Abstract: Due to the complex external conditions and system load fluctuation, the vehicle cable terminal frequently breakdown and even explode during the operation, which seriously influences the stable operation of high-speed trains. In-depth research on the mechanism and the characteristics of partial discharge of internal defects of cable terminals are of vital significance for fundamentally solving the potential hidden dangers in the safe and reliable operation of high-speed trains. To identify the internal partial discharge of the vehicle cable terminals accurately and overcome the problem that the existing methods can not accurately determine the type of defects in the vehicle cable terminals, the influence mechanism of the change of the internal defects of the vehicle cable terminal on the electric-acoustic field by simulation modeling and theoretical analysis of the electric-acoustic field is analyzed, and a two-channel two-dimensional convolutional neural network(CNN) model is proposed based on the MTF mapping and the electric-acoustic fusion features for the recognition of internal defects of the vehicle cable terminals. The results show that the two-channel CNN model based on electro-acoustic fusion features has excellent recognition effect, which makes the proposed model can better understand the feature expression of the input data with the accuracy of up to 99%, improving the accuracy and generalization ability of pattern recognition.