Abstract
At present, information systems are highly susceptible to unauthorized human access. Various techniques have been introduced to secure these information systems. One of the most popular techniques for verifying the user for gaining access to information systems is fingerprint recognition. The popularity of fingerprint biometric is due to its invariant behavior over age and time. A novel and accurate fingerprint identification technique is presented in this paper based on neural nature of a pRAM. The proposed method uses a methodology incorporating the use of data mappings and reinforcement learning in order to maximize the efficiency and accuracy in identifying the scanned user prints. Since, the world is moving in the era of "Internet of things (IoT)," these biometric techniques are integral to the future information securing framework. pRAM-based network is a recently introduced technique employed in pattern recognition and is different from other classical neural network models reason being that a pRAM networks gets trained in relatively less time and can be implemented in minimal hardware setup. Here the application of the permuted mapping is derived using the proposed data-based input mapping with a bit plane-encoding scheme to cover multi-gray level images. Furthermore, binarization is also done using eight binary planes and a high-resolution image is processed by dividing it into sub-images so that it can be handled by several networks in parallel. The current recognition procedure has been applied on realistic fingerprint scans/images and the results drawn have shown significant improvements. The present results drawn here prove that a pRAM structure can provide highly reliable results by introducing the permuted mapping scheme for efficient identification.