The invention provides a movement behavior recognition method based on an axial acceleration sensor. According to the existing movement behavior recognition method based on multiple sensors or multiple axial sensors, data of multiple sensors or data of multiple axial sensors are required to combine, accordingly high computing cost is resulted, cost is increased, and meanwhile, real-time recognition is reduced. The method includes adopting axial data information of one acceleration sensor, extracting human body features, particularly three feature values of threshold in front of the peak of wave, threshold in front of the trough of wave, and time interval between the peak of wave and the trough of wave, in movement by waveforms formed through comparing the axial data information acquired by the sensor, and realizing accurate recognition of movement behaviors like squat, jump, run and the like. By the aid of the method, high recognition accuracy is achieved, computing cost is reduced, real-time recognition is improved, and cost is reduced.