Goal: Fast Fourier transform (FFT), has been the main tool for EEG spectral analysis (SPA). As EEG can show nonlinear and non-stationary behavior, FFT may at times be meaningless. A novel method was developed for analyzing nonlinear and non-stationary signals using the Hilbert-Huang transform. Methods: We compared spectral analyses of EEG using FFT with Hilbert marginal spectra (HMS) with a multivariate empirical mode decomposition algorithm. Segments of continuous 60-sec EEGs recorded from 19 leads of 47 healthy volunteers were studied. Results: HMS showed a reduction of the alpha activity (-5.64%), with increments in the beta-1 (+1.67%), and gamma (+1.38%) fast activity bands, an increment in theta (+2.14%), and in delta (+0.45%) bands, and vice versa for the FFT method. For weighted mean frequencies, insignificant mean differences (lower than 1Hz) were observed between both methods for delta, theta, alpha, beta-1 and beta-2 bands, and only for gamma band values. The HMS were 3 Hz higher than the FFT method. Conclusion: HMS may be considered a good alternative for SPA of the EEG when nonlinearity or non-stationarity may be present.