A Fast and Light Fingerprint-Matching Model Based on Deep Learning Approaches

Nowadays, biometric identification has become very important due to the need to identify people in different places and devices. Among these features, the fingerprint has received more attention than others because of its biometric criteria and the ability to use easily and quickly. Neural network-based methods received considerable attention due to their high accuracy and performance. These methods also do not need data preprocessing and image segmentation. In identification systems, hardware implementation capability is critical. This paper proposes a novel convolutional neural network architecture for identification using fingerprints. The proposed architecture in this paper offers more than 94% accuracy using different databases. Also, by reducing the number of parameters and memory used by more than 75% compared to state-of-the-art counterparts and the number of convolutional layers, the proposed architecture is hardware friendly and offers at least 10% better speed than the state-of-the-art counterparts.

Automated summary

Nowadays, biometric identification is crucial for identifying people in various places and devices. Among these features, fingerprint has gained more attention due to its biometric criteria and ease of use. Neural network-based methods have gained significant attention for their high accuracy and performance, and they do not require data preprocessing and image segmentation. In this paper, a novel convolutional neural network architecture for fingerprint identification is proposed, achieving an accuracy rate of over 94% on different databases. The proposed architecture also reduces the number of parameters and memory usage by more than 75% compared to existing models and offers at least 10% better speed.