Reducing the environmental sensitivity of cepstral features for speaker recognition

This paper investigates the robustness of cepstral based features with respect to additive noise, and details two methods of increasing the robustness with minimal need for a-prioro knowledge of the noise statistics. The first approach is a form of noise masking which adds a fixed offset to the linear spectral estimate. The second is a form of sub-land filtering, again in the linear domain, which estimates the dynamic content of the speech using Fourier transforms. This avoids negative values normally inherent in such filtering and which presents difficulties in deriving log estimates. Both methods are shown to provide useful levels of robustness to additive noise, for example, speaker identification error rates in SNR mis-matched conditions of 15 dB are reduced from 60.5% for standard mel cepstra to 13.8% and 24.1% for the two approaches respectively, a relative reduction in error of 77% and 60.1%.