Laser Doppler electrophoretic light scattering is one of the standard methods for measuring the zeta potential of nanoparticles. However, the signal is usually mixed with the low frequency noises and the deviation from mathematical model of autocorrelation function. We introduced the Hilbert-Huang transform into the data analysis of the measurement in order to improve the measurement accuracy. After decomposing the original signal into several intrinsic modes, we found that by eliminating the low frequency noises and the deviation from mathematical model of autocorrelation function, the reconstructed signal was greatly improved. The simulated and experimental results demonstrate the feasibility and validity of this method.