부산대학교 도서관

We examine a new method for peak localization, the centroid of the data raised to some power, which we call the generalized centroid. We derive the peak localization uncertainty for the generalized centroid and compare it with the Cramer-Rao lower bound for both Gaussian and quadratic fits (with Gaussian signal and noise). We find that the centroid of squares and the Gaussian fit yield the best results in both one and two dimensions. We perform similar analysis with a Lorentz-like signal and find that the centroid of cubes and the nonlinear least squares fit provide the best results. We support our derivations with simulations, and also show simulation results when the maximum function is used to initially estimate the peak location.
Comment: 6 pages,6 figures