COM4511/6511 Speech Technology

Practical Exercise - Task 1
MFCCs
Speech Technology - COM4511/6511
12th March 2021
1 Introduction
The aim of this task is to learn how to use Python for speech processing. There are several online Python tutorials,
and particularly on numpy and how to use it for signal processing. In this task you will implement a Python
program to convert a speech signal to Mel Frequency Cepstral Coefficients (MFCCs).
2 Infrastructure
A speech audio file from the TIMIT corpus, a Python helper script and example output are provided. These can be
found on MOLE in the folder named task1. It is also available for download at https://www.dcs.shef.ac.uk/∼th/campus_only3 Task
To help you start the task, the provided Python script contains all the necessary steps and a few utility functions
for this task. You are asked to fill in the missing parts indicated by ####.
1. From the link above, download the audio file “SA1.wav” and the Python script “ task1_compute_mfcc.py ”
2. Take a look at the script and find out which parts need to be implemented
3. Convert the complete audio file into a sequence of MFCC vectors using the following baseline configuration.
(a) A pre-emphasis filter with coefficient 0.97
(b) 10ms frame step, 25ms frame length (Hamming window)
(c) 26 Mel-filters
(d) 12 cepstral coefficients (C1-C12, omitting C0)
(e) 22-order liftering.
(f) Cepstral mean and variance normalisation (optional)
(g) Plot different  coefficients and consider the implications of what you see
(h) Adopt any good coding style with meaningful comments. Use functions when possible.
You can (and should) use NumPy and SciPy for basic numerical computing and the FFT/DCT computation, but
not any other packages (i.e. the helper script has provided all the necessary imports). If you are not sure, just raise
your hand and ask.
1
4 Assessment:
Note that for any module assignment full marks will only be obtained for outstanding performance that goes
well beyond the questions asked. The marks allocated for each assignment are 20%. The marks will be assigned according the following general criteria - for every assignment handed in:
1. Fulfilling the basic requirements (5%)
Full marks will be 写COM4511/6511 Speech Technology given to fulfilling the work as described, in source code and results given.
2. Submitting high quality documentation (5%)
Full marks will be given to a write-up that is at the highest standard of technical writing and illustration.
3. Showing good reasoning (5%)
Full marks will be given if the experiments and the outcomes are explained to the best standard.
4. Going beyond what was asked (5%)
Full marks will be given for interesting ideas on how to extend work that are well motivated and described.
In order to report this task as complete 3 elements have to be submitted, in gzup form. The file name should have
the following format:
<lastname>_<firstname>_task1.gz
1. The MFCC encoded reference audio file in text format.
Each line contains one MFCC vector, the values are separated by single space.
2. A compiled Latex document (pdf) that contains 4 plots (matplotlib) of the MFFCs in the following configurations (modification to the baseline) and detailed comments on the variation to the baseline configuration
as outlined above. Also briefly comment on the reasons for the effects.
(a) No Hamming window
(b) 40 MFCCs
(c) 80 Filterbanks, 40 MFCCs
(d) No Pre-emphasis.
3. The completed source code for the baseline configuration as outlined in 3.
 WX：codinghelp