Multi-Sensing System for Parkinson’s Disease Stage Assessment based on FPGA-embedded Serial SVM Classifier

The paper proposes a multi-sensing system for the jointly assessment of electromyographic (EMG) and electroencephalographic (EEG) signals for the neuromuscular syndromes progression assessment, such as the Parkinson’s disease (PD). The architecture implemented on Altera Cyclone V FPGA, interfaces 8 wireless surface EMGs (lower limbs) and 7 wireless EEGs (motor-cortex). The acquired signals, digitized and pre-processed, underwent a time-frequency features extraction (FE), making data suitable for a Serial Support Vector Machine (SSVM) real-time classification. The system has been in-vivo tested on 5 subjects (n=3 affected by mild PD and n=2 by severe PD). The experimental results showed an accuracy of ~94% in the pathology stage recognition, and near the 100% in distinguish healthy subjects from PD affected ones. The FPGA resources utilization results in: 31.04% ALMs, 15.87% registers, < 6% memory blocks and the 82.75% DSP blocks.