EMD Empirical Mode Decomposition

Empirical Mode Decomposition (EMD) is a data-driven method for analysing nonlinear and non-stationary signals. Through iterative sifting, it separates a signal into frequency-specific intrinsic mode functions and a residual trend. Combined with the Hilbert transform, EMD provides time-varying estimates of instantaneous amplitude and frequency. Its flexibility supports applications in cardiovascular monitoring, neuroscience, speech recognition, image processing, and finance. In financial markets, EMD decomposes stock anomalies into frequency components that can improve return prediction. An EMD composite based on 20 anomalies and more than four million observations produces a monthly FF6 alpha of approximately 0.82%. The method is also applied to micro-Doppler radar signatures for distinguishing different unmanned aerial vehicles. Complex EMD separates positive- and negative-frequency components while retaining information contained in complex radar signals. Applications to bionic birds and drones show that wing movements can be identified, although rotating propellers remain harder to detect. Future research must address mode mixing, noise robustness, phase-based region selection, competing decompositions, and multi-object environments.