Ear Biometric Identification based on Gabor Filters using Backpropagation Neural Networks

Abstract

The development of reliable security systems is crucial for protecting personal information and access control. Ear biometrics, which utilizes the unique structure of the ear, is a promising method for human identification due to its resistance to forgery. This research aims to design and test an ear biometric identification system using images of the right ear without accessories from five men, totaling 224 images. The preprocessing steps include resizing the images, converting them to grayscale, and applying Gaussian filters. Image segmentation is performed using Canny edge detection, followed by morphological operations such as dilation and hole filling. Features of the ear images are extracted using Gabor filters, and classification is carried out using Backpropagation Neural Networks. The system achieved an average success rate of 88.8% across five testing scenarios, with the highest accuracy of 94% in the first and fifth scenarios. Sensitivity for classes 1, 2, 3, 4, and 5 was 98%, 74%, 92%, 96%, and 82%, respectively. Specificity reached 100% for classes 1 and 3, and 94%, 97.5%, and 94.5% for classes 2, 4, and 5. Based on the results of accuracy, sensitivity, and specificity testing, the ear biometric system using Gabor feature extraction and Backpropagation Neural Network classification demonstrates good performance and potential for security applications.