MyoArmControl

Research of mechanical control via myoelectrical signals using Myo Armband.

Student work in Technical University of Liberec (TUL) under supervision of Dr. Jan Koprnicky, PhD. The work was described in the master's thesis and defended at TUL (full text here).

New results were presented in the article for Student Conferense of Faculty of Mechatronics in TUL (see here).

You can contact me through email: [email protected] or Telegram

Achievements

• a dataset with 7000 samples for 7 different gestures was collected
• effective ML model for gesture recognition was created (Python / TensorFlow / Keras)
• ML model was moved to remote device and control via BLE implemented (Arduino / C)
• a mini game was created to test the model on a PC (Unity3D / C#)

Preview

Requirements

Python: 3.6

• tensorflow==1.4.0
• keras==2.1.3

Unity: 2019.2.10f

• TensorFlow-Sharp (download here)

ML model

Models were created for PC (tested in Unity) and for Arduino (Robot was created)

Model for Arduino

Gestures

Fitting results

The optimal result for the microcontroller was obtained with the model taking the last 8 measurements as input.