EEG-based Emotion Recognition via Neural Networks and Optimised Feature Selection
Accurate recognition of human emotions plays a main role in developing intelligent and adaptive systems for healthcare, interaction, human–computer and affective computing applications. This paper presents an emotion recognition system using electroencephalogram signals in which neural networks and feature selection methods are combined to boost the recognition accuracy. Electroencephalogram (EEG) signals are used as input data because they are more accurate than conventional methods, such as facial and speech recognition, and provide a direct measure of the brain activity. The recorded EEG signals are first preprocessed to reduce noise and artefacts, ensuring quality data. Then, important features are extracted through time domain, frequency domain and statistical methods to represent the emotional information. To address the issue of redundant and irrelevant features, an optimisation-based feature selection method is employed to identify the most informative subset of features, thereby reducing dimensionality and computational cost. These features are then used to train a neural network classifier for emotion recognition, such as happy, sad, angry and neutral. The experimental results show that the proposed method results in higher accuracy, reliability and speed. These findings show the potential of using an optimised feature selection combined with neural networks for accurate and efficient real-time emotion recognition from the EEG.