Prediction based person recognition using face and speech (multi modal) for improved performance

Abstract

In recent World Technological Applications Person Recognition plays a major role in biometric security applications and it is a process of authenticating true identity of a speaker using speech or face image. This automatically recognizes the person speaking based on the speech information which includes the individual speech signals. It’s one of applications is Bio-metric applications and in order to verify each person identity, the speaker’s voice or face images have been used in the database. The importance of it becoming more popular now-a-days for security purpose and identification. In the existing work using visualization eye scan, impressions, expression scan, finger print, speech print, script for individuals of identifying the chances of theft and fraud are increasing. To address these issues, biometric voice recognition and real time face recognition system are proposed, the exclusive speaker physical appearance of a distinct can be identified. In general the individuals have different speed of speaking; therefore the sound should be adjusted, in order to match with the speed of the stored sounds templates in the memory of the proposed system. The proposed work is implemented and simulated using on Mat lab 2014A. The parallel hardware structure of the proposed work significantly reduces the time-consumption. The proposed research work provides maximum False Acceptance Rate (FAR) of 1.1765%, False Rejection Rate of 10% with an accuracy of 98.89%.