To solve the problem of low recognition rate of perimeter intrusion behaviors, an improved empirical mode decomposition algorithm is used in the perimeter intrusion behaviors classification of fiber Bragg gratings. In this algorithm, the intrusion signal is extracted from the overall signal by using the short time average zero-crossing rate algorithm, and the double extreme wave prolongation is used to decompose the end effect of empirical mode decomposition algorithm. The improved algorithm is employed to decompose the intrusion signal and the characteristics of the effective components are extracted. Support vector machine is used to identify the intrusion behaviors. The nonintrusive behavior and four different invasion behaviors such as climbing, shearing, colliding, and touching are used to classify and recognize in outdoor environment. The results show that the proposed method can effectively identify different intrusion behaviors, and the recognition rate is greater than 96%.