Aiming at reducing the large amount of computation of Sub Space Decomposition(SSD) algorithm and improving the poor precision(when signal to noise ratio is below 0 dB) of the Maximum Likelihood Estimation(MLE) algorithm to estimate the signal to noise ratio of the Continuous Phase Modulation(CPM) signal，a new fast algorithm with higher accuracy is presented in this paper. This new algorithm is an improvement based on the frequently used MLE method. It takes the statistic and cross-correlation of the received signal into account, and can estimate the SNR fast and precisely without the initial phase. Error estimate is less than 1 dB when RSN is between -20 dB and 20 dB. A comprehensive deduction of the algorithm is included in this paper. Simulations by MATLAB and FPGA implementation show the validity of this algorithm.