A Novel Signal Feature Extraction Method Based on Wavelet Analysis

Abstract

The frequency characteristics of the Daubechies wavelet series were analyzed. The problem of the frequency band energy leakage (FBEL) of the wavelet analysis was pointed out. The FBEL affects the veracity of the signal feature extraction. By the simulation research, the basic rules of the FBEL were opened out, and a frequency band energy modification method was proposed. On the basis of the frequency band energy increment comparison method, the feature of the signal change are gotten by calculating the FBEL coefficients of the left and right adjacent frequency bands and the modified energy of the feature frequency band. It is shown that the effect of the FBEL can be eliminated by the frequency band energy modification method. This method is suitable to the signal feature extraction