METHOD AND ALGORITHM OF WINDOWED PROCESSING OF EMG AND PIEZOELECTRIC SIGNALS FOR THE FORMATION OF CONTROL ACTIONS IN BIOPROSTHESES

Abstract

The paper presents a window-based method for processing biosignals to generate control actions in upper-limb bioprostheses operating in real time. The core of the method is the segmentation of EMG and piezoelectric signals into short time windows of 40-80 ms with partial overlap, ensuring continuous information updating and minimal system response latency (<50 ms). The EMG signal serves as the primary channel of user intent, while the piezoelectric signal functions as a sensory feedback channel. For each window, noise filtering, root mean square (RMS) calculation, envelope extraction using the Hilbert transform, and exponential smoothing are performed. The piezoelectric signal is introduced into the system with a temporal delay of 30-70 ms, replicating the natural sensorimotor loop «intention – action – contact – correction». Based on the smoothed EMG and delayed piezo signals, a control signal is generated and applied to the actuator motor, taking into account its electrical and mechanical dynamics. The proposed method combines windowed processing and sensory delay in a unified algorithm that enhances the stability, smoothness, and physiological compliance of control, suppresses random spikes, and ensures adaptability under varying conditions. Simulation results confirm the effectiveness of the proposed approach for accurate and reliable control of next-generation bioprostheses.