Motion prediction using brain waves based on artificial intelligence deep learning recurrent neural network

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Electroencephalogram (EEG) research provides valuable insights into human body movements.
• Optimizing motion discrimination prediction from EEG signals is crucial for applications in exercise rehabilitation and understanding brain-motor patterns.

Purpose of the Study:

• To investigate the optimization of motion discrimination prediction using a gated recurrent unit (GRU) deep learning model with unique EEG data.
• To enhance the performance index of machine learning operations for improved motion recognition accuracy.

Main Methods:

• Collected 32-channel EEG data from 10 participants (gymnasts and physical education students) performing tasks with five difficulty levels of postural control.
• Applied spectrum analysis using fast Fourier transform and utilized a GRU network for machine learning on EEG frequency domains.
• Extracted brain-motor patterns using machine learning techniques.

Main Results:

• The GRU network algorithm achieved up to a 15.92% improvement in the performance index compared to existing models.
• Motion recognition accuracy ranged from 94.67% to 99.15% between actual and predicted values.
• Enhanced accuracy and cost function of the GRU network's hidden layers contributed to the optimization outcomes.

Conclusions:

• AI-driven motion identification optimization using EEG signals offers an innovative approach for exercise rehabilitation.
• The study highlights the interconnectedness between brain activity and the science of exercise.
• The developed GRU model demonstrates significant potential for accurate motion recognition in clinical and research settings.