A generalization of the Fourier transform and its application to spectral analysis of chirp-like signals (Q412414): Difference between revisions

The paper extends a well-known fact in Fourier Signal Analysis (FSA): a signal composed of a finite sum of sinusoids is transposed by FSA into a sum of separate spectral lines. Also, it is not known if a sum of signals (other than sinusoids) is carried into a sum of separate spectral images corresponding to single components. The paper provides a similar answer for a sum of chirp-like signals (a prototype of such an object is exp\((it^2)\)). The basis of this interesting result is constituted by De Branges Spaces Theory (DBST). This construction, a generalization of the Fourier Transform (FT), offers the main facts of a Fourier-like Transform: introducing a spectral function which constitutes a Reproducing Kernel Hilbert Space (RKHS); a Parseval-like equality and an inversion formula. The theory involves compact support functions. The basic facts of DBST are contained in Sections 1 and 2. Next, the structure of the associated spectral function for chirp-like signals is considered. The attention is focused on linear chirps and frequency-modulated chirps with the index of modulation equal at most to unity. First, linear chirps are considered and a general form of such objects is derived. For identification purposes, an additive zero-mean additive Gauss noise is provided. Therefore, for this form it is asked to estimate the included parameters based on experimental measurements. Similar problems occur in chirplet decomposition where the maximum likelihood approach is used. The main fact upon which the present approach rests lies in the fact that for a single chirp the spectral function tends a Dirac in frequency when the support of the signal tends to infinity. So the identification leads to maximum modulus identification. As an example, a signal composed of two chirps and additive noise is considered. The experimental results clearly show two spectral lines and diffuse spectral lines centered at the origin. Next, frequency-modulated chirps were considered. Numerical experiments were carried but the conclusions were that this type of chirps was not suitable for all cases. Finally, the authors introduce a new class of chip-like signals where the complex exponentials are replaced by Bessel functions which could be useful to model the chirps with a decaying envelope. For such signals, representation theorems are given. Also an example showing the appearance of spectral lines is presented. Finally, we think the paper gives a new and interesting insight into the DBST and could open new ways in chirp computing analysis.