Turning a knob: deep learning-based prediction of torque and arm angles using force myography.

Area of Science:

• Human-computer interaction
• Biomechanics
• Signal processing

Background:

• Accurate prediction of human motor actions is crucial for intuitive human-machine interaction.
• Force myography (FMG) is an emerging technology for non-invasive muscle activity measurement.
• Understanding limb dynamics in daily tasks is key for adaptive system development.

Purpose of the Study:

• To investigate the efficacy of FMG in predicting knob-turning activities.
• To assess the impact of varying torque and arm angles on prediction accuracy.
• To evaluate the influence of subcutaneous fat on FMG-based prediction.

Main Methods:

• Participants performed knob-turning tasks with varying torque and arm angles.
• Force myography (FMG) data was collected during these activities.
• A hybrid deep learning model (CNN-LSTM) was utilized for data classification and prediction.

Main Results:

• The FMG-based prediction model achieved high accuracy for both torque (95.87 ± 2.59%) and arm angle (94.06 ± 2.44%).
• Subcutaneous fat presence did not significantly impact the prediction accuracy of torque and arm angle.
• The study demonstrated robust performance across different task parameters.

Conclusions:

• Force myography is a viable and accurate method for predicting complex motor actions like knob turning.
• The developed deep learning approach shows significant potential for real-time human-machine interaction.
• FMG technology offers a promising avenue for enhancing assistive technologies and understanding daily living activities.