Tracking and separating non-stationary multi-component chirp signals. (Q1873142)

The authors develop an algorithm which performs a stable reconstruction of a chirp signal from an environment containing several other chirp components, disjoint in the time-frequency domain plus Gaussian additive noise. The algorithm provides the opportunity to iteratively analyze data in the time-frequency domain by tuning its parameters to obtain more accurate results. The algorithm is non-parametric as far as no model assumptions are needed for the chirps while reconstruction follows by direct manipulations of data. It can be easily implemented into computer codes and can be applied to various problems of digital image processing.