Aiming at problem of non-stationary and multi-data decomposition scale of the acoustic signal of pipeline blockage

a method of feature extraction and recognition is proposed based on multivariate empirical mode decomposition（ MEMD）. Multi-channel synchronous adaptive decomposition

acoustic signal of pipeline under three operation state by using MEMD

and the common mode of each signal channel is obtained. According to the correlation coefficient and the variance contribution rate

the intrinsic mode function（ IMF） component which reflects the main information of the pipeline is selected

the energy proportion of the selected IMF component is calculated as the characteristic vector of the pipeline operation state. Support vector machine（ SVM） is used to classify the acoustic signal of pipeline in different operation state. Experimental results show that the proposed method can effectively extract and identify the pipeline blocking acoustic signal.