Color-Based Non-Contact Analysis of Skin Changed by Sweating for Emotion Estimation

摘要

In this study, we analyzed the visual difference between dry skin and wet skin for non-contact sweat detection. Recently, there are many researches on emotion recognition using some indexes related to emotion such as electrocardiogram, respiratory, skin temperature, and so on. This is because using some indexes is said to make the accuracy of emotion recognition higher and to help to recognize more various and complex emotions. Also, non-contact method is needed from view point of infection, simplicity when using, and so on. In this paper, we focused on the sweating because detecting the sweating is expected to give us important information on emotion and sympathetic nerve system. For instance, feeling nervous makes palm sweating. Therefore, we considered sweating detection may be useful for emotion recognition. Sweating is conventionally evaluated based on electrodermal activity as known as skin conductance, and that is obtained with a contact method such as attaching electrodes. However, contact methods can cause infection and prevent subjects from concentrating on stimulus to evoke emotion when sweating is used as index of emotion for emotion recognition. Then, we focused on color change caused by wetness. Because total reflection is occurred at water-air interface by water on a material and it makes the light paths longer, the color of material is changed as getting darker and more saturated by wetness. Therefore, we thought that the color change can be useful index for non-contact sweat detection. In this study, we obtained feature values with respect to color using image processing and compared those feature values of dry palm and wet palm. As a result, we found that wetness makes the mean of luminance component smaller, the mean of saturation larger, and the entropy of hue smaller.