Telephone Speech Recognition with Data-Driven Selective Temporal Filtering based on Principal Component Analysis

  • Jung Sun Gyun (School of Electronic and Electrical Engineering, Kyungpook National University) ;
  • Son Jong Mok (School of Electronic and Electrical Engineering, Kyungpook National University) ;
  • Bae Keun Sung (School of Electronic and Electrical Engineering, Kyungpook National University)
  • Published : 2004.08.01

Abstract

The performance of a speech recognition system is generally degraded in telephone environment because of distortions caused by background noise and various channel characteristics. In this paper, data-driven temporal filters are investigated to improve the performance of a specific recognition task such as telephone speech. Three different temporal filtering methods are presented with recognition results for Korean connected-digit telephone speech. Filter coefficients are derived from the cepstral domain feature vectors using the principal component analysis.

Keywords