EEE-3207
Digital Signal Processing
Assignment task-1
Obtaining CO2
Analyzing Audio Signals of Distinctive Words Using MATLAB
Objective:
This assignment aims to give you an experience with hands-on audio signal processing, Fourier
analysis, and cross-correlation techniques to differentiate between distinct words based on their
frequency characteristics.
You can use any internet sources, tools, resources etc to complete the project. You will be eligible
for bonus marking if you can detect all 9 Bangla digits (check the rubrics at the bottom).
Task Overview:
Record a short audio clip containing two distinct words: "chaar" and "tiin". Analyze the audio
signals in MATLAB to observe their frequency characteristics and differentiate between the words
based on the following methods:
1. Fourier Analysis to extract distinctive frequency components (Fourier coefficient bins).
2. Cross-Correlation to detect the characteristic vowel sounds "aa" and " within the words.
Steps to follow:
1. Audio Recording
1. Record three audio clips (in MP3 format) where you and your family members distinctly
pronounce the two words "chaar" and tiin". Ensure the recording is clear and free from
background noise.
2. Save the audio file with an appropriate name in the same folder as your .m file.
2. Preprocessing the Audio
1. Import the audio file into MATLAB using audioread() or any function you want!
2. Extract the two words separately from the audio signal (you may use MATLAB's audio
visualization tools to identify time intervals for each word, by plotting or toolboxes, this is
NOT necessary for the grading).
3. You can also normalize the signal for consistency if you want.
3. Fourier Analysis
1. Apply the Fourier Transform to the signal of each word using the Fast Fourier Transform
(FFT). You can also use the manual method instead of FFT. Any method will work.
2. Plot the frequency spectrum of both words.
3. Identify the distinctive frequencies corresponding to the vowel sounds "aa" in "chaar" and
"ii" in "tiin".
4. Discuss in your brief report how the observed frequencies differ and how they can be used
to distinguish between the two words.
4. Cross-Correlation Analysis
1. Record two more audios with “aa” and “ii”
2. Perform cross-correlation using MATLAB's xcorr() function to compare:
o The "aa" segment against the words.
o The "ii" segment against the words.
3. Identify peaks in the correlation values and use them to determine which word is being
analyzed.
4. Explain your results, highlighting the higher correlation values observed.
5. Reporting Results
Document following:
o Fourier spectra of each word.
o Key frequencies associated with the vowels "aa" and "ii" (Such as which hertz).
o Cross-correlation plots showing the detected correlations.
Submission Zip file will contain:
1. MATLAB Code: Submit one or multiple script (.m file) containing your entire analysis.
2. Read Me file: You know how to run your script, we do not. Write a small read me script to
tell me which .m file should I run first and how will I load the audio files and so on.
3. Audio Files: All the audios you and your family/friends recorded!
4. Plots: Include all relevant plots (Fourier spectra, cross-correlation results) with proper labels
and legends.
5. Report: Write a concise report (3-6 pages is enough, you can go over but not encouraged)
detailing your process, observations, and conclusions. Include figure of key MATLAB
outputs.
Grading Rubrics:
You can use only one method (either Fourier or Cross-Correlation) to do this assignment.
Criteria Weight
Audio Recording 10%
Fourier Analysis 30%
Cross-Correlation Analysis 30%
Coding Quality 10%
Report 20%
