To address the issue of low localization accuracy of the generalized cross-correlation （GCC） algorithm in strong noise environments, an improved GCC time delay estimation algorithm is proposed. This algorithm combines the least mean squares （LMS） algorithm with the quadratic cross-correlation method to enhance the traditional GCC algorithm, effectively suppressing the impact of noise. Additionally, spline interpolation techniques are used to significantly improve the accuracy of time delay estimation during the secondary peak detection process. The simulation results show that this algorithm reduces the error in sound source coordinate positioning from 17.7% with the traditional GCC algorithm to within 8.36%. In practical application tests, the error in sound source coordinate positioning is also significantly reduced from 42.2% with the traditional GCC algorithm to 15.4%.