End-to-end acoustic modeling using convolutional neural networks for HMM-based automatic speech recognition.
Speech Communication(2019)
摘要
•Novel CNN-based end-to-end acoustic modeling approach is proposed.•Relevant features are automatically learned from the signal by discriminating phones.•Learned features are more discriminative than cepstral-based features.•Learned features are somewhat invariant to languages and domains.•Proposed approach leads to better ASR systems.
更多查看译文
关键词
Automatic speech recognition,Hidden Markov models,Deep learning,Feature learning,Artificial neural networks,Convolution neural networks,Hybrid HMM/ANN
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要