基于小波特征和神经网络的天然地震与人工爆破自动识别

采用福建地区天然地震和人工爆破事件波形记录,通过一维离散小波变换(DWT)及4层小波包变换(WPT)对信号进行分解,提取出用于识别的4种波形小波特征:小波能量比特征、小波包能量比特征、小波包香农熵特征及小波包对数能量熵,此外还提取出P/S震相振幅比;采用BP神经网络对4种小波特征及分别加入P/S震相振幅比的组合特征进行识别效果检验,结果表明,单小波判据小波能量比特征识别效果好;双判据组合P/S震相振幅比和小波包对数能量熵的组合识别效果最好,可考虑作为实际天然地震与人工爆破在线自动识别系统的识别判据。

Automatic Identification of Earthquake and Explosion Based on Wavelet Transform and Neural Network

Adopting natural earthquakes and artificial explosion waveform record events of Fujian region, through one dimensional discrete wavelet transform(DWT) and 4-layer wavelet packet transform(WPT) for signal decomposition, we use the extract to identify four waveform little potter characters: the wavelet energy than characteristics, wavelet packet energy than features, wavelet packet Shannon entropy, logarithmic of wavelet packet energy entropy.Inaddition, we extract the original waveform P/S seismic phase amplitude ratio.We use BP neural network to test the recognition effect of four kinds of wavelet characteristics and add the characteristics of P/S seismic phase amplitude ratio respectively. The results show that the wavelet energy ratio feature recognition is effective. The combination of P/S seismic phase amplitude ratio and wavelet packet logarithmic energy entropy has the best recognition effect, which can be considered as the identification criterion for the online automatic identification system of natural earthquake and artificial explosion.