Study On Some Key Issues Of Ear And Face Multi-Biometrics Recognition

In recent years, the new identification technology are developing rapidly, particularly in biometric technology represented a more quickly. Based on biometric identification technology has been widely applied to public safety, information security and other places, and achieved good social effects.Although biometric identification technology has many other advantages that are not,but it also affected by many factors,such as noise interference,non-universal and anti-fraud capacity to have significant deficiencies.The disadvantage for a single biometric, people put forward a multi-biometric fusion identification. Experiments have shown that the fusion of multiple biometrics will not only improve the final recognition rate, verify accuracy, but also can expand the use of the system. Especially in the last few years, the synthesis and biological features and more integration of the development of biometric identification systems Biometric technology has become a research focus areas of development, but also of difficulties.More than the human ear and face recognition biometric information fusion is an emerging biometric technology, is still in its infancy, there are many critical issues still not resolved, this paper from a single biological feature extraction and multi-biometric fusion stage of consideration.In the more ear and face biometrics process, for a single biometric feature extraction is also a very important part of the biological characteristics of the pros and cons of extraction of the final verdict will have a significant impact. Ear recognition as an emerging biometric technology, current research is not enough depth, and facial features relative to the human ear feature points less.This article feature extraction from human ear paper proposed a tensor based on principal component analysis method for ear recognition. First, the human ear image wavelet transform, to obtain four sub-band signals, then the four sub-band signals tensor PCA feature extraction, and finally the nearest neighbor method to identify, test results show that tensor principal component analysis algorithm for biological feature extraction, can obtain good recognition results.Ear and face in the integration layer, the paper also presents a decision-making based on D-S theory of evidence fusion layer, the conventional DS evidence theory to solve the problems, improved D-S theory of evidence is the same traditional DS evidence theory appropriate. The D-S theory of evidence is modified in the original conventional D-S theory of evidence based on the improvement, through experiments that the use of improved D-S theory of evidence can improve recognition accuracy.The proposed feature extraction algorithm and the improved version of the DS evidence theory, experimental results show the effectiveness of the method, but there are inadequacies in the majority of researchers need to be further improved.