Résumé:
The cardiovascular signals are, in general, quasi-periodic, repeated at each cardiac
cycle with a period that is not strictly constant and also their amplitudes can vary from one
cycle to another. These variations are due to the action of the autonomic nervous system
(ANS). In other words, these changes allow us to draw useful information about everything
controlled by the ANS such as the physiological or pathological condition of an individual, in
a non-invasive way.
The purpose behind this thesis has been divided in three phases i) proposed a solution
to the non-stationary problem of Herat Rate Variability (HRV) signals, ii) Study of a human
physiological problem, trying to find a link between this problem and the features of HRV
and iii) detection of a pathological problem using the HRV signal.
As results, we have proposed a simple and effective solution to the non-stationary
problem and also to estimate the duration of the sympathetic and parasympathetic ANS
behaviours as well as their location in time. For the physiological study, a hybrid method was
proposed to study the evolution of the HRV characteristics during a stressful experience, and
some HRV features were found strongly linked to stress. This study can help to connect the
ANS behaviour to the corresponding stress situations. Finally, for pathological study, a new
approach based on Alpha integration was presented to make an automatic detection of arousalin patients suffering from sleep apnoea, a new optimality criterion was introduced. The proposed algorithm was tested on four real signals in order to validate its effectiveness. Theresults of our algorithm were compared to another study on the same data, our method has provided very good results.
