Problems and Solutions Associated with Prony’s Method for Processing Data with Pulse Noise

Parametric spectral analysis methods such as the Prony’s method can provide precision far greater than Fourier transform-based methods for estimation of signals which conform to their model. One of the the main downsides of such methods can be their robustness when presented with noisy inputs. While classic white noise usually does not pose a problem for the calculations, more exotic types such as impulse noise can lead to a complete breakdown of the model. The paper explores the possibility of using additional algorithms to improve the Prony’s method’s ability to recover the signal parameters in the presence of a mix of white and impulse noise. These methods include the already known signal segmentation approach and the point skipping approach which can be considered the extreme form of the former. Basic auxiliary algorithm for impulse noise identification which relies on matched filtering-like approach and the use of the minimum duration principle is shown and applied during the testing. The methods are tested on digital models of signals and the best is then applied to real-world data obtained during radar scanning procedure.