Time-frequency analysis of speech provides detailed information which may be used for the study of its production mechanisms and for recognition purposes. The smoothed Wigner-Ville distribution can be used as an effective speech analysis tool when a high temporal resolution is required. Smoothing demands a careful balance between resolution and suppression of interferences in the distribution while taking into account the characteristics of the speech signal. The resulting representation is applied to analyze nasal phonemes and transitions in continuous speech and shown to unveil important transitory features not revealed by the spectrogram and to result in accurate segmentation and discrimination of their place of articulation.

Date Published: 1 November 1990
PDF: 9 pages
Proc. SPIE 1348, Advanced Signal Processing Algorithms, Architectures, and Implementations, (1 November 1990); doi: 10.1117/12.23476

Published in SPIE Proceedings Vol. 1348:
Advanced Signal Processing Algorithms, Architectures, and Implementations
Franklin T. Luk, Editor(s)