UWB based dielectric material characterization using hardware/software co-design based ANN

The feasibility of using reflected Ultra Wideband (UWB) waves in Hardware/Software (Hw/Sw) Co-design based Artificial Neural Network (ANN) framework as a non-destructive method for dielectric material characterization by estimating their relative dielectric constant, is discussed in this paper. The property of an electromagnetic wave changes owing to the effects of relative dielectric constant & conductivity of a dielectric material. Depending on the relative dielectric constant & conductivity of a dielectric material, the reflection or transmission signal changes in terms of it´s amplitude and spread. This property can be utilized to estimate the relative dielectric constant of a dielectric material. First, software implementation was carried out for feasibility analysis. In the next step, Hw/Sw co-design implementation was proposed to overcome the limitations of software implementation of ANN. These approaches are discussed and validated using FDTD simulation.