Vision Based Body Dither Measurement for Estimating Human Emotion Parameters

摘要

In this paper, we propose a new body dither analyzing method in order to estimating various kinds of intention and emotion of human. In previous researches for quantitatively measuring human intention and emotion, many kinds of physiological sensors such as ECG, PPG, GSR, SKT, and EEG have been adopted. However, these sensor based methods may supply inconvenience caused by sensor attachment to user. Also, therefrom caused negative emotion can be a noise factor in terms of measuring particular emotion. To solve these problems, we focus on facial dither by analyzing successive image frames captured from conventional webcam. For that, face region is firstly detected from the captured upper body image. Then, the amount of facial movement is calculated by subtracting adjacency two image frames. Since the calculated successive values of facial movement has the form of 1D temporal signal, all of conventional temporal signal processing methods can be used to analysis that. Results of feasibility test by inducing positive and negative emotions showed that more facial movement when inducing positive emotion was occurred compared with the case of negative emotion.