Ear Based Age Recognition System Using ANN

Biometrics involve the study of specific aspects of an individual’s anatomy or physiology, such as the face, fingerprint, ear, palm print, iris, etc. Among these, ear recognition systems are notably powerful tools, particularly in forensic applications. In conventional biometric facial images are employed for age classification of individuals. One drawback of the existing system, which relies on facial images for age recognition, is that it can be less accurate when there are variations in lighting and facial expressions. Additionally, the facial recognition approach may not be as robust in scenarios where individuals may not have readily available or visible facial images, such as in forensic or security applications. In our proposed Age Recognition system, we utilize ear images and an Artificial Neural Network (ANN) to precisely determine an individual’s age. This work presents a comprehensive analysis of the extraction of soft biometric characteristic, such as age from an individual. This study explores the utilization of ear biometrics and artificial neural networks (ANN) for accurate age estimation. A unique dataset encompassing A variety of age groups, spanning from 10 to over 50 years, forms the basis of our research. Gabor features are employed in a Feed Forward Neural Network (FFNN) to facilitate age estimation. The incorporation of seven Gabor features in the input layer aids in capturing vital ear details, enhancing age estimation accuracy. Seven hidden layers within the FFNN enable the network to discern intricate patterns in the data. The output layer classifies individuals into five distinct age groups, simplifying age estimation. Our paper highlights the potential of ear biometrics in age estimation, with practical implications for security, personalization, and beyond. This work is developed with own dataset, which is divided into five age groups. System has achieved 92.05% accuracy for age recognition for 302 samples. It also achieves an accuracy of 92% for verification and 90% for validation.