Image-based measurement of changes to skin texture using piloerection for emotion estimation

Mihiro Uchida, Rina Akaho, Keiko Ogawa-Ochiai, Norimichi Tsumura

Abstract: In this paper, we find effective feature values for skin texture as captured by a non-contact camera to monitor piloerection on the skin to estimate emotion. Piloerection is observed as goose bumps on the skin when a person is emotionally moved or scared. This phenomenon is caused by the contraction of the arrector pili muscles with the activation of the sympathetic nervous system. Piloerection changes skin texture, because of which we think it effective to examine skin texture to estimate the subject’s emotions. Skin texture is important in the cosmetic industry to evaluate skin condition. Therefore, we thought that it will be effective to evaluate the condition of skin texture for emotion estimation. Evaluations were performed by extracting effective feature values from skin textures captured by using a high-resolution camera, where these feature values should be highly correlated with the degree of piloerection. The results showed that the feature value “the standard deviation of short-line inclination angles in texture” was satisfactorily correlated with the degree of piloerection.