A neuro-fuzzy approach for noise tuning in bistable measuring devices

It is well known that several classes of systems (biological, geophysical, electronic) show interesting features when a suitable level of noise is either forced into or naturally present in their working environment, For example, this phenomenon has been used to improve the performance of some sensing devices. The main problem in dealing with noise-added systems is the identification of the relationship between the optimal level of noise to be forced into the system and the system parameters. In previous work the authors suggested both analytical and experimental solutions to the problem of optimal noise level tuning in stochastic systems. In particular an experimental device allowing optimal system performance has been tested. In this paper a different working condition is investigated and a new nonlinear form of the noise tuning law is proposed.