Sixth International Conference on Spoken Language Processing
(ICSLP 2000)

Beijing, China
October 16-20, 2000

Multichannel Signal Separation for Cocktail Party Speech Recognition: A Dynamic Recurrent Network

Seungjin Choi (1), Heonseok Hong (1), Hervé Glotin (2), Frédéric Berthommier (2)

(1) Department of Electrical Engineering, Chungbuk National University, Korea
(2) ICP, INPG, Grenoble, France

This paper addresses the method of multichannel signal separation with its application to cocktail party speech recognition. First, we present a fundamental principle for multi-channel signal separation which describes what spatial independence criterion results in. Second, for practical implementation of the signal separation filter, we consider a dynamic recurrent network and develop a new simple learning algorithm. The performance of the proposed method is evaluated in terms of word recognition error rate (WER). Experimental results show that our proposed method dramatically improves the word recognition performance in the case of two simultaneous speeches.

Full Paper

Bibliographic reference.  Choi, Seungjin / Hong, Heonseok / Glotin, Hervé / Berthommier, Frédéric (2000): "Multichannel signal separation for cocktail party speech recognition: a dynamic recurrent network", In ICSLP-2000, vol.3, 83-86.