July 30, 2015 QuickLogic and Sensory Partner to Provide Always-Listening, Deeply Embedded Voice Recognition at Ultra-Low Power
June 30, 2015 Sensory Releases Embedded Speaker Verification SDK – Enables Devices to Switch User Profiles Based on Who’s Talking
Embedded Software & SDKs
TrulyNatural SDK– scalable large vocabulary continuous speech recognition for embedded and constrained platforms
Sensory’s TrulyNatural™ technology is capable of recognizing millions of phrases. Designed to create a natural man-machine interface, TrulyNatural™ employs grammar based and language modeling approaches to avoid the limitations of discrete word command-and-control. TrulyNatural™ is targeted for low-cost, large volume applications such as mobile devices, automobiles, home automation, and medical and industrial instrumentation.
The TrulyNatural™ SDK enables you to integrate revolutionary high-quality, advanced speech recognition technology into your high-end embedded consumer electronics, while using moderate amounts of memory and CPU resources. The TrulyNatural™ phonetic recognition engine can build new vocabularies based on text input, even on the device…and it can do it quickly. The SDK is available for Android, iOS, Linux, QNX and Windows as well as ultra-low power Deeply Embedded ports on leading DSP IP cores and integrated circuits.
TrulyNatural has acoustic models for over a dozen languages, tools for text parsing and grammar creation, and Sensory supports common language models for functions like phone dialing, music/media search, and destination entry for navigation. Other language models can be easily imported from open source and standard formats such as OpenFST, Kaldi, and ATT FSM.
TrulyHandsfree SDK– Small vocabulary, ultra-low power speech recognition using state of the art key-word spotting
Sensory’s TrulyHandsfree™ technology is capable of recognizing many dozens of words in command and control settings. Designed to create a flexible and fault tolerant user experience the TrulyHandsfree engine employs word spotting and phrase spotting technologies to avoid the limitations of discrete word command-and-control, and excellent performance in high noise situations. With phrase spotting, the user can say any speech including stuttering, “ums” and many of the other common accidents or sounds that occur while talking yet the recognizer pulls the right words out from all the background. TrulyHandsfree enable always-on, always listening devices that work reliably in real-world conditions and at power consumption as low as 1ma for the complete system. Vocabularies can be built with grammar tools from Sensory, or Sensory can build custom models for the ultimate in accuracy. TrulyHandsfree is supported by over a dozen common languages.
The TrulyHandsfree SDK is available for Android, iOS, Linux, QNX and Windows as well as ultra-low power Deeply Embedded ports on leading DSP IP cores and integrated circuits.