Sensory releases v5 of TrulyHandsfree speech recognition system
Sensory has released version 5.0 of its TrulyHandsfree speech recognition system, which includes significant updates to the embedded AI to significantly increase its performance and accuracy while staying small and low power.
Incorporating Sensory’s smartest and most efficient deep neural network technologies to date, TrulyHandsfree 5.0 delivers an on-device voice user interface experience that is more natural and intuitive than previous versions.
The product’s new shallow learning approach reduces the model sizes to operate in ultra-low power and with minimal memory and MIPS.
Sensory’s latest deep neural network models for embedded AI have allowed the company to deliver a five-times reduction in false accepts compared to version 4.01, which nearly eliminates any chances of the speech recognition system activating when it hasn’t actually been triggered by the user.
A new shallow learning approach reduces the biggest speech models by a factor of 5-10 without impacting the accuracy.
Additionally, the latest neural network models offer greater reliability for user-defined triggers, providing the option for users to select the wake word they prefer, while still having the same accuracy and performance offered with specialized fixed triggers.
TrulyHandsfree 5.0’s embedded speaker verification technology is highly flexible, allowing users to enroll their voice and their own custom trigger or passphrase.
In the event that an unauthorized person learns the trigger or passphrase, the company’s voice biometrics technology will decipher that it’s not the enrolled user speaking and not authenticate the person, preventing access to the device.
The new voice barge-in feature is built on Sensory’s proprietary Acoustic Echo Cancellation (AEC) technology, which is configured specifically to maximize speech recognition system accuracy.
Users can interrupt devices while playing voice prompts, music or other sounds by simply stating the trigger phrase to control music playback by voice, or provide any other kind of supported speech commands.
AEC technology increases the performance of the embedded TrulyHandsfree speech recognizer, as well as any cloud-based speech recognition system that the speech requests are sent to.
The new version incorporates new deep learning noise suppression algorithms that decrease the level of ambient noise provided to the speech recognizer.
As a result, the system is able to clearly recognize wake words and voice requests, which further improves its recognition hit rate. This is particularly useful in home, automotive and mobile applications where background noise can interfere with the volume of the user’s voice.
TrulyHandsfree offers a high resource- and power-efficiency, and is compatible with ports ranging from powerful applications processors and low-power DSP platforms.
For ultra-low power devices with limited battery capacity, Sensory offers its Low Power Sound Detector (LPSD) hardware component for DSPs and smart microphones that can decrease low-power configurations of TrulyHandsfree to operate at an average battery draw of less than a 1mA.