Robust hybrid adaptive perceptual wavelet packet threshold to enhance speech in adverse noisy environment

A new robust hybrid method has been proposed by combining the Ephraim-Malah filtering method with subband adaptive time adaptive perceptual wavelet packet based threshold technique for enhancement of speech from adverse noisy environment. It is obtained by applying Ephraim-Malah filtering (EM) method for preprocessing the signal to initially reduce the noise. Because EM method enhances the signal but still remains musical noise. Hence, to further enhance the signal, a new approach sub-band adaptive time adaptive perceptual wavelet packet based threshold technique has been proposed (SATA-PWPT). This method is very simple and estimates the noise level from noisy signal without using complicated V/UV detection algorithm. Experimental results show that the proposed method works better than the other wavelet based threshold approaches for enhancing the speech signal. This method is analyzed in frequency domain analysis. The objective results are measured by global Signal to Noise Ratio (SNR) and Mean Square Error (MSE) with NOIZEUS database sentences corrupted by various types of colored noise. Spectrograms are also used to show the visual results.