Hardware

The g.tec g.Nautilus Research – 32-channel, active dry electrodes – wearable headset was used to collect the EEG data. The headset was placed on the scalp using the ear lobe openings and chin strap as a guide. To establish proper contact between the electrodes and scalp, the headset was gently pulled down and tightened as needed, as shown in Figure 1. In doing this, the headset’s netting places the electrodes in the correct positions. In performing a proper check of electrode placement, another team member reviewed the cap’s placement by verifying that the c3 and c4 electrodes were in line with the wearer’s ears and pressing down on all electrodes to make sure that they were in contact with the wearer’s scalp. To ground the headset, the wearer held the ground electrode and reference electrode directly to the bone behind their ears. This method of grounding was used, as opposed to connecting them to disposable electrode adhesive pads, to allow for improved contact between the scalp and grounding electrodes.

Figure 1. Proper fitting of the g.tec headset, where the ground electrode and reference electrode are manually pressed down.

Software

The data acquisition software g.Recorder was used to display and review the EEG raw data as it was being collected on the Brain Dynamics Laptop. The connection of the g.Nautilus hardware to the g.Recorder software was verified by looking at the Data Viewer, where 32 rows (representing the 32 channels for the headset) appeared on the screen. The channel sensitivity was set between -30 and 30uV with an offset of 0 to properly display data on the data viewer. A frequency band between 1Hz and 30Hz was used to assist with the visualization of the EEG data. The Data Viewer was used to observe if any of the channels were flat-lining. If there was such a channel, the corresponding electrode on the headset was readjusted until the channel was no longer flat-lining. Figure 2 demonstrates an example of an annotated Data Viewer that was displaying EEG data after being properly connected to the headset.

Figure 2. The g.Recorder Data Viewer displaying the 32 channels of EEG data. Note that the signals are not flat-lining, but have small amplitudes due to the natural difficulty of recording brain activity using non-invasive means.

Data Acquisition

The collection of EEG data required the wearer to sit still in a relaxed position, face a non-distracting view, have their eyes open, and blink at a normal pace. The recording began with 10 seconds of sitting still to allow for the headset to initialize the recording. This helped separate the high amount of noise that was generated when the headset began recording. At the 10 second mark, the wearer was told to move to create an identifying event that would be used to account for any time delay in data collection. Afterwards, the wearer waited (without moving) an additional 10 seconds. Therefore, the actual training period began at 20 seconds. At that point, the wearer imagined the following five events in five second duration intervals back-to-back: (1) silence, (2) soft right hand squeeze, (3) soft left hand squeeze, (4) hard right hand squeeze, (5) hard left hand squeeze. Wagner performed this cycle for 180 seconds. Cheung and Fleites performed this cycle for 200 seconds. It is important to note that the wearer did not move their hands as they imagined the scenarios. This is because physical movements can introduce artefact into the signal, while the actual EEG signal reading should be nearly identical between physically moving and imagining the movement. After three minutes of recording data, the recording was stopped and exported as a HDF5 file, which was later converted to a MATLAB file to be processed.