A feature information extraction method based on wavelet packet transform with fuzzy threshold

When the wavelet packet is used to extract signal, the result is greatly affected by the coefficient quantization threshold. It is difficult for the threshold to reflect the current state of the signal accurately, which leads to the failure of signal extraction. A novel method of extracting signal based on fuzzy threshold for wavelet packet is proposed in this paper. The wavelet packet decomposition coefficient is quantified by fuzzy threshold method. The concept of the maximum value of noise to signal ratio is given to characterize the signal´s strength compared to the noise in the simulation experiment. The frequency distribution probability is computed. The maximum value of noise to signal ratio and the frequency distribution probability are used as fuzzy input. To make the computing of fuzzy threshold conveniently, the fuzzy output is set by fixed value. The simulation results show that the method can meet the needs of different frequency signals, compared to conventional filtering and matched filtering, the novel method can inhibit the influence of the noise on the signal effectively, enhance the signal extraction ability and improve the detection distance.