当前位置 :

新闻中心

关于微软亚洲研究院首席研究员俞栋做学术报告的通知

  讲座题目:Structured Computational Networks for Speech Recognition

  主讲嘉宾:Dr. Dong Yu (Microsoft Research)

  时间:2015年4月13日9:30-10:30

  地点:科大西区科技实验西楼一楼117会议室

  主办单位:语音及语言信息处理国家工程实验室

  Abstract:

  In this talk, I argue that to further improve the performance of speech recognition systems we need to build structured computational networks. I will use our recently proposed prediction-adaptation-correction RNN (PAC-RNN) as an example. In the PAC-RNN the classification DNN estimates the state posterior probability based on both the current frame and the prediction made on the past by a prediction DNN. The result from the classification DNN is fed back to the prediction DNN to make better predictions for the future frames. In the PAC-RNN, we can consider that, given the new, current frame information, the classification DNN makes a correction on the prediction made by the prediction DNN. Alternatively, it can be viewed as adapting the classification DNN’s behavior based on the prediction DNN’s prediction. Experiments indicate that the PAC-RNN outperforms both DNNs and LSTMs on TIMIT phone recognition and Babel keyword spotting.

  Biography:

  Dr. Dong Yu is a principal researcher at Microsoft Research and a guest professor at the University of Science and Technology of China. His current research focuses on applying computational networks to speech processing. He has published over 140 papers and is the co-inventor of more than 50 granted/pending patents. His work on context-dependent deep neural network hidden Markov model (CD-DNN-HMM) has helped to shape the new direction on large vocabulary speech recognition research and was recognized by the IEEE SPS 2013 best paper award.