Detection of Sleep Spindles from Electroencephalogram (EEG) Signals Using Auto Recursive (AR) Model

Detection of sleep spindles in EEG was commonly performed inefficiently by doctorpsilas eye inspection. In this paper, a new approach is presented for analysis of EEG signals and detection and localization of sleep spindles. By estimating auto recursive (AR) models on short segments the EEG is described as a superposition of harmonic oscillators with damping and frequencies varying in time. Most of the oscillatory events are detected, whenever the damping coefficients of one or more frequencies fall below a predefined threshold. The algorithm works well for the detection of sleep spindles and in addition identifies delta and alpha waves.