Phase term modeling for enhanced feature-space VTS
Autor: | Maxim Korenevsky |
---|---|
Rok vydání: | 2017 |
Předmět: |
Linguistics and Language
Computer science Gaussian Feature vector Speech recognition 02 engineering and technology Language and Linguistics 030507 speech-language pathology & audiology 03 medical and health sciences symbols.namesake Distortion 0202 electrical engineering electronic engineering information engineering Feature (machine learning) Taylor series Minimum mean square error Communication 020206 networking & telecommunications Computer Science Applications Term (time) Noise Computer Science::Sound Modeling and Simulation symbols Computer Vision and Pattern Recognition 0305 other medical science Software |
Zdroj: | Speech Communication. 89:84-91 |
ISSN: | 0167-6393 |
Popis: | HIghlightsVector Taylor Series (VTS) is a popular approach in robust speech recognition.Speech distortion model taking phase term into account is more accurate.Phase term can be modeled as a Gaussian random vector.Phase term modeling improves speech recognition accuracy on Aurora2 and Aurora4 database.Combination of VTS with phase term modeling and CMN is effective. This paper proposes a generalization of the Vector Taylor Series (VTS) approach for the compensation of speech feature distortions. It uses a phase term aware representation of the speech distortion model. It considers this term as a Gaussian random vector with unknown parameters in the same manner as it is conventionally done for additive noise. These parameters are estimated by means of the EM-algorithm. The explicit expressions for parameters update are derived. The minimum mean square error (MMSE) estimate of clean speech features is also obtained. Experiments carried out on the Aurora2 and Aurora4 databases show that the proposed approach outperforms the phase-insensitive version of feature-space VTS significantly for both GMM and DNN acoustic models. It is also shown that the combination of the proposed approach with the cepstral mean normalization (CMN) provides additional accuracy gains. |
Databáze: | OpenAIRE |
Externí odkaz: |