Multi-Modal Biometric Recognition Using Human Iris and Dynamic Pressure Variation of Handwritten Signatures
Vinayak Ashok Bharadi, Bhavesh Pandva, Georgina Cosma · 2018
Physiological traits containing information extracted from human organs such as fingers, ears, eyes, palm prints, and knuckles can be used for biometric recognition. In addition, behavioural biometric traits such as speech, gait, handwritten signature, and keyboard dynamics while typing a text can provide reliable biometric features. In the current research, texture information extracted from iris data, and dynamic pressure variation data extracted from online signatures was combined to form a reliable biometric system. In this view, this paper is proposing a multi-modal biometric recognition architecture, which utilises a new feature vector extraction mechanism based on the Webber Local Descriptor and the k-NN machine learning classifier. The proposed architecture was evaluated using multimodal data obtained from 64 users. The TAR-TRR results when using iris recognition alone reached 88.39%. When using dynamic pressure variation data extracted from online signatures TAR-TRR reached 75.89%. When combining the data of the two modalities the recognition TAR-TRR increased to 90.18%. The proposed architecture revealed that biometric systems which perform recognition using two different data channels are more reliable that when using single channel data.