Security Identity: An Audio-Video Biometric Approach

Audio-Video biometrics finds applicability in various sectors such as banking, network security, and e-commerce. Secure audio biometric authentication refers to a system that can verify a person's identity using their unique voice characteristics. Incorporating visual information can complement audio information, by verifying the person's face and hence improve the system's accuracy. The paper aims at developing an audio-video biometric authentication system, capable of overcoming the challenges such as audio replay attacks and audio forgery attacks. A CNN-based facial recognition system as well as a MoE-based audio recognition system is implemented in this work. The combined system achieved an accuracy of 85% in identifying the user based on the audio and video inputs. An app is also developed to serve as a front-end for the classifiers in order to facilitate practical interfacing.