Myo band EMG classifier

This was a project for 'introduction to biomechatronics' course at Ainshams university, I used a recurrent neural network (LSTM) to classify EMG signals recorded from the Myo armband's eight probes without any previous filteration. The dataset was collected from 5 volunteers and consisted of three gestures:

  • open hand
  • closed hand
  • pointing hand
The model showed 100% accuracy when trained with data from 3 people and tested with 2 others that it has not seen, showing a good potential of such simple technique with larger datasets.