Abstract: Recognizing hand movements by using surface electromyography (sEMG) signal in real time and accurately is an important aspect of the application of sEMG signal. Hand movements onset detection from sEMG signals is the precondition of real-time hand movements recognizing. In this paper, The study aims at detecting the hand movement onset based on the consecutive sEMG signal. The sEMG signal is preprocessed by using the teager-kaiser energy (TKE) operator and a sEMG signal state binary function is designed to detect the hand movement onset from the consecutive sEMG signal. Then we design a heuristic filter according to the actual sEMG signal character. The detected results are filtered further by the heuristic filter which can cancel the effects of noise. We compare the application several methods for the simulation model. It proves the validity of TKE operator method. In the end, using sEMG signal acquisition and processing system for experimental verification, the results show that the detection method can rival action initiated for high precision of real-time detection.