Instantaneous frequency estimation and localization for ENF signals

Forensic analysis based on Electric Network Frequency (ENF) fluctuations is an emerging technology for authenticating multimedia recordings. This class of techniques requires extracting frequency fluctuations from multimedia recordings and comparing them with the ground truth frequencies, obtained from the power mains, at the corresponding time. Most current guidelines for frequency estimation from the ENF signal use non-parametric approaches. Such approaches have limited temporal-frequency resolution due to the tradeoffs of the time-frequency resolutions as well as computational power. To facilitate robust high-resolution matching, it is important to estimate instantaneous frequency using as few samples as possible. The use of subspace-based methods for high resolution frequency estimation is fairly new for ENF analysis. In this paper, a systematic study of several high resolution low-complexity frequency estimation algorithms is conducted, focusing on estimating the frequencies in short time-frames. After establishing the performance of several frequency estimation algorithms, a study towards using the ENF signal for estimating the location-of-recording is carried out. Experiments conducted on ENF data collected in several cities indicate the presence of location-specific signatures that can be exploited for future forensic applications.