MIT researchers
MIT researchers create power-efficient chip for dedicated speech recognition

Researchers at the MIT have come up with a power-efficient solution in creating a dedicated speech recognition chip that can cut down power consumption by 90-99% across all real-world implementations. The new chip technology gains precedence considering that always-on speech recognition devices are fully battery operated.

As Android Police reports, the recent research data suggests that any smartphone involved in processing or listening to voice commands will consume around 1 watt of power, while the new chip will barely use up 0.02 to 0.1 milliwatts, depending on its listening period.

The MIT created chip will reportedly filter out ambient noise while listening to voice commands using the language interpretation co-processor that works independent of the main processor. The co-processor identifies and validates a human speech, before activating the speech recognition circuit to process the commands further.

Thereby, it is ascertained that the power-efficient chip could enable devices across various platforms to run on a single charge for months or easily regain energy from the surrounding environment.

Wearable devices and other IoT enabled with smaller battery units could benefit the most out of such power-efficient chipsets for active listening and speech recognition.

[Source: MIT]