Classification of human state emotion from physiological signal pattern using pulse sensor based on learning vector quantization

Many studies have shown that there is a close relationship between emotions and human health. Negative emotions have a bad influence on patients with chronic diseases, otherwise positive emotions have a good effect on human health. Early detection on emotional state of patients with chronic diseases is necessary to avoid more severe consequences. This study is a preliminary approach to achieve that goal. In this research 6 emotional states were identified via a video stimuli by using a pulse sensor, there were disgust, fear, surprised, angry, sad and happy. Thirty participants were involved in this research. Ten parameters were anayzed via pulse sensor such as mean value of the data relative to the baseline threshold, min and max value, standard deviation, number of Onset and Offset, and number and total of amplitude value, total of rising time, and time of arrousal. Optimum parameters for data classification using Learning Vector Quantization was presented ed in this study. The accuracy of LVQ in predicting emotion according to the physiological data was 68, 52% with precentage split 70%. This preliminary study showed that some improvement is needed so that each emotion state can be recognized uniquely by using physiological data from the pulse sensor.