Wavelet-Based Palmprint Recognition Using CNN and Dimensionality Reduction Techniques

Abstract

Palmprint recognition has emerged as a reliable biometric identification method due to its richness in unique features such as lines, ridges, and textures. This paper presents an advanced approach to palmprint-based identity recognition by leveraging wavelet transform techniques—Discrete Wavelet Transform (DWT), Continuous Wavelet Transform (CWT), and Multi-resolution Wavelet Transform (MWT)—combined with Convolutional Neural Networks (CNN) for classification. The proposed system uses wavelet features extracted from the PolyU palmprint database, which are further optimized through dimensionality reduction methods such as Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA). The experiments compare classification accuracies across different wavelet methods with and without dimensionality reduction. Results indicate that the DWT combined with CNN outperforms CWT and MWT in terms of accuracy and robustness. The study also explores sub-band filtering and filter bank architectures for efficient feature extraction. The findings affirm that integrating wavelet-based techniques with CNN and reduction strategies enhances the overall reliability and performance of palmprint recognition systems, making them suitable for high-security applications