The discrimination between the intrusion and the nuisance events without compromising sensitivity is a key performance parameter for any outdoor perimeter intrusion detection system. In certain circumstances, the signals of the intrusion and the nuisance events are almost the same. This is especially the case for the intrusion and the nuisance events which may have a similar impact. In this paper, a series of methods are proposed to extract the fiber-signals: to denoise the input signal based on the wavelet transform; with a new practical algorithm of pre-segmentation of the fiber-signal, based on the hypothesis that the distribution of the energy of the intrusion in the time domain is different from that of the nuisance; to generate the eigenvector extracted from the distribution of the energy in the wavelet space; and to classify the fiber-signal using the support vector machine. With the assumption that different kinds of fiber-signals have different energy distributions in each frequency, the results of this experiment are satisfactory. This method is practical because we may take advantages of the SVM with very small cost.