Speech recognition chip can bring voice commands to more products

Thanks to the success of products like Amazon echo and Google home, not to mention Apple's Siri voice assistant, the idea of talking to our devices has become more and more used to. Low power speech recognition chip is the trend of communication between people, products and technology.

Voice engineers believe that a low-power speech recognition chip can save 90% to 99% of power compared to a typical mobile phone running voice recognition software today. For example, an iPhone running Siri may require about 1 watt of power, while a low-power chip can do the same thing in the range of 0.2 to 10 MW, depending on the amount of vocabulary that needs to be recognized.

In addition to our personal mobile and wearable devices, low-power chips may be critical for adding voice recognition to the many sensors and other power constrained devices that make up the fast-growing internet of things. Such devices may take a long time between charging, or they may collect energy from the surrounding environment, effectively utilizing the power supply critical to their operation.

"Voice input will be a natural interface for many wearable applications and smart devices," said anantha chandrakasan, a professor of electrical engineering and computer science at MIT "The miniaturization of these devices will require a different interface from touch or keyboard. Compared with performing this operation in the cloud, it is very important to embed voice functions locally to save system energy consumption. "

Voice chips use some key strategies to save energy. The first is a relatively simple "voice activity detection" circuit, which monitors ambient noise to determine whether the sound in its environment constitutes speech. Only when it recognizes the actual speech will it start a large, more complex and energy consuming speech recognition circuit.

The second strategy involves compressing data and evaluating a small portion of the audio in on-board memory at a time to maintain low power requirements.

Marian Verhelst, Professor of microelectronics at the Catholic University of Leuven, Belgium, reiterated: "for the next generation of mobile and wearable devices, it is essential to achieve speech recognition with ultra-low power consumption." He is also working on more efficient devices and sensors. "This is because there is a clear trend towards smaller devices Voice provides a very natural way to interact with such devices. "