On linear weak predictability with single point spectrum degeneracy

The paper studies predicting problems for continuous time signals with the Fourier transforms vanishing a certain rate at single point. It shows that, some linear causal predictors, anticausal convolution integrals over future times these processes can be approximated by convolutions past times. corresponding kernels are invariant, and they presented explicitly in frequency domain via their transfer functions. These predictors “universal” meaning that do not require to know details of spectrum underlying signals; same predictor used entire class point degeneracy. feature robustness respect noise contamination.