Abstract: Given of the weakness of SNR estimator based on Empirical Mode Decomposition，as large computation and low accuracy, a new algorithm combined with the probability theory was proposed．Analyzed the power spectrum distribution of Intrinsic Mode Functions (IMF) in the way of discrete fourier transformation, and determined the effectiveness of decomposition three times，simplified operating process. Proposed the normal distribution approximate expression of the power spectral density of components about the characteristic parameters based on statistical data, and analyzed the leakage of energy in the decomposition process，then gave the signal to noise ratio estimate method of characteristic parameters．For different sample length and data size，simulation results show that the new algorithm has a better performance in lower SNR environment，and the normalized mean square error is less than 0.5dB when SNR is 0dB．