
The system should be implemented in Matlab and used to process the speech data that you will
be given via moodle.

Technical Report Structure


The report should include an introduction, methodology, results, conclusions and references sections.
The system should be described in detail, using mathematics and associated English language ex-
planations to describe each stage and equation.
The report should be no more than 6 sides of A4 written with a font size of 11. The 6 page limit
is for everything. The short report format means that a cover, table of contents, table of figures are
not necessary. A good number of references from high quality sources should be included. The
references should be formatted so that they do not occupy too much space. A double column format
should be used, e.g. similar to the IEEE paper format available here:
https://www.ieee.org/publications_standards/publications/authors/author_templates.html

Marking Scheme and Timetable


• This coursework contributes 40% of your total mark towards this unit.

• The marks for this coursework are split according:

– 12 marks will be allocated based on your ability to describe mathematically the different
stages of the MFCC computations. In particular, you should describe the steps in English
and provide equations. Warning: The description needs to be in your own words. Similarly
the mathematics should be your work only. Any similarity with another source including
another student’s work will likely to be considered plagiarism for which there are strict
penalties.
– 8 marks will come from the design of the overall system with appropriate explanations.
– 4 marks will come from your results demonstrating the correct functioning of the developed
system.
Hint: Matlab results comparing appropriate designs with less appropriate design values
could be useful here.

2020/21 Page 1 of 2
– 12 marks will come from your results of applying the system to data that has been provided.
– 4 marks will come from your technical report style and presentation.
– Any report longer than the specified 6 pages will only get marked up to the 6 page limit.

• All code will need to be submitted to enable credit for the scores detailed above, along with a sin-
gle .mat file containing the results of processing your system. There will be separate submission
boxes for the report, code and processed data files. Warning: the code needs to be your own
written code. Any similarity with another student’s work will likely to be considered plagiarism
for which there are strict penalties. Code should be submitted in the separate Moodle submis-
sion box as a single text file. Please note, that there will be no marks for the development of
a Graphical User Interface. Furthermore, the developed code should not use any third party
libraries. The use of Matlab functions, such as fft, dct, hamming is acceptable and indeed
desirable but other higher level functions should not form the core of your system design.

The total out of 40 will be multiplied by 2.5 to obtain a percentage for this work.

General Grading Criteria


The following general grading criteria is used as a guide line in deciding, more generally the quality
of the work, for individual categories and as a whole.

Grade Range Commensurate Level of Work


80-100 Excellent work that goes beyond the normal expectations at this level. This can
include work that is of international publishable quality.
70-79 Excellent work that exceeds the usual requirements of the work, including some
elements of novelty and demonstrates creativity in the solution.
60-69 Work that is well written and formatted, meets all the objectives and demon-
strates some relatively deep insight into the work.
50-59 Work that is reasonably well written and formatted and can be considered a sat-
isfactory attempt.
40-49 Work that adequately attempts and addresses the main objectives of the work.
Some attempt at organising and structuring the work.
30-39 Work that is not complete, has some errors.
0-29 No serious attempt.

2020/21 Page 2 of 2

You might also like