Real time estimation of multichannel power spectrum density for variability analysis of the cardiovascular system

Abstract

Frequency domain analysis by power spectrum density (PSD) estimate has proven to be an effective method of investigation for studying the influence of the automatic nervous system on the systemic and coronary hemodynamics. Since the problem is intrinsically multichannel it should be studied by some proper multichannel PSD estimates. Parametric autoregressive spectral methods were used and in particular the Nuttal-Strand algorithm which guarantees good frequency resolution, even in narrow time window, with low bias and low variance estimate. As real time performance is highly desirable for immediate monitoring of pathophysiological features evolution, both during human exercise tests and in animal experiments, the authors developed a system for real time multichannel spectral estimate. The system delegates the heavy computing of PSD estimate to a specialized digital signal processing board and offers a combination of visual panels for instantaneous checking of signals, parameters and spectral envelopes.<>