A Neural Interface System-on-Chip for Nerve Signal Recording and Analysis of Human Gesture

Abstract

The intensity of surface electromyography (sEMG) reflects the state of muscle activity, and the monitoring of neural activity is accomplished by recording and analyzing sEMG, so as to help the elderly and patients with muscle injury live normally. However, the current products in the literature or on the market cannot support the integration of signal recording and decision making simultaneously. To solve this challenge, a novel neural interface system-on-chip (SoC) is developed, which includes a neural signal recorder, a hardware integrated classifier with signal screening capability and a series of communication interfaces for data transmission. In addition, function circuits consisting of the controller and the clock generator are designed to provide operation instructions and the necessary reference timing. The prototype of the system was developed by using Verilog on FPGA. In the experiments, five volunteer subjects with healthy upper limbs were invited to participate in the verification of model parameter training and real-time gesture recognition. The experimental results show that the average recognition accuracy reach to 98.14%. which is better than the existing model of the same classifier.