I trained a model to recognize gestures for M, N and O using the Google Colab notebook created by Yining.
I first trained the Arduino to collect accelorometer and gyroscope data from my movements in creating the letters M, N and O. I moved left to right with my hand on the breadboard. I logged the data in separate .CSV files that I uploaded onto the Google Colab.
I traied the model on this data then uploaded it back onto the Arduino. Here is a video of the outcome:
Green lights up for M. Yellow lights up for N. Red should light for O. The second video shows the O working, which didn’t work in the first video. The M and the N unsurprisingly were often confused by the model: sometimes I would make an M and the model would identify N; sometimes I would make a N and it would identify an M. In order to get more precise readings, I might need to be cleaner with my training and the starting point. I think also more data points might help with the readings.