Basic Information
- Course Code: EE 679
- Course Name: Speech Processing
- Course Offered In: Autumn 2022
- Instructors: Prof. Preeti Rao
- Prerequisites: Hard pre req: EE 229, soft pre requisite: EE 338
- Difficulty (on a scale of 5): 4
Course Content
- Speech production (vocal chords, vocal tract), kinds of sounds, voicing, articulation, modelling it using Source-filter model
- Time domain and frequency domain (spectral) methods of speech analysis
- Linear predictive analysis
- Cepstral analysis
- Basic ASR (automatic speech recognition)
- Speech compression (kinds of encoding, etc)
Feedback on Lectures
Lectures were highly engaging. Ma’am’s teaching was very good. The pace was quick and one had to be attentive to follow. Lectures were well structured. The link between theory and real life examples/applications was very well presented.
Feedback on Evaluations
There were weekly Moodle quizzes during one of the three lectures based on the last week’s content which made up 15% of the total grade
- Midsem: 20% subjective
- Endsem 25% subjective + 10% objective (Moodle quiz)
- Programming Assignments (3 of them): 30%
Midsem and Endsem had some targetted questions based on a very small part of the entire syllabus, so if you knew it you got the marks, otherwise lost them. The weekly Moodle quizzes, on the other hand, were more towards covering the maximum part of the course, and were often times tricky, and required thorough understanding of the week’s content.
Study Material and Resources
Reference books were mentioned (authors:)
- O’Shaughnessy
- Rabiner and Schafer
Slides and lecture notes were posted on Moodle
Follow-up Courses
- CS 753: Automatic Speech Recognition
- EE 779: Advanced topics in Signal Processing
Final Takeaway
Course is very good if you are interested in signal processing. More on the practical side than on the theoretical. Your Python skills will also improve because of the coding assignments. You will get a feel for WHY we study so much stuff in EE 229 and EE 338.