And ipXchange finally covers something from Arduino! We’ve written about Syntiant’s speech-recognising Neural Decision Processors (NDPs) before, so let’s dive into the Nicla Voice, a 22.86 x 22.86 mm module that promises to make easy work of implementing voice and gesture control in AI-enhanced, battery-operated designs.
For those unfamiliar with Syntiant’s NDP devices, these always-on sensor and speech processors use at-memory computing for direct processing of neural network layers without secondary compilers to enable low-latency AI inferencing at power consumptions small enough for devices to operate for months on a single coin-cell battery. The NDP120 employed on the Nicla Voice module provides a low-cost deep neural network inference engine with a HiFi 3 audio DSP and an Arm Cortex-M0 core at up to 48 MHz. That’s plenty of processing power for many applications, audio or otherwise.
When you pair the NDP120 with Nicla Voice’s onboard microphone, 6-axis IMU, and 3-axis magnetometer, you get an AI/ML module capable of advanced workloads such as noise and vibration detection, low-power speech recognition, and contactless operations, all of which can be used to create speech- and gesture-driven HMIs that can lead to improved safety in applications such as the motorcycle helmet in the cover picture to this article.
What is even more interesting is that the audio-processing capabilities of the Nicla Voice give it the ability to suppress unwanted external noise while simultaneously allowing sounds like fire alarms through, meaning that not only can you use the Nicla Voice to improve safety by keeping hands where they should be when operating complex machinery, you can prevent end-user hearing damage while simultaneously keeping them informed of dangers in their environment; traditional ear defenders compromise between health and safety, rather than promoting both!
You may have now realised that the human voice is not the only sound that the Nicla Voice can be trained to recognise. The sound of shattering glass could be used to trigger break-in alerts in smart security systems, or mechanical sound monitoring might be useful in improving predictive maintenance and process monitoring systems as a means of remote or non-destructive testing.
This module clearly has a lot of potential for AI inferencing applications beyond what has been described above, so learn more about Nicla Voice by following the link to our board page, where you can also apply to evaluate this technology for use in a commercial project.
Keep designing!
You must be a member to leave a comment. Sign up or Sign in