Using Empirical Mode Decomposition to Estimate Amplitudes in Noisy Data

Abstract

Empirical Mode Decomposition, an adaptive data-driven technique which can be used to extract non-stationary signals buried in noise, seldom admits theoretical calcula- tion of the statistical properties of the extracted signals. Instead, numerical experiments are required. In this pa- per we use Monte Carlo simulations to investigate the accuracy of the amplitudes of sinusoids extracted from synthetic noisy signals using Empirical Mode Decompo- sition. We show that even for relatively low signal-to- noise data, the amplitude of the extracted signal is close to true amplitude. We also show that edge effects due to the spline curves which are used to calculate the decom- position do not affect the amplitude estimate beyond the first two oscillations.