Predicting Emotion from EEG Data

This project focuses on predicting emotional states from EEG data using deep learning. Leveraging the well-known DEAP dataset, which captures EEG signals of participants watching emotion-evoking music videos, we aim to build an advanced emotion recognition system. Using MNE Tools, we thoroughly analyze, clean, and prepare the EEG data, ensuring it is ready for deep learning applications. Visualization techniques within MNE Tools allow us to graphically interpret the EEG signals, uncovering crucial patterns. For the model building, we use TensorFlow to design and train deep learning models that can accurately associate EEG patterns with specific emotional states. The ultimate goal of our project is to develop a robust system capable of real-time emotion prediction, opening new possibilities in mental health monitoring and enhancing human-computer interactions.

Upon reviewing the charts, it becomes evident that our model has most successfully learned the pattern of the arousal rating. In the chart comparing actual and predicted arousal ratings, the data points align closely with the ideal line where the x-axis values equal the y-axis values, indicating a strong correlation and accurate predictions by the model. However, the other three charts—representing valence, liking, and dominance ratings—show that the model has not yet fully mastered these patterns. While the predicted values for these three ratings are not as precise as for arousal, they still achieve an approximate accuracy of 70%. This suggests that while the model has made significant progress, further refinement is needed to improve its performance on valence, liking, and dominance ratings.