Comprehensive Evaluation Of Appearance-Based Techniques For Palmprint Features Extraction Using Probabilistic Neural Network, Cosine Measures And Euclidean Distance Classifiers

Abstract

Most biometric systems work by comparing features extracted from a query biometric trait with those extracted from a stored biometric trait. Therefore, to a great extent, the accuracy of any biometric system is dependent on the effectiveness of its features extraction stage. With an intention to establish a suitable appearance based features extraction technique, an independent comparative study of Independent Component Analysis (ICA), Linear Discriminant Analysis (LDA) and Principal Component Analysis (PCA) algorithms for palmprint features extraction is reported in this article. Euclidean distance, Probabilistic Neural Network (PNN) and cosine measures were used as classifiers. Results obtained revealed that cosine metrics is preferable for ICA features extraction while PNN is preferable for LDA features extraction. Both PNN and Euclidean distance yielded a better recognition rate for PCA. However, ICA yielded the best recognition rate in terms of FAR and FRR followed by LDA then PCA