Deep Biometrics: Exploring the Intersection of Deep Learning and Biometric Applications
A. Ashwini, V. Kavitha, S Balasubramaniam · 2024
Recent developments in the field of machine learning, such as image processing, speech processing and computer vision, have shown how deep learning techniques have evolved. Biometrics systems that can analyse and quantify human physical or behavioural attributes, are increasingly common and have a wide range of uses. The enablers of such deep biometric system models are the Internet of Things (IoT) and Artificial Intelligence (AI). When used in conjunction with deep neural networks and machine learning, AI-assisted authentication can be tailored to a variety of goals. The key goal of the research work is understanding the authentication function using deep learning in enhancing the security of biometric-based systems. Eighty-four distinct methods that investigate deep learning for identifying people using multiple biometric modalities are reviewed. Commonly used deep learning architectures include convolutional auto-encoders, stacking auto-encoders, convolutional neural networks, restricted boltzmann machines, recurrent neural networks, and deep belief networks. In recent years, AI-enhanced authentication methods have become increasingly popular, with AI-assisted authentication techniques setting the standard for security and adaptability to as many real-world scenarios as feasible. This chapter summarizes these approaches and explores various challenges with improvements and future works in AI-based biometric applications.