Design of Parallel Rank-Order Filtering System Based on Neural Circuits of Discrete-Time

A rank-order parallel filtering (ROF) system is presented. It is modeled by system of difference equations. There are designed functional block-schemes of the system. In contrast to other analogs, the system has arbitrary high resolution of inputs, linear computational complexity and it can process unknown finite value inputs located in any known confined region. The ROF has moderate computational and hardware implementation complexity. Moreover, it does not need readjusting and appropriate control circuits. Numerical simulations showing performance of the ROF are given.