SoC-based architecture for robotic prosthetics control using surface electromyography

Over the past few years, surface electromyography (sEMG) controlled prosthetics have served many amputees or partially paralyzed subjects. Although the control accuracy and efficiency of such prosthetics have been widely focused, the extendibility in practical scenario is still quite limited, and many kinds of such prosthetics have only several fixed specific functions. Considering the requirements of improving the extendibility and flexibility of the prosthetics, in this paper, we proposed a novel System on Chip (SoC) architecture for the manipulation of sEMG controlled robotic prosthetics. By implementing on Xilinx Zynq FPGA platform, this system fulfills a flexible and economic resources and power consumption design. Within the workflow, firstly, several sEMG features including mean absolute value (MAV), waveform length (WL) and zero crossing rate (ZC) are extracted real time. Then, different movements of arm and hand are identified by a Naive Bayes classifier. Final control commands are transmitted to a humanoid robotic arm to finish specific arm/hand movements. In real time experiments, we successfully control a robotic prosthetics to complete all prescriptive tasks. The results confirm our design and implementation, and also indicate that the proposed SoC architecture is highly flexible and low-cost for the prototype design of sEMG controlled prosthetics.