Abstract | ||
---|---|---|
Rate of speech (ROS) is a very important factor in speech recognition. In this paper we present a new speech rate measure method which first normalizes the duration of different acoustic units to standard duration and then builds a trigram duration model to measure the speech rate of sentence. We propose two methods based on the standard duration to compensate the influence introduced by speech rate variation in data corpus and get 11% error rate reduction in mandarin digit string recognition. |
Year | DOI | Venue |
---|---|---|
2004 | 10.1109/CHINSL.2004.1409627 | 2004 International Symposium on Chinese Spoken Language Processing, Proceedings |
Keywords | Field | DocType |
error rate,natural languages,speech recognition | Pattern recognition,Voice activity detection,Computer science,Audio mining,Trigram,Word error rate,Speech recognition,Natural language,Artificial intelligence,Sentence,Mandarin Chinese,Acoustic model | Conference |
Volume | Issue | ISSN |
null | null | null |
Citations | PageRank | References |
1 | 0.40 | 7 |
Authors | ||
3 |