Even MIT researchers are of the opinion that voice-enabled communication is the future of technology. The advent rise of smart devices in the last one year has given credibility to voice-activated AI assistants such as Google Assistant and Amazon Alexa. The mobile is also one of the few avenues that’ll be completely engulfed with AI-powered services in the coming months. Thus, it would require its own innovations to propagate the said technology.
With regards to the same, MIT researchers have today announced the development of a low-power chip, designed specifically for automatic speech recognition. But, speech recognition is not the highlight of this innovation. Instead, the chip produced by MIT is capable of 90 to 99 percent power savings in real-world applications. And if this is true, then the said chip could have some impressive application in AI-enabled electronic systems.
Speaking about the chip, MIT professor Anantha Chandrakasan, head of the project, said,
Speech input will become a natural interface for many wearable applications and intelligent devices. The miniaturization of these devices will require a different interface than touch or keyboard. It will be critical to embed the speech functionality locally to save system energy consumption compared to performing this operation in the cloud.
Today’s devices are packed with virtual assistant which may be required to continuous speech recognition. Due to the same, it will often be required to stay in always mode to detect commands. This could lead the battery of your electronic device to drain in no time, and charging the device every few hours is a massive chore (and headache). Thus, the researchers at MIT have designed the chip keeping power effeciency in mind.
The chip only uses a fraction (~0.2 to 10 milliwatts) of the power (~1 watt) required to run speech recognition on a mobile device. The power usage of this new chip has been reduced to such an extent by not running complete neural networks in the background at all times. Instead, a “voice activity detection” circuit has been integrated into the chip to activate “the larger, more complex speech-recognition circuit” when human voice is spotted.
Further, Michael Price, the MIT graduate student who led the design of the chip and now works for chipmaker Analog Devices, continues to add,
I don’t think that we really developed this technology for a particular application. We have tried to put the infrastructure in place to provide better trade-offs to a system designer than they would have had with previous technology, whether it was software or hardware acceleration.
This new chip will also help improve the internal structure of Internet of Thing (IoT) devices, wearables, smart home speakers and mobile devices as well. Voice (alongside touch) are expected to become the only prominent mode of communication with electronic devices in the future.
