In the time delay estimation of chaotic spread spectra, the main peak to side lobe ratio obtained by using basic cross-correlation is not ideal, and the peak values of misjudged points are too large. Based on the advantages of generalized cross correlation (GCC) which can realize pre-filtering of signals and properly eliminate the influences of noise and interference in the received signal, we analyze three commonly used weighting functions of GCC, and we find that SCOT weighting function can suppress the noise of the transmitted signal and the received signal at the same time. Three commonly used weighting functions are applied to the Simulink simulation model of chaotic time delay estimation. The simulation results show that the results of generalized cross-correlation using SCOT weighting function are better than that of the basic cross-correlation and the other two commonly used weighting functions. Compared with the basic cross-correlation results, the main peak to side lobe ratio obtained by the generalized cross-correlation using SCOT weighting function increases by more than 0.9 dB, and the ratio of the peak value of the misjudgment point to the peak value of the fault point decreases by more than 0.2.