Design Win

Motion sensing application using AI to recognize human activities

Solution Description

A solution detecting human motions in real-time using artificial intelligence (AI) and machine learning (ML) algorithms for pattern recognition and then display activity information on a smartphone or battery-operated handheld device via a wireless connection.

The application uses an ultra-low-power, high-accuracy LSM6DSL 6DoF inertial measurement unit (IMU) to collect motion data that is sent to an ultra-low-power STM32L475VG microcontroller with a single-precision floating point unit (FPU) and ST's adaptive real-time accelerator (ART Accelerator™) running one of three different pattern recognition algorithms (FP-AI-SENSING1) based on artificial neural networks for real-time human activity recognition (HAR).

An ultra-low-power, dual-core, multi-protocol STM32WB5MMG wireless module with a royalty-free Bluetooth® Low Energy 5.2 protocol stack sends the results of the pattern recognition algorithms to a smartphone or any other battery-operated device where it can be integrated with a dedicated mobile app. This app can be developed based on our ready-to-use BLE sensor mobile application for Android and iOS (STBLESensor).

This type of application is useful across a wide variety of domains including wearables, safety, environmental monitoring, healthcare & fitness, and transportation.

  • All Features

    • Complete ready-to-use firmware featuring an implementation of neural networks for real-time human activity recognition (HAR)
    • All motion data is processed on the STM32 microcontroller, leveraging modern computing capabilities at the edge
    • 3 different artificial neural network (ANN) models with various properties to fit with solution requirement and able to recognize up to 5 different classes (walking, jogging, climbing stairs, biking and driving)
    • Ultra-low-power implementation based on the use of a real-time operating system (RTOS)
    • Compatible with our ST BLE Sensor mobile app (Android/iOS) to display HAR algorithm results
    • Easy portability and scalability across different STM32 MCU series thanks to STM32Cube
    • Compliant with Bluetooth® Low Energy (BLE) SIG specification v5.2