Sixth International Conference on Spoken Language Processing
(ICSLP 2000)

Beijing, China
October 16-20, 2000

A Novel Feature Extraction Using Multiple Acoustic Feature Planes for HMM-based Speech Recognition

Tsuneo Nitta, Masashi Takigawa, Takashi Fukuda

Graduate School of Eng., Toyohashi University of Technology, Japan

This paper describes an attempt to extract multiple peripheral features of a point x(ti,fj) on a timespectrum (TS) pattern by observing nīn neighborhoods of the point, and to incorporate these peripheral features (MPFPs: multiple peripheral feature planes) into the feature extractor of a speech recognition system together with MFCC parameters. Two types of peripheral feature extractor, MPFP-KL and MPFP-LR, are proposed. MPFP-KL adopts the orthogonal bases extracted directly from speech data by using KLT of 7x7 blocks on TS patterns. In MPFP-LR, the upper two primal bases are selected and simplified in the form of Δt-operator and Δfoperator obtained by linear regression calculation. MPFP-KL and MPFP-LR show significant improvements in comparison with the standard MFCC feature extractor in experiments with the HMM-based ASR system.

Full Paper

Bibliographic reference.  Nitta, Tsuneo / Takigawa, Masashi / Fukuda, Takashi (2000): "A novel feature extraction using multiple acoustic feature planes for HMM-based speech recognition", In ICSLP-2000, vol.1, 385-388.