Sixth European Conference on Speech Communication and Technology
(EUROSPEECH'99)

Budapest, Hungary
September 5-9, 1999

The Analysis and Application of a New Endpoint Detection Method Based on Distance of Autocorrelated Similarity#

Jie Zhu, Fei-li Chen

SJTU & Bell Labs Communications and Network Joint Laboratory, Shanghai Jiao Tong University, Shanghai, China

Endpoints detection play the important role in speech recognition. But the performance of some often used algorithms will decrease under low SNR conditions. A new concept as distance between autocorrelated functions is brought fully, and a new endpoint detecting method based on it is discussed in this paper. Results from experiment show that this new method has higher performance in endpoint detection. Especially in low SNR circumstance, it still can detect endpoint of speech accurately.


Full Paper (PDF)   Gnu-Zipped Postscript

Bibliographic reference.  Zhu, Jie / Chen, Fei-li (1999): "The analysis and application of a new endpoint detection method based on distance of autocorrelated similarity#", In EUROSPEECH'99, 105-108.