FPGA implementation of neural network for linearization of thermistor characteristics

Abstract

This paper presents an FPGA (Field Programmable Gate Array) implementation of an artificial neural network (ANN) for linearization of nonlinear characteristics of a thermistor. A feed forward ANN is used for linearization. The network is trained in MATLAB with back propagation algorithm; weights and biases are determined and then implemented in Spartan-III FPGA. Subroutines are developed for single precision floating point arithmetic in IEEE-754 format.