Human emotion recognition by analyzing facial expressions, heart rate and blogs using deep learning method

Development of automated systems to recognize human emotions can enhance the quality of delivery of public health service to a great extent. Due to this reason, extensive researches have been started in the recent years to recognize human emotions. This investigation proposes a novel method of human emotion recognition using analyses of both physiological and text-based features. Two physiological features (facial expressions and heart rate variability) have been considered in the present investigation. Variability in facial expressions and heart rate corresponding to different emotions have been collected by showing a bunch of emotion motivation sample movie clips corresponding to each type of emotions. In text-based analysis, various subjective information related to each of the emotional categories have been analyzed from different blogs written by persons in different emotions. The physiological and text analysis-based features have been combined and studied in Recurrent Neural Network (RNN)-based classification platform, one of the deep learning methods. Two different models of RNN-Long Short-Term Memory and Bidirectional Long Short-Term Memory have been used for this purpose. The performance of the proposed system has been evaluated using one public as well as one self-generated dataset and it yields a high recognition accuracy.