Classification of Speech Feature Signals Based on DE-BP Neural Network Model

摘要

When using BP model to classify speech feature signal, the initial weight and threshold of BP model randomly causes the model to fall into local minimum. A differential evolution algorithm is proposed to optimize the weight and threshold parameters of BP model, then establishing a classification model of speech feature signal based on differential evolution BP neural network algorithm. The classification models are trained and tested using the speech feature signals of four different kinds of music, folk songs, zither, rock and pop music. The test results show that compared with the traditional BP model, the improved BP algorithm model, the speech feature signal classification model based on differential evolution BP neural network is better in operation stability and classification accuracy.