The Design and Application of an Acoustic Front-End for Use in Speech Interfaces

Gerber, Christoph (1996) The Design and Application of an Acoustic Front-End for Use in Speech Interfaces. MSc(R) thesis, University of Glasgow.

Abstract

This thesis describes the design, implementation, and application of an acoustic front-end. Such front-ends constitute the core of automatic speech recognition systems. The front-end whose development is reported here has been designed for speaker-independent large vocabulary recognition. The emphasis of this thesis is more one of design than of application. This work exploits the current state-of-the-art in speech recognition research, for example, the use of Hidden Markov Models. It describes the steps taken to build a speaker-independent large vocabulary system from signal processing, through pattern matching, to language modelling. An acoustic front-end can be considered as a multi-stage process, each of which requires the specification of many parameters. Some parameters have fundamental consequences for the ultimate application of the front-end. Therefore, a major part of this thesis is concerned with their analysis and specification. Experiments were carried out to determine the characteristics of individual parameters, the results of which were then used to motivate particular parameter settings. The thesis concludes with some applications that point out, not only the power of the resulting acoustic front-end, but also its limitations.

• Item Type: Thesis (MSc(R))
• Qualification Level: Masters
• Additional Information: Adviser: Keith van Rijsbergen
• Keywords: Computer science
• Date of Award: 1996
• Depositing User: Enlighten Team
• Unique ID: glathesis:1996-76110
• Copyright: Copyright of this thesis is held by the author.
• Date Deposited: 19 Nov 2019 16:38
• Last Modified: 19 Nov 2019 16:38
• URI: https://theses.gla.ac.uk/id/eprint/76110