Uncertainty Management in Brain Data for Olfactory Perceptual-Ability Assessment of Human Subjects by General Type-2 Fuzzy Reasoning

Abstract

The paper introduces an interesting approach in assessing olfactory perceptual-ability and its gradual degradation over months for both healthy persons and people suffering from early olfactory ailments. Functional near infrared spectroscopic (f-NIRs) data acquired from the experimental subjects’ brain are first pre-processes and then fed to a novel general type-2 fuzzy regression unit to predict the subjective olfactory perceptual-ability. The model parameters are corrected using subjective feedback about the olfactory stimuli concentration. During the test phase, the model is used to predict perceptual degradation in olfaction for patients suffering from olfactory ailments. The prediction error computed with respect to subject's self-assessment of stimulus concentration is used as a metric of performance of the proposed prediction algorithm. The original contribution of the work lies in the formulation to handle uncertainty in multi-trial and multi-session experimental brain data using general type-2 fuzzy logic. The proposed technique outperforms traditional type-2 fuzzy techniques both with respect to percent success rate and run-time complexity.