Multimodal biometric system based on feature source compaction and the proposed VCG (Virtual Center of Gravity) feature

Abstract

The multimodal biometric system is a biometric system that uses more than one biometric characteristic and vice versa for unimodal biometric systems. There are also two types of multimodal systems regarding their authenticity techniques: serial and parallel authentication. In this study, the proposed multimodal biometric system uses a different technique, namely the feature source compaction technique. Each recorded biometric characteristic data (voice, face, and fingerprint) is converted into a grayscale image (called a feature source image). All feature source images of the same sample are arranged into a 3D image. The proposed VCG feature is extracted from the texture image resulting from applying the feature source compaction technique. In this way, only one generated feature for each data sample (as opposed to other serial or parallel multimodal systems). This study uses three genuine users with five data samples for each and 25 fake users. The authentication stage has tested using all data samples from all users as guest users. It means that there will be a total of 35 guest users for each of the three prototypes of VCG features. There are three test scenarios (in addition to the proposed method) used to determine the compaction sub-stage's effect on the resulting VCG feature prototype. The study results showed that the proposed method has a FAR range with the smallest limit value (0 – 17.33%) and the highest accuracy (85.56% - 100%). It has proven that the proposed method is much better than the other three scenarios.