Conventional maximum likelihood algorithm based on cyclic prefix has plateau effect and has no obvious peak correlation value in multi path channels,because some of the cyclic prefixes have been polluted by Inter Symbol Interference(ISI)due to the multi path delays and Doppler effects.The algorithm can not be used in multi path channels without optimization.An optimized method based on the cyclic prefixes is proposed in this paper,which reduces the number of cyclic prefixes in correlation calculation according to the signal noise ratio and multi path delays and uses average correlation values in multi sampling points.The proposed method proves to be able to get sharp peak values and improve the symbol precision compared with the existing methods both in Additive white Gaussian noise(AWGN) channels and in multi path channels.In multi path channel,plateau effect decreases significantly.