The present invention provides a face expression identification method based on a video time sequence, and relates to the image processing technology field. The method comprises: performing division of image materials obtained after video framing, finding out a complete dynamic expression, employing a face calibration tool to extract the face feature points of each set of expressions, performing geometric normalization processing of the obtained face feature points, calculating the maximum value, the minimum value, the average value, a peak, a skewness, the DFT (Discrete Fourier Transform) peak and the frequency of coordinate sequence of each feature point, employing a PCA (Principal Component Analysis) to perform dimension reduction, remove redundant data, taking as expression feature factors, and finally employing an SVM (Support Vector Machine) or k-NN (k-Nearest Neighbor) to perform learning and identification of the feature vectors to obtain a final expression result. The identification of a learner's dynamic expression is introduced in online learning to identify the most common five expressions such as amazing, carefulness, sleeping, talking and laughing, the identification rate is high, the interaction problem of MOOC (Massive Open Online Course) is effectively solved, and teaching feedback practicality is high in classroom.