EEG Signals Based Emotion Prediction to Implement AI Therapeutic Bot

Emotion recognition using Electroencephalogram (EEG) signals has become an essential tool in mental health monitoring, human-computer interaction, and affective computing. EEG provides a non-invasive method for capturing electrical activity in the brain with high temporal resolution, making it highly suitable for real-time emotion analysis. Traditional emotion recognition methods relying on facial expressions or voice can be biased or manipulated, whereas EEG-based analysis offers a more objective and direct understanding of emotional states. Existing systems typically implement EEG-based emotion classification pipelines used Multi-Scale Principal Component Analysis (MSPCA) for denoising. Feature extraction methods like Second-Order Difference Plot (SODP) and Summation of Distance to Coordinate (SDC) are commonly used, followed by spatial transformations like Equidistant Azimuthal Projection (EAP). Advanced models integrate Convolutional Block Attention Module (CBAM) and Generative Adversarial Networks (GANs) for refinement and data augmentation, respectively. This system, often limited by fixed emotion categories, insufficient temporal modelling, and lack of real-time interaction. To overcome these limitations, the proposed system introduces an advanced end-to-end EEG-based emotion recognition framework. It enhances preprocessing using MSPCA and performs Welch’s Power Spectral Density (PSD) estimation and Differential Entropy (DE) calculation across five EEG bands. A Linear Dynamic System (DE_LDS) with Kalman filtering effectively models temporal feature dynamics. The system integrates GANs for data augmentation and utilizes a Bi-Directional Long Short-Term Memory (Bi-LSTM) network for capturing complex temporal dependencies in EEG signals. The model is further integrated to a Flask-based AI therapeutic chatbot, which receives real-time emotion predictions via an external API and uses win32com.client for speech synthesis, enabling empathetic, voice-enabled interactions.