Machine Learning Models for Classification of Human Emotions Using Multivariate Brain Signals

Humans can portray different expressions contrary to their emotional state of mind. Therefore, it is difficult judge humans’ real simply by judging physical appearance. Although researchers are working on facial analysis, voice recognition, and gesture recognition; the accuracy levels such analysis much less results not reliable. Hence, becomes vital have realistic emotion detector. Electroencephalogram (EEG) signals remain neutral external appearance behavior human help in ensuring accurate The EEG from various electrodes scalp regions studied for performance. has gained attention over time obtain classification states beings human–machine interaction as well design a program where an individual could perform self-analysis his state. In proposed scheme, we extract power spectral densities multivariate sections brain. From extracted density (PSD), features which provide better feature selected classified using long short-term memory (LSTM) bi-directional (Bi-LSTM). 2-D model considered frontal, parietal, temporal, occipital studied. region-based performed considering positive negative emotions. performance our previous model’s artificial neural network (ANN), support vector machine (SVM), K-nearest neighbor (K-NN), LSTM was compared 94.95% received Bi-LSTM four prefrontal electrodes.