Emotion induction and emotion recognition using their physiological signals

This study is related with emotions and their recognition. Three different emotional states (boredom, pain and surprise) are evoked by emotional stimuli, physiological signals (EDA, ECG, PPG and SKT) for the induced emotions are measured as the reactions of stimuli, and 27 features are extracted from their physiological signals for emotion recognition. The stimuli are used to provoke emotions and tested their appropriateness and effectiveness. Audio-visual film clips used as stimuli are captured originally from movies, documentary, and TV shows with the appropriateness of 86%, 97.3% and 94.1% for boredom, pain and surprise, respectively, and the effectiveness of 5.23 for happiness, 4.96 for pain and 6.12 for surprise (7 point Likert scale). Also, for the three emotion recognition, we propose a Fuzzy c-means clustering based neural networks using the physiological signals. The proposed model consists of three layers, namely, input, hidden and output layers. Here, fuzzy c-means clustering method, two types of polynomial and linear combination function are used as a kernel function in the input layer, the hidden layer and the output layer of neural networks, respectively. To evaluate the performance of emotion recognition of the proposed model, we use the 10-fold cross validation and a comparative analysis shows that the proposed model exhibit higher accuracy when compared with some other models that exist in the literature.