IMU-Based Real-Time Biofeedback Wristband With Automatic Sensor-to-Segment Calibration for Arm Swing Training

Arm swing during walking is significant as it enhances gait stability and reduces energy expenditure. Reduced arm swing in patients with Parkinson’s disease (PD) or the elderly causes poor mobility and higher fall risk. This article develops an IMU-based real-time biofeedback wristband with an automatic sensor-to-segment calibration algorithm for arm swing training. The real-time algorithm detects key frames and calculates the sensor-to-segment alignment transformation, then monitors arm swing motion, and provides vibration feedback when insufficient arm swing is detected. Validation experiments were performed to assess the accuracies of arm swing angle, angular velocity estimations, and the effectiveness of real-time biofeedback. Seventeen subjects wore wristbands and optical markers on their arms while performing slow, normal, and fast walking on a treadmill. Arm swing angle, angular velocity, and range of motion were extracted with and without the proposed calibration method. An additional wristband manually aligned with the forearm was used for comparison and an optical motion capture (OMC) system was utilized for reference. For the proposed method with automatic calibration, the root-mean-square errors (RMSEs) of the estimated continuous swing angle and angular velocity were 2.1° and 12.8°/s, respectively, which were 46% and 58% of those estimated with the manually aligned wristband. In real-time feedback experiments, subjects exhibited a notable increase in arm swing amplitude given the stimulus from vibration feedback. The developed wristband could serve as an effective training and easy-to-use device, which has the potential for PD patients to increase their arm swing amplitude during walking.