Multi-channel EEG signal feature extraction and pattern recognition on horizontal mental imagination task of 1-D...

Area of Science:

• Neuroscience
• Computer Science
• Signal Processing

Background:

• Brain-computer interfaces (BCIs) leverage electroencephalogram (EEG) signals for machine control, offering novel human-computer interaction pathways.
• Extracting electroencephalographic activity is key to translating brain signals into actionable commands.
• This research focuses on classifying horizontal mental task patterns for 1-D cursor movement using EEG.

Purpose of the Study:

• To extract and classify features from EEG signals corresponding to horizontal mental task patterns.
• To evaluate the efficacy of different neural network classifiers in recognizing these mental task patterns.
• To enable cursor control through the mental imagination of movement.

Main Methods:

• EEG signals were analyzed for hemispherical power changes in alpha and beta frequencies.
• Feature extraction utilized the average signal power or power difference method.
• Principal Component Analysis (PCA) reduced feature dimensions, followed by classification using Learning Vector Quantization (LVQ), Multilayer Neural Network (MLP), and Probabilistic Neural Network (PNN) with k-fold cross-validation.

Main Results:

• The study successfully extracted features related to mental imagination of cursor movements.
• Neural network classifiers, particularly LVQ, MLP, and PNN, demonstrated acceptable accuracy in recognizing horizontal mental task patterns.
• K-fold cross-validation confirmed the robustness of the classification results.

Conclusions:

• EEG-based BCIs can effectively translate mental imagination into cursor control commands.
• The applied feature extraction and neural network classification methods show promise for BCI applications.
• This work contributes to advancing BCI technology for enhanced human-computer interaction.