Phonetic signals and their coding
- Department of Circuit Theory
The subject introduces the processing of speech signals. Within the subject students should manage from basic to advanced and modern algorithms of speech analysis, synthesis, coding or enhancement. Further reasonable part is focused on speech recognition, where students will get to know modern and advanced technique in task as small and large vocabulary speech recognition or speaker recognition. Special attention is devoted to usage of classification techniques based on GMM, DTW, HMM, ANN/DNN, WFST, JFA, i-vectrors, etc.
Digital signal processing are supposed as preliminary knowledge.
- Syllabus of lectures:
1. Speech production and perception model, phonetic description of speech
2. Spectral characteristics of speech (DFT, LPC, filter banks)
3. Cepstral reprezentation of speech and possible applications
4. Voice activity detection and speech enahncement.
5. Speech synthesis
6. Speech coding
7. Basic and advanced feature extraction techniques (PCA, LDA)
8. Classification approaches for particular ASR tasks (GMM, HMM, VQ)
9. Modern methods of speaker verification and identification (UBM-GMM, JFA, i-vectors)
10. DTW- and HMM-based speech recognition
11. Continuous speech recognition, language modelling, WFST
12. Adaptation techniques in speech recognition
13. Modern ASR systems based on ANN/DNN, methods of deep learning
- Syllabus of tutorials:
Seminars are organized as common consultations of registered students. The main focus is paid on individual work of students during the semester and a their work on chosen individual topics. The solutions of individual projects are discussed at common consultations.
- Study Objective:
- Study materials:
 Deller, J.R. - Hansen, J.H.L. - Proakis, J.G.: Discrete-time processing of speech signals, New York: IEEE Press 2000, 908 s., ISBN
- Further information:
- No time-table has been prepared for this course
- The course is a part of the following study plans: