A TinyML creative project
Bop it game that uses accelerometer learning.

Accelerometer classification

Edge impulse and accelerometer learning with the BLE sense.

I noticed that you can do lots of things with edge impulse. You can try speech recognition. But there was also a way to utilize accelerometer data. So the robot could detect the movement you do with it. Eg. Picking the robot up. Letting the robot fall. shaking the robot, etc. The possibilities are endless. So to utilize this learning at its max. I'm making a fun bop it game, to illustrate that it can detect different kind of gestures.

Edge impulse?

It was almost the same process as the speech recognition learning. I only had to do minor changes. You can check the edge impulse documentation to learn more about this type of learning.

https://docs.edgeimpulse.com/docs/continuous-motion-recognition

The Arduino sketch

The Arduino sketch was way different than the speech recognition one. This time interference did not wait 2 seconds but it was continuous. You can check this github to know more about my methods:

https://github.com/yorbenGoor123/PersonalPassionProject/blob/master/src/official%20demos/Speech%20%2B%20accelerometer/Brains/accelerometer_works/accelerometer_works.ino

I also needed to send the classification results to my Arduino Uno using the Wire library. This way I could make the LCD and speakers do stuff! You can check this code in this link:

https://github.com/yorbenGoor123/PersonalPassionProject/blob/master/src/official%20demos/Speech%20%2B%20accelerometer/Arduino%20Uno%20code/LCD_Show/LCD_Show.ino

The result

There were three kinds of classification results:

  • Resting
  • Up and down
  • Back and forward

Back and forward:

Up and down:

The feedback system and output are applied with the speakers and LCD module. If you did not do the gestures in time. A sad tune started playing and the music resets. If you hit the gesture in time a victory tune started playing and the delay time substracts with a 20 milliseconds delay. So the song will speed up and you will have to show how fast your reflexes are!

Future:

  • Adding more classification results like:
    • snake
    • wave
    • shake
    • etc

In the final week this will be implemented.