Sensory is trying to make it easier for developers to build speech recognition applications for children. To that end, the company has released new speech recognition models that have been specifically trained to recognize children’s unique vocal patterns.
In that regard, Sensory noted that many speech recognition algorithms do not perform as well for minors as they do for adults, largely because the sentence structure and vocabulary of children tends to differ from that of their parents. The new Sensory models are designed to address that problem, and provide developers with a reliable solution for younger audiences. Sensory spent several years collecting and analyzing children’s speech samples to achieve that goal, eventually delivering a product that has a word error rate that is 33 percent lower for children when competing against an algorithm designed for adults.
The new recognizer is compatible with Sensory’s TrulyHandsfree and TrulyNatural solutions, and is available now through the Sensory VoiceHub developer portal. TrulyHandsfree allows the platform to identify common phrases, while TrulyNatural fills out the rest of the vocabulary. TrulyNatural can also run locally on a device, which is noteworthy because the resulting solutions do not require any data sharing, and therefore respect the privacy of minors.
“We challenged the team to create a private and accurate recognizer for kid’s speech and they delivered,” said Sensory CEO Todd Mozer. “This opens up new and fun voice enabled products for kids of all ages.”
On that front, Sensory believes that the kid-friendly speech technology can be used in educational apps, and to enable voice control in toys and wearables geared toward a younger market. The new speech model is already gaining traction with developers, and has now been deployed in integrated circuits from Generalplus Technology. Sensory itself recently updated the TrulyHandsfree engine to offer support for Apple’s Siri voice assistant.
October 5, 2021 – by Eric Weiss