Previous [1] [2] [3] [4] [5] [6] [7] [8]

Journal of Inforamtion Science and Engineering, Vol.15 No.4, pp.615-628 (July 1999)
A Segmentation Method for Continuous Speech
Utilizing Hybrid Neuro-Fuzzy Network

Ching-Tang Hsieh, Mu-Chun Su, Eugene Lai
and Chih-Hsu Hsu*
Department of Electrical Engineering
* Graduate Institute of Information Engineering
Tamkang University
Taipei Hsien,Taiwan 251, R.O.C.

This paper presents a hierarchical neuro-fuzzy network system for segmenting continuous speech into syllables. The system formulates speech segmentation as a two-phase procedure. In the first phase, the Hybrid Neuro-Fuzzy network (HNFN) is utilized to classify the speech signal into three different types. The hybrid model (HNFN) composed of a distributed representation of a fuzzy system (DRF) and a hyperrectangular composite neural network (HRCNN) is proposed and used to cluster frames. This special hybrid system may neutralize the disadvantages of each alternative. The parameters in the trained HNFN are utilized to extract both crispy and fuzzy classification rules. In the following phase, the self-tuning back-propagation neural network (STBNN) is utilized to solve the coarticulation effects of vowel-vowel (V-V) concatenation. In our experiments, a database containing continuous reading-rate Mandarin speech recorded from newscasts was utilized to test the performance of the proposed speaker-independent speech segmentation system. The effectiveness of the proposed segmentation system is confirmed by the experimental results.

Keywords: speech segmentation, hybrid neuro-fuzzy network, self-tuning back-propagation neural network, clustering analysis, neural network

Full Text () Retrieve PDF document (199907_08.pdf : 122,937 bytes)

Received February 14, 1997; revised July 10, 1998; accepted September 3, 1998.
Communicated by Jhing-Fa Wang.