Gender Identification from the Nose Shape by the Statistical Methods

• Main Authors: You-Jyun Jian, 簡佑峻
• Other Authors: "作者未提供"
• Format: Others
• Language: zh-TW
• Published: 2011
• Online Access: http://ndltd.ncl.edu.tw/handle/82005672412034143576

碩士 === 銘傳大學 === 電子工程學系碩士班 === 99 === Identification technology of biological features is in progress. Scholars have studied fingerprint and iris, but the nose discrimination is still open. Nose is an important feature of face. The shape of nose can be divided into six type, including Roman style, Greek type, Nubian hooknose, snub nose, turn-up nose. Usually, people do not cover their noses. Thus, the recognition of the nose shape is applicable. Our methods can be divided into two steps. One is the statistics of the whole shape of nose. We use 2DPCA, LDA, and 2DLDA. The other step is to compare the features of noses. We adopt some image processing methods such as gray level, binarization, and the detection of the edge. It can find some distinguishable features such as wing of the nose, wide of the nose, foot of the mountain, wing of the nose curve. We then use logistic regression to analyze these features. Experimental results show that female is usually with smaller shape than male is. Finally, the best recognition by our simulation results is about 76 % to distinguish the human gender.