Arrival time difference or time delay estimation is critically important for detection of leak location in buried water supply pipes. Because the exact leak locating depends upon the precision of the arrival time difference estimation between leak signals measured by sensors and the propagation speed of the leak-related elastic wave, the research on the estimation of time delay has been one of the key issues in leak locating. The arrival time difference was estimated with the peak time of cross correlation functions of the measured signals. In this study six different window functions, including the basic rectangular, Roth, Wiener, SCOT, PHAT and maximum likelihood windows were applied. Experimental results against an actual buried pipe made of cast iron showed that the introduction of the window functions improved the precision of time delay estimation. In this paper, a new statistical approach, that combines all results of each window function, is suggested for better leak locating. Apart from the experiment, an intensive theoretical analysis in terms of signal processing is described.