Video: Google Pay users can now use Assistant to send money
In the past, running an always-on, always-listening wake word recognizer as software meant constantly running the applications processor, which consumed too much battery power.
This has meant that hands-free voice UI technology has not yet made its way into mainstream mobile apps. Fortunately, there has been a breakthrough in low-powered wake words technology.
Read also: ReSpeaker releases its 4-microphone Raspberry Pi software for voice applications[1]
Santa Clara, Calif-based vision and voice technology company Sensory[2] has announced a new low-power wake word AI for apps.
Its TrulyHandsfree[3] wake word and speech recognition AI engine can run on Android and iOS platform at low power.
It can be used for any app without special hardware or DSP (Digital Signal Processor).
Previously, creating an always-listening wake word solution for apps would take too much resource and power to be practical -- especially for apps that remain open and active in the background. Your mobile device battery would drain far too quickly.
Sensory's implementation and coding process enables developers to deploy apps with power-efficient, always-listening wake word and command set capabilities across mobile and PC operating systems.
It produced independent actions that, when combined, could lower power consumption on a mobile app using a wake word by more than 80 percent -- around 200mAh per 12-hour day.
That reduction means that a mobile app wake word will consume around 1 percent of the smartphone battery in 12 hours.
Its "little-big" approach uses a very small model to identify an interesting event, which then revalidates the event on a large model processed on the Application Processor.