S. Kim, J. Kim, and K. Bae (Korea)
Speech interface, telematics, barge-in, AEC, speech recognition, speech enhancement, car environment
The interior of an automobile is a very noisy environment where both stationary cruising noise and nonstationary music or speech coming from an audio system in the car are present. For robust speech recognition in the car environment it is necessary to remove these background noises and extract a driver’s voice command well. In this paper, we propose a new speech interface system that has echo cancellation and barge-in functionalities to remove those noises. The proposed system employs an acoustic echo canceller (AEC) to remove the speech or music from an audio system or telematics system in the car, and exploits the double-talk (DT) detection result from the AEC to provide barge-in functionality. The MMSE-STSA speech enhancement algorithm and center clipping method are used as post-processing for further reduction of residual noise components in the extracted driver’s voice command. Experimental results demonstrate that the proposed speech interface system extracts a driver’s speech signal from the background noises quite well, which could improve the speech recognition performance in the car environment.
Important Links:
Go Back
