The intrusion events in the optical fiber pre-warning system (OFPS) a re divided into two types,which are harmful intrusion event and harmless interference event.At present,the sig nal feature extraction methods of them are usually designed from the view of time domain.Howe ver,the differences of time-domain characteristics for different harmful intrusion events are not obvi ous,which cannot reflect the diversity of them in detail.In our statistical study,we find that the spectrum distributions of different intrusion signals have obvious differences,so that the intrusion signal is transformed int o the frequency domain for analysis. In this paper,an identification method of harmful intrusion event based on ener gy ratio is proposed according to the analysis method of acoustic signal.Firstly,the intrusion signals are pre- processed and the power spectral density (PSD) is calculated.Then the energy ratios of different frequency bands are calculated and the corresponding feature vector of each type of intrusion signal is further formed,which is inp ut into the classifier for training and recognition.The linear discriminant analysis (LDA) classifier is used to identi fy the signal,which can make use of the information of original data and distinguish different intrusion signals efficiently.Experimental results show that the algorithm improves the recognition accuracy of the intrusi on signal,which verifies the feasibility and validity of the algorithm.