Title :
Segmental reliability weighting for robust recognition of partly corrupted speech
Author :
Cho, H.Y. ; Kim, L.Y. ; Oh, Y.H.
Author_Institution :
Dept. of Comput. Sci., Korea Adv. Inst. of Sci. & Technol., Taejon, South Korea
fDate :
6/6/2002 12:00:00 AM
Abstract :
A new method for the recognition of speech that is partly corrupted by noise is proposed. For this purpose, the reliability of each speech segment is measured based on a log likelihood ratio. A segment that has higher reliability is given more importance in the decision making by a modified Viterbi algorithm. Experimental results showed that the method significantly improves the performance of isolated word recognition under various burst noise situations
Keywords :
Viterbi decoding; burst noise; reliability; speech recognition; white noise; Korean phoneme balanced words; decision making; isolated word recognition; log likelihood ratio; modified Viterbi algorithm; partly corrupted speech; robust recognition; segmental reliability weighting; signal-to-impulsive noise ratio; speech recognition; speech segment; white burst noise; word recognition experiments;
Journal_Title :
Electronics Letters
DOI :
10.1049/el:20020374
