| Paper: | SP-P12.8 | ||
| Session: | Acoustic Modeling: Model Complexity, General Topics | ||
| Time: | Thursday, May 20, 09:30 - 11:30 | ||
| Presentation: | Poster | ||
| Topic: | Speech Processing: Acoustic Modeling for Speech Recognition | ||
| Title: | STUDIES IN MASSIVELY SPEAKER-SPECIFIC SPEECH RECOGNITION | ||
| Authors: | Yu Shi; Microsoft Research Asia | ||
| Eric Chang; Microsoft Research Asia | |||
| Abstract: | Over the past several years, the primary focus for the speechrecognition research community has been speaker-independentspeech recognition, with the emphasis of working on databaseswith larger and larger number of speakers. For example, the mostrecent EARS program which is sponsored by DARPA calls forrecordings of thousands of speakers. In this paper, however, weare interested in making speech interface work well for oneparticular individual. For this purpose, we propose using massiveamounts of speaker-specific training data recorded in one's dailylife. We call this Massively Speaker-Specific Recognition (MSSR).As a pre-research, we leverage the large corpus we have availablefrom speech-synthesis work to study the benefit of MSSR only fromacoustic-modeling aspect. Initial results show that by changingthe focus to MSSR, word error rates can drop very significantly.In comparison with speaker-adaptive speech recognition system,MSSR also performs better since model parameters can be tuned tobe suitable to one particular individual. | ||
| Back | |||
Home -||-
Organizing Committee -||-
Technical Committee -||-
Technical Program -||-
Plenaries
Paper Submission -||-
Special Sessions -||-
ITT -||-
Paper Review -||-
Exhibits -||-
Tutorials
Information -||-
Registration -||-
Travel Insurance -||-
Housing -||-
Workshops