Parallel Computation Of The Dwdf Based On Neural Schemes

Abstract

The Wigner Distribution Function is a joint representation of signals that offers excellent properties to be applied to signal processing. Unfortunately, it requires high computational cost to generate, making this distribution function less than adequate for use in real-time processes. The paper proposes parallel computational schemes based on neural models to compute the kernel of the Discrete Wigner Distribution Function (DWDF), enhancing its inherent parallelism in oreder for it to be implemented with VLSI technology. The specification of these schemes implies the existence of two types of neurons: input neurons and computational neurons. The specification also determines the connections, their associated weights, and how the input values are presented to the models.