A novel method for nonstationary power spectral density estimation of cardiovascular pressure signals based on a Kalman filter with variable number of measurements

Abstract

We present a novel parametric power spectral density (PSD) estimation algorithm for nonstationary signals based on a Kalman filter with variable number of measurements (KFVNM). The nonstationary signals under consideration are modeled as time-varying autoregressive (AR) processes. The proposed algorithm uses a block of measurements to estimate the time-varying AR coefficients and obtains high-resolution PSD estimates. The intersection of confidence intervals (ICI) rule is incorporated into the algorithm to generate a PSD with adaptive window size from a series of PSDs with different number of measurements. We report the results of a quantitative assessment study and show an illustrative example involving the application of the algorithm to intracranial pressure signals (ICP) from patients with traumatic brain injury (TBI).

A Novel Adaptive Fading Kalman Filter-Based Approach to Time-Varying Brain Spectral/Connectivity Analyses of Event-Related EEG Signals

“…Two similar components, namely, the early ERP component (with increased low-frequency power) and the ERD or alpha/beta-blocking component (with decreased alpha-beta bands power) were also observed in the analysis of the ERP spectrum [32], [33]. However, the associated connectivity information has not been revealed before.…”

“…The proposed AF-KF-VNM algorithm aims to identify the MVAR models from nonstationary EEG signals. It is an extension of our previous KF algorithm for single channel AR model estimation using a variable number of measurements [32], [33] instead of using a single or fixed number of measurements as in the conventional KFs [18]. The use of variable number of measurements enables better adaptation to the nonstationary EEG signals.…”

DOA Estimation and Tracking of ULAs with Mutual Coupling

“…Choosing L adaptively in KFVM has the advantage of achieving a better bias-variance tradeoff at each time instant. The scheme previously proposed in [35], [36] can be utilized here to select L at each time t. First, we definê e(t) =ẑ(t ¡ 1) ¡z(t ¡ 1) (49)…”

“…In the MPAST and MOPAST algorithms, we find that a better tracking performance can be obtained by repeating the respective PAST and OPAST iteration one or more times, since the "projection approximation" will be further improved with the subspace estimates. In the adaptive Kalman filter-based algorithm, the signal subspace is regarded as the system state and an adaptive Kalman filter with variable number of measurements (KFVM) [35,36] is employed for tracking the fast-varying subspace. Hence, the fast-varying DOAs can be estimated using the framework previously developed in the paper.…”

The Nightingale Award 2008

“…Bio-electricity and related studies on the brain and nervous system is an important topic for the journal and papers from these categories are on the short list of this year as well [ 1 , 3 – 5 , 7 , 13 ]. Also, the heart and the cardiovascular system are core topics of our journal [ 8 , 14 , 19 ]. We are also happy to publish papers on biomechanics and especially the spine forms a subject on which we have regular submissions [ 2 , 11 , 12 ].…”