以臉部器官特徵作多媒體性別分類之研究

近幾年，由於安全上的需求，所以利用人臉來進行身分辨識的應用越來越廣泛。在人臉辨識的相關研究中，所要加強和克服的不外乎辨識的準確率，而輸入影像的前處理步驟及輸入影像的內容也將明顯影響辨識的準確性。之前許多的研究都是拿整張人臉去做辨識。在本論文中是將人臉分成不同的區域，然後分別將這些區域去做辨識。因為整張人臉包含了許多膚色區域，而這些膚色區域對於表示人的特徵上並沒有太大的幫助，反而會因為光線或環境的變化去影響這些膚色區域，所以為了能減少這些膚色區域對辨識的影響，提出以人臉五官區塊取代整張人臉的辨識方法。本實驗採用網路蒐集的臉部圖片所構成的資料庫，進行實驗測試。系統架構主要包含三個模組：影像前處理、特徵萃取與分類辨識模組。輸入一人臉部特徵，由影像前處理模組利用影像處理演算法，自臉部影像中取得所需的特徵影像。再經由特徵萃取模組以二維主成份分析（two-dimensional principal components analysis， 2D-PCA）運算出男女性別的臉部特徵碼（face feature codes）。最後進入分類辨識模組，利用特徵碼進行比較分類的步驟，來完成本論文所需要辨識的目的。

關鍵字

生物辨識；臉部辨識；性別分類；二維主成份分析

並列摘要

Recently, because of the necessity of social security, the research about personal identification using face recognition is widely studied. The relevant research of face identification has to strengthen and overcome the accurate rate of recognition. The previous treatment steps and content of input image affect accuracy of recognition obviously. The so-called content of image means the environment and the changes of lighting on face, and this factor affects the rate of recognition deeply. Many research focus on taking the whole face image for recognition. This thesis focuses on dividing the face image into different areas, then taking these different areas to recognize. The face image is often full of skin areas which have less help to recognition but damage. Furthermore, the color of skin areas is easily influenced by the lighting condition or changes of the environment. In order to prevent the above drawbacks which may decrease the accuracy of face recognition, the paper proposes a method which fetches different areas of the face as features. Therefore this paper proposes with the five senses of person face sub-area substitution whole piece person face identification method. This experiment uses the database which collected the face picture by the network to do the experiment. The system construction mainly contains three mold trains: image pre-processing, feature extraction and classified recognition module. Image preprocessing module uses some image processing algorithms to localize the region of interest of face from the input image. And then the extraction module use the characteristics of Two-dimensional Principal Component Analysis (2D-PCA) computing a signature facial to identify men and women. Finally, the classified recognition module compares the signature classification using the steps to complete this paper which required for the purpose of identification.