Detection of the Intention to Grasp During Reaching in Stroke Using Inertial Sensing.

Area of Science:

• Rehabilitation Engineering
• Biomedical Signal Processing
• Wearable Robotics

Background:

• Wearable soft-robotic gloves are crucial for assisting stroke survivors with daily activities.
• Early and accurate detection of movement intent is essential for seamless glove actuation.
• Distinguishing grasp intention from other movements is key for effective assistance.

Purpose of the Study:

• To investigate the classification of reach and grasp movements in stroke survivors using inertial sensing.
• To evaluate the potential for early detection of grasp intention.
• To assess the efficacy of Support Vector Machine (SVM) classifiers for this application.

Main Methods:

• Inertial sensing was used to capture hand and wrist movements of 10 stroke survivors.
• A Support Vector Machine (SVM) classifier was employed to analyze movement data.
• Classification accuracy was evaluated for single- and multi-user scenarios, and at varying movement lengths.

Main Results:

• High mean accuracies of 96.8% (single-user) and 83.3% (multi-user) were achieved.
• Accuracies up to 90% were obtained using only 50% of the movement data after optimal kernel selection.
• Early detection of 300-750ms was possible, albeit with a potential trade-off in accuracy.

Conclusions:

• Inertial sensing combined with SVM classification is a promising approach for actuating grasp-supporting devices.
• This method can significantly aid stroke survivors in performing daily living activities.
• Further research into online implementation is recommended for practical application.