This page is part on a series on the Unicorn EEG system, see also the other posts on the design and prototyping of alternative wet “sponge” electrodes, and a 3D-printed case for the Unicorn Naked with connectors for EEG, EMG and ECG..

I am exited to report on the Unicorn Hybrid Black EEG headset that I received a few weeks ago and that I have now been able to explore. Together with colleagues from the Donders Institute, the Baby and Child research center and the Body Brain Digital Musical Instrument project we are going to explore this system, both for non-intrusive EEG measurements in young children and for EEG biofeedback for creative musical applications.

The Unicorn is a low-cost 8-channel wireless head-mounted EEG device developed by Gtec and first released in 2019. Gtec has been developing research-quality EEG systems for Brain-Computer Interfaces for a long time, and in recent years have been organizing a series of BR41N.IO hackathons that target hackers/makers and creative applications. I have worked with Gtec systems at a number of BCI2000 workshops, and we used the Gtec Nautilus for the Cogito in Space project (see also here) Given this background, I had high expectations for this new system and the Unicorn does not disappoint.

Unboxing

It arrived in a nicely designed and environmentally friendly cardboard packaging, including dry electrodes, a pack of 50 stick-on electrodes for the reference and ground, a medium-sized cap, an USB Bluetooth dongle and a micro-USB cable for charging. The Bluetooth dongle was not needed to get it working, it also works fine with the built-in Bluetooth of the Lenovo Yoga and Lenovo Thinkpad laptops that I tested it with.

The electrodes are made of a ductile conductive polymer that feels a bit rubbery. The electrodes seem to be injection molded and have the Gtec logo on them. They seem to be identical to the g.SAHARA Hybrid electrodes.

Here are closeups of the electrode and the clip. The leads from the amplifier click on the electrodes with a custom U-shaped PCB design.

The wireless amplifier is about 63x45x17 mm large and has magnets that hold it firmly in place at the back of the cap. The amplifier has one button and a RGB status led. Besides the 8 leads for the dry electrodes, the amplifier has two leads with standard snap electrodes and comes with Kendall H124SG stick-on electrodes for the reference and ground behind the left and right ear.

Besides the 8 EEG channels, the amplifier includes a IMU sensor with three accelerometer and three gyroscope channels; these are sampled and transmitted at the same 250 Hz rate as the EEG.

On the right side of the amplifier there is something that looks like the microphone boom of a headset, it extends over the right ear and has 8 LEDs. The LEDs can be controlled by the software, but are not used by default and I don’t really see any practical use for them besides implementing some silly neurofeedback system that allows someone sitting next to you to see how much “brain power” you have 🤷 ?!

It comes with a short “getting started” information brochure, and there are online video tutorials.

The EEG cap and electrodes

The photo below shows the cap mounted inside-out on a styrofoam head with the electrodes pointing outward. The electrodes are quite high and stick out 17 mm from the cap. This allows the pins to be wiggled through the hair and touch the scalp. Note that the electrodes have a hole in the center: it is possible to combine them with electrode gel to improve the electrode-skin contact, or to bridge the hair in case the electrode does not have sturdy contact with the skin.

Below it is shown with the electrodes oriented the right way, i.e., inwards.

This is with all the leads attached to the electrodes and with the amplifier mounted at the back of the head. The amplifier mount uses magnets with a sturdy click to a counter piece attached to the cap, i.e., not with velcro which gets worse over time.

Note that by default it comes with a medium-sized cap and if you have a larger head like me, you should definitely also get an additional large-sized cap. My head is slightly over 60 cm in circumference and does not fit the medium cap, but the large cap fits me well. Although not specified in the online shop, I suspect the small cap to be 54-56, the medium cap 56-58 and the large cap 58-60 cm. The dry electrodes need some pressure; with a cap that is too large you would not get good signals, but a cap that is too small will push the electrodes painfully into your scalp. I recommend using a tailor’s measuring tape to measure the circumference of your head or that of your participants.

The software suite

Upon ordering the system I received an email with a download link for the software suite. The software includes a Recorder application to visualize the signals and record them in a CSV file to disk (in a format supported by the FieldTrip toolbox). The Bandpower application does spectral analysis and allows sending these as UDP over the network. Perhaps the most useful is that it comes with an application that streams the data over LSL. The software suite also includes the P300 Speller and Blondy Check applications, but these require an additional license. Furthermore, it comes with a number of SDKs for C/C++ and C#, but regretfully the Python SDK again requires an additional license. Since I don’t develop on Windows anyway, the LSL application suffices to stream the data to another computer. It would have been nice had there been native software support for different Linux and macOS systems; the online documentation on GitHub does include details of the Bluetooth connection, so it should be possible to implement that from scratch.

Evaluating the EEG signal

Here you can see a screenshot of the Recorder software, with me wearing the cap, not moving and with my eyes open. The horizontal scale is 10 seconds, the vertical scale is 100uV, and a 1-30 Hz bandpass filter plus a 50 Hz notch filter were used.

Below is another 10-second recording in which I made some eye blinks and clenched my teeth; it uses the same filter settings and scale. Channel 1 shows the eye blink the strongest and corresponds to the most frontal electrode placed approximately at location FCz. The artifact at the end and the begin of the window (which rolls around) is due to me moving my hands to the keyboard, touching the laptop (which is plugged into power) and pressing the print screen button. It shows a transient increase in the 50 Hz line noise which is not immediately suppressed by the notch filter. Without the power cable plugged in, there is no artifact upon touching the laptop.

I have also uploaded a short recording UnicornRecorder_20220625_121622.csv. It has about 2 minutes of data with me sitting quiet and occasionally closing and opening my eyes. Channel 6, 7, and 8 correspond to the most posterior electrodes placed approximately at location O1, Oz, and O2. Channel 8 (electrode O2) shows the alpha band activity the clearest, which suggests that the electrode behind the left ear is the reference and that consequently the one behind the right ear must be the ground. On the Unicorn GitHub page there are a few more datasets. Besides having 8 channels of EEG data, these recordings also include the accelerometer and gyroscope values and some status channels. You can view and analyze these datasets with the FieldTrip toolbox, for example with the following code:


cfg = [];
cfg.dataset = 'UnicornRecorder_20220625_121622.csv';
cfg.channel = 1:8;
cfg.blocksize = 10;;
cfg.ylim = [-100 100];
ft_databrowser(cfg)


Summary

I have a few minor annoyances with the system: I don’t see the point of the LED “microphone boom” over the right ear, the software needs an online-activated license to work, and the software is Windows-only and closed-source. However, the software comes with an UnicornLSL application, so it is easy to stream the data to a Linux or macOS computer using LabStreamingLayer and to process the data using other software, such as our Python-based EEGsynth realtime software.

Overall the system works great and the EEG signal from the dry electrodes is good; note that it is not as good as gel-based electrodes would give, it is more sensitive motion artifacts and to electrostatic noise such as 50Hz line noise and the participant rubbing their feet on the carpet. But that is to be expected, and when the participant sits still, the signal is good enough for many applications. See for example also these publications from Heis et al. (2022), Fiedler et al. (2022), and Fiedler et al. (2015). I have not yet tried to combine the Unicorn Hybrid Black electrodes with gel, but am confident that this will further reduce the noise of the electrode-scalp interface and make the signal more robust against external noise.

Perhaps the most exiting is that there is also a naked PCB-only version of this amplifier, which invites to adapt it to your own needs. This is exactly what I’ll be doing and I will report on it here in follow-up posts.