© 2016 ACM.Seismic signals which are generated by human footsteps could be used in intrusion. Feature extraction is the most important step of the target detection and classification. In this study, vibration signals, which are generated by human steps, are analyzed for intrusion detection. A test set-up is established for measuring seismic signals equipped with a seismic accelerometer and experiments are performed according to different measurement scenarios. Crest factor and kurtosis values of raw data and discrete wavelet transform (DWT) analysis results are examined. Three different features matrices are generated using crest factor and kurtosis value of raw data and DWT analysis data. Detection and no-detection cases are divided into two groups and three support vector machine (SVM) classifiers are modeled using data. The results show that, detection distance increases with improved feature matrix generated with DWT crest factor and kurtosis values of data, which also decreases the error rate of the classifier.