A sEMG-Based Hand Motions Recognition System With Dimension-Reduced FFT

Surface electromyogram(sEMG) signals have been widely used in the control of upper limb rehabilitation robot for their easy collection and acquisition. However, the medical sEMG devices are expensive and the signal processing is complicated. Unlike previous research, our goal is to identify as many motions as possible with as few channels as possible. This paper designed a four-channel hand motion recognition system that could accurately identify eleven common hand movements. The cores of the sEMG acquisition system are CC2530 and ADS1299. The signal processing includes three parts:signal acquisition, feature extraction, and classification. The dimension-reduced Fast Fourier Transformation(FFT) is selected as the feature extraction method to extract the frequency domain information from sEMG signal. In addition, the linear discriminant analysis (LDA) is selected as the classifier. The whole system is easy to promote and the algorithms proposed by this paper take less time, even the accuracy rate can reach 93%. It has the potential of cooperating with robot-aided rehabilitation and being the robot-human interface.