Abstract: Time-delay estimation algorithm is one of the key technologies in array signal processing, used in the microphone array sound localization system. Many researvhes have been focused on the time-delayestimation in recent years. Time-delay estimation algorithm is roughly divided into two steps to orientate the sound source , to deal with the microphone receives the signal, estimate the relative time delay in the microphone array,then, according to the spatial geometric relationships of microphone array in each, combined the relative time delay with the first step to get the location of the source in the reference field.There are many factors that can influence the properties of the microphone array sound localization system, reverberation is one of them. Usually a microphone array sound positioning system is in the indoor environment, in addition to the influence of other interfering noise and environmental noise, the sound source itself through the environment reflection caused by reverberation will cause great influence to its performance. The size of the indoor environment and the location of sound source can also lead to different reverberation. In this paper, based on the principle of cross-correlation, comprehensive two kinds of different frequency weighted algorithm, the PATH and the ML . The new algorithm can also robust noise and reverberation, and get the optimal q value at the variety of environments, reducing the influence of the multipath effect arrays to estimate the relative time delay, improves the accuracy of time delay estimation and sound source localization accuracy. The experimental results also proved the new method of time delay estimation accuracy, the effectiveness of the optimal q value.