DWT-Based Application of SVD and SVM Techniques to Seismic Signal Classifications

Seismic event and quarry blast classification is one of key issues in seismic data processing. However, automatic microseism classification is still a challenging work in practice. In this paper, based on discrete wavelet transform (DWT), a novel approach combining singular value decomposition (SVD) and support vector machine (SVM) techniques was proposed to categorize seismic signals. Firstly, the DWT was used to obtain wavelet coefficients, then correlation coefficient and variance contribution ratio were proposed to choose main wavelet coefficients. After that, the SVD was employed to compress data and extract features of the selected main wavelet coefficients. Finally, the BP neural network, Bayes and SVM classifiers were implemented to identify seismic events and quarry blasts of Yongshaba Mine, China. The results show that SVD can effectively reflect signal features and classification results of SVM is better than BP neural network and Bayes in accuracy, specificity and sensitivity, being the value of 95.5%, 97.0% and 94.0%, respectively. Thus DWT-based SVD and SVM techniques provide a useful method for seismic event and quarry blast classification.