A New Approach for Physiological Time Series

An understanding of physiological time series such as the heart-beat intervals is important to many areas, like heart-attack prediction, cardiovascular health, sport and exercise, etc. The study of time series can reveal underlying mechanisms of the physiological system, which usually contains both deterministic and stochastic components. Therefore the analysis of time series is very complicated because of the nonlinear and non-stationary characteristics of physiological time series data. Over the past years, time series analysis methods are applied to quantify physiological data for identification and classification (see [7, 12]). The application of physiological time series analysis commonly focus on measuring different aspects of time series data such as complexity, regularity, predictability, dimensionality, randomness, self similarity, etc. The tools used in these techniques include but not restrict to the mean, standard deviation, Fourier transform, Wavelet, entropy, fractal dimension, pattern detection (see [8, 13]). Recently a new mathematical tool, empirical mode decomposition (EMD), was proposed by Norden Huang et al (see [5, 6]). It decomposes a time series into a finite sum of intrinsic mode functions (IMF) that generally admit wellbehaved Hilbert transforms. This decomposition is based on the local characteristic time scale of the data, which makes EMD applicable to analyze nonlinear and non-stationary signals. EMD and Hilbert transform together, called the HilbertCHuang transform (HHT), usually allow to construct meaningful time-