FEATURE-EXTRACTION USING A MATRIX COEFFICIENT FILTER FOR SPEECH RECOGNITION

This paper provides a new interpretation of the so-called ''delta-ceps trum'' and extends the formulation of the conventional delta-cepstrum towards an optimal design of the filter, which extracts important spec tral dynamics from a cepstrum sequence. The algorithm to obtain new fe ature parameters is unified to a formulation using a matrix coefficien t filter and is tested through Japanese speech recognition experiments . The average recognition error rate in a Japanese 24 phoneme recognit ion experiment for four speakers was reduced from 12.2% to 10.3%.