Abstract: To capture gesture motions and subsequently process human-machine interactions, we design an identification system to improve the classification precision and rapidity performances using a surface ElectroMyoGraphy(sEMG)signal system. First, we establish an intelligent interactive platform based on the wireless sEMG measurement system and a quadrotor. Second, we design a transient energy threshold to determine the beginning and the end of the active signal segment of raw sEMG signals by sliding time window methods; this effectively suppresses the bad effect of the trend segment on the identification result. Third, we analyze the characteristics of sEMG signals using statistical analysis in time domain. We propose a scheme that combines the acceleration feature information with sEMG signals to build a classification model of five gesture motions. Compared with using a single sEMG information source, the recognition accuracy is improved. Finally, the experiments of the quadrotor control by gesture motions verify the feasibility and effectiveness of the proposed method.