Skip to main content
D
Denise SANFILIPPO
ST Employee
June 27, 2025

Getting started with human activity recognition on ST AIoT Craft

  • June 27, 2025
  • 0 replies
  • 1681 views

Introduction

Human activity recognition (HAR) is a crucial aspect of wearable technology, enabling the classification of various physical activities such as sitting, standing, running, and walking. STMicroelectronics offers a solution for HAR using the machine learning core (MLC). It's integrated into the smart MEMS sensor LSM6DSV16X, available on the SensorTile.box PRO board kit. For wireless tracking, the ST AIoT Craft mobile app, available both on Google Play Store and Apple Store, can be used.

1. Getting started

To begin, click on the human activity recognition project available inside the project examples. A pop-up window appears with a brief description, as shown in Figure 1.
Next, click on the three vertical dots next to the [Try out] button and select [View details].

Figure 1: Human Activity Recognition projectFigure 1: Human Activity Recognition project

 

This reveals that the project uses a single AI model named human_activity_recognition, targeting the SensorTile.box PRO and the IMU LSM6DSV16X with a machine learning core. The model classifies data into four categories: running, walking, standing, and sitting.
 
Figure 2: Human Activity Recognition: a ready-to-run use caseFigure 2: Human Activity Recognition: a ready-to-run use case

 

 

You can evaluate the AI model either through the web application or via the mobile app. If you choose the mobile app, QR codes for installation are provided.

 

Figure 3: Choose your environment: either Web browser or ST AIoT Craft mobile appFigure 3: Choose your environment: either Web browser or ST AIoT Craft mobile app

 

 

For now, let's focus on evaluating the AI model directly on the web application using the SensorTile.box PRO.

1.1 Firmware update and connection

  • Update the firmware: Connect the SensorTile.box PRO to your PC using a USB Type-C® cable. Turn on the board in DFU mode by holding down user button 2 (BT2) and powering on the board via the power switch (SW).
Figure 4: Turn on the board in DFU modeFigure 4: Turn on the board in DFU mode

 

 

  • Flash the firmware: Click the [Flash firmware] button to download and stream the firmware binary to the board. Ensure that your browser detects the connected SensorTile.box PRO. Once the firmware update is complete, establish a connection with the board by clicking the [Connect device] yellow button.
  • Enable sensors: The AI model is downloaded and programmed onto the SensorTile.box PRO. The accelerometer settings are configured as follows:
    • ODR = 30 Hz
    • FS = 4 g
    • Machine learning core enabled

1.2 Evaluation phase

Click the [Start] button to begin evaluating the AI model. The SensorTile.box PRO recognizes the following states based on its placement and movement:

  • Running
  • Walking
  • Standing
  • Sitting

To end the evaluation, click the [Stop] button.

Conclusion

You can effectively classify the state of your activity using the human activity recognition project example on the ST AIoT Craft platform: share your experience!

Create your own solution exploring the "My datasets" section and the "My projects" section. 

Figure 5: Explore My datasets and My Projects sectionsFigure 5: Explore My datasets and My Projects sections

 

 

Stay tuned for more insights on how to leverage ST AIoT Craft for your projects!

Related links

Terms & ConditionsAccessibility statement