Equivalent Bandwidth of Modes Obtained by EMD from White Gaussian Noise

In this paper the mean-square equivalent bandwidth of modes received by applying Empirical Mode Decomposition procedure to the white Gaussian noise time series data is estimated. Several functional dependencies have been confirmed. The estimated bandwidth of wGn modes can be approximated as the exponential function of the mode number with the base power of 2. The mean-square equivalent bandwidth also decreases with less total length of time series data points (direct proportionality), and, vice versa, increases with reduced time span (inverse proportionality). It means that the effective bandwidth of those modes is directly proportional to the sampling frequency. The empirical value of this dependence has been also approximated. Moreover, the value of this estimated frequency bandwidth for modes is constant for the different samples of white Gaussian noise with various means and standard deviations. These parameters only modify the instant values of modes, but not their frequency bandwidths. In other words, the mean-square equivalent bandwidth is a function only of a sampling frequency and a mode number.