Security Enhancement in Multimodal System Fusion with Quantile Normalization for Speech and Signature Modalities

Multimodal systems, which merge two or more biometric qualities to increase security, are utilized as effective biometric systems. With the use of offline signature and voice signals, in this innovation the effectiveness of multimodal systems for fusion at feature level is tested. Modified SIFT and MFCC are employed in an effective multimodal system to extract the signature and voice features, respectively. During testing and validation, Gaussian noise was added to both samples and contrast the outcomes of the system using FAR, FRR, and system accuracy. The experimental outcomes of this research demonstrate that the fusion at feature level with msum system produces reliable results regardless of whether the data contain noise. Quantile Normalization was further used to enhance the results. This indicates that a multimodal system is effective even when the biometric data is vulnerable or corrupted to noise with the use of normalization.