[Objective] collecting athlete training data and formulating an aided decision model for athletes training based on multi-feature analysis. [Methods] the training images of athletes were collected by camera, and the moving image features were extracted by scale invariant feature transformation; on this basis, the preliminary evaluation results of athletes’ training state were obtained by support vector machine; the sensor is used to collect athletes’ EMG signal data, the wavelet change is used to extract the characteristics of EMG signal, and the athletes’ fatigue state is identified through the time window to obtain the preliminary evaluation results of sports fatigue state. The decision-making level integrates the two preliminary evaluation results for the design of the athlete training auxiliary decision-making model to realize effective evaluation of athletic training fatigue state. [Results] after experimental analysis, the average training scores of all sports items exceed 90 points after using the model in this paper. [Conclusion] the model can accurately extract the characteristics of sports image and athletes’ EMG, and accurately evaluate athletes’ sports fatigue state, which lays a solid foundation for formulating athletes’ training plan.