Sensory develops low-power wake word capability for TrulyHandsfree
Sensory has announced a breakthrough in the development of its TrulyHandsfree wake word and speech recognition AI engine, allowing it to run directly on Android and iOS smartphone applications at low power. The company says TrulyHandsfree can be adapted to any app without special hardware or digital signal processors.
TrulyHandsfree was launched in 2009, and is featured in many voice assistants, Android smartphones and wearables with specialized hardware. It remains a popular voice control engine, as seen by its integration with several products and solutions announced at CES earlier this year.
The power requirements of wake word solutions for apps, particularly those running in the background, have previously prevented their development, according to the announcement. The complexity of the challenge was increased by the need to deliver the same user experience to multiple operating systems. Sensory began experimenting on Qualcomm and ARM processors to create lower power consumption for wake words on mobile platforms in 2017, and discovered a series of independent actions which reduce the power consumption of an app using wake words by 80 percent, to approximately 1 percent of a smartphone battery charge over a 12-hour period.
“Hands-free operation for voice control has become the norm, and application developers are now looking to create hands-free wake words for their own apps,” said Todd Mozer, CEO of Sensory. “For example, we recently helped Google’s Waze accept hands-free voice commands by supplying them with Sensory’s ‘OK Waze’ wake word that runs when the app is open. With previous versions of TrulyHandsfree, having our always-on wake word engine listening for the OK Waze wake word during a short trip would have had minimal effect on a smartphone’s battery, but for longer trips a more efficient system was desired – so we created it. Sensory is excited to now offer TrulyHandsfree with excellent low-power performance to all app developers!”
Approaches to reducing the power consumption of TrulyHandsfree for app wake words include the “little-big” method of identifying an event of interest before revalidating it with a more power-intensive model, as well as frame stacking and multithreading.
TrulyHandsfree supports independent fixed wake words, user enrolled fixed wake words, and user defined wake words, and Sensory offers off-the-shelf models for all major Virtual Personal Assistant (VPA) services, and for third-party devices that support cloud AI systems from Baidu, Alibaba, and Tencent. The system currently supports more than 20 languages and dialects, and an SDK is available for Android, iOS, Linux, Mac OS, QNX, and Windows.