Video: ‘The hidden background noise that can catch criminals’

Posted on 17 January 2022

Amelia Gully shows L&LS graduate Tom Scott how audio and video recordings may contain concealed patterns revealing when and where they were made

‘The hidden background noise that can catch criminals’ featured L&LS graduate Tom Scott and Dr Amelia Gully, British Academy Postdoctoral Research Fellow in the Department of Language and Linguistic Science, University of York.

The video focuses on Electrical Network Frequency analysis (ENF analysis) which matches background hum against power grid logs.

The UK national power grid distributes electricity at a frequency of 50Hz, or rather 50Hz on average. It's important to say 'on average', because there are small, random fluctuations above and below the 50Hz frequency over time. The 50Hz signal, though very quiet, can be detected in sound recordings, including the soundtrack of video footage, because the recording device is either connected to an electrical power source or sits close to one. And because our 50Hz signal fluctuates randomly and never repeats, we can compare what we detect in a recording against a separate, continuously-recorded log of the 50Hz fluctuations. That can tell us quite precisely when in time the recording was made, and maybe where it was made as well. For instance, if the recording is claimed to have been made abroad, we can compare it with the national grid log data to see if there is a match that would suggest the recording was in reality made here. We can also tell whether a recording has been edited, if we find jumps or gaps in the embedded 50Hz signal that don't line up with what is present in the reference data. So a recording that is produced as 'alibi evidence' may reveal that an accused person is not telling the truth; it may also show that a recording has been redacted or otherwise tampered with for criminal purposes. In the video, Tom Scott talks to Amelia Gully to find out more about how acoustics experts can make use of ENF data in the fight against crime.

Dr Amelia Gully is currently a British Academy Postdoctoral Research Fellow in the Department of Language and Linguistic Science with the research project Anatomy, acoustics, and the individual: investigating inter-speaker vocal tract variation for forensic speaker comparison (2019-2023). This project makes use of cutting edge acoustic and morphometric techniques to understand how the unique physiology of a person's vocal tract affects the acoustics of their voice, with applications to forensic speaker identification.

Dr Gully is interested in what makes voices unique, in particular the physiology of the vocal anatomy, captured using medical imaging techniques. Her background is in acoustics and signal processing and her research sits at the junction of phonetics and numerical acoustic modelling. In her research Amelia uses MRI and other imaging technologies to capture the shape of the vocal tract, informing a 3D model of acoustic wave propagation within the airway. She is interested in how differences in vocal anatomy contribute to making a speaker’s voice unique, both for forensic speaker comparison applications and in order to inform personalised speech synthesis for people who have lost their voice. Amelia also supervises student projects on a range of topics related to speech science and signal processing, including projects on electrical network frequency analysis.

I always love talking about ENF analysis because it feels like science fiction, when in reality it’s quite a simple application of well-established digital signal processing techniques. I was delighted to have the opportunity to demonstrate to Tom Scott how this technique can be used to provide evidence for criminal cases. - Dr Amelia Gully, Dept. Language and Linguistic Science