| Paper: | SLP-P20.4 | ||
| Session: | Acoustic Modeling and Adaptation | ||
| Time: | Friday, May 19, 14:00 - 16:00 | ||
| Presentation: | Poster | ||
| Topic: | Speech and Spoken Language Processing: Clustering and novel modeling algorithms | ||
| Title: | Trajectory Clustering of Syllable-length Acoustic Models for Continuous Speech Recognition | ||
| Authors: | Yan Han, Annika Hämäläinen, Louis Boves, Radboud University, Netherlands | ||
| Abstract: | Recent research suggests that modeling coarticulation in speech is more appropriate at the syllable level. However, due to a number of additional factors that can affect the way syllables are articulated, creating multiple acoustic models per syllable might be necessary. Our previous research on longer-length multi-path models in connected digit recognition has proved trajectory clustering to be an attractive approach to derive multi-path models. In this paper, we extend our research to large vocabulary continuous speech recognition by deriving trajectory clusters for 94 very frequent syllables in a 37-hour dataset of Dutch read speech. The resulting clusters are compared with a knowledge-based classification. The comparison results show that multi-path model for longer-length units is difficult to build based on phonetic and linguistic knowledge. By applying trajectory clustering based multi-path model topologies, the performance on speech recognition accuracy was significantly improved. Thus, it is concluded that data-driven trajectory clustering is very effective approach to develop multi-path model. | ||
©2018 Conference Management Services, Inc. -||- email: webmaster@icassp2006.org -||- Last updated Friday, August 17, 2012