Using the Myo Armband with Bela

Here's a little Bela project that uses the Myo armband as a control interface. The Myo armband contains 8 EMG (muscle) sensors, and an IMU (movement) sensor. Here's a demo.

This idea for this project is to connect the Bela to one particular Myo, and pass data from these sensors to a Pd patch so that we can have fun using them to drive synthesis processes.

To use this project, you have to install an extra library on your Bela (instructions below).

How to use:

0. Plug in a Myo bluetooth adapter to your Beaglebone black. Does this work with built-in BLE on the new beaglebones? not sure.
1. Install MyoLinux on your Bela. This involves copying over the source code and building according to the instructions given on the MyoLinux readme. At this point you can test the connectivity using the myolinux_example program that is built when installing MyoLinux.
2. Copy the MyoLinux shared object file to where it will be seen by the build environment: cp /usr/local/lib/libmyolinux.so /usr/lib/
3. Move the project files from this repo to your bela.
4. To connect to a specific Myo you have to specify the address in line 16 of render.cpp, e.g., #define MYO_ADDRESS "c8:2f:84:e5:88:af". You can find this address when running myolinux_example in step 1.
5. Alternatively, you can replace client.connect(MYO_ADDRESS); with client.connect()); from line 465 and the Bela will connect to the first Myo it sees. In a multi-myo situation this could be very bad (or... very good?).
6. Make some cool sounds in the _main.pd file. I've put in an example that uses the EMGs to control the volume of eight oscillators while the pitch (angle) of your arm changes the base pitch (frequency). Fun!

Data you get:

All the sensor data is scaled to [-1,1] and packed in lists that are sent to outlets Pd. The scaling can be adjusted in render.cpp.

• EMGs: 8 values in a list, goes to outlet emg.
• Quaternions (x,y,z,w) -- probably not very useful in Pd -- goes to outlet ori.
• Euler angles (roll, pitch, yaw) -- much more useful -- goes to outlet euler.
• Acceleration (x,y,z) goes to outlet acc.
• Gyroscope (x,y,z) goes to outlet gyr.
• Bonus: magnitude of acceleration (sqrt(x*x + y*y + z*z)) goes to outlet accmag as a float (not a list).

I think the Myo can do some on-board gesture classification, but I don't think the MyoLinux library can extract this data which is a shame. Maybe somebody can update it?

Todo:

• make some cool music
• figure out why the shared object file isn't seen when it's in /usr/local/lib
• see if we can get the classifier data

PS:

Try typing with my example Pd patch running. Fun times with EMGs!

Stillness Under Tension

The _main.pd file is for our work at NIME2018: "Stillness Under Tension" - try it out!