表情表出時の顔特徴点の変化を用いたアンサンブル学習による表情認識

摘要

様々な表情が表出されている顔画像から，眉，目，口の端点や中心点など，表情の長州こ閑適する顔特徴点を抽出し，表情表出時の顔特徴点の変化に基づいて表情を認識する手法を提案する．特徴点間の距離変化や特徴点周囲の画素の輝度値変化など，単純な特徴量を用いることにより，効率的に表情認識を行う．また，少数のわ用な特徴量を用いてAdaBoost に基づくアンサンブル学習を行い，認識精度の向上を図る．%We propose a facial expression recognition method using several facial feature points such as end points and centroids of eyebrows, eyes and mouth, extracted from facial images including various facial expressions. The proposed method performs efficient facial expression recognition by utilizing simple facial features including the distance between two facial feature points and the intensity distributions around several facial feature points. In addition, the recognition accuracy can be enhanced by introducing an ensemble learning based on AdaBoost using a small number of useful facial features.