In the first of a few ipXchange interviews with AONDevices at CES 2024, Guy chats with Mouna about AONDevices’ chip- and AI-algorithm-based solutions for providing always-on voice, sound, and sensor-fusion data recognition, which is great for wearables, home automation devices, in-cabin automotive HMIs, and more.
As Mouna explains, the key differentiator here is the super-low power consumption, less than 260 µW for a 12-voice-command recognition workload, as an example. This makes AONDevices’ solution extremely competitive against alternative chipsets from manufacturers like Syntiant, or much more efficient than similar AI workloads running on hardware that is not purpose-built for the task, such as an AI-capable general-purpose microcontroller.
Another advantage of AONDevices’ solution is user privacy; the chip does not require a connection to the cloud to operate, and the high performance of the AI algorithms means that individual users can be recognised by voice alone – including their snoring patterns – without the requirement for expensive, invasive facial-recognition cameras.
The broad application base for AONDevices’ solution is enabled by the flexibility of the AI algorithm, which can be trained using many different datasets to suit a target use case, using the AONx360 software platform. This streamlines the AI training process and provides the required model to be uploaded to the chip, which is built around a RISC-V processing architecture.
Learn more about AONDevices’ solutions by following the link to the board page, where you can apply to test the technology for use in a commercial project – there’s a very nice evaluation board to get you started.
Stay tuned for the follow-up interviews, which go into more detail about AONDevices’ solutions for active noise reduction, keyword recognition, and more.