Fingerprint Identification Techniques Based On Support Vector Machines And Realization

Biological recognition is one kind of technique to identifying objective individual through his/her own biological characteristics. Compared to the traditional biological recognition methods based on face or iris, the automatic fingerprint recognition is more practical and ideal. The uniqueness and stability of one person’s fingerprint are usually used as the main identification principle. Based on the computer technology, automatic fingerprint identification system presents its convenience, efficiency, safety and reliability in various applications such as the financial security, dataset encryption, e-commerce, etc. As concluded, this technique has played more and more important role in our production and daily life.Support Vector Machines (SVMs) are one of the classic classifiers in pattern recognition. In general, the whole pattern recognition system is composed of the following four main steps:1) Collection and preprocessing of samples;2) Feature extractions for samples generated from the preprocessing;3) Classification for samples obtained from the feature processing;4) Correct decisions making according to the results of classification. The key process of one pattern recognition system is the classification, in which the core is to design the classification model, namely the classifier design. The purpose to design classifiers is to train the computer automatically through the training data to learn a set of optimal structural parameters, in order to classify the input test data into the correct classes. The typical classifiers include:the k-Nearest Neighbor (k-NN) rules, the Bayesian classifiers, the Artificial Neural Networks (ANNs), the Decision Trees, and the SVMs.This paper mainly focuses on the feasibility and further potential of inplementation of SVMs on the fingerprint recognition. The main contributions of this paper are listed as follows:(1) Investigating the results on both the accuracy and efficiency of SVMs on condition of the same dataset with the different kernel parameters;(2) Suspecting the results on both the accuracy and efficiency of SVMs on condition of the various sizes of dataset with the fixed kernel parameters;(3) Comparing SVMs with the classic k-Nearest Neighbor classifiers;(4) Simulating the fingerprint recognition system based on SVMs by using Matlab software.In this paper, identification experiment was carried out by a practical fingerprint attendance system. The experimental results show that the support vector machines (SVMs) is fast and high recognition rate in this a fingerprint identification method, SVMs have great certain practical value.