Multi-Dimensional Spectrum Data Denoising Based on Tensor Theory

In this paper, we propose a novel multi-dimensional spectrum data denoising scheme from the perspective of tensor theory. The spectrum data is organized into spectrum tensor comprehensively from multiple dimensions. The optimal low rank approximation of the noisy spectrum tensor can be calculated by TUCKALS3 algorithm to reduce noise. Estimating the n-rank of tensor more accurately is necessary to improve the denoising performance of the TUCKALS3 algorithm. Therefore, we further improve the existing minimum description length (MDL) algorithm. Experimental results show that the signal-to-noise ratio (SNR) of the spectrum tensor can be increased by 15dB averagely by applying the enhanced algorithm, even at a higher noise level. The enhanced TUCKALS3 algorithm can effectively denoise multi-dimensional spectrum data and improve the corresponding system performance.