An intrusion signal recognition method of optical fiber perimeter security system based on adaptive empirical mode decomposition（ AEMD） is proposed. Firstly

the time-domain characteristics including the shorttime energy and the over-threshold rate are extracted to identify and preprocess vibration signals

some external disturbance signals are excluded by the statistical characteristics of over-threshold ratio and the ratio difference coefficient. Then

AEMD method is used to decompose the vibration signal and obtain several intrinsic mode functions（ IMF）

IMF components that are highly correlated with the original signal are selected to obtain the energy entropy and kurtosis characteristics. Finally

all the features are constructed as feature vectors to identify the intrusion signals by support vector machine（ SVM）. Experimental results show that the method can effectively identify the types of intrusion signals

improve the recognition speed and rate.