Finite-state vector quantization over noisy channels and its application to LSP parameters

Yunus Hussain, Nariman Farvardin

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In this paper we describe a noisy channel finite-state vector quantizer (FSVQ) for quantizing speech LSP parameters. First we show that under noiseless channel conditions, the ordinary FSVQ of [1], when used to encode the LSP parameters, achieves a 1 dB average spectral distortion at 24 bits/frame; each LSP vector is split into 3 subvectors and an 8-state FSVQ is used for each subvector. However, when channel noise is present, the ordinary FSVQ performance degrades drastically. Therefore, we describe a modified FSVQ system that is designed taking into account the channel noise. It is shown by means of simulations that the new system is robust to channel noise and outperforms the channel-optimized VQ [2] and the channel-matched multi-stage VQ of [3] by saving 1.5-4 bits/vector, depending on the level of noise in the channel.

Original languageEnglish
Title of host publicationICASSP 1992 - 1992 International Conference on Acoustics, Speech, and Signal Processing
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages133-136
Number of pages4
ISBN (Electronic)0780305329
DOIs
StatePublished - 1992
Event1992 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 1992 - San Francisco, United States
Duration: Mar 23 1992Mar 26 1992

Publication series

NameICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
Volume2

Conference

Conference1992 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 1992
Country/TerritoryUnited States
CitySan Francisco
Period3/23/923/26/92

ASJC Scopus subject areas

  • Software
  • Signal Processing
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'Finite-state vector quantization over noisy channels and its application to LSP parameters'. Together they form a unique fingerprint.

Cite this