Wavelet in conjunction with Neural Network method for speech enhancement quality evaluation

Abstract

Wavelet Neural Network Evaluation method WNNEM is proposed as a powerful tool for enhanced speech signal evaluation. This objective evaluation measure utilizes Feed forward back Propagation Neural Network FFBNN to train the free of noise signal, and then enhanced signal is simulated to the training output results taken for given target. The distance between simulation and the target, over different wavelet sub bands is studied. Four known speech enhancement method for studying the performance of WNNEM are utilized. The advantage of this method is the evaluation of different band passes of frequency based on wavelet transform by neural network, which is very influential tool for non stationary signals processing. Several objective measures are used to investigate the WNNEM compatibility. Results proved the validity of the proposed method.