Abstract
Wearable orthoses can function both as assistive devices, which allow the user to live independently, and as rehabilitation devices, which allow the user to regain use of an impaired limb. To be fully wearable, such devices must have intuitive controls, and to improve quality of life, the device should enable the user to perform Activities of Daily Living. In this context, we explore the feasibility of using electromyography (EMG) signals to control a wearable exotendon device to enable pick and place tasks. We use an easy to don, commodity forearm EMG band with 8 sensors to create an EMG pattern classification control for an exotendon device. With this control, we are able to detect a user's intent to open, and can thus enable extension and pick and place tasks. In experiments with stroke survivors, we explore the accuracy of this control in both non-functional and functional tasks. Our results support the feasibility of developing wearable devices with intuitive controls which provide a functional context for rehabilitation.
Introduction
Stroke often leads to hemiparesis and spasticity, where involuntary muscle contractions leave the hand clenched, severely limiting Activities of Daily Living (ADLs). While robotic orthoses offer a solution, traditional control systems often require precise, professional placement of medical-grade sensors, making home use difficult. This research presents a wearable hand orthosis that utilizes a commodity EMG armband and machine learning to provide intuitive, functional hand control for stroke survivors.
System Design and Methods
The system integrates a soft "exotendon" orthosis with a consumer-grade Myo Armband. The armband features eight surface electromyography (EMG) sensors that capture muscle activity patterns across the forearm.
Control Algorithm
Rather than simple signal thresholding, the system employs a Random Forest classifier to analyze patterns across all eight sensors. A key innovation is the "in-the-loop" training protocol: because the physical presence of the orthosis alters EMG signals, users are trained while wearing the active device. To ensure smooth operation, a 0.5-second median filter is applied to the classifier's output, preventing "jitter" from momentary misclassifications.
Experimental Evaluation
The system was tested with four stroke survivors exhibiting chronic hemiparesis and varying levels of spasticity. Calibration involved a brief 45-second session to collect training data. Evaluation covered both "non-functional" (static) and "functional" (pick and place) tasks.
Key Insights
- Sensor Quantity over Quality: Using multiple commodity sensors provides a holistic "image" of forearm activity, making the system robust to slight variations in placement.
- Addressing Spasticity: Since "relaxed" and "closed" intents often produce similar EMG signals in stroke patients, these states were grouped together. The device only triggers hand extension when a clear "open" intent is detected.
- Mechanical Interaction: Training must occur with the device on, as the mechanical assistance of the orthosis significantly changes the user's muscle activation patterns.
Results
In non-functional testing, the classifier achieved high accuracy, ranging from 82.2% to 93.6%. During functional "pick and place" tasks, subjects successfully moved objects using their impaired hands. For example, one subject achieved a 100% success rate (6/6).
The results indicate that EMG-based control is as effective as manual button control while offering a more intuitive, hands-free experience. Subjects were able to operate the device effectively after minimal training.
Conclusion and Future Work
This study demonstrates the feasibility of using commodity EMG sensors and pattern classification to control wearable orthoses for stroke rehabilitation. While limited by a small sample size and some mechanical issues, future work will focus on improving robustness to device donning/doffing, expanding to different grasp types, and better handling complex muscle coactivations.
Summary updated