Wavelet Network with OLS Optimization for Speech Signal Processing

Abstract

Taking advantage of both the scaling property of wavelets and the high learning ability of neural networks, wavelet networks have recently emerged as a powerful tool for many applications in the field of signal processing, such as data compression and function approximation. It is an implementation of wavelet transform, decomposing a signal into a series of scaled and translated wavelets. In the construction of wavelet network, the architecture must be pruned to reduce the system complexity, as well as to increase the generalization capability. Thus, the Orthogonal Least Square (OLS) algorithm is incorporated into the wavelet network to achieve the goal. Our experiments show that the OLS algorithm achieves the best results among all the network pruning algorithms, and the optimal wavelet network outperforms other traditional speech synthesis methods.