Analysis on Temporal Sparsity of Human Face for Various Emotion Stimuli

Recently, many researchers have been focused on partial region of single frame image such as mouth, eyes in order to recognize facial expression. However, according to the recent psychological theory, the other regions of face included important factors in terms of facial expression. To solve the problem of previous works, our proposed method verifies facial temporal movement in the face region. For that, one hundred regions (10×10) are defined onto the detected face region. In addition, temporal sparsity of each region was calculated on specific frequency band well known to express micro-movement of human body. To perform the calculation, adjacent frame subtraction method was used. Consequently, sparsity data against four emotion stimuli was obtained. At result, we confirmed that spatial characteristics of temporal sparsity were different accordance with different emotions.