To enhance the Signal to Noise Ratio(SNR) estimation performance of unknown type signals, a novel algorithm based on noise-assisted is proposed, in which the boundary of the signal and noise is determined according to the average period curve of Intrinsic Mode Functions(IMF). The algorithm principle and its flow chart are presented, and the performance of noise-assisted method is also analyzed. Simulation results show that, noise-assisted method is adapted to unknown signals SNR estimation, and the mean square error is below 0.2 dB under SNR of 0 dB.