| Presentation | 2012/12/13 An Ivestigation of Clustering Methods using Speaker-Class Models in Lecture Speech Recognition , |
|---|---|
| PDF Download Page | PDF download Page Link |
| Abstract(in Japanese) | (See Japanese page) |
| Abstract(in English) | In this paper, we have examined speaker clustering method using more than 100 clusters in order to improve the performance of spontaneous speech recognition. In this method, we use a soft clustering algorithm that allows a speaker to belong to more than one cluster in order to prevent a decrease in amount of training data per cluster. In the recognition procedure, the system needs to select one recognition result from the results of each speaker-class model. The selection can be conducted on the basis of the maximum likelihood among speaker-class model. In this work, we carry out two types of selection method; one is the method that selects the model every speaker and the other is the method that selects the model every utterance. The evaluation is conducted on CSJ (Corpus of Spontaneous Japanese). As the results, a word error rate of 21.08% was obtained in the baseline experiment. Meanwhile, 20.59%(selection every speaker) and 20.69%(selection every utterance) were obtained by using the proposed method. The results showed that the proposed method was effective for spontaneous speech recognition. |
| Keyword(in Japanese) | (See Japanese page) |
| Keyword(in English) | LVCSR / speaker-class model / hard clustering / soft clustering |
| Paper # | SLP-94 |
| Date of Issue |
| Conference Information | |
| Committee | SP |
|---|---|
| Conference Date | 2012/12/13(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 | Speech (SP) |
|---|---|
| Language | JPN |
| Title (in Japanese) | (See Japanese page) |
| Sub Title (in Japanese) | (See Japanese page) |
| Title (in English) | An Ivestigation of Clustering Methods using Speaker-Class Models in Lecture Speech Recognition |
| Sub Title (in English) | |
| Keyword(1) | LVCSR |
| Keyword(2) | speaker-class model |
| Keyword(3) | hard clustering |
| Keyword(4) | soft clustering |
| 1st Author's Name | |
| 1st Author's Affiliation | Yamagata University() |
| Date | 2012/12/13 |
| Paper # | SLP-94 |
| Volume (vol) | vol.112 |
| Number (no) | 369 |
| Page | pp.pp.- |
| #Pages | 6 |
| Date of Issue | |