Time-frequency peak filtering: analysis and implementation details

Time-frequency peak filtering is a novel technique for signal filtering and enhancement. The method is based upon encoding the signal as an instantaneous frequency (IF) and using IF analysis techniques to recover a filtered version of it. A general iterative time-frequency peak filtering (TFPF) scheme based on this method has been formulated. In this contribution we give details of bias and variance analysis of the method as implemented with the Wigner-Ville distribution (WVD), in particular providing conditions under which the technique will be unbiased. It is then shown that white Gaussian noise satisfies the constraints placed upon the random (additive) component of the signal for unbiasedness to hold. Additionally, details of the frequency encoding in the finite sampling rate case and window length considerations for implementation with the WVD are given. Simulations are given which show extremely clean recovery of signals in high levels of noise