Automated posture adjustment system for immobilized patients using EEG signals

Area of Science:

• Neuroscience
• Computer Science
• Biomedical Engineering

Background:

• Brain-Computer Interfaces (BCIs) enable communication and control through brain activity.
• Electroencephalography (EEG) is a non-invasive technique for measuring brain electrical activity.
• Accurate human posture identification via BCI is crucial for various applications.

Purpose of the Study:

• To develop and evaluate a BCI system for human posture identification using EEG signals.
• To compare the performance of different AI models for classifying motor imagery tasks.
• To introduce and validate a novel Graph Transformer All Attention (GTAA) model.

Main Methods:

• Utilized a five-step process for classification, including filtering and feature extraction.
• Employed a Convolutional Recurrent Denoising Autoencoder (CRDAE) for feature extraction.
• Compared Gated Recurrent Unit (GRU) with Attention, Temporal Transformer (TT), Bidirectional Long Short-Term Memory (Bi-LSTM) with Attention, and the proposed Graph Transformer All Attention (GTAA) models.

Main Results:

• The proposed Graph Transformer All Attention (GTAA) model achieved the highest classification accuracy.
• The system demonstrated reliability and efficiency through validation against BCI Competition IV 2a datasets.
• Ten-fold subject cross-validation confirmed the robustness of the proposed BCI system.

Conclusions:

• The integration of advanced AI techniques with EEG offers significant potential for practical BCI applications.
• The developed BCI system provides an accurate and efficient method for human posture identification.
• The GTAA model represents a promising advancement in EEG-based BCI classification.