An Efficient Nonstationary Wiener Filter Hardware Implementation
An efficient multi-cycle hardware design of a non-stationary (Time-Varying (TV)) Wiener filter, based on Time-Frequency (TF) analysis, is considered. It is developed by following the idea of a new method for Filter’s Region of Support (FRS) real-time estimation, also proposed here. Quite general estimation method, based on cross-terms-free time-frequency representations, provides multiple detection of the local filter’s regions of support in the practically only important case of a known single noisy signal realization. In this paper, a very efficient real-time filtering of mono- and multicomponent non-stationary signals is enabled.