A time-space adapted wavelet de-noising algorithm for robust automatic speech recognition in low-SNR environments

This paper addresses the problem of noise robustness of automatic speech recognition (ASR) systems, using a pre-processing speech enhancement approach based on wavelet-thresholding speech enhancement algorithm that does not require an explicit estimation of the noise level or of the a-priori knowledge of the SNR. This algorithm adapts the thresholds in both space and time which allows the removal of various environmental noises. The time-space adapted (TSA) wavelet de-noising algorithm is integrated in the front-end of an ASR system in order to evaluate its robustness in severe interfering car noise environments. The hidden Markov model toolkit (HTK) was used throughout our experiments. Results show that the proposed approach, when included in the front-end of an HTK-based ASR system, outperforms that of the conventional recognition process in severe interfering car noise environments for a wide range of SNRs varying down to 0 dB using a noisy version of the TIMIT database