Bias Compensation for Cepstro-Temporal Smoothing of Spectral Filter Gains

Abstract:
Many speech enhancement algorithms that modify shortterm spectral magnitudes of the noisy signal are plagued by annoying spectral outliers that are perceived as musical noise. Recently, we presented techniques that reduce these outliers by means of a temporal smoothing in the cepstral domain. This cepstro-temporal smoothing increases the quality of the enhanced output signal, as it affects only spectral outliers caused by estimation errors, while the speech characteristics are well preserved. However, due to the cepstral transform, the temporal smoothing is done in the logarithmic domain rather than the linear domain, and hence results in a certain bias. In this paper we derive a general bias compensation for a cepstro-temporal smoothing of spectral filter gain functions that is only dependent on the lower limit on the spectral filter-gain function. We show that the proposed bias-compensation increases the performance in terms of instrumental measures.