A MUSIC Spectrum Combining Approach for ENF-based Video Timestamping

Abstract

Electric Network Frequency (ENF) fluctuates over time around the nominal value because of power load changes. Such fluctuations have a subtle influence on the artificial light emitted by sources that are supplied by the electric grid. Captured by camera sensors, this information can be used in digital forensic investigations to timestamp, geolocate or authenticate video recordings. However, extracting an ENF signal of sufficient quality from digital videos is still challenging, especially in presence of rolling shutter mechanisms and scene movements that interfere with the ENF extraction. In this paper, we present a novel method for the extraction of ENF signals in video recordings by exploiting parametric and non-parametric spectral estimation, and combining them to limit ENF distortions and increase frequency resolution. The experimental results obtained evaluating our technique using a publicly available dataset demonstrate its effectiveness in timestamping video recordings with a high level of accuracy.