Robust Multi-Dimensional Model Order Estimation Using LineAr Regression of Global Eigenvalues (LaRGE)

The efficient estimation of an approximate model order is very important for real applications with multi-dimensional data if the observed low-rank corrupted by additive noise. In this paper, we present a novel robust method noise-corrupted based on LineAr Regression Global Eigenvalues (LaRGE). LaRGE uses multi-linear singular values obtained from HOSVD measurement tensor to construct global eigenvalues. contrast Modified Exponential Test (EFT) that also exploits exponential profile noise eigenvalues, does not require calculation probability false alarm. Moreover, achieves significantly improved performance in comparison popular state-of-the-art methods. It well suited analysis biomedical data. excellent illustrated via simulations and results EEG recordings.