Sonomyography-Based Proportional Control For Isokinetic Movements: A Comparative Analysis Of Features And Algorithms

Area of Science:

• Biomechatronics
• Biomedical Engineering
• Rehabilitation Engineering

Background:

• Sonomyography (SMG) offers real-time ultrasound imaging of muscle activity, surpassing traditional methods for deeper muscle visualization.
• Wearable ultrasound devices are advancing real-time muscle sensing for proportional control systems.
• Precise control of biomechatronic interfaces, like robotic prosthetics, relies on accurate muscle movement intensity and dynamics tracking.

Purpose of the Study:

• To explore proportional control for target tracking using sonomyography (SMG) signals.
• To evaluate and compare three distinct feature extraction methods for SMG signal analysis.
• To assess the effectiveness of trained regression models in controlling real-time gestures via SMG.

Main Methods:

• Utilized a sensor array to capture sonomyography (SMG) signals for proportional control.
• Compared three feature extraction techniques: peak location-based, amplitude histogram, and curve-fitting features.
• Trained an ensemble regression model using extracted features to evaluate real-time gesture control accuracy.

Main Results:

• The amplitude histogram feature set demonstrated superior performance in the target tracking task.
• An accuracy of $88.72 \pm 3.27 \%$ was achieved using the amplitude histogram features with an ensemble regression model.
• This indicates the effectiveness of SMG signal analysis for precise proportional control.

Conclusions:

• Amplitude histogram features are highly effective for sonomyography-based proportional control.
• Wearable sonomyography systems show significant potential for enhancing real-time control in biomechatronic applications.
• The study validates SMG as a viable technique for precise user-friendly control of advanced interfaces.