A Wearable Platform for Real-Time Control of a Prosthetic Hand by High-Density EMG

This study presents a wearable platform for high-density surface electromyography (HD sEMG) and real-time myoelectric control of a hand prosthesis. The system features two 64-channel RHD2164 analog front-ends (totaling 128 channels) integrated with a Zynq UltraScale+ MPSoC for heterogeneous computing. The PYNQ-based framework offers a Python/Linux environment for designing scalable and flexible multi-purpose algorithms. Experiments with 21 healthy subjects performing eight motor tasks (finger flexion/extension, thumb opposition, and prehension patterns) at two frequencies (0.50 Hz and 0.75 Hz) demonstrated the system’s capabilities in both HD sEMG recording and real-time prosthetic control for a single degree of freedom (1-DoF). Signal quality metrics exceeded recommended thresholds (Signal-to-noise Ratio: 13.93 ± 7.51 dB; Signal-to-motion-artifact Ratio: 25.18 ± 5.18 dB), validating the dual-front-end architecture. The processing pipeline combining reinforced electrode signal adaptation (RESA), non-negative matrix factorization (NMF), and Kalman filtering achieved high similarity between estimated and reference signals for the test data, assessed using the maximum normalized cross-correlation (XC _max ), with values from 0.54 ± 0.22 to 0.80 ± 0.16. The coefficient of determination (R ² ) for HD sEMG signal reconstruction ranged from 0.87±0.09 to 0.95±0.03, with higher values observed in tasks involving prehension patterns. The platform demonstrated scalable real-time performance, with total latency from acquisition to command output ranging from 63.3±1.0 ms (30 ms buffer) to 219.1 ± 4.5 ms (150 ms buffer), while maintaining temporal alignment (lag at XC _max : 0.02 ± 2.09 The platform’s heterogeneous architecture enabled local execution of all processing stages (at 2052.52 Hz sampling rate/channel), with the FPGA handling signal acquisition and Arm Cortex-A53 cores performing motor intention decoding, supporting future integration of adaptive multi-DOF control for advanced prosthetic applications.