arXiv subjects

Thomas Schreiber

Publications and source records attributed to Thomas Schreiber.

24 records · Page 2Linked to original sources

arXiv1998

Phase space embedding of electrocardiograms

We study properties of the human electrocardiogram under the working hypothesis that fluctuations beyond the regular structure of single cardiac cycles are unpredictable. On this background we discuss the possibility to use the phase space embedding method for this kind of signal. In particular, the specific nature of the stochastic or high dimensional component allows to use phase space embeddings for certain signal processing tasks. As practical applications, we discuss noise filtering, fetal ECG extraction, and the automatic detection of clinically relevant features. The main purpose of the paper is to connect results of embedding theory which had not been previously applied in practise, and practical applications which had not yet been justified theoretically.

chao-dyn↗

arXiv1998

The human ECG - nonlinear deterministic versus stochastic aspects

We discuss aspects of randomness and of determinism in electrocardiographic signals. In particular, we take a critical look at attempts to apply methods of nonlinear time series analysis derived from the theory of deterministic dynamical systems. We will argue that deterministic chaos is not a likely explanation for the short time variablity of the inter-beat interval times, except for certain pathologies. Conversely, densely sampled full ECG recordings possess properties typical of deterministic signals. In the latter case, methods of deterministic nonlinear time series analysis can yield new insights.

chao-dyn↗

arXiv1998

Interdisciplinary application of nonlinear time series methods

This paper reports on the application to field measurements of time series methods developed on the basis of the theory of deterministic chaos. The major difficulties are pointed out that arise when the data cannot be assumed to be purely deterministic and the potential that remains in this situation is discussed. For signals with weakly nonlinear structure, the presence of nonlinearity in a general sense has to be inferred statistically. The paper reviews the relevant methods and discusses the implications for deterministic modeling. Most field measurements yield nonstationary time series, which poses a severe problem for their analysis. Recent progress in the detection and understanding of nonstationarity is reported. If a clear signature of approximate determinism is found, the notions of phase space, attractors, invariant manifolds etc. provide a convenient framework for time series analysis. Although the results have to be interpreted with great care, superior performance can be achieved for typical signal processing tasks. In particular, prediction and filtering of signals are discussed, as well as the classification of system states by means of time series recordings.

chao-dyn↗

arXiv1998

Constrained randomization of time series for hypothesis testing

We propose a general scheme to create time sequences that fulfill given constraints but are random otherwise. Significance levels for nonlinearity tests are as usually obtained by Monte Carlo resampling. In a new scheme, constraints including multivariate, nonlinear, and nonstationary properties are implemented in the form of a cost function.

chao-dyn↗

arXiv1998

Testing for nonlinearity in unevenly sampled time series

We generalize the method of surrogate data of testing for nonlinearity in time series to the case that the data are sampled with uneven time intervals. The null hypothesis will be that the data have been generated by a linear stochastic process, possibly rescaled, and sampled at times chosen independently from the generating process. The surrogate data are generated with their linear properties specified by the Lomb periodogram. The inversion problem is solved by combinatorial optimization.

chao-dyn↗

arXiv1998

Nonlinear projective filtering in a data stream

We introduce a modified algorithm to perform nonlinear filtering of a time series by locally linear phase space projections. Unlike previous implementations, the algorithm can be used not only for a posteriori processing but includes the possibility to perform real time filtering in a data stream. The data base that represents the phase space structure generated by the data is updated dynamically. This also allows filtering of non-stationary signals and dynamic parameter adjustment. We discuss exemplary applications, including the real time extraction of the fetal electrocardiogram from abdominal recordings.

chao-dyn↗