Instantaneous frequency estimation based on empirical mode decomposition

When a monocomponent signal with slowly oscillating instantaneous frequency is embedded in a noise, its instantaneous frequency is usually added by many high peaks and low troughs. But real instantaneous frequency is a trend in estimated instantaneous frequency if signal has a small SNR. Though Maximum Likelihood Estimator can smooth some higher crests and lower troughs, real instantaneous frequency is still not very obvious. A new nonparametric instantaneous frequency estimation method is established based on Empirical Mode Decomposition (EMD). Firstly, smoothed version of the Phase Difference Estimator-Maximum Likelihood Estimator is adopted to estimate the instantaneous frequency. Then more accurate instantaneous frequency can be shifted as a trend by EMD. A simulation is used to test this new method. The result shows the method described is proved to be more accurate and less computation.