Time Delay Estimation Method based on Improved Frequency Sliding Generalized Cross-correlation

The frequency sliding GCC (FS-GCC) has been recently proposed, which can achieve robust GCC performance. However, the performance of FS-GCC decreases due to the influence of the weighting function and low signal-to-noise ratio sub-bands. To resolve the related issues, an improved FS-GCC algorithm has been proposed. The FS-GCC is improved from two aspects, one of which is inspired by the concept of blind acoustic masking. It classifies the sub-band signals into reliable and unreliable ones based on the posterior SNR and highlights the reliable parts in the sub-band GCC matrix by assigning different weights. Subsequently, a new weighting function is developed to further enhance estimation performance. Extensive experiments validated the superiority of the proposed algorithm. Moreover, ablation experiments demonstrated that the two improvements reduced estimation errors and lowered the probability of anomalies and higher FSPR.