Physiological signals based day-dependence analysis with metric multidimensional scaling for sentiment classification in wearable sensors

Abstract

Affective recognition has emerged in implicit human-computer interaction. Given the physiological signals in the process of affective recognition, the different positions where the physiological signal sensors are installed on the body, along with the daily habits and moods of human beings, influence the affective physiological signals. The scalar product matrix was calculated in this study, based on metric multidimensional scaling witha dissimilarity matrix. Subsequently, the matrix of individual attribute reconstructs was obtained using the principal component factor. The method proposed in this study eliminates day dependence, reduces the effect of time on the physiological signals of the affective state, and improves the accuracy of sentiment classification.