Optimized EEG based mood detection with signal processing and deep neural networks for brain-computer interface.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Biomedical Engineering

Background:

• Electroencephalogram (EEG) is a key tool for studying brain activity, measuring electrical potentials from the scalp.
• While EEG is used for diagnosing neurological conditions, its potential for mood recognition remains underexplored.
• Developing a smart decision-making model to link EEG signals with subject mood is a significant research gap.

Purpose of the Study:

• To investigate the relationship between electroencephalogram (EEG) signals and human mood states.
• To develop and optimize a computational model for accurate mood classification using EEG data.
• To assess the feasibility of using EEG as a biomarker for emotional states.

Main Methods:

• EEG data were collected from 28 healthy human subjects under ethical consent.
• Data preprocessing involved Savitzky-Golay band-pass filtering and Independent Component Analysis (ICA).
• Mood classification was performed using various neural network algorithms, optimized with Blackman window-based Fourier Transformation to extract significant frequencies.

Main Results:

• The implemented model achieved a high detection accuracy of up to 96.01% in classifying moods from EEG data.
• Signal processing techniques effectively filtered noise and extracted relevant features from the EEG signals.
• Neural network models demonstrated significant capability in differentiating between various mood states based on brain activity patterns.

Conclusions:

• EEG signals contain discernible patterns that correlate with human mood states.
• The developed computational model, leveraging advanced signal processing and machine learning, shows high accuracy in mood detection.
• This research highlights the potential of EEG-based systems for non-invasive mood monitoring and assessment.