Scribd
Upload
20 views2 pages

Speech Processing: Noise & Reverberation Effects

This document provides instructions for a speech processing research assignment involving analyzing the effects of noise and reverberation on identifying vowels. Students will analyze vowels in clean and noisy conditions using time domain, magnitude spectrum, short-time Fourier transform, and continuous wavelet transform representations. They will then explain the effects of noise and reverberation on vowel formants and fundamental frequency based on their analysis.

Uploaded by

RAGHAV BHATIA
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
20 views2 pages

Speech Processing: Noise & Reverberation Effects

This document provides instructions for a speech processing research assignment involving analyzing the effects of noise and reverberation on identifying vowels. Students will analyze vowels in clean and noisy conditions using time domain, magnitude spectrum, short-time Fourier transform, and continuous wavelet transform representations. They will then explain the effects of noise and reverberation on vowel formants and fundamental frequency based on their analysis.

Uploaded by

RAGHAV BHATIA
Copyright
© All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd

BIRLA INSTITUTE OF TECHNOLOGY AND SCIENCE, PILANI

FIRST SEMESTER 2019-2020


EEE G613 - Advanced Digital Signal Processing

Research Oriented MATLAB Assignment: Speech Processing

Effects of noise and reverberation on vowel identification

Project Submission: 18th September 2019

Total Marks: 20

Note: This assignment should be done in a group of two students. Each group must submit their
project report (in a PDF format) with MATLAB codes (at the end) for the following questions.
Please make sure you add a title, axis labels, x-axis limit and y-axis limit, grid on, and legend (if
required) to each of your figures.

You have been given three different vowels: /i/ (“beet”), /ɑ/ (“father”) and /æ/ (“bat”). The
fundamental and formant frequencies for each vowel are shown below in the Table.

Vowel Fundamental First Second Third Fourth Fifth


Frequency Formant Formant Formant Formant Formant
(F0) (F1) (F2) (F3) (F4) (F5)

/i/ 125 250 2250 3050 3350 3850

/ɑ/ 200 750 1050 2950 3350 3850

/æ/ 200 750 1450 2450 3350 3850

The .wav file for each vowel is provided to you and is created based on the values shown in the
above table.

To obtain noisy vowel, add white Gaussian noise to it. Select two different SNR conditions. Overall,
you will have three different versions of the same vowel: clean vowel and two noisy vowels.

Q1) Plot and analyze these vowels (i.e., 3 vowels x 3 conditions) individually in the time-domain
representation.

Q2) Plot and analyze these vowels using the magnitude spectrum. Can we estimate F0, F1, F2
and F3 using this approach? If yes, then explain otherwise why not?

Q3) Plot, analyze and estimate F0, F1, F2 and F3 of these vowels using the short-time fourier
transform. Make your observations.

Q4) Plot, analyze and estimate F0, F1, F2 and F3 of these vowels using the continuous wavelet
transform. Comment on your observations.
Q5) Explain the effect of noise on vowels’ F0 and formants (especially to F1 and F2) based on the
findings you get from above representations (Q3 and Q4).

To obtain reverberant vowel, convolve the original vowel with the reverberant impulse responses
(see the attached [Link]; [Link] and [Link]).

Q6) Compute the magnitude spectrum and short-time fourier transform of the reverberant vowels
(3 vowels x 4 conditions). Comment on the feature extraction (F0 and F1, F2, F3) of these vowels.

Q7) Repeat Q6) using the continuous wavelet transform.

Q8) Explain the effect of reverberation on vowels’ F0 and formants (especially to F1 and F2) based
on the findings you get from above representations (Q6 and Q7).

You might also like