Régression linéaire crédibiliste pour l’analyse douce des fluctuations redressées (soft DFA)

Detrended Fluctuation Analysis (DFA) provides insights on signal complexity which have shown to be relevant and effective for distinguishing healthy and nonhealthy persons through different physiological signals. This method is based on different steps involving linear regression. This paper proposes an evidential linear regression model and its application on DFA in order to take into account limitations of DFA due to uncertainties associated with crisp linear regression estimates. Three R packages that contain a DFA implementation have been compared with the proposed method in experiments realised on sinusoidal signals, noises and Hausdorff  famous dataset that illustrates the interest of the method.