Auditory Behavioural Estimation from Stochastic Properties of Neural Response

Auditory human behavioural phenomena can be estimated from the stochastic properties of the neural response. In this paper we estimate the precision a human can localize sound sources. The neural activity is modelled as a non-homogenous Poisson process, and human ability is determined by optimal lower bounds. Both Cramer-Rao (CRLB) and Barenkin (BLB) bounds are calculated. They both predicts the psychoacoustical data if loss of synchrony in the neural response is considered. Only BLB that accounts for the ambiguous perception of sound sources matches experimental data also when the loss of synchrony is ignored.