Short-time multifractal analysis: application to biological signals

Some signals obtained from biological systems evince a great complexity. Recently, new tools which allow the extraction of information from them have been proposed. In particular, multifractal analysis gives a quantification of the degree and distribution of irregularities in a signal. A possible approach for this analysis is the one based on wavelet leaders. In this work, the use of wavelet leader based multifractal analysis in short-time windows is proposed in order to analyze the evolution of the multifractal behavior of biological signals. In particular, applications of this technique to the detection of ischemic episodes in heart rate variability signals and to voice activity detection are examined. It is shown that the study of the time evolution of indexes obtained with the proposed new method gives useful information hidden in the signals.