An LDA-Based Approach for Real-Time Simultaneous Classification of Movements Using Surface Electromyography

Area of Science:

• Biomedical Engineering
• Neuroscience
• Human-Machine Interaction

Background:

• Myoelectric control offers intuitive human-machine interaction but typically decodes single movements sequentially.
• Human motor control naturally involves simultaneous actuation of multiple degrees of freedom (DOFs).
• Simultaneous decoding for myoelectric control is a growing research area.

Purpose of the Study:

• To develop a novel classification strategy for simultaneous decoding of individual and combined movements.
• To simplify control complexity and reduce computational cost in myoelectric control systems.
• To achieve accurate decoding using data from individual motions only.

Main Methods:

• Proposed a novel classification strategy for myoelectric signal decoding.
• Utilized linear discriminant analysis for low-dimensional signal representation.
• Collected data from individual motions to train the decoder for combined movements.
• Tested the decoding algorithm in an online, two-DOF (hand and wrist) task.

Main Results:

• Achieved an overall classification accuracy of 88.02% for decoding both individual and combined movements.
• Demonstrated the feasibility of simultaneous decoding using data from individual motions.
• Showcased the effectiveness of linear discriminant analysis in simplifying myoelectric signal complexity.

Conclusions:

• The proposed strategy enables simultaneous decoding of individual and combined movements, enhancing natural human-machine interaction.
• The method simplifies control and reduces computational load, making it practical for real-world applications.
• This approach represents a significant advancement in myoelectric-based movement control, achieving high accuracy.