A Continuous Control Scheme for Multifunctional Robotic Arm with Surface EMG Signal

This paper applies real time pattern recognition into the control of robotic hand with surface electromyographic (sEMG) signal. We focus on the hardware system design and the control strategy implementation. Time domain statistic methods are employed to extract the features, which have good effects on the pattern recognition. After the feature dimension reduction by Fisher linear discriminant (FLD), the feature vector is classified by a multi-layer perception (MLP) neural network. At last the data of several subjects is analyzed, and it shows good recognition accuracy. Using this system, the subjects can control a robotic arm to perform desired movements intuitively.