Curve-Fitting Algorithms Versus Neural Networks When Applied for Estimation of Wavelength and Power in DWDM Systems

摘要

This paper is on optical performance monitors for applications in dense wavelength-division multiplexing (DWDM) communication systems. Two algorithms for estimation of central wavelength and signal power of DWDM channels, on the basis on raw measurement data provided by a low-resolution spectrometric transducer, are compared. The first algorithm is based on the use of a curve-fitting and constrained-optimization technique; the second on application of a superposition of simple feedforward neural networks. The comparison is carried out using semisyn-thetic data. Conclusions are drawn concerning the applicability of compared algorithms in engineering practice.