Characterisation and recognition of biometric authentication using human naturalistic facial expressions

Abstract

Tremendous growth in the use of biometric technology in high security applications has motivated the requirement for highly dependable face recognition systems with expressional robustness. Current standards demonstrate that the accuracy of state-of-the-art algorithms is sufficient only for certain application but they would not perform significantly for images exhibiting facial expressions variations. In this research work, an attempt is made to evolve a Genetic Algorithm (GA) based on characteristics classification of human behaviour. This research work tries to achieve insensitivity to facial variations that lead to the motivation in classifying the physiological properties of the facial expressions and mapping it to the behavioural characters underlying the expression. The proposed model also incorporates the asymmetric cryptosystem into the advantages of biometrics in the domain of asymmetric cryptosystems. The use of invariant features as a key to produce a hierarchical security system, the same key (face recognition) can be used to generate encrypted messages at different levels of security. The proposed system describes and evaluates a resilient face expressional recognition system using a geometric structural representation of the various expressions like happiness, sadness, anger, fear, surprise, disgust and mapping it to the behavioural traits stored in the form of either hidden or exposed property in the genes adopted in genetic algorithm model. The experimental evaluations are conducted taking 10 samples of six different level of emotions to human being using the faces of 63 persons (21 men and 42 women). The geometric structures related to the different expressions are identified and stored in the template of training samples. The behaviour pattern mapping of different expressions are associated with genetic properties. The matching is performed with the template of training sample to evaluate the recognition and its rate of matching.