In this study, we developed stimuli and questionnaires to create a protocol designed to evoke a variety of emotion. Seven audio-visual film clips (joy, sadness, anger, fear, disgust, surprise, stress) were selected and surveys were taken for the verification of the suitability and effectiveness of the selected stimulation set. Using this protocol, we conducted experiments over total of 5 times. During this process, we measured the following bio signals; GSR, ECG, PPG, SKT. To extract meaningful bio signal parameter, we analyzed them in each emotion between the baseline and emotional state for 30 seconds. Through this analysis, 26 meaningful parameters were extracted. For the pattern recognition against the above mentioned emotions with these parameters, we operated the neural network, a decision tree and the discriminant analysis. In conclusion, we have found the differences of the degree of accuracy among 3 kinds of emotion classifiers.
View full abstract