Robot Control Approach Based On Surface Electromyography(sEMG)

Surface Electromyography(sEMG) is a regular neural signal which represents the contractions of muscle activity, motion state of different joints, and muscles fatigue, etc. This paper established a novel robot control system using the sEMG as the input signal, purposing to demonstrate the significance of s EMG to improve the controllability of traditional human-machine system.In this paper, the signal processing algorithm of s EMG was designed firstly. Four subjects were participated in our research, and the sEMG samples of different joints’ actions were collected for the algorithm design. The on-line identification algorithm for various movements and relaxations of joint muscles was presented subsequently by comparing different methods of feature extraction for sEMG. To test the practicability of sEMG, a robot control experiment was established using multi-degree of freedom robot(NAO). To perform the complicated control tasks, we designed an effective combined strategy. In further research of the robot walking control, a new gait cadence detection method was also proposed by using the sEMG signals of lower limb muscles, which could inform the statistical correlation characteristics between the human gait information and sEMG. To detect the gait cadence, sEMG signals were collected from Tibialis Anterior(TA) and Gastronomies Lateral(GL) of lower limb muscles when subjects were walking at different gait cadences. Through peak detection and evaluating the different rhythms of sEMG activities, different gait cadences and step lengths of subjects could be detected and evaluated respectively. After data processing analysis of 4 subjects who participated in on-line gait cadence experiments, the averaged error from sEMG signals is 1.7%, which shows ideal effect, high accuracy and good reliability.The recognition results by analyzing and identifying sEMG of human actions were used as input commands to realize the real time control of the humanoid robot. All the experimental results of the real time control of NAO and gait cadence detection based on sEMG revealed that sEMG signals could reflect human motion state accurately and effectively, which showed a potential application prospect in the field of neural human-machine interaction.