Classification of Mandarin consonants based on wavelet transforms

This paper describes a new approach to classify the Mandarin consonants. Based on the wavelet transforms, the proposed method could divide the Mandarin consonants into five classes by using the product function. The product function is generated from the appropriate wavelet and scaling coefficients of input speech signal, and the classification criterion is dependent on the product function and its energy profile as well as zero-crossing rate (ZCR). In general, the duration, ZCR, and energy ratio of different consonant-vowel transitions have different representations. Hence, with the additional verification of energy profile and ZCR, the Mandarin consonants can be accurately classified into five types. An overall accuracy rate of 90.2% for first selection is achieved.

[1]  Chung-Hsien Wu,et al.  A hierarchical neural network model based on a C/V segmentation algorithm for isolated Mandarin speech recognition , 1991, IEEE Trans. Signal Process..

[2]  Ronald N. Bracewell,et al.  The Fourier Transform and Its Applications , 1966 .

[3]  O. L. Clubb,et al.  A robust C/V segmentation algorithm for Cantonese , 1996, Proceedings of Digital Processing Applications (TENCON '96).

[4]  Alejandro Murua,et al.  Classification and clustering of stop consonants via nonparametric transformations and wavelets , 1995, 1995 International Conference on Acoustics, Speech, and Signal Processing.

[5]  Jhing-Fa Wang,et al.  A C/V segmentation algorithm for Mandarin speech signal based on wavelet transforms , 1999, 1999 IEEE International Conference on Acoustics, Speech, and Signal Processing. Proceedings. ICASSP99 (Cat. No.99CH36258).