A Wireless EEG Acquisition Platform based on Embedded Systems

F. Pinho

, J. H. Correia

and N. S. Dias

2,3,4

Department of Industrial Electronics, University of Minho, Guimarães, Portugal

Life and Health Sciences Research Institute (ICVS), School of Health Sciences, University of Minho,

Campus Gualtar, 4710-057 Braga, Portugal

ICVS/3B’s, PT Government Associate Laboratory, 4710-057 Braga/Guimarães, Portugal

DIGARC, Polytechnic Institute of Cavado and Ave, 4750-810 Barcelos, Portugal

Keywords: EEG, Wireless, Dry-Electrodes, Embedded Systems.

Abstract: This paper proposes a wireless EEG acquisition platform based on Open Multimedia Architecture Platform

(OMAP) embedded system. A high-impedance active dry electrode was tested for improving the scalp-

electrode interface. It was used the sigma-delta ADS1298 analog-to-digital converter, and developed a

“kernelspace” character driver to manage the communications between the converter unit and the OMAP’s

ARM core. The acquired EEG signal data is processed by a “userspace” application, which accesses the

driver’s memory, saves the data to a SD-card and transmits them through a wireless TCP/IP-socket to a PC.

The electrodes were tested through the alpha wave replacement phenomenon. The experimental results

presented the expected alpha rhythm (8-13 Hz) reactiveness to the eyes opening task. The driver spends

about 725 μs to acquire and store the data samples. The application takes about 244 μs to get the data from

the driver and 1.4 ms to save it in the SD-card. A WiFi throughput of 12.8Mbps was measured which results

in a transmission time of 5 ms for 512 kb of data. The embedded system consumes about 200 mAh when

wireless off and 400 mAh when it is on. The system exhibits a reliable performance to record EEG signals

and transmit them wirelessly. Besides the microcontroller-based architectures, the proposed platform

demonstrates that powerful ARM processors running embedded operating systems can be programmed with

real-time constrains at the kernel level in order to control hardware, while maintaining their parallel

processing abilities in high level software applications.

1 INTRODUCTION

Electroencephalography (EEG) signals have been

used in numerous clinical applications, ranging from

sleep studies, epilepsy, to brain function monitoring

in intensive care units, and in several research

applications used alone or in conjunction with other

neuroimaging techniques (Niedermeyer and da Silva

2004); (Hung-Yi et al., 2006); (Campbell et al.,

2010); (Lin et al., 2010); (Castellaro et al., 2011);

(Pichiorri et al., 2011). Ambulatory EEG recordings

with high-electrode density are often challenging. In

applications such as epilepsy studies, which require

long-term recordings, the electrolyte gel starts to

dry, and the electrode’s impedance increases causing

the reduction of the signal-to-noise ratio.

The high-cable density that usually connects the

patient to the acquisition platform also limits EEG

applicability. Due to the large number of channels

(32 channels or above) needed to record EEG

signals with high spatial resolution, the setup

comprises a minimum of 34 cables (considering

reference and ground) connecting the EEG channels

to the acquisition platform. These cables should

remain stable in order to avoid artifact

contamination into the EEG signals. Thus, all the

applications are extremely limited to “non-

movement paradigms”, and so, the signals are

recorded in abnormal environment for the patients.

Wireless technology applied to the EEG

recording has been developed in the past few years

(Modarreszadeh and Schmidt, 1997); (Hsieh et al.,

2006); (Matthews et al., 2007); (Lin et al., 2008);

(Penhaker et al., 2010); (Usakli, 2010); (Park et al.,

2011). There are some interesting platforms that can

fulfill some applications, such as drowsiness

detection and epilepsy monitoring (Modarreszadeh

and Schmidt, 1997); (Hsieh et al. 2006). The already

reported EEG wireless platforms, present

190

Pinho F., Correia J. and Dias N..

A Wireless EEG Acquisition Platform based on Embedded Systems.

DOI: 10.5220/0004325601900196

In Proceedings of the International Conference on Biomedical Electronics and Devices (BIODEVICES-2013), pages 190-196

ISBN: 978-989-8565-34-1

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

communication platforms based on Zig-Bee (Shao-

Yen et al., 2009) and Bluetooth (Hsieh et al., 2006)

protocols, channel density ranging from 2 to 14

channels, sampling frequencies below 256 SPS,

resolutions usually below 12 bits (Hsieh et al.,

2006); (Penhaker et al., 2010), wet or dry electrodes

(Matthews, McDonald et al. 2007; Sellers, Turner et

al. 2009).

Despite all of these developments, the

ambulatory monitoring of epilepsy or sleeping

disorders requires a new technology paradigm: long-

term monitoring with a large number of channels

and online data selection.

With the development of embedded systems and

real-time signal processing techniques, there is a

growing interest in applying them to these novel

platforms (Chin-Teng et al., 2006); (Hsieh et al.,

2006); (Penhaker et al., 2010). Features like low-

power consumption, small size and high-speed

performance, the embedded devices can produce

signal-acquisition based applications that may be

easy to migrate to other platforms with smaller and

more powerful devices.

The hardware of these platforms must face the

best of two requirements, low-power consumption

and enough speed to acquire, process and transmit

the signals as close as possible to real-time. To

accomplish these requirements, some platforms are

based in ASIC (Application-Specific integrated

circuit), but a low-cost COTS (Commercial off-the-

shelf) EEG platform is becoming more feasible.

There have been made some advances in

embedded systems multi-core processing (Chin-

Teng et al., 2006); (Majoe et al., 2012). Because

OMAP (Open Multimedia Architecture Platform)

architecture has both an ARM and DSP cores, some

studies have explored the parallel use of the ARM to

interface and communicate with peripherals, with

DSP to manage and to process signal data

recordings.

This work proposes an EEG acquisition platform

based on a OMAP embedded system, low-power

COTS hardware components, with dry electrodes,

signal resolution of 16 bits, sampling rate of 1000

samples per second, and wireless TCP/IP-Socket

transmission.

2 SYSTEM ARCHITECTURE

The developed system can be divided into three

major functionality blocks: the active dry electrodes,

the analog-to-digital converter (ADC) and the

central processing and transmission unit (CPTU).

The diagram of the proposed system is given in the

Fig. 1.

Figure 1: Diagram of the EEG acquisition platform.

The EEG electrodes are connected to the ADC

ADS1298. The developed software driver

establishes a bidirectional communication between

the ADC ADS1298 and the CPTU both for ADC

configuration and data acquisition. Once the

acquired data arrives to the CPTU memory buffer,

they are remotely transmitted to a PC by wireless

TCP/IP (Fig.2).

Figure 2: Diagram of the whole EEG acquisition platform

and remote PC for signal visualization.

2.1 Processing and Communications

Unit

The Processing and Communications unit is

composed by Gumstix’s Overo

Fire development

platform that includes an OMAP3530-based

computer-on-module that delivers ARM cortex-A8

running at 720 MHz, DSP (Digital Signal

Processor), 3D graphics acceleration and 802.11b/g

wireless communication on-board. Its software

development is given by a high-level Open-source

Linux operating system, the Angstrom Distribution.

The main functions of this unit are:

 Acquisition data samples from the ADS1298;

 Store data conversions in a SD-card for offline

analyses and backup;

 Data wireless transmission by TCP/IP-socket to a

PC or a mobile platform.

To accomplish these functions, we developed a linux

ActiveDry

Electrode

Sigma‐Delta

Analog‐to‐Digital

Conversion

Processing&

transmission

Driver

AWirelessEEGAcquisitionPlatformbasedonEmbeddedSystems

191

character driver in the “kernelspace” environment

that implements the following tasks: bi-directional

communications with ADS1298 and access to the

data conversion readings from a “userspace”

application. To communicate with ADS1298 we

used a SPI Bit-Bang technique to control the

required sequence and GPIO (General Purpose Input

Output) toggle timings. For synchronization

purposes, we developed an interruption in a GPIO

that manages the sampling frequency conversions.

After programming the ADC’s 25 registers, the

driver synchronizes readings with an interrupt

routine given by ADS1298’s DRDY (Data Ready)

signal (Fig.3).

Figure 3: Diagram of the interaction between ADS1298

and communication driver.

The driver also allocates two equally sized blocks of

memory: one is used to manage the data readings in

every interruption; the other one is a mapped

memory block to be used as a “shared memory

buffer” with a flag to sign an event occurrence (Fig.

4).

It was developed also a “userspace” application

that accesses the “shared memory buffer”, saves the

data readings in a SD-card and sends them

wirelessly by TCP/IP-socket.

The diagram of the application and driver

interaction is presented in Fig. 4.

Figure 4: Diagram of application and driver’s interaction.

When the number of readings reaches the size of

the memory block, the driver locks that process,

copies the data to the “shared memory buffer”, turns

the flag “on” and unlocks again the readings process,

all just in time to not compromise the sampling

frequency. At the same time, the application checks

the flag’s state and when it’s turned “on”, accesses

the “shared memory buffer´s” data, saves it in the

SD-card, sends it by a socket structure and turns the

flag “off”. The flag’s double access from the

application and driver allows them to be

synchronized.

The memory blocks can be resized according to

the transmission data rate, and restart the whole

process.

2.2 Analog-to-Digital Conversion

Sigma-Delta Texas

ADS1298 performed the

analog-to-digital conversion. This component was

chosen due to its low power and specific design for

bio-signal application features. The ADC features 8

channels with 24 bits resolution, sampling

frequencies from 250 SPS to 32 kSPS,

instrumentation amplifier with CMRR of -115dB,

Programmable Gain Amplifier (PGA) of 1 to 12,

Driven-Right-Leg (DRL) input, input noise of

4μVpp (150Hz BW, G=6) and power consumption

of 0.75 mW/Channel.

All ADS1298 configurations, sampling rate,

PGA, reference, DRL properties, lead-off detection,

and others, are accessible through its 25 registers.

Through the SPI protocol, we managed to program

the ADC with 1000 SPS sampling rate, PGA of 6,

internal reference of about ½ of VCC, and DRL

switched on.

In continuous reading mode, the ADS1298

provides an output signal timer interrupt (DRDY-

Data Ready), which allows the synchronization with

the processing unit.

2.3 Active Dry Electrodes

The active dry electrodes were assembled in two

parts: the mechanical interface with the scalp and the

signal conditioning circuit.

A set of 25 probes, each consisting of a head, a

plunger, a spring and a barrel (Fig. 5), interfaces

with the patient’s scalp. The probes were soldered to

a round copper plate of 1cm radius. On the copper

plate’s top, we soldered a spring clip that is attached

to the electronic part of the electrode (Fig.5).

The proposed conditioning signal circuit (Fig. 6)

uses a bootstrap technique based on a unity gain

buffer (TLC272), first order low-pass filter at 160

kHz and high-input impedance at low frequency

signals (<100 Hz) of 900 GΩ as shown in the

impedance spectrogram in Figure 7.

BIODEVICES2013-InternationalConferenceonBiomedicalElectronicsandDevices

192

Figure 5: Active dry electrodes schematic.

Figure 6: Electrode´s electronic schematic.

Figure 7: Impedance spectrogram simulated in TINA

SPICE for the proposed circuit.

3 RESULTS AND DISCUSSION

The presented platform was tested in respect to

driver and application performance, communication

throughput, power consumption and overall system

functionality.

3.1 Performance of Embedded System

The driver showed a SPI clock frequency of

2,5MHz, as a result of a bit-bang technique. The

driver spends about 420 μs to read 512 kbits (32

Channels x 16 bits x 1000 samples) for each

interruption. The time that the driver needs to copy a

set of data from de “readings memory buffer” to the

“shared memory buffer” is about 305 μs. In overall,

for each interruption given by DRDY, the driver

took about 725 μs to read the 512 kbits and copy

them to the “shared memory buffer”.

The time required for the “userspace” application

to access the 512 kb in the “shared memory buffer”

is about 244 μs. The read/write speed in the SD-card

showed a 470 Mbps, and took respectively 1.4ms to

save 64Kb.

Then, the synchronization flag was configured to

be turned on after the acquisition of 250 samples

(128 kb), 500 samples (256 kb) and 1000 samples

(512 kb). For each configuration, we measured the

time required for the application to access the

“shared memory buffer” and send the data through

wireless TCP/IP-Socket. It resulted respectively in 2,

3 and 5 ms, with an average throughput of

12.8 Mbps. In our case the sample block is 1000

samples so the total time of acquisition and

transmission is 5 ms.

Other studies applied different wireless

communication protocols, like Bluetooth and

Zigbee. Usually, these systems provide less than 8

channels, less than 12 bits resolution per channel,

and sampling frequencies of less than 256 SPS

(Hsieh et al., 2006); (Shao-Yen et al., 2009);

(Christoforou et al., 2010). In these cases, the

bandwidth necessary to transmit wirelessly the

signal data is quite small in comparison to our

platform (12.8 Mbps). In studies of epilepsy or

sleeping disorders, a higher number of channels,

higher resolution (at least 16 bits) and higher

sampling rates (at least1000 SPS) are often required.

The power consumption of the proposed

platform was measured. As expected, the WiFi

communication module is the most expensive part.

Acquiring and saving data to a SD-card with WiFi

turned off consumes about 200 mAh. When the

wireless feature is turned on, the consumption

increases up to 400 mAh. Therefore, a 6000 mAh

lithium ion polymer battery will be able to power the

proposed platform during 30 hours (WIFI off) or 15

hours (WIFI on). In comparison to other systems

already reported (Parthasarathy et al., 2006);

(Sullivan et al., 2007); (Yates and Rodriguez-

Villegas, 2007); (Yazicioglu et al., 2009), our

platform consumes more power, due to the usage of

a more expensive communication protocol.

Although the other platforms may consume less

energy, they also provide fewer channels, less signal

resolution and lower sampling rates, which renders

in a data package a lot smaller than the data package

transmitted by the proposed platform. The

requirements of high channel density, signal

resolution and sampling rate drove the decision of

selecting an 802.11g communications infrastructure,

even considering the power consumption limitations.

In applications such as long-term monitoring of

AWirelessEEGAcquisitionPlatformbasedonEmbeddedSystems

193

epileptic patients, the EEG acquisition platform must

read the EEG data for periods as long as 24 hours

continuously. Because long-term monitoring

produces large amounts of data, an event detection

algorithm is mandatory. The transmission of selected

EEG segments with identified epileptic-like activity

allows the reduction of bandwidth and power

consumption requirements.

3.2 EEG Data Acquisition Signals

A simple TCP server application that receives data

from the platform and saves it in a file was

developed. At the same time, KST2 software was

configured to open the file and process the

visualization in real-time as the data arrives. KST2 is

an Open-source software platform for visualization

and processing of large-dataset signals.

With the ADC inputs shorted to ground, an input

noise of 5,7 μVpp was measured. According to

(Usakli, 2010), EEG noise amplitude should be less

than 2 μVpp. To overcome this limitation, a new

version of the proposed platform will be developed

based on an EEG-specific ADC (ADS1299 with 1

μVpp noise input feature) instead of the ADS1298,

more suitable for electrocardiogram monitoring

applications.

Then the active dry electrodes were plugged to

the scalp of a normal subject for EEG signals

acquisition. First the subject was asked to be relaxed

and keep eyes closed. As a preliminary evaluation in

EEG recordings (Webster, 2009), in awaken relaxed

subjects, a paradigm of the visual alpha rhythm was

employed. The alpha rhythm appears on occipital

regions of the scalp when the eyes are closed, and

disappears when the eyes are opened. In this

experiment, two bipolar dry electrodes (both with

and without the electronic active circuit) were placed

on the scalp of the subject, on the Cz and O1

positions, forming one EEG channel. The signal was

band-pass-filtered (fourth order FIR-Finite Impulse

Response) at 0.5-40 Hz and the power spectrum

amplitude was calculated. The Figure 8 presents

three signals, eyes open (EO), eyes closed (EC) and

the respective power spectra. As shown in Fig. 8, the

data is comprised in an amplitude envelope of 40

μV.

In Fig. 8.c. we can observe the differences in

power spectra of EO and EC signals, specifically in

the 8-13 Hz frequencies (alpha rhythm). As

expected, the spectrum from the EC segment shows

higher power on alpha frequency range than the

spectrum of the EO segment.

The time spent to mount the cap and electrodes

was significantly less then the traditional wet

electrodes.

Finally, the results showed that the proposed

system seems to be feasible and satisfies the

requirements established at first.

Figure 8: EEG data acquired with a) eyes open (EO), b)

eyes closed (EC) and e) (i) EO versus (ii) EC power

spectra.

4 CONCLUSIONS

The proposed system integrates the development of

EEG dry electrodes, linux embedded system

programming of “kernelspace/userspace” and

wireless transmission techniques.

The proposed platform integrates an ARM core

of OMAP embedded system and interfaces with a

sigma-delta ADCs with 16 bits resolution and

sampling rate of 1000 SPS. Once the data arrives in

the processing unit, they are saved in SD-card and

sent by 802.11g TCP/IP-socket protocols with

minimum delay.

Although the proposed monitoring system

presents higher power consumption than others

BIODEVICES2013-InternationalConferenceonBiomedicalElectronicsandDevices

194

already reported, it provides higher channel density

with more signal resolution and state-of-the-art

sampling rate, which establishes an interesting

tradeoff between power consumption and system

flexibility.

The development of monitoring platforms such

as the one proposed challenges the traditional usage

of microcontrollers to interface with the ADCs and

implement low level hardware operations. Currently,

the powerful ARM processors running embedded

operating systems can be programmed with real-

time constrains at the kernel level to control

hardware, while maintaining their parallel

processing abilities in high level software

applications.

ACKNOWLEDGEMENTS

This work was supported by FCT with the reference

project FCOMP 01 0124-FEDER-010909

(FCT/PTDC/SAU-BEB/100392/2008), FCOMP 01

0124 FEDER 021145 (FCT/PTDC/SAU-

ENB/118383/2010) and by FEDER funds through

the Programa Operacional Fatores de

Competitividade – COMPETE and National Funds

through FCT – Fundação para a Ciência e

Tecnologia with the reference Project: FCOMP-01-

0124-FEDER-022674.

This work is also supported by ADI Project "DoIT -

Desenvolvimento e Operacionalização da

Investigação de Translação" (project nº 13853,

PPS4-MyHealth), funded by Fundo Europeu de

Desenvolvimento Regional (FEDER) through the

Programa Operacional Factores de Competitividade

(POFC).

REFERENCES

Campbell, A., T. Choudhury, et al. (2010). NeuroPhone:

brain-mobile phone interface using a wireless EEG

headset, ACM.

Castellaro, C., G. Favaro, et al. (2011). Hardware for

seizure prediction: Towards wearable devices to

support epileptic people, IEEE.

Chin-Teng, L., H. Hung-Yi, et al. (2006). Development of

portable wireless brain computer interface with

embedded systems. Biomedical Circuits and Systems

Conference, 2006. BioCAS 2006. IEEE.

Christoforou, C., N. Mavridis, et al. (2010). Android tele-

operation through Brain-Computer Interfacing: A real-

world demo with non-expert users.

Hsieh, H. Y., S. F. Liang, et al. (2006). Development of a

real-time wireless embedded brain signal

acquisition/processing system and its application on

driver's drowsiness estimation, IEEE.

Hung-Yi, H., L. Sheng-Fu, et al. (2006). Development of a

Real-Time Wireless Embedded Brain Signal

Acquisition/Processing System and its Application on

Driver's Drowsiness Estimation. Systems, Man and

Cybernetics, 2006. SMC '06. IEEE International

Conference on.

Lin, C. T., Y. C. Chen, et al. (2008). "Development of

wireless brain computer interface with embedded

multitask scheduling and its application on real-time

driver's drowsiness detection and warning."

Biomedical Engineering, IEEE Transactions on 55(5):

1582-1591.

Lin, C. T., F. C. Lin, et al. (2010). "EEG-based Brain-

computer Interface for Smart Living Environmental

Auto-adjustment." Journal of Medical and Biological

Engineering 30(4): 237-245.

Majoe, D., L. Widmer, et al. (2012). A Reconfigurable

Multi-core Computing Platform for Robotics and e-

Health Applications. Computer and Information

Science (ICIS), 2012 IEEE/ACIS 11th International

Conference on, IEEE.

Matthews, R., N. McDonald, et al. (2007). "Novel hybrid

bioelectrodes for ambulatory zero-prep EEG

measurements using multi-channel wireless EEG

system." Foundations of Augmented Cognition: 137-

146.

Modarreszadeh, M. and R. N. Schmidt (1997). Wireless,

32-channel, EEG and epilepsy monitoring system.

Engineering in Medicine and Biology Society, 1997.

Proceedings of the 19th Annual International

Conference of the IEEE.

Niedermeyer, E. and F. L. da Silva (2004).

Electroencephalography: basic principles, clinical

applications, and related fields, Lippincott Williams &

Wilkins.

Park, J., L. Xu, et al. (2011). Wireless dry EEG for

drowsiness detection. Engineering in Medicine and

Biology Society, EMBC, 2011 Annual International

Conference of the IEEE, IEEE.

Parthasarathy, J., J. Hogenson, et al. (2006). Battery-

operated high-bandwidth multi-channel wireless

neural recording system using 802.11 b, IEEE.

Penhaker, M., M. Stankus, et al. (2010). Design and

Application of Mobile Embedded Systems for Home

Care Applications, IEEE.

Pichiorri, F. and et al. (2011). "Sensorimotor rhythm-

based brain–computer interface training: the impact on

motor cortical responsiveness." Journal of Neural

Engineering 8(2): 025020.

Sellers, E., P. Turner, et al. (2009). "A novel dry electrode

for brain-computer interface." Human-Computer

Interaction. Novel Interaction Methods and

Techniques: 623-631.

Shao-Yen, T., T. Chung-Han, et al. (2009). A wireless

biomedical sensor network using IEEE802.15.4. Life

Science Systems and Applications Workshop, 2009.

LiSSA 2009. IEEE/NIH.

Sullivan, T. J., S. R. Deiss, et al. (2007). A low-noise,

AWirelessEEGAcquisitionPlatformbasedonEmbeddedSystems

195

non-contact EEG/ECG sensor. Biomedical Circuits

and Systems Conference, 2007. BIOCAS 2007. IEEE,

IEEE.

Usakli, A. B. (2010). "Improvement of EEG signal

acquisition: An electrical aspect for state of the art of

front end." Computational intelligence and

neuroscience 2010: 12.

Webster, J. (2009). Medical instrumentation: application

and design, Wiley-India.

Yates, D. C. and E. Rodriguez-Villegas (2007). A Key

Power Trade-off in Wireless EEG Headset Design.

Neural Engineering, 2007. CNE '07. 3rd International

IEEE/EMBS Conference on.

Yazicioglu, R., T. Torfs, et al. (2009). Ultra-low-power

wearable biopotential sensor nodes, IEEE.

BIODEVICES2013-InternationalConferenceonBiomedicalElectronicsandDevices

196