WEARABLE EXOSKELETON REHABILITATION SYSTEM

Overview

A wearable exoskeleton rehabilitation system developed at Zhejiang Fuzhi Technology Co., Ltd., designed for motor function recovery in patients with movement disorders. The system integrates three core modules: EMG signal acquisition, adaptive movement prediction, and real-time motor control.

Technical Architecture

EMG Signal Acquisition

The system uses a multi-channel EMG sensor array to capture muscle activity in real time. Signals pass through an analog front-end for filtering and amplification, then are sampled and digitally processed by an STM32 microcontroller.

// EMG signal processing pipeline
void emg_process_sample(emg_channel_t *ch, uint16_t raw) {
    // Bandpass filter: 20-450Hz
    float filtered = biquad_filter(&ch->bpf, (float)raw);
    // Full-wave rectification
    float rectified = fabsf(filtered);
    // Moving average envelope
    ch->envelope = moving_avg_update(&ch->mavg, rectified);
}

Adaptive RNN Control

An adaptive control algorithm based on Recurrent Neural Networks (RNN) learns patient movement patterns to predict motor intent. The model runs fixed-point inference on the MCU, achieving <10ms control response latency.

Firmware Architecture

The MCU firmware uses a bare-metal real-time architecture with interrupt-driven coordination of three task pipelines: EMG acquisition, RNN inference, and motor control. Precise timing management ensures deterministic control loop response.

Key Metrics

Outcome

Contributed to a patent filing for a novel electromechanical architecture that improves upon traditional exoskeleton drive-train design, enhancing wearability and motion tracking precision.