Enhanced Signal Decomposition Using Golay-Optimized Empirical Mode Decomposition (Go-EMD)

Detection of neurological diseases with the aid of computers using EEG requires an efficient signal decomposition algorithm. EMD is the widely used method, in which, the mode-mixing problem, sifting stopping rule problem and envelope interpolation issue were involved. Considering this, Go-EMD, proposed in the current work, applies the Savitzky-Golay filter with calculated noise for eliminating mode-mixing. While, other EMD variants did not take computational cost and reconstruction iterations into consideration,the proposed algorithm identifies it and works on it. The quantitative analysis of the results shows Go-EMD outperforms EMD and EEMD with the lower RMSE value of 5.454 $\mu \mathrm{V}$ in comparison with 6.102 $\mu \mathrm{V}$ for EMD and 7.321 $\mu \mathrm{V}$ for EEMD; it is accompanied by near-zero NMSE and much better SNR, SSIM, and CC values, meaning its superior performance in signal reconstruction and noise reduction across EEG, ECG, and synthetic signals. Thus, compared to EMD and EEMD, Go-EMD surpasses these methods in noise attenuation and signal reconstruction. Furthermore, features obtained from the IMFs of Go-EMD can be used to enhance the diagnostic interpretation of EEG and ECG signals that are helpful in cardiology and cardiovascular research.