Discover innovative Neuton solutions at Embedded World 23
Modern always-on devices either have limited functionality or discharge very fast. Developers have to compromise between analysis/ functionality complexity and energy capacity. At Neuton.AI, we have managed to establish an innovational approach that helps us develop extremely compact neural networks able to recognize complex activities and keywords with the minimum energy and memory required.
We actually step into a new-gen era of always-on device operation making it possible to build complex functionality that consumes a minimum amount of energy and memory.
Audio and Human Activity Recognition is becoming one of the most sophisticated and promising areas in the AI field. This technology generates great demand in many industries and drives AI providers to create more optimal sound, voice and gesture recognition experiences, like voice assistants, smart robots, home appliances, wearable devices, etc.
The evolution of smart sensors is bringing IIoT devices to the next level, bringing more sophisticated ways to interact with the world. Ultra-low power sensors with built-in AI operate at the microwatt level. They open up never-before-seen possibilities for intelligent devices letting them sense, process and act on their own while running on a single coin battery for years.
But how to create an ultra-tiny KWS model that would fit in a low-power MCU and smart sensors?
We’ve developed a unique approach embodied in our patented Neural Network Framework, Neuton. It empowers users to build audio and sensor-data-based models:
- with excellent generalizing capability
- with minimal size, often less than 1 Kb
- without loss of accuracy
- without compression techniques.
Neuton brings intelligence into the tiniest devices and sensors! and enables users to:
- Have several models on one MCU or sensor
- Wake the device up with tiny audio or HAR models
- Get more business logic on one MCU
- Spend less energy on calculations and boost productivity
- Embed models into tiny pieces of HW, such as smart sensors
- Run models on MCUs even with 8-bit capacity.
Join us to learn how to automatically create and deploy a super compact and accurate ML model without deep data science knowledge, high costs, or extraordinary efforts. The practical session is dedicated to applying the best Tiny ML practices. We will demonstrate two live demos that perfectly illustrate the uniqueness of our approach:
- Keyword Spotting –> 40 Kb in total footprint
- Human activities recognition –> 3 Kb in total footprint.
You will also learn how these models can be created automatically without special knowledge of data science.
HAR model embedded into the intelligent sensor - a live demo in Human Activity Recognition with a model size of only 0.3 Kb in size and 98% accuracy. This model can recognize 6 classes of activity of the most optimal size to-accuracy ratio with the never-before-seen tiniest footprint that allowed to embed it into a low-power resource-constrained Intelligent sensor.
Key Word Spotting model with the tiniest footprint - a solution that can recognize keywords and demonstrates an incredibly small total footprint of 40 kb at one time.
The model has excellent generalization characteristics. It boasts 98% accuracy when recognizing most speakers and produces not more than 0.4 false positives per hour.
We are convinced that advanced approaches to building compact neural networks will eventually open new horizons for modern IoT products. What’s more, always-on devices will get a more complex functionality along with enhanced energy-efficient AI solutions. On the other hand, simpler hardware will let producers optimize their HW costs.