earticle

영어

Human face is generally viewed as most flexible model when it comes to the field of biometric applications. The reason being it is a cost effective one due to its easy data acquisition fundamentals as well as it being invariable when considered over a fixed time period. There have been various algorithms that have been invented for the purpose of facial recognition as well as demonstrated over last three decades which are mainly categorized into kernel based technique s, Vector Methods and so on. A unique binary string generated out of facial features of a person is called a template. This template is used in several applications like network security, public key cryptography and so on. The main challenge in this regard is that no two instances of the faces acquired at different time instance can be same. The templates are similar in regard to the hamming distance but are not same. Therefore it is important to propose an algorithm that presents a face recognition system with the invariant template generation. In this work we propose a unique security architecture where face features and subsequently the generated templates are used as the key for document security. A framework needs users to register with the system along with their face instances. The instances are used for training samples. Once a user selects a folder for encryption, all the files of the folders are encrypted with the templates generated from users face data. A decryption request needs to be authenticated through the face data and the template generated at the time of decryption is used for decrypting the encrypted files. Rinjdal method is used for the cryptographic framework. Frames are acquired in real time from camera and face part is segmented based on skin segmentation. Further Eigen face based template generation and matching is used for face recognition. Results show significant low FAR in comparison to FRR and improved performance in encryption process and recognition rate.