Design and Implementation of a RISC-V Soc for Real-Time Epilepsy Detection on FPGA

Abstract

Electroencephalographic (EEG) signals record brain activity and are commonly used to make clinical diagnoses of brain dysfunction. However, traditional inspection methods greatly limit patient movement and are extremely inefficient. Therefore, it has become a popular research direction to make the detection simpler, more accurate, and more efficient. In this paper, a RISC-V SoC is designed to automatically capture and process EEG signals for real-time epilepsy detection. And, the epilepsy detection program is based on Support Vector Machine (SVM) algorithm and implemented with C language. The SoC consists of a RISC-V core, a custom bus, an ADS1299 controller, a FFT computing module, and a UART module. Further, the SoC is tested and verified on FPGAs and compared with Hummingbird E203 SoC and Tinyriscv SoC, two leading open-source RISCV SoC platforms. The results show that the design consumes 36% and 19% fewer hardware resources, compared to the two SoC platforms mentioned above, respectively. Finally, the design achieves more than 80% epilepsy detection sensitivity, demonstrating that it can be effectively applied to EEG signal processing scenarios.