A Novel Approach to Multimodal Biometric Authentication Using Ear and Palmprint

Abstract

Biometric recognition has become a vital field of study due to the growing demand for precise personal identification across various sectors, from entertainment to security. Biometrics can be categorized into two main types: unimodal and multimodal. Unimodal biometric systems rely on a single biometric trait for identification, but their effectiveness can be hampered by factors such as intra-class variations and the fact that some individuals may not possess the required trait (non-universality) [1]. On the other hand, multimodal biometric approaches employ more than one trait and render the systems more accurate and less prone to spoofing [2]. Traditional feature extraction techniques face challenges such as illumination variations, pose differences, and aging effects, which reduce the accuracy of recognition systems [3]. However, convolutional neural network (CNN)-based feature extraction techniques can learn to overcome these challenges by training on a large and diverse dataset, enabling CNNs to generalize better across variations [4]. We propose a multimodal biometric system with horizontal feature-level fusion of face, ear, and periocular region modalities, where CNN is used for feature extraction [5]. A custom dataset is introduced to account for intra-class variations and improve robustness. The system's performance is evaluated using this dataset, and the results demonstrate significant improvements in key performance metrics, including accuracy, precision, recall, and F1-score, compared to existing biometric systems [6].