Predicting Human Intention-Behavior Through EEG Signal Analysis Using Multi-Scale CNN.

Area of Science:

• Neuroscience
• Computer Science
• Signal Processing

Background:

• Electroencephalogram (EEG) signal classification is crucial for human intention-behavior prediction in brain-computer interface (BCI) research.
• Convolutional Neural Networks (CNNs) have improved EEG classification, but accuracy remains a challenge due to reliance on last-layer features.
• Existing methods may miss vital local and detailed information for precise EEG signal classification.

Purpose of the Study:

• To propose a novel multi-scale CNN model for enhanced EEG signal classification.
• To address the limitations of existing CNN-based methods in capturing comprehensive feature information.
• To improve the accuracy of human intention-behavior prediction using EEG signals.

Main Methods:

• Preprocessing EEG signals and converting them into time-frequency images using the Short-Time Fourier Transform (STFT).
• Designing and implementing a multi-scale CNN model that integrates both local and global feature extraction.
• Validating the proposed method on the benchmark dataset 2b from BCI competition IV.

Main Results:

• The proposed multi-scale CNN model achieved an average accuracy of 73.9% on the benchmark dataset.
• This represents a significant improvement over traditional methods, including Artificial Neural Network (ANN), Support Vector Machine (SVM), and Stacked Auto-Encoder (SAE).
• Accuracy gains of 10.4%, 5.5%, and 16.2% were observed compared to ANN, SVM, and SAE, respectively.

Conclusions:

• The multi-scale CNN model effectively captures both local and global information in time-frequency images of EEG signals.
• This approach offers a substantial improvement in EEG signal classification accuracy for BCI applications.
• The proposed method demonstrates a promising direction for advancing human intention-behavior prediction through enhanced EEG analysis.