Automatic Time Skew Detection and Correction
In this paper, we propose a new approach for the automatic time skew detection and correction for multisource audiovisual data, recorded by different cameras/recorders during the same event. All recorded data are successfully tested for potential time skew problem and corrected based on ASR-related features. The core of the algorithm is based on perceptual time-quefrency analysis with a precision of 10 ms. The results show correct time skew detection and elimination in 100% of cases for a real life dataset of 32 broken sessions and surpass the performance of fast cross correlation while keeping lower system requirements.
Korchagin_Idiap-RR-42-2010.pdf
openaccess
932.31 KB
Adobe PDF
adb0a02f84b7843656d4c33b596e8878