Everyone’s using always-on voice-activated devices everywhere: in their homes, in their cars, in their offices — even on their bodies. While voice-activated devices may make people’s lives easier, they are also collecting — and often sharing — users' personal data.To avoid a consumer revolt against massive data collection and sharing, manufacturers are designing a new generation of voice-first products that integrate more of the voice data analysis into the device itself, rather than sending data off to the cloud for keyword- and command-recognition.
Innovating Edge Processing with Analog ML
Over the past few years, remarkable advances in machine learning technology have introduced on-device data processing (also called edge processing) to voice-activated products. While edge processing addresses the most critical privacy vulnerabilities in these devices, there’s an elephant in the room: The most common edge-processing approach often comes at a high power cost, making it impractical for battery-operated devices.
Aspinity, on the other hand, now makes it possible for engineers to develop always-on, portable, voice-first devices — without compromising on battery life. That’s because our RAMP technology incorporates machine learning with analog processing to provide a new architectural approach to system design that we call “analyze-first.” A RAMP
chip analyzes and detects events from raw analog sensor data in the original domain in which it is sensed, eliminating the inefficiency of digitizing and processing irrelevant data. Let that DSP sleep until you really need it, we say! With Aspinity RAMP, you can run always-on voice-first devices for weeks or months instead of hours or days on a single charge.
Meet Us at tinyML Summit
At the tinyML Summit
this week in Silicon Valley, Aspinity will present its analyze-first approach to always-on sensing. If you’re going to tinyML Summit, we hope you’ll join us for our poster
session. Or, if you want to discuss our new approach to edge processing, email us today: firstname.lastname@example.org
Stay up-to-date on Aspinity news by joining our mailing list.