Towards a Lognormal Model of Perception

In this contribution, an information theoretic approach to human perception is proposed. Initially, a model is derived to characterize the stimulus se, as perceived by the brain, in response to an event e. Then, an optimal estimation algorithm is conceived to infer the information ie brought by e, knowing the stochastic properties of se. In this way, it becomes possible to mimic the short term brain response to an exogenous stimulus. In addition, a Bayesian framework is supposed to learn the parameters that characterize se from past experiences, thus emulating the long term response of the brain. Finally, a generalization of the proposed approach is discussed to demonstrate its flexibility and adaptability to different contexts.