Hardware for the muscle, blink and jaw-clench switch¶

This page helps you pick and connect the hardware for the biosignal switch. It is deliberately modest: YazSes has been tested on synthetic and replayed signals, and no specific board, electrode or placement has been shown to work for a particular person. Treat everything here as a starting point and check it with yazses biosignal doctor and yazses biosignal monitor.

Choose a path¶

The research page explains why the DIY sensor is the better trigger: the same muscle signal is cleaner measured directly on the muscle than through the skull.

The DIY serial path¶

Connect the sensor to a microcontroller that speaks YESP, plug it in over USB, and set the port:

[emg]
device_port = "/dev/ttyUSB0"   # empty means disabled; a Windows port looks like COM3

An empty device_port means the serial switch is off. Run yazses restart after you set it. This path is documented in the configuration reference under [emg]. On Linux your user may need to be in the dialout group to open a serial port. The yazses biosignal commands do not read this path; they work on the BrainFlow, file and synthetic sources.

The BrainFlow path¶

1. Install the extra into the environment YazSes runs from (yazses[emg-band]; see the install section).
2. Find your board's integer id with yazses biosignal boards.
3. Set the source, for example brainflow:0:/dev/ttyUSB0 for an OpenBCI Cyton on a serial dongle, and run yazses biosignal doctor.

A board that is unplugged or busy is logged and leaves the switch unavailable; YazSes carries on without it.

Other devices and recordings¶

These sources do not go through BrainFlow. Each one is opened only when you name it, and each one keeps its library optional.

yazses biosignal record --out session.edf (or .bdf) writes a file that EDFbrowser, MNE and most vendor tools open. Before you rely on a new device, run yazses biosignal check-source <source>: it opens the device for a few seconds and checks that the stream is well-formed (no dropped packets, a steady sample rate, consistent channels). While the switch runs, a stream that breaks those rules closes the switch the same way a bad electrode does.

None of these devices has been tested with YazSes on real hardware yet. They were built and tested against their published protocol documents.

Which channels are read¶

What YazSes can detect depends on which channels your hardware gives it:

If you do not have a channel for an event, that detector has nothing to read. yazses biosignal doctor reports the channels it found, and calibrate skips a step your source cannot support and says so.

Where to put the electrodes¶

This repository documents the kind of site for each event (the table above) and nothing more specific. For the exact position, electrode type, skin preparation and reference or ground placement, consult your board's manual. Two practical points that apply whatever the hardware:

• Good contact matters more than the exact spot. A flat, clipped or noisy channel is ignored by the quality gate, on purpose, and doctor and monitor tell you which.
• Re-calibrate after you move an electrode. A profile made for one placement is wrong for another.

Set the mains frequency¶

Powerline hum is far larger than the signals YazSes wants, so the filters place a notch on it. Set mains_hz in [biosignal] to the frequency where you are:

• 50 for most of Europe, Africa, Asia and Oceania (the default);
• 60 for the Americas, South Korea, Taiwan and western Japan.

A wrong value leaves real hum in the signal and raises the false-positive rate. yazses biosignal doctor and monitor report interference they see, so if they flag mains noise on an electrode that is well seated, check mains_hz first. Keep leads away from power cables and power supplies.

Check it before you rely on it¶

yazses biosignal doctor      # chain, sample rate, mains, flat or clipped channels, profile
yazses biosignal calibrate   # your own thresholds
yazses biosignal monitor     # what makes it fire
yazses biosignal field-test  # how often it fires by mistake, by activity

Start with doctor and read its last line, the Verdict:: it says either that you are ready or the one thing to fix next (a missing library, a source that will not open, no calibration). It is the quickest way to tell a hardware problem from a configuration one. The quality gate behind it also recognises an electrode that has lifted off (an abrupt step followed by a flat line) and movement shared by every channel at once, which usually points at the cable or your posture rather than one badly seated electrode.

Measure your own rate¶

No board, electrode or placement has been shown to give a low false-activation rate for any particular person. That is the honest state of the evidence, and it is why the last command above exists: yazses biosignal field-test walks you through sitting still, talking, chewing, typing and looking around with your real hardware connected, counts what fires when you did not mean it, and reports it per hour for each activity, with an upper bound. Nothing is typed or acted on and no signal is saved. A short session bounds a rate only loosely, so repeat it on different days and after re-seating the electrode before you believe either a good or a bad result, and treat what you get as true of your board, your placement and your day. Details, and what the numbers do and do not mean, are in Measure your own rate.

If you would like your result to count towards a wider picture, you may share the JSON the test writes (--out) together with a sentence on your board and where the electrode sat. It holds counts and durations, no signal. Do not share a recording.

Then keep judging the switch on your own numbers with yazses biosignal report. A recording of your signal is sensitive data; see Privacy.

See also¶

• Use a muscle, blink or jaw-clench switch: configuration, calibration, the speech guard and the false-positive budget.
• Muscle and brain control: the honest hierarchy: what EMG and EEG can and cannot do.
• Hands-free computing with a muscle switch.