Improved generalized cross correlation-phase transform based time delay estimation by frequency domain autocorrelation

There are several methods for estimating the time delay between incoming signals to two sensors. Among them, the GCC-PRAT (Generalized Cross Correlation-Phase Transform) method, which estimates the relative delay from the signal whitening and the cross-correlation between the different signal inputs to the two sensors, is a traditionally well known method for achieving stable performance. In this paper, we have identified a part of GCC-PHAT that can improve the periodicity. Also, we apply the auto-correlation method that is widely used as a method to improve the periodicity. Comparing the proposed method with the GCC-PRAT method, we show that the proposed method improves the mean square error performance by 5 dB similar to 15 dB at the SNR above 0 dB for white Gaussian signal source and also show that the method improves the mean square error performance up to 15 dB at the SNR above 2 dB for the color signal source.