VERY DEEP CONVOLUTIONAL NETWORKS FOR END-TO-END SPEECH RECOGNITION

被引：0

作者：

Zhang, Yu ^{[1
]}

Chan, William ^{[2
]}

Jaitly, Navdeep ^{[3
]}

机构：

[1] MIT, Cambridge, MA 02139 USA

[2] Carnegie Mellon Univ, Pittsburgh, PA 15213 USA

[3] Google Brain, Mountain View, CA USA

来源：

2017 IEEE INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING (ICASSP) | 2017年

关键词：

Automatic Speech Recognition; End-to-End Speech Recognition; Very Deep Convolutional Neural Networks;

D O I：

暂无

中图分类号：

O42 [声学];

学科分类号：

070206 ; 082403 ;

摘要：

Sequence-to-sequence models have shown success in end-to-end speech recognition. However these models have only used shallow acoustic encoder networks. In our work, we successively train very deep convolutional networks to add more expressive power and better generalization for end-to-end ASR models. We apply network-in-network principles, batch normalization, residual connections and convolutional LSTMs to build very deep recurrent and convolutional structures. Our models exploit the spectral structure in the feature space and add computational depth without overfitting issues. We experiment with the WSJ ASR task and achieve 10.5% word error rate without any dictionary or language model using a 15 layer deep network.

引用

页码：4845 / 4849

页数：5