Proc. SPIE. 9794, Sixth International Conference on Electronics and Information Engineering
KEYWORDS: Signal to noise ratio, Principal component analysis, Wavelets, Interference (communication), Feature extraction, Signal processing, Signal analysis, Target recognition, Electronic filtering, Time-frequency analysis
This paper presents a target type recognition method based on local mean decomposition (LMD) and support vector machine (SVM) using the seismic signal caused by the ground-moving target. The wavelet packet filter is used for improving signal noise ratio (SNR). Then, the seismic signal is decomposed into several production function (PF) components. The feature vector is composed of the energy of each principal PF. SVM is used as classifier which discriminate the human, car and truck. The experiment result shows that, the average discrimination accuracy of proposed method is over 92.0%.
This paper presents a novel ground surface seismic source location method based on time difference of arrival (TDOA) and local mean decomposition (LMD). The wavelet packet filter is used for environmental noise reduction of seismic signal. Because of the non-stationary and randomness of seismic signal, LMD is applied to analyze seismic signal. The production function (PF) components can be obtained after the local mean decomposition to the seismic signal. Then, the principal PF component is selected based on the cross correlation coefficients. The instantaneous frequency distribution of principal PF component can be acquired by taking a derivative of pure frequency modulated signal with respect to time. In frequency spectrum graph, the frequency corresponding to the maximum of amplitude is selected as characteristic frequency. Finally, the time difference of arrival can be got according to the moment of characteristic frequency first appearance in instantaneous frequency distribution. The results of experiment show that the proposed method is effective.