Convolutional Neural Network Approach for Multimodal Biometric Recognition System for Banking Sector on Fusion of Face and Finger
Sandeep Kumar, Shilpa Choudhary, Swathi Gowroju, Abhishek Bhola · 2023
In the last 10 years, fingerprint recognition has become popular because it is now a standard function on most mobile devices, tablets, and PCs. In addition to the security benefits this kind of biometric scanner offers at work, more and more businesses are substituting passwords, ID cards, and door access codes to track attendance and manage their staff. It continues to be plagued by variances, such as print traits, including alignment, edge orientation shift, arches, swirls, and whorls. The face is almost unaffected since it has a solid 3D structure compared to the finger. More application areas can use the face and finger because they can be taken from a distance without being obtrusive. Due to its physiological makeup and placement, the finger may easily replace the face for biometric identification. Combining the face and finger has become famous for nonintrusive multimodal recognition to increase security, durability, and accuracy. A multimodal system achieves a better result than a unimodal system because of the fusion rule. This article describes a machine learning-based multimodal biometrics fusion method. Data pretreatment is accomplished by data transformation. A 2D filter was utilized to examine the texture of local subblocks to extract the phase information of multimodal biometrics data. A proposed algorithm was created for the multimodal integration of biometrics. We evaluated the effectiveness of the suggested approach using the finger and face data sets. Following the findings, the quality of fused images is higher, the accuracy of feature extraction is between 91% and 95%, the average accuracy is 97%, the multimodal biometric impact is positive, and the practicability is reasonable.