In this paper， a tool wear stage identification method based on empirical mode decomposition （EMD） and support vector machine （SVM） is proposed based on cemented carbide tool milling Ti-6Al-4V. Firstly， the original acceleration signal and force signal are decomposed into a series of intrinsic mode function （IMF）， and an effective IMF is selected to combine a new signal. Then the multi evaluation index matrix of the new signal is calculated. Taking the multi index matrix as the input feature vector， a tool wear recognition model based on linear classifier is established. Finally， the detection signal is input into the model for recognition， and the recognition accuracy of tool wear stage reaches 99.17%. Compared with SVM， back propagation （BP） and wavelet packet-SVM model， EMD-SVM has less operation time and higher precision. The experimental results show that the model has a good recognition effect on tool wear in titanium alloy milling process， and provides a new method as a reference for tool wear monitoring.