Image reasoning for long-term spectrum prediction

Abstract

The invention discloses a long-term spectrum prediction method of image reasoning. Taking each monitoring time slot of a complete monitoring day as the horizontal axis and the multi-frequency points / bands as the vertical axis, the complete spectrum situation of a monitoring day is formed into an image, and a set of graphs is formed according to the time sequence of the monitoring day. Like, infer the next image. The steps of the method are as follows: (1) Model the original spectral data of the image, and establish a spectral data tensor suitable for the idea of ​​image reasoning; (2) Based on the tensor completion technology, the completion of historical missing data and the spectral data to be predicted are completed accurately and efficiently at the same time. ③Use the root mean square error to evaluate the performance of the long-term spectrum prediction algorithm. The present invention can complete long-term spectrum prediction more accurately and efficiently when large-scale historical data is missing.

Description

technical field [0001] The invention belongs to the field of cognitive radio of wireless communication technology, in particular, it is a long-term frequency spectrum prediction method based on image reasoning. Background technique [0002] Spectrum prediction is a technology that infers the future spectrum status from known spectrum occupancy data, and infers the complete spectrum situation from sparse spectrum samples, by effectively mining the inherent correlation and regularity between spectrum data. As an important application of prediction technology (and also data mining technology) in the field of spectrum data analysis, spectrum prediction is an effective supplement to spectrum sensing, helping to reduce delay and energy consumption in the process of spectrum sensing and increase spectrum access. system throughput. More and more researchers have begun to devote themselves to the research and development of efficient prediction techniques in the field of cognitive r...

AI Technical Summary

Problems solved by technology

[0006] The purpose of the present invention is to propose a long-term spectrum prediction method based on image reasoning, which solves the problem of long-term spectrum prediction under large-scale data loss from the perspective of image reasoning, and can use a large number of missing historical spectrum data of multiple days and multiple frequency points , to predict the complete spectrum situation of the next day (time resolution remains consistent with historical data)

Images