Place: VIVA Room, Level 7, School of Computer Sciences
Since past three decades, the world is transformed into a digital society, where every individual is living with a unique digital identifier. The purpose of digital identifier is to distinguish one from the other as well as to work with digital machines which are surrounding the world. Simultaneously, the security level is the most important part in the modern digital society. Several user authentication systems have been proposed to verify the claimant and prevent imposters using password, token key, MAC address, Id card...etc. In contrast, the digital identifier which will be used for authentication should be based on physical and behavioural characteristics of the person, where this kind of authentication is called Biometric authentication. There are several types of biometrics such as facial recognition, iris recognition, fingerprint, voice recognition...etc.
A new biometric user authentication system that is based on graphical recording of the brain electrical activity is called Electroencephalogram (EEG) signals is proposed in this dissertation.
In general, this dissertation involves four contributions which are i) adapting a meta-heuristic algorithm to find the optimal wavelet parameters for EEG signal denoising. ii) proposing a new solution for decomposition level representation. iii) designing a multi-objective function to obtain the efficient feature extraction. v) hybridization between flower pollination algorithm and β-hill climbing algorithm to select the optimal EEG channels which are providing the highest accuracy in the authentication process.