Sensory Increases Wake Word Accuracy with New High-Res Voice Recognition and Authentication
September 27, 2018
Sensory raises the bar for always-listening wake word and speech recognition, with upgraded turn-key voice interface solutions for devices of all shapes, sizes and power requirements.
Sensory announced that it has made major improvements to its sixth generation of TrulyHandsfree, increasing wake word performance and accuracy by more than 65 percent. TrulyHandsfree’s improved deep-neural network training will provide better near- and far-field speech recognition performance in a variety of room conditions.
Version 6.0’s new high-resolution speech feature front-end boasts a higher resolution digitized representation of the speech audio, supplemented with on-device wake word post-qualification, which will prevent false positives.
TrulyHandsfree will be able to support a multitude of wake words including “Okay Google” “Alexa,” “Hey Cortana,” “Hey Siri” and ““Xiaodu, Xiaodu” in a single implementation. As a result, new products with a user-friendly voice interface can function with more than one digital assistant technology.
The upgraded machine learning technology’s algorithms are trained to anticipate conditions associated with wake word performance, such as word pronunciation, acoustics, device placement, room size, reverb and echo.
One of the drawbacks in the IoT industry is concern for privacy and data security. TrulyHandsfree 6.0 is designed to process all information on the device, keeping voice data completely safe by not storing information or sending it to the cloud.
Version 6.0 also has an upgraded AEC solution that allows users to interrupt their devices by saying the wake word while the device is in use and being able to hear voice prompts from other rooms.
For more information, please visit www.sensory.com.