The invention discloses a tensor decomposition-based multi-feature fusion 4D expression recognition method. The method comprises the steps of obtaining human face 4D expression data; preprocessing thehuman face 4D expression data, and calculating to obtain three normal vector components, a shape index and a depth map of the 4D human face data; tensor decomposition is carried out on three normal vector components, the shape index and the depth map of the 4D face data, and dynamic face expression information is extracted; and classifying the dynamic facial expression information by using a dynamic image network, and performing score fusion on a classification result to obtain a final classification result. According to the method, the information of the 4D face data is fully utilized, the three components, the shape index and the depth map of the normal vector of the face are calculated for the sequence face data, the 3D geometrical information of the face is fully utilized, for different people, the features of the face are more representative and discriminative, and the accuracy of face recognition and expression recognition is higher.