Rocksus / EMG-DroneControl

A project made to complete my undergraduate studies in computer engineering.

EMG RNN CNN Gesture Recognition

This is a project that I made to complete my undergraduate studies in computer engineering. I used muscle contraction data (EMG) as inputs for a CNN RNN hybrid neural network. I also provided several other models located in the Models folder.

Overview

Model is trained using 8 EMG points, here's an example of a gesture data:

Data was then processed into a several Neural Network Model with different configurations. Working model was then tested using a real drone / simulated environment.

Model Results

Training results from each model can be seen in the Results folder.

Requirements

• Python 3.5 or more
• Keras
• Myo
• Dronekit
• Dronekit-SITL
• Pymavlink
• Pymavproxy
• Mission Planner

Setup

1. Install the required libraries in Python

    pip install dronekit
    pip install dronekit-sitl
    pip install Keras
    pip install myo-python
   

2. Download the Myo SDK and edit the sdk_path accordingly.
3. Test the connection to your drone using DroneControl/connect_check.py
4. If everything is fine and well you are good to go!

Usage

Run

emg_control.py

Located in the DroneControl folder. Note: Please edit the velocity of your drone accordingly inside the code.