空間周波数を用いた表情認識の検討(コミュニケーション支援及び一般)

概要

論文の詳細を見る
We investigated facial expression recognition using spatial frequency components of facial images (happiness, sadness, surprise, fear, anger, and disgust). Spatial frequency components in low and middle range were used for recognition. The results show that "surprise", "happiness", "anger", "sadness" and "disgust" were recognized very well, however, the correct recognition rate of "fear" was low. This paper described t the way how to processed the facial images and the criteria of facial expression recognition using the spatial frequency components. We tried to test various ways to process facial images and then we found that the following method was the best, 1) size and horizontal direction adjustment, 2) clip the eye and mouth from the image, 3) differentiate the expression face image from the neutral face, 4) binarization. And then, we discussed the algorism of facial expression recognition by using spatial frequency components. We found that the best recognition rate was achieved when the intersections of horizontal and vertical spatial components of 1, 2, 4, 8, 16 and 32 c/f were used.
社団法人電子情報通信学会の論文
2006-05-22