Analog ML voice eval kit for battery-operated devices
The EVK2 enables the integration of analog machine learning and analog data compression on Aspinity’s RAMP chip (reconfigurable analog modular processor) into voice-enabled devices, such as hearables/wearables, smart speakers and smart TV remotes.
In contrast to alternative always-listening system architectures – which digitize all sound data, relevant or not, before wake-word analysis – the RAMP chip analyzes raw, unstructured analog microphone data at the start of the signal chain to determine if voice is present prior to triggering the wake word engine. Since up to 90 percent of the sound data captured within a day is not voice, the RAMP chip’s analyze-first approach minimizes the power-on time of the analog-to-digital converter (ADC) and wake word engine (WWE), increasing battery life by up to 10x.
The RAMP chip is also the first analog voice wake up solution to continuously collect and compress (into ~2kbytes of memory) the 500ms of sound prior to the wake word (preroll) that is required by most WWEs in order to accurately determine that a command has been spoken.
The Voice-First EVK2 features:
The latest-generation RAMP chip, along with voice activity detection and preroll collection, compression, and reconstruction algorithms.
Audio test files for quick start-up and an audio testing option that uses a MEMS microphone from Infineon for flexible testing
Integration with STM32H743ZI MCU from STMicroelectronics allows the testing of analog voice activity detection with or without preroll collection and delivery to a third-party WWE.
News articles:
Aspinity raises funds for analog ML roadmap, expansion
Infineon gets ‘always-on’ with Aspinity deal
MEMS market decline in 2020 includes options for long-term growth
Swiss neuromorphic startup changes name, raises money
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
