Abstract:In this paper, we introduce the Reconstructive Ideal Fractional Transform (RIFT), an entropy-based probabilistic filtering algorithm formulated to reconstruct the Ideal Time-Frequency Representation (ITFR). RIFT surpasses the limitations imposed by the Gabor uncertainty principle for linear transforms, achieving the bilinear transform accuracy present in the Wigner-Ville Distribution (WVD) while effectively suppressing cross-terms. The algorithm utilises a hierarchical fractional wavelet-based scheme to account for localised time-frequency trajectory curvature. This scheme is optimised through an entropic-based filtering method that probabilistically extracts auto-terms while retaining the resolution of the WVD. This is achieved by employing a spatially varying, positively constrained deconvolution algorithm (Lucy-Richardson) with regularisation for adequate noise suppression. Additionally, the optimisation yields an Instantaneous Phase Direction field, which allows the localised curvature in speech or music extracts to be visualised and utilised within a Kalman tracking scheme, enabling the extraction of signal component trajectories. Evaluation demonstrates the algorithm's ability to eradicate cross-terms effectively and achieve superior time-frequency precision. This advance holds significant potential for a wide range of applications requiring high-resolution cross-term-free time-frequency analysis.