An efficient multi-band spectral subtraction method for robust speech recognition

In this paper we present a novel approach for adjusting a multi band spectral subtraction filter coefficients based on speech recognition system results. Currently most speech enhancement techniques are designed according to various waveform level criteria such as maximizing SNR or minimizing signal error. However improvement in these criteria does not necessarily result in increasing speech recognition performance. Only if these methods generate sequence of features that maximize or increase the likelihood of the correct transcription relative to other incorrect competing hypotheses, speech recognition performance will increase. Here we use an utterance with a known transcription and optimize multi-band spectral subtraction filter coefficients according to increasing difference between likelihood of the correct transcription and most likely incorrect hypothesis. We show that by incorporating the speech recognition system into filter design process, word recognition rate are improved, significantly.