Adaptive microphone array for speech enhancement

Abstract

This paper describes the design of an adaptive microphone array for speech enhancement in teleconference environments by using spatial filtering. Matlab simulations of the system allowed determining the most suitable algorithms for the final realtime implementation. This implementation was made on a Texas Instruments digital signal processor TMS320C6701 attached to a sensor array with two elements. Implementation details, optimizations and performance tests of the implemented system are presented. The signals captured by the sensors are thereinafter processed by the following algorithms: computation of the number of sources present in the environment using the MDL criterion; computation of the direction of arrival by means of a RootMUSIC algorithm, and finally, the beamforming toward the source of the desired signal.