Abstract: A novel algorithm based on correlation of symbols was proposed to solve the problem of high complexity and poor adaptability of traditional blind SNR estimators for OFDM signals in cognitive radio. Firstly the testing region of correlation of cycle padding correlation is expanded and got a signature sequence. Then use wavelet noise reduction to overcome the effect of multipath fading channel. Finally compute the SNR through the relationship between the max and min of the sequence and signal power and noise variance. Simulation results show that the proposed method can estimate the SNR of ZP-OFDM signals accurately as well as CP-OFDM signals, and gain better adaptability and lower complexity, which has better applicability for cognitive radio.