Modeling of Olfactory Brainwaves for Odour Independent Biometric Identification

Brainwave captured through electroencephalogram (EEG) is a promising potential biometric for subject identification. EEG can be acquired, when the subject is exposed to external stimuli such as visual triggers or when the subject is resting. During biometric identification, expecting a person to be in a resting state, is not realistic; also external stimuli introduce artifacts in acquired EEG signals resulting in poor performance of EEG-based biometric systems. Odours evoke natural emotions in humans by associating strong memories with a particular smell. As a result, odours can be a potential stimulus for generating strong brainwaves; and unlike other triggers, odours can produce EEG without prominent artifacts. In this paper, an olfactory brainwave-based biometric system is proposed, where we model the subject-specific characteristics from the EEG signals captured using different odours as the trigger. We extract, from EEG signals, a set of handcrafted and diversified spectro-temporal features to train a biometric model. Experiments conducted on a publicly available dataset, show that it is possible to build an odour-independent biometric system with high-performance accuracy. Additionally, using odour-chemistry literature, we show that a small set of carefully chosen odours are sufficient to build a high-performance biometric system.