Predicting quadriceps muscle activity during gait with an automatic rule determination method

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Control Systems

Background:

• Skill-based expert systems show promise for gait restoration.
• Rule-based systems offer fast, comprehensible, and implementable control.
• A key challenge is the difficulty users face in determining system rules.

Purpose of the Study:

• To describe an automatic method for deriving production rules from example data.
• To develop a rule base induced from a model using external sensor signals and electromyogram (EMG) patterns.
• To estimate muscle activity patterns using sensor information via production rules.

Main Methods:

• An automatic rule induction method based on entropy minimization was employed.
• A model was used to induce the rule base from sensor inputs and EMG outputs.
• The algorithm was validated using data from able-bodied individuals during walking with and without ankle-foot orthoses.

Main Results:

• The developed method successfully induced production rules for estimating muscle activity patterns.
• Gait variability increased in subjects with restricted ankle joint motion, validating the system's generalization test.
• Experimental results demonstrated the production rule's performance in estimating quadriceps muscle activity during walking.

Conclusions:

• The automatic rule induction method provides a viable approach for developing skill-based expert systems for gait restoration.
• The system effectively estimates muscle activity patterns, contributing to advancements in rehabilitation technology.
• The findings highlight the potential of entropy minimization for rule generation in biological control systems.