Sensory Expands Its Partnerships with ARM
Sensory, a provider of embedded artificial intelligence (AI) technologies, continued to expand its support for ARM Cortex processors, bringing embedded AI to even more product categories. With broad ARM platform support, Sensory offers device manufacturers embedded AI-enabling solutions for on-device voice wake up, speech and natural language recognition and processing, chatbots, computer vision, and voice biometrics recognition, and more, all of which do not require internet connectivity.
Sensory's TrulyHandsfree speech recognition and wake word technology with deep neural nets is available on a wide range of ARM Cortex-M and Cortex-A based processors deeply embedded on-chip or at the OS level. TrulyNatural enables large-vocabulary speech recognition with phrase-spotting techniques for natural language recognition and chatbots on-device via Cortex-A cores running at the OS level. TrulySecure provides face and voice recognition with user identification and biometric authentication via Cortex-A cores at the operating system level.
"ARM has an increasing focus on bringing intelligence from the cloud to the device," said Laurence Bryant, vice president of personal mobile compute at ARM, in a statement. "Sensory has developed some of the best demonstrations of how AI solutions, such as natural language, computer vision, biometric wake words, and more, can be implemented on the edge. ARM's IP engines are well-suited for utilizing these embedded technologies and driving compute innovation forward in this space."
Together, Sensory and ARM are allowing device manufacturers to reduce their reliance on the cloud through deeply embedded AI solutions that run on the applications processor. Sensory's technologies can be used for either completely cloud-free operation or with cloud-based systems.
Embedded AI applications from Sensory that currently support ARM processor-based platforms include the following:
- Deep neural networks;
- Speech recognition;
- Always-listening wake word;
- Statistical language models;
- Chatbots - natural language processing;
- Computer vision; and
- Voice biometrics.
"ARM is uniquely positioned to support all of Sensory's AI-based technologies on multiple platforms," said Todd Mozer, CEO of Sensory, in a statement. "Our common vision of AI distributed to the edge makes for a valuable offering for the ARM developer community and end users."
Billions of devices feature ARM processor-based platforms, ranging from low-power wearables to smartphones and complex IoT hubs. To date, Sensory's deeply embedded technologies and operating system-level software have shipped in more than 2 billion devices from some of the world's largest consumer electronics, toy, and robotics manufacturers. Sensory also provides ports of its solutions to numerous chip providers in the ARM ecosystem, including NXP, QuickLogic and ST Micro, and offers OS-level support for ARM-based applications processors running Android, Linux, and others.
Sensory has boosted the performance and accuracy of its embedded wake word and speech recognition with shallow learning.
Sensory's voice recognition software powers a voice-processing microphone for waking up smart devices.