講演抄録/キーワード |
講演名 |
2014-03-17 11:00
不完全な顔情報からの表情識別 ○平澤諒一(電通大) |
抄録 |
(和) |
顔の一部が隠れる状態,特にマスクをして口と鼻が隠れている状態を想定し,不完全な顔情報を用いた表情識別を行った.目の周辺の8個の特徴点をActive Appearance Modelsを用いて抽出し,各特徴点同士の距離の二乗や角度といった特徴量を求め,特徴量をもとにSupport Vector Machineによる機械学習を行った結果,37.58%の適合率および37.12%の再現率を得た. |
(英) |
The author tried facial expression identification in supposed state, that some part of the face is hidden, especially the nose and the mouth is hidden by the mask. 8 feature points around the eyes were extracted by using Active Appearance Models. The author defined feature values as square of the distance between the feature points, and angle. Machine learning was carried out by Support Vector Machine based on the feature value. The recall was 37.58% and the precision was 37.12%. |
キーワード |
(和) |
画像処理 / 表情識別 / 特徴点 / 識別器 / / / / |
(英) |
Image processing / Facial identification / Feature points / Classifier / / / / |
文献情報 |
映情学技報, vol. 38, pp. 101-102, 2014年3月. |
資料番号 |
|
発行日 |
2014-03-10 (AIT) |
ISSN |
Print edition: ISSN 1342-6893 |
PDFダウンロード |
|