Aug 10 2015
Sensory Inc., a Silicon Valley-based company focused on improving the user experience and security of consumer electronics through state-of-the-art embedded voice and vision technologies, today announced TrulyHandsfree 4.0, the next evolution of the company's extremely successful embedded small-footprint voice user interface platform.
Compared to previous versions of TrulyHandsfree, version 4.0 with deep learning offers improved performance and up to an 80 percent increase in word accuracy1. TrulyHandsfree 4.0 offers filterbank features as an alternative to MFCC features, along with new phrase spotting techniques. These additions have contributed greatly to the improvement of the technology's overall performance. It also contains Sensory's new neural nets engine that supports deep learning acoustic models, dramatically improving speech recognition accuracy in real-world noise.
TrulyHandsfree 4.0 was designed to take advantage of deep learning and other cutting edge techniques to improve the TrulyHandsfree platform's overall performance and accuracy. Sensory uses a unique form of a neural network with deep learning to achieve acoustic models an order of magnitude smaller than the present state-of-the-art. These neural networks employ the most recent breakthroughs in speech feature extraction to produce superior accuracy in noisy environments. Combining this with the latest in filterbank speech features, overall accuracy is improved across the board, while new spotting techniques ensure that the core part of the user's spoken request can be recognized in the middle of speech, or when surrounded by ambient noise. Most impressively, the technology is able to fully operate with less than a 2mA draw on the battery, and as low as 1mA with Sensory's low power sound detection intellectual property.
Summary of TrulyHandsfree 4.0's Performance Improvements:
- Addition of larger deep learning acoustic models providing substantially higher accuracy
- New algorithms that overcome reverb and harsh acoustic environments
- Advanced filterbank features that improve front-end speech feature extraction
- Fully compatible with TrulyNatural; enables seamless handoff from TrulyHandsfree to TrulyNatural processors
- Enhanced architectural scalability allows for low-power DSP implementations with secondary accuracy improvements at the operating system level
"TrulyHandsfree was an industry-first that literally changed the way people interact with devices of all kinds," said Todd Mozer, CEO of Sensory Inc. "Many thought it was impossible to create an always-listening voice user interface, that had both accuracy and low power, but Sensory did it, and the industry has followed but trailed in performance. TrulyHandsfree 4.0 takes performance to a whole new level with an accuracy, footprint and power consumption that others just can't touch."
Sensory's TrulyHandsfree trigger technology has quickly become the most widely adopted keyword spotting technology in the speech industry. Over a billion products utilizing Sensory's TrulyHandsfree have shipped over the past several years from manufacturers such as BlueAnt, Hallmark, Huawei, LG, Mattel, Motorola, Plantronics, Pantech, Samsung and TomTom. The TrulyHandsfree trigger is a low power, high accuracy, phrase recognition technology that listens for a special "wake up" phrase, yet ignores all other conversations.
TrulyHandsfree 4.0 supports US English, UK English, French, German, Italian, Japanese, Korean, Mandarin Chinese, Portuguese, Russian, and Spanish. The TrulyHandsfree SDK is available for Android, iOS, Linux, QNX and Windows. Additionally, ultra-low-power deeply embedded ports of TrulyHandsfree are available for leading DSP/MCU IP cores from ARM, Cadence, CEVA, NXP CoolFlux, Synopsys and Verisilicon, as well as for integrated circuits from Audience, Avnera, Cirrus Logic, Conexant, DSPG, Fortemedia, Intel, Invensense, NXP, Qualcomm, QuickLogic, Realtek, STMicroelectronics, TI and Yamaha.
For more information about this announcement, Sensory or its technologies, please contact [email protected]; Press inquiries: [email protected]
1. Internal testing shows the new features of TrulyHandsfree offer a 60-80% decrease in word error rate compared to the previous version of TrulyHandsfree, which is already considered best in class.