性別辨識在近年來電腦視覺領域中,是一個有趣的研究課題,在現今生活數位化的時代,人們為了追求更便利的生活方式,使得以人臉辨識為基礎之識別技術為人們所重視。人臉辨識相關技術不僅僅能用於身分辨識上,甚至在影像及多媒體的相關應用上都能看到蹤跡。 由於在影像中人臉隨著性別、年齡、體型等有很大的變異性存在,然而他們又有著大致相同的形狀,因此如果能套用在一個可變的模型,而這模型又能兼顧觀察他們大致的形狀,那樣將能達到最佳的形狀定位結果。主動形狀模型(Active shape models, ASM)被認為是一個有彈性可變化又能吸收影像資訊的一個方法,因此我們採用這個方法,有效定位人臉以及擷取人臉五官輪廓、膚色及毛髮特徵,建立Component-based辨識系統。實驗中利用毛髮特徵直接辨識男女之間較獨特特徵,如鬍鬚及禿頭,再利用人臉五官及人臉輪廓幾何特徵透過支持向量機(Support vector machine, SVM)進行辨識。 論文採用FERET人臉資料庫作為測試資料庫,其中包含了不同人種、年齡、表情、人臉傾斜角度以及亮度變化,從中取出900張(男女各450張)正面人臉影像進行實驗測試。實驗結果顯示,在不同人種、年齡、表情、人臉傾斜角度以及亮度變化下能夠有效的辨識人臉之性別。
Gender classification is a branch of face recognition can be used as pre-treatment or in combination with other identification to improve recognition results. In addition, the human face recognition technology not only can be used for identity, even in the images and related multimedia applications can see traces. Because of face in image exists very large variation that is followed sex, age and figure. But they have almost the same shape. So if we apply it to a changeable model to observe the rough shape of the model. That must achieve the best result of shape of position. ASM is considered a method that is elastic to absorb the information of image. So we take the method to posit face, fetch outline, beige and character of hair to build Component-based recognition system. We take hair to recognize character between man and woman in experiment, like beard and bald, then take SVM to recognize face and outline of human being. We use FERET face database, contains different race, age, expression, face rotation, and brightness changes, in order to compare with other methods, we use the same method to pick out 900 (450 males and 450 females) frontal face image Experimental test. According to the experimental results, the different race, age, expression, face rotation and brightness changes, can be effective of face gender recognition.