ON THE CORRECTNESS OF THE APPLICATION OF TENSOR DECOMPOSITION FOR EEG SPECTRA ANALYSIS

In our previous work, it was suggested that the subject's EEG spectra in different functional states have a third-order tensor structure, and the PARAFAC tensor decomposition can be used to isolate physically and physiologically meaningful components from them. However, the correctness of using tensor decomposition to analyze EEG spectra in different physiological states has been substantiated neither experimentally nor theoretically. In this paper, we used the residual of data approximation by a low-rank tensor and proved the correctness of the application of the PARAFAC tensor decomposition for the analysis of human EEG spectra.