Channel normalization using pole-filtered cepstral mean subtraction

Devang K. Naik, Richard J. Mammone

Research output: Contribution to journalConference article

5 Scopus citations

Abstract

In this paper, we introduce a new methodology called Pole Filtering to remove the residual effects of speech from thecepstral mean channel estimate, for extracting features robust to transmission channel degradations. The approach isbased on filtering the eigenmodes of speech that are more susceptible to convolutional distortions caused by transmissionchannels. Poles and their corresponding eigenmodes for a frame of speech are investigated when there is a channelmismatch for speaker identification systems.Linear Predictive(LP) cepstra of speech has been found to be a useful feature set for recognition systems [1,2,10].The relation between the LP cepstral coecients and eigenmodes of speech has been exploited to develop a robustfeature set [14]. En this paper an algorithm based on Pole-filtering has been developed to improve the cepstral featuresfor channel normalization. Experiments are presented in speaker identification using speech in the TIMIT databaseprocessed through a telephone channel simulator and on the San Deigo portion of the KING database. The techniqueis shown to offer improved recognition accuracy under cross channel scenarios when compared to conventional methods.

Original languageEnglish (US)
Pages (from-to)99-110
Number of pages12
JournalProceedings of SPIE - The International Society for Optical Engineering
Volume2277
DOIs
StatePublished - Oct 25 1994
EventAutomatic Systems for the Identification and Inspection of Humans 1994 - San Diego, United States
Duration: Jul 24 1994Jul 29 1994

All Science Journal Classification (ASJC) codes

  • Electronic, Optical and Magnetic Materials
  • Condensed Matter Physics
  • Computer Science Applications
  • Applied Mathematics
  • Electrical and Electronic Engineering

Fingerprint Dive into the research topics of 'Channel normalization using pole-filtered cepstral mean subtraction'. Together they form a unique fingerprint.

  • Cite this