Human hearing focuses on the familiar, but the new system works differently: it catches any sound burst, even those never encountered before. This is audio diarization — not just searching for specific noises, but slicing the entire audio into fragments, marking the start and end of each event. This approach requires no list of known sounds; it’s like a vigilant sentry reacting to the slightest movement.
Technically, the method relies on the same idea as spectroscopy: complex sound is broken down into simple waves, like light into rainbow bands. This possibility was discovered two centuries ago by Joseph Fourier. And entropy, a measure of surprise, helps separate important signals from the background.
Such an "all-hearing" algorithm will be useful in smart homes to detect appliance malfunctions, or in the wild for monitoring animals. And it’s also a step toward assistants for people with hearing impairments. After all, what truly matters is not only recognizing the familiar but also catching the unfamiliar in time.
Even in sleep, our brain instantly reacts to unfamiliar sounds: a sudden rustle will wake you faster than a familiar noise. This principle is embedded in the new system.