Abstract
Technological resource assessments security is a major concerned and Biometric is one of the most robust identification techniques. The common approach for the biometric identification process is to compare the extracted feature vectors of query imposter with feature vectors of rest imposters. In biometric recognition, the datasets have very large number of imposters and this imposes the condition on the identification process. To make the identification process fast, dimensionality reduction is required at either dataset or in feature vectors. This paper proposes the palmprint identification algorithm with dimensionality reduction at datasets as it reduces feature vector size too. One Dimensional Principle component analysis (1DPCA) cannot correlate the neighbor pixels and transformation from 2-dimension-to-1-dimension increases the computation cost. Therefore, two Dimensional PCA (2DPCA) is employed to process the dataset fast in comparison with 1DPCA. For classification, Supervised learning-based classifier provides higher accuracy and hence Support Vector Machine (SVM) classifier is used for recognition. The success of the classifier depends on the extracted features to be matched. The proposed algorithm uses Histogram of Gradient (HOG) features which is the best combination with SVM. Accuracy of the proposed algorithm is compared with the accuracy of other models. The experiment results and comparative analysis on PolyU datasets reveal that the proposed algorithm achieves 96.36% accuracy which is best amongst all.