Description

1. The talk introduces Empirical Mode Decomposition (EMD) as a fully data-driven tool for analyzing non-stationary and nonlinear signals without a priori model assumptions .

2. EMD decomposes a signal into Intrinsic Mode Functions (IMFs), each representing a distinct time-varying frequency component .

3. The methodology relies on an iterative sifting process using local extrema and cubic spline envelopes to enforce IMF conditions .

4. Combined with the Hilbert Transform, the approach yields instantaneous amplitude and frequency for mono-component signals .

5. EMD is highlighted as flexible and applicable in neuroscience, speech, image processing, finance, and molecular dynamics .

6. Applications to micro-Doppler radar show how IMFs capture vibrations from UAV propellers and bird wings .

7. A complex-valued extension (CEMD) enables decomposition of I/Q radar data via frequency separation in the Fourier domain .

8. Financial examples illustrate EMD decomposition of stock factors and construction of an EMD composite return predictor .

9. Limitations such as mode mixing are addressed through Ensemble EMD and alternative methods like VMD .

10. Future challenges include rotation-invariant complex EMD, noisy environments, and multi-object radar scenarios .