The restoration of motor functions in patients with severe upper-limb paresis, particularly those resulting from brachial plexus trauma, remains a critical challenge in neurorehabilitation. Traditional myoelectric control often fails due to the lack of viable signals from atrophied distal muscles [1]. To overcome this, the sternocleidomastoid (SCM) neck muscle is utilized as a control proxy, as it is innervated by the accessory nerve (cranial nerve XI) and remains functional even in cases of complete brachial plexus injury [2]. However, processing low-amplitude SCM signals while ensuring safe mechanical actuation requires a sophisticated computational framework capable of handling high-speed data acquisition and complex classification tasks without compromising real-time safety [3].
This study evaluates a distributed, multi-threaded embedded control architecture for an active elbow orthosis. The objective is to demonstrate that a modular master-slave topology can provide the necessary responsiveness (low latency) and deterministic safety required for clinical applications.
The proposed control architecture is physically and logically decoupled into three distinct operational layers.
● The Acquisition Node (STM32F407) interfaces with an AD8232 analog front-end to perform 1 kHz digitization and local preprocessing of surface EMG signals.
● The Supervisory Node (Jetson Nano) acts as the central intelligence, managing high-level system states and executing motion intent recognition algorithms.
● The Actuator Node (ESP32) manages the BTS7960 motor driver and processes feedback from the AS5048A 14-bit absolute magnetic encoder.
Communication between nodes is achieved via TCP/IP over Wi-Fi, eliminating physical tethers that hinder patient mobility. The software on the Jetson Nano & STM32F407 is implemented as a multi-threaded system governed by a Finite State Machine (FSM) [4], as illustrated in Fig.1. To ensure data integrity, all shared buffers are protected by mutex primitives. The pipeline consists of an Acquisition Thread (filling a 500-sample circular buffer), a Processing Thread (digital filtering and feature extraction), and a Classification Thread. Simultaneously, the software of Actuator Node on the ESP32 performs local safety checks, ensuring that joint motion is halted if anatomical limits (e.g., 145°) are reached, regardless of network status.
Figure 1. Multi-threaded EMG-controlled orthosis software architecture
Evaluation of the multi-threaded software of the proposed solution revealed highly deterministic performance. The average total processing time per EMG window was 1.69 ms, while the most intensive stage—Motion Classification—averaged 46.14 ms. Despite this, the parallelized execution allowed the overall sample-to-result latency to remain at a mean of 6.02 ms (with a standard deviation of 8.52 ms). This timing is significantly faster than the 100 ms human perception threshold, ensuring a seamless user experience. The AS5048A absolute encoder provided precise angular feedback with a resolution of 0.0219°, which was critical for maintaining kinematic safety during flexion and extension cycles.
The distributed master-slave architecture successfully addresses the latency and safety requirements for EMG-driven rehabilitation robotics [5]. By isolating acquisition, supervisory logic, and actuation into concurrent threads, the architecture provides a robust platform for assistive technology. Future research will focus on integrating brushless DC (BLDC) motors to increase low-speed responsiveness and reduce mechanical noise.
References:
1. M. Asghari Oskoei and H. Hu, “Myoelectric control systems—A survey,” Biomed. Signal Process. Control, vol. 2, no. 4, pp. 275–294, Oct. 2007, doi: 10.1016/j.bspc.2007.07.009.
2. A. V. Kotsiubailo et al., “Designing Minimalistic Powered Arm Orthosis for Brachial Plexus Injuries,” Visnyk NTUU KPI Seriia - Radiotekhnika Radioaparatobuduvannia, no. 96, pp. 50–61, Jun. 2024, doi: 10.20535/RADAP.2024.96.50-61.
3. J. R. . Cram, G. S. . Kasman, and J. Holtz, Introduction to surface electromyography. Aspen Publishers, 1998.
4. A. Williams, C++ concurrency in action : practical multithreading. Manning, 2012.
5. A. M. Dollar and H. Herr, “Lower Extremity Exoskeletons and Active Orthoses: Challenges and State-of-the-Art,” IEEE Trans. Robot., vol. 24, no. 1, pp. 144–158, Feb. 2008, doi: 10.1109/TRO.2008.915453.
|
Голосування 
Ця робота ліцензується відповідно до Creative Commons Attribution 4.0 International License
Знайшли помилку? Виділіть помилковий текст мишкою і натисніть Ctrl + Enter