A fast and high accurate initial values obtainment method for Brillouin scattering spectrum parameter estimation

Abstract

To improve accuracy and arithmetic efficiency of the parameter estimation for Brillouin scattering signal, on the basis of systematic analysis of the Lorentzian function, a fast and high accurate initial parameter estimation algorithm is proposed and it decreases the computational burden of the following least-squares fit. A three-parameter least-squares fit with a Levenberg–Marquardt optimization scheme is implemented in Matlab for the Brillouin scattering parameter estimation, and the random variable, the results from the particle swarm optimization (PSO) algorithm and the proposed algorithm are taken as the initial values of the above least-squares fit. Two numerically generated signals with considerable noise and one real signal are selected. The results reveal that the least-squares fit based on the random variable method almost never converges. However, both the PSO method and the proposed method can converge in various situations. The computation time of the PSO method ranges from 500ms to 700ms and that of the proposed method ranges from 1ms to 3ms, that is, the proposed method can guarantee convergence and at the same time, the arithmetic efficiency is greatly improved. The proposed method fixes the problem of fast and high accurate parameter estimation for Brillouin scattering.