After countless awkward interactions and botched commands, we’re finally getting to the point where it feels natural to have a conversation with our devices. From the rumored Samsung Bixby and Apple’s OG personal assistant Siri in phones to Amazon Alexa and Google Assistant in smart home hubs, using at least one form of voice control tech is becoming the norm.
But this is just the start for smart devices — and a new breakthrough from MIT researchers could make it even easier to include voice command systems in everyday electronics. The MIT team has developed specialized chips designed to make automatic speech recognition far less power-consuming.
The MIT chip requires between 0.2 and 10 milliwatts of power to work, depending on the number of words it’s processing. To put that in perspective, MIT claims most current smartphones running speech-recognition software require about 1 watt of power. That efficiency swing could result in power savings of up to 99 percent, which is sure to have heavy Siri users salivating over all that extra battery life.
MIT’s chip is so efficient because it runs on a much simpler circuit setup than other designs. Instead of always running on power-draining full-scale neural networks — the processors based on the human brain that typically handle speech recognition — the chip depends on a simpler “voice activity detection” circuit to pinpoint words directed at the device. Once the chip is tipped off by someone speaking, the neural network kicks into gear and gets to work.
This type of efficiency could be especially useful as our devices continue to shrink, according to researchers.
“For the next generation of mobile and wearable devices, it is crucial to enable speech recognition at ultralow power consumption,” professor of microelectronics Marian Verhelst said in the MIT press release. “This is because there is a clear trend toward smaller-form-factor devices, such as watches, earbuds, or glasses, requiring a user interface which can no longer rely on touch screen. Speech offers a very natural way to interface with such devices.”
While finding another shortcut to save smartphone battery life and shrink our smart devices is exciting, introducing the chips in basic electronics, which currently don’t have the power capacity for these smart systems, could have even more potential.
The researchers envision a future filled with voice command-enabled devices everywhere as a major feature of the Internet of Things — the term used to describe a near future full of smart, connected everyday objects working together to make our lives more efficient.
If the chips wind up in our devices, you’ll be able to talk to more than just your Echo speaker: Alexa, Siri, Bixby and all the rest could be accessible everywhere.
