The Research On Palmprint Recognition Using Gabor Wavelet And Support Vector Machine

As an emerging biometric technology, palmprint recognition technology has been developed quickly because of the advantages of its large recognition region, easy collection, high precision and reliability. There are four parts in palmprint identification algorithms, including palmprint image preprocessing, palmprint feature extraction and feature matching.Palmprint feature extraction and feature matching were studied in this paper. The main works are as follows:(1) Having reviewed related papers published in recent years, this paper introduced the four methods in the palmprint feature extraction: structure based, statistics based, subspace based and texture&transform domain feature based methods. Then advantages and disadvantages of various methods were compared.(2) This paper proposed a palmprint identification algorithm using Gabor wavelet and support vector machine(SVM). Three steps were involved in the algorithm. First, the palmprint image was filtered by a bank of 2DGabor wavelets with five scales and four directions and downsampled to form Gabor feature matrix. Then two-dimensional principle component analysis(2DPCA) was used to extract the features into a lower dimension space. Last, SVM was used to classify the feature vectors.(3) Simulation experimental results showed the algorithm could solve small sample recognition problem, by which high recogition rate could be still kept with relatively smaller number of palmprint samples.