Myo Arm

December 2021

For my embedded systems class, my team developed a low-cost, 3D-printed prosthetic hand controlled by EMG. Traditional prosthetics can cost upwards of $20k, and we aimed to develop a open-source alternative for <$100. Unlike other prosthetics, our final prosthetic can also be used as a mouse/keyboard for a computer.

We used a Myo armband to record 8-channels of EMG data and transmit the data over bluetooth to an ESP32 microcontroller. The microcontroller processes the signals, and controls the hand using 2 servos or can act as a mouse/keybaord for a computer. The user can switch between “hand” and “computer” modes with a button.

I specifically developed the machine learning training and real-time control software for the project (code here). I made a python app that guides the user through collecting training data with 5 different hand gestures, and then trains an LDA classifier. The classifier weights are exported to C code that runs on the ESP32 in real time. I also implemented the training code on the microcontroller so that the user can train the classifier without needing a computer. To let the user type characters, I created a tertiary morse code where the user uses flex/extend/fist gestures to select letters.