关键词: 虹膜识别, 特征提取, 经验模式分解, 奇异值分解

Abstract:

During the extraction of iris features by the wavelet transform and Gabor filter,the wavelet basis function is fixed and Gabor filter parameters should be optimized.In order to solve these problems,a new extraction method of iris features is proposed based on the empirical mode decomposition（EMD） and the singular value decomposition（SVD）.In this method,first,the preprocessed iris image is decomposed via EMD,and a series of intrinsic mode functions and residual components are obtained to construct the initial feature vector matrixes.Then,the matrixes are decomposed via SVD,and the corresponding singular value is taken as the iris feature vector.Finally,iris features are identified by using a Modest AdaBoost classifier.Experimental results show that the proposed method helps to obtain low-dimension feature vector with higher recognition rate and lower time complexity of feature extraction and matching.

Key words: iris recognition, feature extraction, empirical mode decomposition, singular value decomposition