Aiming at the traditional wavelet transform threshold, the soft threshold and hard threshold method are adopted. The two threshold methods have their own limitations. There is a deviation in the processing of the soft threshold method, which affects the stability and continuity of the signal. The hard threshold method can easily lead to Gibbs-pseudo in processing speech signals. According to the empirical formula, the threshold method is uncertain, so the fusing algorithm based on neural network genetic algorithm and wavelet transform is put forward. The feasibility of the fusion algorithm is demonstrated by experiments.