Neural-network-based method of correction in a nonlinear dynamic measuring system

This paper addresses the problem of improving the quality of measurement calibration and reconstruction using an artificial neural network (ANN) for a linear and nonlinear dynamic measuring system. The reconstruction consists of a regularized inversion of the operator of conversion, i.e., finding an operator of reconstruction. A recurrent multilayered neural network structure is used to model the operator of reconstruction. We present numerical results from synthetic and real world data in spectrometric problems. The ANN method studied has been used for correcting the data acquired by means of the optical spectrum analyzer. However, a broadfield of engineering applications including channel equalization, metrology, biomedical engineering, echography and seismology can be considered. A comparison is carried out to test the robustness of the method regarding noise level added to the measured samples and VLSI implementation properties with popular methods of correction.