ADS 1299-based Open Hardware Amplifier from OpenBCI.com:

Signal Quality for EEG Registration and SSVEP-based BCI

M. Zieleniewska

, A. Chabuda

, M. Biesaga

, R. Kuś

and P. Durka

Faculty of Physics, University of Warsaw, Pasteura 5, Warsaw, Poland

Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering,

AGH University of Science and Technology, A. Mickiewicza 30, Cracow, Poland

1 OBJECTIVES

Owing to the recent progress in brain-computer

interfaces (BCI), there is a growing interest in

affordable solutions for EEG recording. Several

competitively priced consumer products appeared on

the market in the last decade, however they do not

seem to offer neither the signal quality of

professional systems nor the control required for

scientific experiments. The appearance of the Texas

Instruments ADS 1299 chip, containing an 8

channels, 24-bit analog-to-digital converter, paved

the way for high quality, low-cost and low-energy

solutions, like the OpenBCI recording board. In this

study we compare its performance to a top-class

commercial EEG amplifier from TMSi.

2 METHODS

There seems not to be any generally accepted

method for comparing the overall performance and

signal quality of EEG recording systems. We

propose a direct comparison of signals recorded

simultaneously by the two systems from one subject.

We compare spectra of resting state EEG (eyes open

/ closed) and the quality of the signal used for BCI.

As for the latter, the most common approach is

P300-BCI (De Vos, 2014); however, this paradigm

is the least demanding in terms of the signal quality.

Therefore we choose the calibration procedure from

the SSVEP-based (steady state visual evoked

potentials) BCI, designed for assessing the expected

performance of the BCI.

2.1 Hardware and Software

As the reference system we used the TMSi Porti 32

amplifier (resolution 22 bits, input impedance

>10

Ω, active shielding of electrode cables) from

Twente Medical Systems International and Open

Source software from the OpenBCI.pl project

(Durka, 2012). The tested system consisted of a 8-

channel board and the software provided by the

openbci.com project

2.2 Experimental Setup

To provide possibly straightforward comparison, we

recorded signals simultaneously from two separate

sets of electrodes and grounds, placed close to each

other in alternating occipital locations above the

visual cortex. We used cup electrodes, Ag/AgCl on

shielded cables for TMSi, while those included in

OpenBCI Kit were gold-plated and connected with

unshielded cables. We ran the recordings with all the

available filters turned off, including the notch

filters.

Sampling frequencies were 256 Hz for TMSi and

250 Hz for the OpenBCI board. It was not possible

to use exactly the same sampling on both systems,

so the signal from OpenBCI board was offline

resampled up to 256 Hz before further processing

and comparisons. Both signals were high-pass

filtered at cutoff frequency 3 Hz. Power spectral

analysis was performed by means of Welch’s

estimate.

To assess the potential performance in SSVEP-

based BCI application we used the following

procedure: different frequencies f

…f

were

simultaneously flashing on the neighboring fields of

the BCI Appliance (Durka, 2012); subject was asked

to concentrate on the square indicated by asterix,

flashing with the target frequency f

, while the other

fields were flashing with different frequencies from

the testing set of 10 frequencies from range 13-

22Hz. 8 trials, each 5 seconds-long with 1 second

break, were recorded for each frequency.

Spectral powers in all epochs for each frequency

(except for the target frequency f

) formed the

distribution for the null hypothesis of a random

guess of the target frequency. Z-score was computed

Zieleniewska, M., Chabuda, A., Biesaga, M., Kus, R. and Durka, P..

ADS 1299-based Open Hardware Ampliﬁer from OpenBCI.com: Signal Quality for EEG Registration and SSVEP-based BCI.

for the target frequency as the difference between

the mean of this distribution and the mean power

computed for the non-target frequencies distribution,

divided by the variance of the non-target

distribution.

3 RESULTS

3.1 Technical Remarks

The first and most noticeable advantage of the TMSi

system is the immunity to movement artifacts

stemming from the active shielding of the

electrodes. In case of the tested system from

openbci.com, neither the board nor the electrodes

were shielded, so at the first attempt the signal was

severely contaminated with the 50 Hz artifact from

the mains. To cope with this problem we took

advantage of it’s small footprint, allowing to keep it

close to the body with entwined cables.

3.2 Spectra

Figures 1 presents the power spectral density of 60-

sec epoch of the resting state EEG with eyes open

and closed, recorded simultaneously from the same

subject.

Figure 1: Spectra from the same 60-sec epoch of EEG

recorded simultaneously. Upper panel - eyes open, lower

panel - eyes closed.

3.3 SSVEP

To asses the potential performance in real world

application of SSVEP-based BCI systems, we

recorded the calibration session from the openbci.pl

system. Figure 2 presents the Z-scores (Section

“Signal Processing”) obtained from the signals

recorded simultaneously by both amplifiers.

Figure 2: Z-scores for the differentiation of the SSVEP

responses recorded by the OpenBCI (upper panel) and

TMSi (lower panel) systems. The line indicates 5%

significance of a correct determination of the target

frequency by the BCI.

4 DISCUSSION

In this brief study, a reference EEG system from

TMSi showed performance superior to the OpenBCI

board in terms of suppression of the 50 Hz

interference without notch filter and immunity to

movement artifacts. Spectra of recorded signals and

performance in laboratory recordings of SSVEP had

similar properties. These preliminary results

indicate, that ADS 1299-based Open Hardware

systems may provide signal quality comparable to

the top-class commercial EEG amplifiers,

potentially sufficient for research and advanced BCI

applications.

REFERENCES

Durka, P.J., Kuś, R., Żygierewicz, J., Michalska, M.,

Milanowski, P., Łabę

cki, M., Spustek, T., Laszuk, D.,

Duszyk, A., Kruszyński, M. (2012). User-centered

design of brain-computer interfaces: OpenBCI.pl and

BCI Appliance, Bulletin of the Polish Academy of

Sciences, 60(3), 427-433.

De Vos, M., Kroesen, M., Emkes, R., Debener, S. (2014).

P300 speller BCI with a mobile EEG system:

comparison to a traditional amplifier, Journal of

Neural Engineering, 11(3).