Third International Conference on Spoken Language Processing (ICSLP 94)

Yokohama, Japan
September 18-22, 1994

Contributions of Selected Spectral Regions to Vowel Classification Accuracy

Parham Mokhtari, Frantz Clermont

Department of Computer Science, University College, University of New South Wales, Australian Defence Force Academy, Canberra, ACT, Australia

In this paper we describe the results of several vowel classification experiments, which shed light on the relative importance of certain spectral regions for achieving optimum classification accuracy in either a speaker-dependent or a speaker-independent task. The methodology adopted to conduct this investigation is based on a new formulation [3] of the quefrency-weighted cepstral distance measure, which allows specification of any frequency band within the available spectral range. This more flexible approach to distance computation was used to study the behaviour of classification accuracy as a function of increasing spectral range, and to identify spectral regions which predominantly carry either phonetic or speaker-specific information.

Full Paper

Bibliographic reference.  Mokhtari, Parham / Clermont, Frantz (1994): "Contributions of selected spectral regions to vowel classification accuracy", In ICSLP-1994, 1923-1926.