High-dimensional data completion method for wireless sensor networks based on Hankel transform

Abstract

The invention discloses a method for complementing high-dimensional data of a wireless sensor network based on Hankel transformation, comprising: S1, modeling the received high-dimensional data of a wireless sensor network into a tensor form χ; S2, counting tensor data χ The effective element positions of the tensor data constitute the tensor data support set Ω; S3, perform TUCKER decomposition on the tensor data χ to obtain its kernel tensor: χ=C× 1 u (1) ×……× i u (i) ……× N u (N) ; S4 establishes a simple nuclear tensor nuclear norm tensor completion model: S5, based on the simple nuclear tensor nuclear norm tensor completion model, performs Hankel transformation and variable separation, and optimizes the missing The data is calculated and filled, and the completion result is output. The invention can realize fast and accurate data completion.

Description

technical field [0001] The invention relates to a high-dimensional data completion method for wireless sensor networks based on Hankel transformation, and belongs to the technical field of data completion. Background technique [0002] With the development of artificial intelligence technology, sensor applications are also being popularized, such as various wearable devices, wireless communication devices, intelligent control devices, etc. Wireless sensor networks composed of various sensors have been widely used in national defense, military, environmental monitoring , biomedical, intelligent transportation, emergency and disaster relief, industrial safety and supporting commercial applications. The wireless sensor network completes the perception, collection, processing and transmission of data information in the monitoring area through the cooperation of sensor nodes, and finally sends the relevant information to the data service center to provide data support for the sta...

AI Technical Summary

Problems solved by technology

The file performs Hankelization operations on the observation data and the correlation matrix. If the data volume is large or the dimension is high, there will be problems such as complex operation and long time consumption.

The patent document with the application number "201810932079.X" discloses a "uniform area array tensor reconstruction method and signal source location method with partial array element damage", which uses a tensor completion method based on CP decomposition to realize the lack of received signals Data completion, its disadvantage requires a lot of calculation of the outer product of vectors and the sum of rank-1 tensors

The patent document with the application number "201811648489.8" discloses a "3D tensor completion method based on Capped nuclear norm". Completed low-rank tensor data, but in the process of constructing the capped nuclear norm, it needs to set the capped parameter, calculate the singular value of the tensor, compare the singular value and the size of the parameter and sum them up, etc., which has high calculation the complexity

The patent document with the application number "201810824663.3" discloses a "long-term spectrum prediction method based on image reasoning". By modeling the monitoring information of different frequency bands obtained from a complete monitoring day as a picture, the tensor is composed according to the time series of the monitoring day. The high-precision low-rank tensor completion method performs missing data completion and prediction, so it cannot consider the influence of noise on the data

Images