cancel
Showing results for 
Search instead for 
Did you mean: 

Animal Activity detection and classification using MEMS Sensors

Deepak Gupta
Associate II

Hello, we are working on a product, to detect, classify and monitor the activity of a cattle, using a custom built ear tag, that will among other things have a MEMS sensor. I am an expert at embedded development, but dont really have any background on activity tracking.

I am looking for some help with deciding the hardware architecture for the application. Is a 6 axis/9axis (Acc / Gyro / Mag) + a STM32 MCU capable of handling this application? or will I need something more like a small DSP for example?

Any pointers on how I should approach this will be useful.

3 REPLIES 3
Miroslav BATEK
ST Employee

Hello,

yes, this is definitely possible to achieve with STM32 and 6 or 9 axis sensor.

The complexity depends on what exactly you want to detect and classify.

I think the best approach is to capture some set of date using existing device (e.g. SensorTile, https://www.st.com/content/st_com/en/products/evaluation-tools/solution-evaluation-tools/sensor-solution-eval-boards/steval-stlkt01v1.html) and then develop the clasification algorithm offline. This will give you an information how complex is the algorithms and according to the complexity you can decide which MCU to use in your final design.

Deepak Gupta
Associate II

Miroslav,

Thanks for the quick response. My aim is to detect and classify activities like (sitting / standing / walking / eating / chewing), I understand it might be difficult for you to give a binary answer if this can be done with a MCU or we will need a DSP processor for this, but still if you can, that will help immensely.

Miroslav BATEK
ST Employee

I'm sure it can be achieved with STM32, DSP is not needed.

You can check our X-CUBE-MEMS1 package where we offer 20 libraries for STM32 Cortex-M3, and M4. One of the library is for example activity recognition for human, this algorithm doesn't require much computation power of the MCU. Even much complex algorithm can be executed on the Cortex-M3 or M4.

https://www.st.com/content/st_com/en/products/embedded-software/mcus-embedded-software/stm32-embedded-software/stm32cube-expansion-packages/x-cube-mems1.html