Research On Recognition Of Hand Feature Fusion Based On Contourlet And Subspace Analysis

Palmprint recognition is an emerging biometric technology and one becoming of hot issues because of its characteristics of stability and unique.Now it has been used in the identification and user authentication. Palmprint recognition includes four steps,these are palmprint image acquisition and preprocessing, feature extraction, feature matching and decision-making, and the feature extraction algorithm is the key of palmprint recognition, it decides the result of recognition. This paper bases on learning existing palmprint recognition algorithm, on the superiority of Contourlet transform and subspace analysis, bases on the fusion methods, paper proposes a recognition of hand feature fusion based on Contourlet and subspace analysis. The main contributions can be exhibited by the following aspects:1. Palmprint recognition algorithm based on Contourlet transform and NMF. Three advantages of Contourlet Transform as follow:first, it directed well, anisotropy, capture the information of edge image well. Second, it has features of shift-invariant that can be effectively reduce the registration error on fusion results influence in the fusion process. Third, base on these characteristics what each scale of transformed subband images are same as the size of source images, the algorithm is easy to find the correspondence between each sub-band image. The NMF can decompose the matrix when all the non-negative elements are constraints. It can obtain the local characteristics of source data by selecting the dimension of feature space appropriately in NMF. Combining the advantages of Contourlet transform and NMF, the algorithm of this paper transformed by Contourlet wavelet on source the recognition rate of the palmprint.2. This paper proposes a palmprint recognition algorithm by Contourlet transform and ICA. It can improve the recognition rate.3. This paper proposes a palmprint recognition algorithm of multi-mode hand feature fusion based on Contourlet and ICA. Firstly this algorithm extraction palmprint and knuckprint features based on Contourlet and ICA,then fusion two hand features used a simple integration method,finally recognize classify by nearest neighbor classification. Contrasting the recognition effect of palm, knuckleprint and fusion of two modal in experiment, results display that recognition algorithm of multi-mode hand feature fusion has validity.