The present invention relates to a gesture recognition method based on interval distribution probability features, comprising: S1, using infrared sensing technology to collect infrared reflection signals and infrared temperature signals of gestures at a set frequency f; S2, judging whether the infrared temperature difference changes Target action; S3, automatic detection of gesture movement segments, and denoising and normalization, interpolation and normalization processing to obtain new data segments; S4, feature extraction, and at the same time, use standard gesture raw data to establish gesture feature templates as training Sample; S5, at last, utilize KNN method to identify test sample gesture type, the method of the present invention is simple and intuitive, unsupervised learning process, high to the recognition rate of common dynamic gesture, real-time performance is good, and to the change of environmental factor or gesture process The difference is inclusive and has good robustness.