Adaptive spectral analysis of sleep spindles based on subspace tracking

Abstract

A method to track the spectra of human sleep electroencephalogram (EEG) spindles is presented. This method uses a low-rank approximation of the covariance matrix and offers a compromise between numerical complexity and convergence. In the first part of the article, the authors describe the method briefly. In the second part, they apply it to filtered spindles to find an adequate agreement with a model of spindles that they put forward. Finally, it is concluded that there are different sorts of spindles according to frequency variation.