Geek Answers Handbook

How to work with machine learning algorithms in embedded systems?

I am doing a project to detect (classify) human activity using the ARM cortex-m0 microcontroller (Freedom - KL25Z) using an accelerometer. I intend to predict user activity through machine learning.

The problem is that cortex-m0 is not capable of processing training or forecasting algorithms, so I will probably have to collect the data, train it on my computer and then somehow embed it into it, which I really don’t know how to do.

I saw a message on the Internet saying that you can create a weight matrix and integrate it into the microcontroller, so it would be a simple task to predict something based on the data that you provide for this function. Would this be the right way?

In any case, my question is: how could I embed the classification algorithm in a microcontroller?

I hope you guys can help me and give me some guidance, I'm lost here.

Thanks in advance.

+4
source share
2 answers

I thought about doing it myself in order to solve a problem that made it difficult for me to develop a heuristic manually.

, , , , MCU.

, , , MCU. , / .

, :

  • () ( UART, SD- , ).
  • (scikit-learn, weka, vowpal wabbit ..). , , .
  • . , .
  • , . , , . FPU, .
  • . -, , , ..
  • .
  • , (, ), , (, ), MCU. / , - (, const).
+4

, . , - . M ARM. , , , . ARM :

  • M -
  • R -
  • A -

, . A-. , A , , NEON-FPU.

TI Discovery - , , Pi ( , )?

, M0, , , - , ROS-C. , ROS, , , .

ROS

+2

More articles:


All Articles