查看文章

Personal security has become an important aspect recently in many organizations. To provide a secured authentication, several biometric techniques are used. In all these techniques, details about the physiological or behavioural characteristics of a person are entered in a database. When a person uses the system, the characteristics required for the system are scanned and a template is formed. In this paper, we consider human ear biometrics since ears are the novel biometric with major advantages. Ears tend to maintain their structure and shape over time even with increasing age. The main objectives of this research are to analyse local and global feature extraction of ear using feature level fusion and matching score level fusion, to analyse the performance and to find out which approach best suits for the extraction of features with improved accuracy and security. Pre-processing is an initial step which includes noise removal using median filter, finding standard deviation, converting the input image into greyscale. The next stage involves extracting the features of ear in which first the edges are detected using zero-cross edge detection method and from the detected edge the local and global features are extracted. Thus by using these fusion methods, the performance is analysed with entropy values based on hamming distance matching technique.