A robust TDOA estimation method for in-car-noise environments

Time difference of arrival (TDOA) estimation is one of the key techniques in array signal processing and has wide applications in hands-free speech interface. TDOA is used to derive the time difference of signal propagation from source to the spatially separated microphones. The acoustic transfer functions ratio (ATF-s ratio) method is one of the robust approaches to get TDOA estimate in presence of strong background noise and in reverberant environment. Although the effectiveness is proved by various simulations in related studies, the anti-noise or anti-reverberation capability of ATF-s ratio is limited in real acoustic environment. In this work, a frequency domain partialwhitening ATF-s ratio method is proposed for detecting the time delay in an in-car mobile system. To test the efficacy of the proposed method, the experiments were conducted in a car running at a speed of about 100km/h on the highway, and the results show that the percentage of TDOA estimation anomaly is reduced by 15% ∼ 25%.