A New Algorithm for Adaptive Smoothing of Signals in Speech Enhancement

Speech signals are non-stationary and nonlinear in nature. They are affected by background noise and this affects the performance of a speech recognition system. This paper deals with smoothing a signal there by removing noise from speech signals for the development of an efficient speech recognition system for recognizing isolated spoken words in Malayalam. Here, speech signals are pre-processed using wavelet denoising methods which uses Soft Thresholding (ST). A new algorithm is proposed which is used to smooth the signal before applying thresholding. This increases the Signal to Noise Ratio (SNR) indicating that adaptive smoothing improves the quality of the speech signal by reducing more noise.