Segmentation of 'rhythmic' and 'noisy' components of sleep EEG, heart rate and respiratory signals based on instantaneous amplitude, frequency, bandwidth and phase

This paper describes the application of instantaneous amplitude, frequency, phase and bandwidth measures to separate narrow-band rhythmic bursts from broad-band noisy parts in EEG, HR and breathing signals. Ongoing EEG is analyzed with constant bandwidth filters, evoked potentials and HR variability with 'constant Q-factor' logarithmic filters. Fast time frequency transform (FTFT) can reliably detect central (EEG bursts) and autonomic (HR increase) microarousals and gives an improved classification of sleep over FFT techniques. Instantaneous phase discerns the central or obstructive nature of apnea and even hypopnea.