In order to solve the problems that the gear vibration signals in reality is difficult to obtain a large number of typical fault samples and the low accuracy in classification,a gear fault diagnosis method based on EMD(empirical mode position) decomposition and multiple features fusion was presented. This method first extract the indicators that reflect the characteristics of the signal as characteristic vector.Then EMD was used to decompose the original signal and extract the Energy index as characteristic vector.To realize intelligent diagnosis of bearing , it combined with the characteristic vector of signal features of various parameters, and the fusion feature vectors used as the training samples of SVM multifault classifier were trained. The result of research on Lab of bearing fault diagnosis shows that this method has a good classification accuracy, can be applied to fault diagnosis of gear very well.