MIT Researchers Create an All-Seeing EarBy
Be careful, the plants have ears. Or, more accurately, they are ears. In a boon to eavesdroppers, researchers at Massachusetts Institute of Technology have figured out a way to reproduce speech by analyzing the surface vibrations of everyday objects. In one experiment, for example, the researchers shot high-speed video through soundproof glass of a potato chip bag sitting on the floor while a person spoke. While to the naked eye the bag was just a piece of litter, it was actually working like a microphone, infinitesimally vibrating in rhythm with the sound waves hitting it. Using their algorithm, the researchers were then able to reproduce the speech. They got similar results by examining the vibrations in a glass of water, the leaves of potted plants, and a box of tissues.
The science is pretty straightforward: Speaking is a matter of making our vocal cords vibrate, which makes the air vibrate in turn. Those vibrations are translated to nearby objects, some of which, if we’re in a conversation, are the eardrums of our interlocutors. The MIT team isn’t the first to think of reproducing sound from surface vibrations—others have designed “laser microphones” that can pick up sounds from the reflection patterns of focused light beams trained on distant objects. What’s new about the MIT team’s approach is that it’s passive. It doesn’t require a laser, or even special lighting. In the potato chip study, the only lighting came through a window. The audio signal they were able to create wasn’t high-fidelity, but it was easily good enough to understand what the person in the soundproof room was saying: Mary Had a Little Lamb, a nod to Thomas Edison, who recited the poem into his new invention, the phonograph, in 1877.
Having dispensed with the need for lasers, the researchers sought to find out whether they even needed a high-speed camera. They found they did not, as they were able to take advantage of a quirk in how most cell phone camera sensors capture images: not in one take, but pixel row by pixel row, top to bottom. By recovering the vibration pattern from each row, the researchers were able to effectively speed up the capture rate of the sensors. All of which makes it easier for amateur eavesdroppers to adopt the method. They don’t need advanced equipment—they just need a mobile phone and a very smart algorithm.