EMD Toolbox and Usage Guide for Empirical Mode Decomposition

The EMD toolbox and usage methodology for Empirical Mode Decomposition (EMD) mentioned in this text is a signal analysis technique proposed by Dr. Norden E. Huang at NASA. It performs signal decomposition based on the intrinsic time-scale characteristics of the data itself, requiring no predefined basis functions. This represents an essential difference compared to Fourier decomposition and wavelet decomposition methods that rely on predetermined harmonic basis functions and wavelet basis functions. Because of this feature, EMD method can theoretically be applied to any type of signal decomposition, thus showing significant advantages when processing non-stationary and nonlinear data. Consequently, EMD method gained rapid and effective application across various engineering fields upon its introduction, including oceanography, atmospheric sciences, astronomical observation data analysis, seismic record analysis, mechanical fault diagnosis, damping identification in dense-frequency dynamic systems, and modal parameter identification in large civil engineering structures. This guidebook is primarily divided into the following sections: Introduction, Non-stationary Signal Overview, One-dimensional Time-Frequency Signal Decomposition, Two-dimensional Time-Frequency Signal Decomposition, and Time-Frequency Image Information Extraction. The implementation typically involves coding the sifting process algorithm, which iteratively extracts intrinsic mode functions (IMFs) by identifying local extrema, constructing upper and lower envelopes using spline interpolation, and calculating the mean envelope. Additionally, the content covers further expansion of EMD application areas and comparisons/integration with other signal analysis methods, such as combining EMD with Hilbert Transform for Hilbert-Huang Spectrum analysis or integrating it with machine learning algorithms for enhanced pattern recognition.