摘要
人耳识别作为一种新的生物特征识别技术,正在逐步成为该领域中的热点。本文针对眼、嘴、外耳轮廓的结构关系,提出一种多模态生物特征识别方法。该方法提取的特征向量包括八项特征,比如外耳轮廓宽度与两瞳孔之间水平距离之比、外耳轮廓高度与瞳孔和嘴之间的垂直距离之比、脸的宽度与两瞳孔之间水平距离之比,然后通过模式识别对样本进行分类。由于提取的特征均以相对量为主,所以该算法具有"缩放性",同时对噪声具有较强鲁棒性,识别效果较好。
As a new subject of biometrics, automatic ear biometrics became a focus in this field step by step. According to the relation of structure of eyes, mouth and ear contour, this paper discusses the multimode biometrics method, The feature vector includes eight characters, for instance, it consists of the proportion of ear＇s width compared with pupils＇ horizontal distance and the propor- tion of ear＇s height compared with vertical distance between eyes and mouth,, then it will classify stylebook by pattern recognition, The algorithm is zoomv because the vector is composed of relative values. In addition, This method is robust for removing the noise, and the results of experiment show that it＇s rather successful in recognition.
出处
《微计算机信息》
北大核心
2007年第34期226-228,共3页
Control & Automation
基金
教育部"春晖计划"科研合作项目
基于人耳识别的个人身份小型快速鉴别系统关键技术研究(Z2005-2-11009)
