Navigation

13M054OPG - Digital Speech Processing and Recognition

Course specification
Course title Digital Speech Processing and Recognition
Acronym 13M054OPG
Study programme Electrical Engineering and Computing
Module
Type of study master academic studies
Lecturer (for classes)
Lecturer/Associate (for practice)
Lecturer/Associate (for OTC)
ESPB 6.0 Status elective
Condition non
The goal Objective of the course is for the students to become capable of applying basic techniques for digital processing and speech recognition.
The outcome Learning outcomes should be for the students to gain the following skills: endpoints detection in the recorded speech sequence, estimation of pitch frequency, design of different speech quantizers,design of speech recognition systems based on spectral, cepstral coefficients or hidden Markov chains.
Contents
URL to the subject page https://automatika.etf.bg.ac.rs/sr/13e054opg
URL to lectures https://teams.microsoft.com/l/team/19%3ashf1lndPz5RMVczUgquDSknqvzsGfhoVe6eOmDqUwF01%40thread.tacv2/conversations?groupId=c480bfab-9bef-4be8-9adc-a22388360494&tenantId=1774ef2e-9c62-478a-8d3a-fd2a495547ba
Contents of lectures Modelling of acoustic waveform, Model of uniform tube, Time-domain speech signal processing, Different techniques of quantizers, Evaluation of quantizers, Homomorphic processing of speech signal, LPC analysis of speech signals, Hidden Markov chains and application in speech recognition.
Contents of exercises Within the course, students have obligation to solve three practical problems: 1. estimation of their pitch frequency and segmentation of their speech, 2. modelling and evaluation of particular quantizer, 3. design of a hidden Markov model. In addition, during computer exercises students also learn how to build a vowel synthesis system, conduct LPC analysis, etc.
Literature
  1. Digital processing of speech signals, L. Rabiner, R. Schafer,Prentice Hall, Englewood, 1979.
  2. Speech Processing, A dynamic and optimization oriented approach,Li Deng, Douglas O'Shaughnessy, Marcel Dekker, 2003.
Number of hours per week during the semester/trimester/year
Lectures Exercises OTC Study and Research Other classes
3 1 1
Methods of teaching lectures, auditory exercises and computer exercises
Knowledge score (maximum points 100)
Pre obligations Points Final exam Points
Activites during lectures 0 Test paper 70
Practical lessons 30 Oral examination 0
Projects
Colloquia 0
Seminars 0