FNET was developed at Virginia Tech and has achieved several successes. Read more about it:
- “FNET system sparks development of power-system analysis tools” (Annual Report 2006)
- “Power grid monitoring network” (Annual Report 2005)
Using frequency measurements from grid to analyze audio
Data from Virginia Tech’s frequency monitoring network (FNET), which monitors the entire U.S. power grid, may help improve the analysis and verification of audio recordings.
With the advent of digital recording, forensic methods of examining audio recordings for editing and authenticity need to be updated, according to Richard Conners, an associate professor emeritus and research associate professor. Traditional methods of analyzing analog audiotapes depend on mechanical and physical interactions with the hardware. One of the most promising new methods was first proposed in Europe a few years ago and suggests that the electric network frequency (ENF) can be used to identify location and time, he says.
“It turns out there is a hum that can be detected in recordings,” Conners explains, “and it matches the frequency of the grid.” The ENF signal is embedded into the audio in recorders that are plugged directly into the wall, as well as in battery-operated devices that pick up the signals from nearby power lines or unshielded electronic devices. Like the frequency of the grid, the ENF signal is a random pattern that varies within a 59.5-60.5 Hz range.
Conners is principal investigator on a $864,000, three-year grant from the National Institutes of Justice to investigate this method. He is working with Yilu Liu of the University of Tennessee-Knoxville, who developed the FNET system while on Virginia Tech’s faculty. FNET is a system of 60 frequency disturbance recorders (FDRs) spread across all three U.S. power grids, that uses GPS timing to synchronize data. The FDRs are plugged into standard 120V outlets at universities and offices across the country. Since the units do not need to be installed at substations, FNET provides an independent observation system of the U.S. power grids.
Conners says the team is doing “basic, not applied research: we’re figuring out questions such as, can you match patterns, and if so, how long do they have to be?” The team will look into the best methods for extracting the ENF from the surrounding background noise and voice patterns of a recording. Right now, Conners says, every known method exhibits a “tragic flaw” under certain conditions.
The team will identify best practices, in order to show how different techniques might be combined or used depending on the recording’s nature and place of origin. Ultimately, the goal is to create a body of evidence, supported by good, peer-reviewed science, that a prosecutor or defense attorney could call upon in a court of law to uphold or disprove the authenticity of recorded audio as evidence. The key focus of the research, Conners remarks, will be improving the “accuracy, reliability, and validity” of the signal processing and analysis procedures.