Speech enhancement using auditory spectral attenuation

This paper presents a new speech enhancement approach for noise reduction based on non-uniform multi-band analysis. The noisy signal is divided into a number of sub-bands using a gammatone filterbank with non-linear ERB resolution, and the sub-bands signals are individually weighted according the power spectral subtraction technique and the Ephraim and Malah spectral attenuation algorithm. Subjective evaluation tests demonstrate significant improvements results over classical subtractive type algorithms, when tested with speech signal corrupted a posterior by various noises at different signal to noise ratios.