Tensor-Guided Interpolation For Off-Grid Power Spectrum Map Construction

Abstract

This paper addresses the off-grid tensor-guided interpolation problem, aiming to reconstruct a 3D power spectrum map from sparse observations. A segmented polynomial model is employed to handle off-grid measurements, while a nuclear norm regularization is incorporated to account for the inherent low-rank characteristics of signals. An alternating regression and singular value thresholding algorithm is developed to solve the proposed method. The numerical results demonstrate the superiority of the proposed method, showcasing a remarkable improvement of over 10% in power spectrum map reconstruction accuracy when the sampling rate exceeds 6%, as compared to state-of-the-art approaches.