Peformance Analysis Of Speech Enhancement Using Adaptive Wavelet Thresholding

Speech is an acoustic signal produced by human vocal chords. The major concern in speech processing is to preserve the message content of the speech signal for storage and convenient transmission. Speech enhancement is the method of improving the intelligibility and perceptual quality of a speech signal and automatic speech recognition is converting speech into text equivalent message. In this paper, a speech enhancement technique by processing of wavelet coefficients is presented. The wavelet packet decomposition and wavelet packet reconstruction are used for analysis and synthesis of speech respectively. The performance of the method is evaluated for the speech data that is created in a noise free environment, by comparing the Signal to Noise Ratio (SNR) of the noisy speech utterance and reconstructed speech utterance. The scheme can be used for robust automatic speech recognition applications.