Spectrogram Contrast Enhancement Improves EEG Signal-Based Emotional Classification

Neuroscience adopts a multidimensional approach to decode thoughts and actions originating inside the brain, also called Brain Computer Interface (BCI). However, achieving high accuracy in the electroencephalography signal-based decoding remains a challenge and an open research topic in BCI research. This study aims to enhance the accuracy of signal classification for identifying human emotional states. We utilized the publicly available EEG–Audio–Video (EAV) dataset that comprises EEG recordings from 42 subjects across five emotional categories. Our key contribution is to exploit the two-dimensional contrast enhancement applied to the spectrogram for feature extraction, followed by classification using the EEGNet model. As a result, 12.5% improvement in classification accuracy over the baseline was achieved. This contribution demonstrates a potential advancement in BCI-based EEG signal processing in neuroscientific research.