EEG-Based Emotion Recognition Using Machine Learning

Abstract

The ability to recognise and categorise emotions is now being manifested through the knowledge in the Brain Machine Interface. Numerous research have suggested that EEG brain waves may be used to detect emotions. Here, new learning method is proposed in identifying the emotions through the use of latest deep learning concepts using long short term memory (LSTM) recurring neural networks. The acquired brain wave signals are processed for classification using convolution neura l network (CNN) and then given to the proposed algorithm for specific emotion recognition. First, we reduce the noise in EEG signal using pre-processing and the processed EEG signal will be directly sent into the input layer of the CNN network structure. The LSTM is one of the widely used deep learning methods which is based on RNN that are efficient on prediction performance. An LSTM contains a crucial hidden module known as the memory module. It has three main gates in the structure of RNN that is input gate, candidate gate, forget gate and Output gate respectively. The Input gate is used for reading the data from the dataset and the forget gate is used for the purpose of storing the data. Then the output gates are used for the writing purpose. Then, we use the network structure based on LSTM and CNN, fully extracting the features and dependency relationship in EEG signal. At last, classification will be realized through the softmax classifier.