HTR-CTO algorithm for wireless data recovery

Abstract

This paper presents a novel tensor completion method based on row-wise Hankel transformation for wireless missing data recovery problems. The received wireless data is modeled as a three-way tensor according to its time, space, and sensing information dimensions, which allows the proposed approach to fully capture the hidden space correlation and multi-attribute information of the received data. Furthermore, the wireless data recovery problem can be formulated as a simple core nuclear norm tensor completion problem based on Tucker decomposition, so that the computational complexity can be reduced. Furthermore, a Hankel transformation is introduced to enrich the structure of the core tensor, so that the new core tensor structure can exploit the underlying structure of tensor data to capture multiple correlations for data recovery. Hence, the recovery performance can be further improved. The proposed method offers significant advantages over the Tucker-based tensor completion methods. For example, (1) theoretical analysis shows that the proposed method has a lower computational complexity; (2) compared with the existing methods, it has better recovery performance even with a larger missing ratio or lower signal-to-noise ratio (SNR) scenario, etc. Simulation results demonstrate the efficiency of the proposed algorithm.