Sixth International Conference on Spoken Language Processing
October 16-20, 2000
A New Speech Classifier Based on YinYang Compensatory Soft Computing Theory
Dong Chen (1), Jingming Kuang (1), Yan Zhang (2)
(1) Department of Electronic Engineering,
(2) Department of Vehicle Engineering,
Beijing Institute of Technology, China
In future mobile communication system, it is very important to
expand the limited radio resources through exploiting the
silence compression technology In this paper, a new speech
classifier based on YinYang Compensatory soft computing
theory is described. The classifier is used to divide the noisy
speech signal input to the mobile phone into the active speech
(including voiced, unvoiced and transition) and inactive speech
(including silence or noise) during conversation. The novel
speech classifier algorithm is discussed and verified at the GSM
discontinuous transmission platform with an adaptive multi-rate
speech codec. We found the accuracy of the classifier is
approximately 98% in clean background acoustic environment.
Moreover, we analyzed its classification accuracy in eight types
of noise environment with varying signal-to noise ratio.
Chen, Dong / Kuang, Jingming / Zhang, Yan (2000):
"A new speech classifier based on Yinyang compensatory soft computing theory",
In ICSLP-2000, vol.2, 186-189.