Ready for a wearable that listens for your snoring — or your stomach? Meet audio machine-learning tech.

You’re jogging down a hectic city street, cranking tunes to your smartphone, oblivious to the realm around you. The intersection ahead looks clear, and you’re ignorant of loud sirens signaling that a speeding ambulance is coming your way. But before disaster strikes, your smartphone shuts off the music and warns you of the coming vehicle.

This is only one of many potential uses of real-time acoustic processing, a machine-learning system that analyzes ambient audio to foretell near-future outcomes. Inside the example above it saved a clueless jogger from being squashed like a bug, however the technology has other potential uses too. It would, as an instance, detect when industrial equipment is set to fail, alert deaf people to alarms and other auditory warnings, helping ornithologists analyze bird calls, or even monitor bodily sounds — equivalent to heartbeats, stomach rumblings, and snoring — to be used by mobile medical apps.

Rapid improvements in mobile devices, most notably faster processors and longer battery life, are helping audio machine-learning technology go mainstream, says One Llama Labs, a brand new York City-based developer of acoustic-processing software.

[There’s more to wearable tech than simply smartwatches. Read Wearables To look at At CES 2014.]

“Wearable technology is now powerful enough to do serious machine learning, even on the audio level. And that technology will change the sector with regards to monitoring,” said David Tcheng, One Llama Labs’ cofounder and chief science officer, in a phone interview with InformationWeek.

The company’s Audio Aware machine-learning app is in a position to analyzing hundreds of sounds, including music, from its surroundings. It will likely be available this month within the Google Play store; One Llama Labs plans to develop iOS and Windows Phone versions too, but no timetable was given.

The audio technology relies on research started a decade ago on the National Center for Supercomputing Applications’ Automated Learning Lab (which Tcheng cofounded) on the University of Illinois at Urbana-Champaign. One Llama Labs’ original focus was on music recommendation technologies — “kind of what like Pandora does but using supercomputers,” explained company cofounder and EVP of commercial development Hassan Miah, who joined the decision.

“The core acoustic, artificial-intelligence machine learning could apply to lots of items,” said Miah. “And now with the emergence of wearable technology, the cloud, and other factors, [our] technology can be utilized well past music. So that is the genesis of ways we came out with the… Audio Aware system.”

The company sees three primary markets for Audio Aware on mobile devices. The 1st: deaf users. “They cannot hear alarms and other alerts,” said Tcheng. “With my previous work with audio recognition and bird-call analysis and speech recognition — as a matter of fact, machine learning — I knew shall we detect these sounds with most of the audio machine-learning software I’ve created.”

The second group: music lovers wearing headphones. “There’s a plague of folks just walking around — just like zombies — attached to their cellphones,” said Tcheng with a chuckle. “And inside the worst case [they’re] cranking music so loud that they cannot hear common threats.”

The third group: those who need to be notified of specific sounds — as an example, nature lovers or users who study birds and other wildlife in outdoor settings.

Medical applications have potential in addition, although identifying bodily sounds may present its own set of technical challenges. “We’ve been serious about doing a snooze apnea application, because each of the system should learn is the right way to recognize a breath,” said Tcheng. “But once you place the microphone on a body, you decide up all kinds of bodily sounds, from heart rate to the digestion system. If you have ever heard someone’s tummy, it makes every type of noise.”

In industrial settings, audio machine-learning technology is perhaps used to tell apart between normally functioning machines, those wanting maintenance, and people about to fail, Tcheng said.

Engage with Oracle president Mark Hurd, NFL CIO Michelle McKenna-Doyle, General Motors CIO Randy Mott, Box founder Aaron Levie, UPMC CIO Dan Drawbaugh, GE Power CIO Jim Fowler, and other leaders of the Digital Business movement on the InformationWeek Conference and Elite 100 Awards Ceremony, to be held along with Interop in Las Vegas, March 31 to April 1, 2014. See the complete agenda here.

Jeff Bertolucci is a technology journalist in La who writes mostly for Kiplinger’s Personal Finance, The Saturday Evening Post, and InformationWeek. View Full Bio

More Insights