The sense of hearing is probably as powerful to a human as being able to see. Hearing helps us match a sight with a sound to evoke the right judgement — a slammed door and an angry face is a negative experience, for instance. It’s no surprise, then, that researchers are trying to add hearing to robots then.
Researchers over at Honda developed a robot named Hearbo (that stands for HEARing Robot, get it?) that not only registers sounds, but also knows how to interpret them. This “super hearing” understanding is part of a research field called Computational Auditory Scene Analysis.
Tools like Siri and Google Talk use a method of voice recognition called beamforming, where your phone records your voice, performs some background-noise reduction on the audio, then tries to figure out what you said. . There are limitations though: you need to be reasonable close to the microphone, and you must speak clearly into it.
Hearbo’s method, called HARK, lets the robot listen to what’s going on around it, pinpoint sounds in space, understand what they are, and respond to them appropriately. And it can do all this in real time—no need for it to stop listening and analyze what somebody said. HARK also gives Hearbo a wider listening range than beamforming allows, distance, and lets it take in audio from more than once source at one time.
For example, Hearbo could be in a crowded room, but still understand and follow specific voices in the room, and then know how to respond. It can even listen to a piece of music and understand it in the same way we do
HARK isn’t just an improvement over what you have on your smartphone; it could one day end up in rescue robots that can find people trapped in earthquake-damaged buildings, for example, or in devices designed to aid the deaf.
We don’t yet know when this new listening technology will come to robots and other types of gadgetry, but Honda and Kyoto University are working to iron out the kinks. This robot already makes me obsolete, though, as I sometimes struggle to separate too many different noises at once!