Low power implementation of a wavelet based noise reduction system

Recently most attentions have been given to wavelet based speech enhancement systems rather other approaches like Fourier transform. Wavelet transform is capable of providing both time and frequency information at the same time, properties which make it very useful in non-stationary systems like speech signals. But the problem of wavelet transform is its computational complexity which is several times higher than traditional Fourier transform that increases power consumption. In order to solve this problem a flexible and efficient structure based on Weighted OverLap Add (WOLA) is used and its computational complexity is examined. This technique significantly reduces the complexity while providing a SNRI performance close to the previous wavelet transform. The simulation results are promising and decrease the computations about 22% in comparison to the wavelet transform.