On the Grunbaum Commutor Based Discrete Fractional Fourier Transform

The basis functions of the continuous fractional Fourier transform (FRFT) are linear chirp signals that are suitable for time-frequency analysis of signals with chirping time-frequency content. In the continuous--time case, analytical results linking the chirp rate of the signal to a specific angle where the FRET of the chirp signal is an impulse exist. Recent efforts towards developing a discrete and computable version of the fractional Fourier transform (DFRFT) have focussed on furnishing a orthogonal set of eigenvectors for the DFT that serve as discrete versions of the Gauss--Hermite functions in the hope of replicating this property. In the discrete case, however, no analytical results connecting the chirp rate of the signal to the angle at which we obtain an impulse exist. Defined via the fractional matrix power of the centered version of the DFT, computation of this transform has been constrained due to the need for computing an eigenvalue decomposition. Analysis of the centered version of the DFRFT obtained from Grunbaum's tridiagonal commuter and the kernel associated with it reveals the presence of both amplitude and frequency modulation in contrast to just frequency modulation seen in the continuous case. Furthermore, the instantaneous frequency of the basis functions of the DFRFT are sigmoidal rather than linear. In this report, we define a centered version of the DFRFT based on the Grunbaum commutor and investigate its capabilities towards representing and concentrating chirp signals in a few transform coefficients. We then propose a fast algorithm using the FFT for efficient computation of the multiangle version of the CDFRFT (MA-CDFRFT) using symmetries in the computed eigenvectors to reduce the size of the eigenvalue problem. We further develop approximate empirical relations that will enable us to estimate the chirp rate of the multicomponent chirp signals from the peaks of the computed MA-CDFRFT. This MA-CDFRFT also lays the ground work for a novel chirp rate Vs. frequency signal representation that is more suitable for the time-frequency analysis of multicomponent chirp signals