Abstract: Aiming at the problem of poor noise immunity and low estimation accuracy of generalized cross-correlation time delay estimation algorithm in time-difference localization of non-stationary audio signals of unmanned aerial vehicle (UAV), an improved algorithm based on generalized quadratic correlation time delay estimation is adopted in this paper. The algorithm performs generalized quadratic correlation on audio signal with actual noise (such as wind sound, rain sound, car whistle sound, etc.), effectively suppresses noise interference and integrates the fine interpolation of correlation peak (FICP) algorithm to improve the resolution of cross-correlation function, which makes the peak delay more obvious. The simulation results show that the improved generalized quadratic correlation estimation method has higher accuracy of time delay estimation and better stability than the generalized cross-correlation and generalized quadratic correlation algorithms under different signal-to-noise ratios. The improved generalized quadratic correlation algorithm has better performance advantages for time delay estimation in UAV positioning, and has strong practical application.