The best part of CES each year isn't seeing the latest and greatest gadgets -- it's seeing the burgeoning developments that are going to change how you use those devices in the future. Audience's eS700 series of voice-processing chips is one of those technologies.
A voice processing chip? Yes, a voice processing chip. Think about Siri. It's useful to be able to ask her questions or have her check your calendar appointments when your hands are full. But in a public environment, talking loud enough for her to understand usually means you're disturbing other folks around you or at least calling attention to yourself. And if there's tons of noise around you, like on a busy street with construction, voice recognition and control can be rendered completely useless.
"Our vision is that there is coming a time when sensory computing takes on greater and greater prominence in consumer devices," Audience CEO Peter Santos told WIRED. "Advanced voice and auditory input computing -- we think we're at the very beginning of that process."
Audience's eS700 chips are a key step toward that goal. They offer improved voice recognition in loud environments -- useful for both making phone calls and for performing tasks on your phone using your voice. The chips are also able to passively listen for keywords, so even if the device is asleep, a simple command like "OK phone, play Talking Heads" will be executed. This passive listening is ultra-low power, and for voice commands, you don't need to pause between the initial wake-up command and whatever other commands follow. A new API will let third-party apps take advantage of these features as well, instead of just services built into the OS.
A demo of the chip's capabilities with Google Now were impressive. In a room with a significant amount of background noise (think a conference room where everyone is talking at normal volumes at the same time), a phone with Audience's current model chip was unable to pick out the user's voice from the din -- or if it did, the words were mangled. The new eS700 chip was not only able to identify the user's voice when spoken at equal volume, it was able to do so accurately. The next step: Being able to do this with sub-ambient voice levels, that is with you talking at a volume that's measurably quieter than the noise around you. That's when Audience really thinks speech recognition is going to take off.
The company's past chips are already in a number of smartphones and tablets, including older versions of iPhones, Samsung Galaxy devices, and the Nexus 10. Devices using these latest eS700 chips will start appearing in the latter half of this year.
