Nonstationary interference suppression using adaptive overdetermined frame representations

Abstract

Transform-based interference suppression works well for stationary signals, but when applied to nonstationary interference, it is often beneficial to use a larger overdetermined library of transforms. We propose a scheme which uses the same project/threshold technique used in traditional transform-based excision applied to an overdetermined frame. Such a frame can consist of a variety of elements, including time and frequency invariant sets of wavelets and chirps, to help characterize a larger variety of interferers. For thresholds within certain bounds, the proposed method guarantees improvement in the effective signal-to-interference ratio (SIR). Analytical and simulation results illustrate the effectiveness of the technique for a variety of interference types, including combinations of non-orthogonal interferers. The study is of interest in DS-SS communications.