TR2016-115

Recent Advances in Distant Speech Recognition



Automatic speech recognition (ASR) is being deployed successfully more and more in products such as voice search applications for mobile devices. However, it remains challenging to perform recognition when the speaker is distant from the microphone, because of the presence of noise, attenuation, and reverberation. Research on distant ASR has received increased attention, and has progressed rapidly due to the emergence of 1) deep neural network (DNN) based ASR systems, 2) the launch of recent challenges such as CHiME series, REVERB, ASpIRE, and DIRHA, and 3) the development of new products such as the Microsoft Kinect and the AMAZON Echo. This tutorial will review the recent progresses made in the field of distant speech recognition in the DNN era, including single and multi-channel speech enhancement front-ends, and acoustic modeling techniques for robust back-ends. The tutorial will also introduce practical schemes for building distant ASR systems based on the expertise acquired from past challenges.