Bottleneck linear transformation network adaptation for speaker adaptive training-based hybrid DNN-HMM speech recognizer
Tsubasa Ochiai, Shigeki Matsuda, Hideyuki Watanabe, Xugang Lu, Hisashi Kawai, Shigeru Katagiri
Abstract
Recently, a Hybrid DNN-HMM recognizer trained with the Speaker Adaptive Training (SAT) concept was successfully modified to a more effective speaker-adaptation-oriented recognizer whose DNN front-end adopted a Linear Transformation Network (LTN) Speaker Dependent (SD) module. However, the size of SD modules is still large, which incurs high storage costs and the risk of over-training. To alleviate this problem, we analyze the characteristics of an LTN module by focusing on the relation between its size and its feature-representation capability. Moreover, we propose a new SAT-based scheme for reducing the LTN size using SVD-based matrix compression. Evaluation experiments on the TED Talks corpus prove that our LTN size-reduction scheme not only maintains the adaptation performance of the original LTN-embedded, SAT-based DNN-HMM recognizer but also further increases it especially in cases where the speech data available for adaptation training are severely limited.
BibTeX
@inproceedings{icassp2016_bottlenecklinear,
title = {Bottleneck linear transformation network adaptation for speaker adaptive training-based hybrid DNN-HMM speech recognizer},
author = {Tsubasa Ochiai and Shigeki Matsuda and Hideyuki Watanabe and Xugang Lu and Hisashi Kawai and Shigeru Katagiri},
booktitle = {ICASSP 2016},
year = {2016}
}