2.2 Connection between Signals and Systems
|
17
Blood pressure
in mmHg
Volume
in ml
120
100
80
60
40
20
0
130
90
50
ECG
Systole
Diastole
Aorta
Atrium
PCG
Venticle
Fig. 2.8: Vital signals of a human heart: Sequence of three heart cycles in the electrocardiogram
(ECG), phonocardiogram (PCG) and the corresponding pressure sequence in the left ventricle and the
aorta (centre), as well as the ventricular volume (bottom).
ricle will be delayed, resulting in a change in the ECG signals. Thus there exist a more
or less pronounced connection between the signal-generating system "heart" and the
respective biosignal. In mathematical modelling, this signal-system-relation is called
sensitivity of a state variable (or measured variable) of the system to the change of a
physical system parameter (e.g. the conductivity of the conduction system). In signal
processing, such dependencies are made clear with the help of correlation. As will be
shown later in the book in subsection 2.3.6, one speaks of a strong/weak correlation
of the signal quantities.
One of the main concerns of biosignal processing is the analysis and diagnostic
use of these signal-system correlations. Due to the complexity and variability of liv-
ing systems, the correlations are usually not clear, not pronounced enough, incom-
plete, or covered by artefacts and affected by uncertainties. Mathematical models are
helpful in searching, as they greatly improve the basic understanding of the system
interrelationships. For example, model systems can be parameterised by exact quant-
ities and their synthetic signals can be analysed possessing the "ground truth". Of-
ten, new signal-system-coherences or signal components are found which could not
yet be measured on the physiological system for technical reasons (for example, due
to unsuitable sensors or measurement locations, too low time resolution, too strong
artefacts or filtering of the signals, etc.). A combination of mathematical modelling,
sensitivity analysis and adaptation of the measurement technology and evaluation
algorithms often proves to be very profitable in the search for the correct parameters
or measured variables in the biosignal.