Key Research Highlights > Cybersecurity > Dynamic Facial Signature
Terence SIM |
Dynamic Facial Signature
Objective To explore whether computers can recognise people using a psychology concept known as “dynamic facial signature”, which refers to the characteristic facial motions of individuals that are used by humans to recognise familiar faces. Results We have developed a method to recognise individuals based on dynamic facial signature. Our method, known as Local Deformation Profile (LDP), is a combination of displacement vectors and skin deformation patterns collected from tracker points on a face. LDP is a cross-expression feature, which means the identity evidence can be collected from any kind of facial motions. The performance of LDP increases as more different kinds of facial motions are covered in training. LDP can even recognise faces under heavy face makeup. |