Fingerprint Classification And Fingerprint Matching Based On Embedded Hidden Markov Model Model

With the advent of electronic banking, e-commerce and smart cards, accurate automatic personal identification becomes an active research topic recently, which possesses with a wide range of potential applications. Biometrics, which refers to identification based on human physical or behavioral characteristics, is being increasingly adopted to provide positive identification with a high degree of confidence. Among all the biometric techniques, fingerprint-based authentication system is one of the most mature and proven techniques.The existing popular fingerprint identification techniques can be broadly classified into two categories: (1) minutiae-based and (2) statistics-based. Minutiae are extracted from the fingerprints and stored as sets of points in the 2-D plane. The minutiae-based matching essentially consists of finding the alignment between the fingerprints and evaluating the input minutiae sets that results in the maximum number of matching minutiae pairings. One of the major disadvantages in the minutiae-based approach is that it is very difficult to reliably extract minutiae in a poor quality fingerprint image. A number of image enhancement techniques are used to improve the quality of the fingerprint image prior to minutiae extraction, which cost the extra computation time and reduce the speed of fingerprint identification to some extents. The statistics-based fingerprint matching is based on the fact that the fingerprint images from the same finger are usually different, but the global pattern of ridges and furrows of the fingerprint images is very alike. Statistics-based techniques are less sensitive to the noise in fingerprint images, however they have problems of their own. The matching precision is not very well since the statistics-based fingerprint matching is merely based on the global pattern of ridges and furrows of the fingerprints instead of the minutiae of the fingerprints. Here, we show that the matching performance can be improved by combining the minutiae-based approaches and the statistics-based approaches. Broadly speaking, three main contributions of the research work are listed below:1) This dissertation deals with the problem of constructing statistical models of fingerprints using Embedded Hidden Markov Model (E-HMM), and put E-HMM into the fields of fingerprints identification. For a fingerprint, the different local regions vertically from top to bottom and then horizontally from left to right can be described as the state sequences of E-HMM. Fingerprint images under different conditions can be identified as observations of the sequences of states of the E-HMM for this fingerprint. One E-HMM encodes a fingerprint features. E-HMM can better encode the structure features of fingerprints in two dimensions with global and local characteristics, so fingerprint identification based on E-HMM has achieved better performance.2) Automatic fingerprint classification provides an important indexing scheme to facilitate efficient matching in large-scale fingerprint databases for any Automatic FingerprintIdentification System (AFIS). A novel method of fingerprmt classification, which is based on E-HMM that is used for analyzing the fingerprint's orientation field, is described in this dissertation. The accurate and robust fingerprint classification can be achieved by matching E-HMM parameters which were built after the processing of extracting features from a category of fingerprints, forming the samples of observation vectors and training the E-HMM parameters.3) In most fingerprint-based verification or identification systems, a reference point is one of the global structure features of a fingerprint. a reference point is established and used to extract fingerprint features robust to translation or rotation. Usually, a core point, one of the singularities of a fingerprint, is used as such a reference point. This dissertation presents an accurate and robust reference point detection algorithm based on orientation pattern labeling. In the process of minutiae matching, an input fingerprin...