Brando Koch is the owner of Intelligentia, a company focusing on building end-to-end AI and data solutions. He has 5+ years of experience in machine learning research and development, having been a part of projects covering ML, NLP, CV.
He obtained his Master’s degree in Computer science at FERIT, Josip Juraj Strossmayer University of Osijek.
Exploiting Physical and Behavioral Features During Speech for Visual Biometric Authentication
The ability to authenticate an individual for the purpose of granting access rights is a fundamental aspect of any security system. This is accomplished by utilizing something the person possesses, knows, or inherently is. Biometric authentication operates on the premise of what a person “is”, leveraging their distinct set of characteristics that set them apart from others.
Facial features are one of the most researched biometric features; however, they still exhibit vulnerabilities, particularly to presentation attacks. These vulnerabilities require developing methods to detect such attacks (presentation attack detection) or to verify the genuine presence of a live individual using liveness detection techniques.
A novel approach to robust biometric authentication is visual lip-based biometric authentication (LBBA). This method focuses on video features around the mouth region of the face during speech. This is for two primary reasons. First, there is the physical uniqueness of a person’s mouth region, such as the shape and texture of lips. Second, behavioral uniqueness is observed from visual patterns represented by movements during speech.
In this talk, we will explore the latest deep learning research in this area, as well as the method’s strengths and weaknesses.