EEG-based emotion classification Model: Combined model with improved score level fusion

Background: It is far more appealing to use biological brain signals to determine human emotions. For measuring brain activity, electroencephalography (EEG) is a reliable and reasonably priced method. Conventionally, the process of detecting emotion from EEG signals involves eliminating artefacts, retrieving temporal characteristics, and finally developing a classification technique. However, to effectively detect emotional states, there is a rising need for additional stages, such as the recognition of epochs and the election of electrodes, which exhibit appreciable fluctuation in brain activity throughout emotional states. Methods: In this work, a new technique for categorizing emotions based on EEG signals is proposed. The first step is pre -processing, which makes use of suggested median filtering. After that, the feature extraction is done, where entropy, improved Stockwell transform (ST),Fourier Transform (FT)and Common Spatial Pattern(CSP)features are extracted. Further, improved recursive feature elimination is done to get rid of dimensionality issues. The step to classify emotions in EEG signals is done with Deep Convolutional Neural Network (DCNN) and Bidirectional Gated Recurrent Unit (Bi-GRU) models. The final classification result on emotions is portrayed by the improved score -level fusion (SLF). Results: The proposed HC (DCNN + Bi-GRU) achieves 0.9624 accuracy in 90 th LP over existing models. Thus the suggested work has better efficiency for the emotion classification from EEG signals. Conclusion: This proposed EEG -based emotion classification model offers a powerful tool for understanding human emotions, bridging the gap between neuroscience and practical applications.