| Presentation | 2001/12/13 Syllable recognition using syllable-segmental statistics and syllable-based HMM Nobutoshi TAKAHASHI, Seiichi NAKAGAWA, |
|---|---|
| PDF Download Page | PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | In our previous research, we demonstrated the validity of segmental unit input hidden Markov model (HMM), which regards successive four frame MEL-cepstrum coefficients as a feature vector. The vector is compressed into 20 dimensions using the KL transform. However, the model considers only the correlation between frames in a short section, but not the correlation between the frames over a long section. In this paper, in order to represent the correlation over a long distance, we use the syllable-segmental statistics that are calculated by the concatenation of feature vectors, corresponding to each state in a syllable based HMM. As this concatenated feature vector consists of a high dimension, the dimension is reduced using the K-L transform. The statistics are modeled by a GMM. The use of syllable-segment statistics allows the model to express the correlation between the frames over a long distance (e.g., the correlation between a vector in the first state and a vector in the fourth state in a syllable-based HMM). For modeling and estimating, we conducted a forced Viterbi alignment against continuous speech using a conventional HMM, and then we segmented continuous speech into syllable segments. By combining this approach with a segmental-unit input HMM, the syllable recognition rate was improved to 87.7% from 83.7% for syllables taken from continuous speech, without using a language model. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | syllable recognition / segment model / HMM |
| Paper # | NLC2001-51,SP2001-86 |
| Date of Issue |
| Conference Information | |
| Committee | NLC |
|---|---|
| Conference Date | 2001/12/13(1days) |
| Place (in Japanese) | (See Japanese page) |
| Place (in English) | |
| Topics (in Japanese) | (See Japanese page) |
| Topics (in English) | |
| Chair | |
| Vice Chair | |
| Secretary | |
| Assistant | |
| Paper Information | |
| Registration To | Natural Language Understanding and Models of Communication (NLC) |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | Syllable recognition using syllable-segmental statistics and syllable-based HMM |
| Sub Title (in English) | |
| Keyword(1) | syllable recognition |
| Keyword(2) | segment model |
| Keyword(3) | HMM |
| 1st Author's Name | Nobutoshi TAKAHASHI |
| 1st Author's Affiliation | Information and Computer Sciences, Toyohashi University of Technology() |
| 2nd Author's Name | Seiichi NAKAGAWA |
| 2nd Author's Affiliation | Information and Computer Sciences, Toyohashi University of Technology |
| Date | 2001/12/13 |
| Paper # | NLC2001-51,SP2001-86 |
| Volume (vol) | vol.101 |
| Number (no) | 520 |
| Page | pp.pp.- |
| #Pages | 6 |
| Date of Issue | |