Direct estimation of time delay using third-order cumulants

A new method is proposed when the signal is non-Gaussian with nonzero skewness and noises are Gaussian with unknown correlation function. The proposed method finds the time shift which minimizes an error function in the TOC domain where the correlated Gaussian noises are eliminated. Compared with Nikias and Pan´s Parametric Bispectrum Method (1988), the proposed method not only has less computational complexity but also exhibits better performance when signal is low-pass and noises are white. Performance comparison between PBM and the proposed method is made through theoretical discussion and computer simulations