The University of Sheffield
Department of Computer Science

COM3502 Speech Processing

Summary This module aims to demonstrate why computer speech processing is an important and difficult problem, to investigate the representation of speech in the articulatory, acoustic and auditory domains, and to illustrate computational approaches to speech parameter extraction. It examines both the production and perception of speech, taking a multi-disciplinary approach (drawing on linguistics, phonetics, psychoacoustics, etc.). It introduces sufficient digital signal processing (linear systems theory, Fourier transforms) to motivate speech parameter extraction techniques (e.g. pitch and formant tracking). Students should be aware that there are limited places available on this course.
Session Autumn 2023/24
Credits 10
Assessment

This module is assessed by a programming assignment and a formal exam

Lecturer(s) Dr Stefan Goetze
Resources
Aims
  • to describe speech production and perception in humans;
  • to teach computational techniques for analysing speech signals.
Learning Outcomes  By the end of the unit, a student will have acquired:
  • Discuss the basic mechanisms of speech production and perception.
  • Analyse acoustic-phonetic descriptions of the speech signal.
  • Apply signal analysis techniques to speech data. 
  • Apply discrete-time signal processing.
Content Half of the course is devoted to the nature of speech and half to speech signal processing. Topics covered include: speaking, hearing, sounds and symbols, articulatory and acoustic phonetics, phonology, prosody, speech spectra, sampling, fourier transform, linear filters, linear prediction cepstral analysis.
Teaching Method Lectures & scheduled access to a Laboratory for practical work
Feedback
  • Verbal interaction during lectures and labs.
  • Comments attached to the marks given to the formative and summative assignments.