Summary:
At any location where propeller aircraft regularly pass (which includes essentially all of the US), normal video soundtracks includes Doppler-shift events that have distinctive sound signatures, which depend on aircraft speed, heading, altitude, distance from observer and other things. Given a date, full data about all US flights that day (from FlightRadar24), and a mind-bending amount of computation, it’s possible to determine the location from which that video’s sound was recorded.
This is honestly one of the most impressive techniques I’ve ever seen in my life. Absolutely blew my mind, and well worth watching.
TLDR they correlate plane flight paths with recordings from a specific location, creating a trilateration model where the doppler shifts in time and space across tens of thousands of points are simulated and brute-forced until a perfect match to the spectrogram are found. It creates an incredibly precise geolocation, down to a city block or so.
The doppler shift of any one plane from any one observer’s standpoint will depend on how far away the plane is, what kind of engine it has, how high and fast it’s flying, where it’s headed, how it turns, etc. Between detailed flight path data and elevation models and a good-enough (meaning perfectly average) sound recording, you can pretty precisely find matches. With 2+ planes, the window between them becomes another data point (especially once you filter out the silence, i.e., exclude areas that also heard other planes between them). But in the video, he showed that often just a single sub-audible plane recording was enough (humans can’t hear it, but frequency-boosted spectrograms could still show it and the technique still geolocated it).
Anyway, I’m probably not paraphrasing all the details right. It’s well worth a watch. Good explainer and really cool stuff.
It is astounding how many ways there are to geolocate people. Electrical network frequency fluctuations work too. So many things that you can simpy not switch off at will. Anonymity seems no longer feasible except by forgoing communication completely. Blackmailers will have to go back to cutting letters from newspapers and gluing them on paper, provided they can find printed newspapers. And then they will leave DNA traces.