Neural decoding of expressive human movement from scalp electroencephalography (EEG)

Area of Science:

• Neuroscience
• Human Movement Analysis
• Machine Learning in Biomechanics

Background:

• Electroencephalography (EEG) characterizes neural activities for limb control and movement kinematics.
• The capacity of EEG to discern expressive movement qualities remains largely unexplored.
• Laban Movement Analysis (LMA) provides a framework for describing expressive movement qualities.

Purpose of the Study:

• To investigate the extent to which EEG can differentiate expressive movement qualities.
• To explore the neural basis of expressive movement using EEG and machine learning.

Main Methods:

• Five skilled dancers performed whole-body movements categorized as Neutral, Think (imagined expression), or Do (enacted expression).
• EEG and inertial sensors recorded brain activity and movement kinematics.
• Delta-band EEG data were processed using locality-preserving Fisher's discriminant analysis (LFDA) and Gaussian mixture models (GMMs) for classification.

Main Results:

• Classification accuracy for Action Type was 59.4 ± 0.6%.
• Classification accuracy for Laban Effort quality was 88.2 ± 0.7%.
• Motion-related artifacts did not significantly impact classification accuracy.

Conclusions:

• EEG contains valuable information for decoding the expressive qualities of human movement.
• These findings advance the understanding of the neural underpinnings of expressive movement.
• Potential applications include neuroprosthetics and enhanced human-computer interfaces for movement restoration.