An aggregate group approach for AR spectral estimation of noisy signals

A least-squares estimate of parameters of an overdetermined Yule-Walker equation can contain very large errors in the low signal-to-noise ratio. This sensitivity to noise can be seen in parameter space, where each of the equations appears as a line. The least-squares method is equivalent to finding the center of a circle inscribed in the lines. As an alternative the authors examine the center of gravity as a new objective function called the weighting gravity center method. The method is made increasingly robust by eliminating wild points attributed to noise and the use of the overdetermined Yule-Walker equation