Microphone Arrays

Such techniques—known collectively as beamforming—have been the subject of intense interest in recent years within the acoustic array processing research community. Unfortunately, such techniques have been largely ignored in the mainstream automatic speech recognition field, although this may rapidly change given the recent release and widespread popularity of the Microsoft Kinect R © platform. The simplest of beamforming algorithms, the delay-and-sum beamformer, uses only this geometric knowledge—i.e., the arrangement of the microphones and the speaker’s position—to compensate for the time delays of the signals arriving at each sensor and then additively combine them. More sophisticated adaptive beamformers minimize the total output power of the array under the constraint that the desired source must be unattenuated. The conventional adaptive beamforming algorithms attempt to minimize a quadratic optimization criterion related to signal-to-noise ratio. Recent research has revealed, however, that such quadratic criteria are not optimal for acoustic beamforming of human speech. Hence, we also present beamformers based on non-conventional optimization criteria, that have appeared more recently in the literature. In particular, recent research has revealed that useful optimization criteria can be devised by attempting to restore the non-Gaussian statistical characteristics present in uncorrupted or “clean” speech. As these characteristics are diminished through the introduction of noise or reverberation, the use of adaptive beamforming techniques to restore the original statistical characteristics reduces the effect of these distortions, and hence improves speech recognition performance.