pygesture is a collection of code for recording multi-channel EMG for the purpose of experimenting with myoelectric gesture recognition.

Several signal input sources are supported:

• Delsys Trigno wireless EMG system, via TCP connection to the Trigno Control Utility.

• Measurement Computing USB data acquisition unit(s). Right now, only the USB-1608G is implemented/tested, but additional MCC DAQs can be easily supported (thanks to pyusb and pydaqflex).

The top-level code is primarily for handling EMG data -- filtering, feature extraction, classification, etc. There is also a Qt GUI for recording data, viewing and processing it, and using it for real-time gesture recogition. Finally, there is also code for interfacing with v-rep for robot control simulation. This is an area that I would like to expand on a bit. morse seems like a good candidate for addition to the simulation options.

The direct dependencies are all Python packages (yay!), though many of them have non-Python dependencies. These are the direct dependencies. You will have to find out and choose how to install them yourself. I don't know the minimum version requirements, but I expect any Linux distro package manager to have sufficiently up-to-date versions. Python 2 and 3 should both be supported. I work primarily with Python 3, so if something breaks Python 2 support, I might not notice immediately.

• PyQt: PyQt5 is required. I'd eventually like to relax this (in the style of pyqtgraph). It's not necessary if using pygesture as an analysis library (i.e. not using the ui subpackage at all).
• scipy and numpy: Integral to the package.
• scikit-learn: Just about as important as scipy and numpy.
• pyqtgraph: Install from the git repo to use PyQt5, as PyQt5 support is fairly new.

Optional:

• pydaqflex: Needed for use of a Measurement Computing DAQ.
• PyOpenGL: If installed, the processing widget in the UI takes advantage of pyqtgraph 3D plotting functionality, so clusters of points in feature space are more easily explored.

The recommended approach is to set up a virtual environment with system site packages (because the dependencies don't pip install nicely) and install in develop mode:

virtualenv --system-site-packages env
source env/bin/activate
python setup.py develop

Now you should be able to run the main GUI by just running pygesture.

The test coverage is pretty terrible, but you can run tests anyway:

python setup.py test

Note that it isn't really feasible to test everything pygesture does with unit tests, such as data acquisition and simulation, since they assume you have additional software running or hardware connected. For these things, there are test scripts in the examples/ directory. These "test-like" scripts are all named test_*.py. Specific instructions for running them are included in the scripts themselves.

You can also run static code checks (requires flake8):

make lint

To get v-rep working, you need to set an environment variable VREP with the full path to v-rep's installed location. On Linux, this is wherever you extract the tarball v-rep distributes. You might want to put an appropriate entry in your bashrc, for example: export VREP='~/usr/vrep/vrep-3.2.2'. On Windows, it installs to somewhere like C:\Program Files (x86)\V-REP3\V-REP_PRO_EDU. Find this location (you should see lots of dlls) and add a VREP system environment variable.

You can test communication with v-rep by loading up the vrep_scenes/mpl_tac_test.ttt scene and running examples/test_vrep.py. If that script runs smoothly, everything should be set up correctly. It also provides some example usage of the vrepsim module.

The Measurement Computing (MCC) DAQ relies on the DAQFlex library and pydaqflex. Start by installing DAQFlex on your chosen platform. On Windows, that should be all that's needed. On Linux, you'll need to install the udev rule file (tools/60-mcc.rules). Once that's done, install pydaqflex. Finally, try running the examples/test_mccdaq.py script. If no errors occur, the device should be set up correctly.

The Trigno system only has USB drivers for Windows, and interfacing with the device goes through the Trigno Control Utility. It is essentially a program that runs a TCP server to which you can connect to control the device and retrieve data. To use the Trigno system, start up Trigno Control Utility before running pygesture or some other script. Try running examples/test_trigno.py to see if things are set up correctly -- if no errors occur, it should be ready to go.

• v1.0 : Initial version, including real-time testing (TAC test). Used for the first TAC test experiment.
• v2.0 : Second version, used for the second TAC test experiment published in IEEE Transactions on Neural Systems and Rehabilitation Engineering.