Speech enhancement with beamforming enhanced by parametric spectral subtraction

Abstract

This paper describes a novel method of speech enhancement that combines multichannel beamforming techniques with single channel spectral subtraction. Spectral subtraction methods have been widely used for speech enhancement because they require only a single microphone and provide moderate improvements in the signal-to-noise ratio (SNR). But a major drawback of spectral subtraction is that it inevitably introduces musical noise, which is nonstationary and very annoying. Generally, the musical noise is reduced by smoothing its spectral variance in the frequency domain. Multimicrophone beamforming techniques, on the other hand, enhance the SNR by using spatial and temporal filters to introduce spatial selectivity towards the desired signal. It is shown that these two techniques are complementary, and can be combined to provide better performance than either method alone. One approach, for example, is to use spectral subtraction as a preprocessor in each channel before beamforming. In this case, the summing process of beamforming reduces the musical noise generated by spectral subtraction. We provide examples of using this approach in an automotive application.