In December 2019 Aspinity demonstrated a power-efficient end-to-end voice wake-up solution for voice-first devices using a couple of STMicroelectronics microcontrollers and the Sensory wake word engine.
Now Infineon plans to combine Aspinity’s Reconfigurable Analog Modular Processing (RAMP) technology with its Xensiv family of sensors, which includes MEMS microphones, pressure, magnetic, current and image sensors.
The RAMP chip uses analog machine learning to recognize patterns in sensor data and infer what is relevant and requires a wake-up. The RAMP chip then triggers an analog-to-digital converter and downstream digital signal processor or microcontroller to perform more complex analysis only on the relevant data, eliminating the power inefficiencies typical of other systems. The RAMP chip can be programmed for application-specific inferencing.
Tom Doyle, founder and CEO of Aspinity, said the approach was suitable for always-on products worn on the body, embedded in a smart home system, or in a smart factory.
Aspinity was technically founded in 2012 and officially spun out from West Virginia University in Morgantown three years later.
Related links and articles: