Counting Finger and Wrist Movements Using Only a Wrist-Worn, Inertial Measurement Unit: Toward Practical Wearable...

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Wearable Sensors

Background:

• Monitoring daily finger and wrist movements is crucial for hand-related healthcare, including stroke rehabilitation and carpal tunnel syndrome management.
• Existing methods often require obtrusive sensors like rings with magnets or inertial measurement units (IMUs).

Purpose of the Study:

• To develop and validate a nonobtrusive, ringless method for detecting finger and wrist flexion/extension movements using a wrist-worn IMU.
• To assess the feasibility of using a convolutional neural network (CNN) for analyzing IMU-generated spectrograms to recognize hand movements.

Main Methods:

• Developed the Hand Activity Recognition through using a Convolutional neural network with Spectrograms (HARCS) approach, utilizing CNNs trained on velocity/acceleration spectrograms from wrist-worn IMUs.
• Validated HARCS against a magnetic sensing algorithm (HAND) using data from twenty stroke survivors during daily activities.
• Further assessed HARCS accuracy using optical motion capture with unimpaired participants.

Main Results:

• HARCS demonstrated a strong positive correlation (R² = 0.76, p < 0.001) with the established HAND algorithm for daily finger/wrist movement counts in stroke survivors.
• The HARCS system achieved 75% accuracy in identifying movements when validated against optical motion capture in unimpaired individuals.
• The study confirmed the feasibility of ringless sensing for detecting the occurrence of finger and wrist movements.

Conclusions:

• Ringless sensing of finger and wrist movement occurrence using a wrist-worn IMU and HARCS is feasible.
• The HARCS approach shows potential for nonobtrusive monitoring in hand-related healthcare applications.
• Further accuracy enhancements may be needed for widespread real-world clinical adoption.