Computation of exponentials in transient data

Abstract
Prony's method is widely used to identify the poles of a time series. A numerical implementation of Prony's method based on the properties of the least-squares solution of the transient data matrices is described, and a practical indicator of the accuracy of the computational results that is based on the singular values of the transient data matrices is derived. This algorithm can be applied to time series which contain simple or multiple poles.