K. FELIX NAVARRO 1
E. LAWRENCE 2
D. MARTIN 3
1, 2 Faculty of Information Technology, University of Technology Sydney, Sydney, Australia
3 Faculty of Science, University of Technology Sydney, Sydney, Australia
ABSTRACT: The multidisciplinary nature of Brain Computer Interfaces (BCIs) technologies demands scientists and researchers in the area to develop a homogeneous framework and classification of the existent BCI technologies in order to identify, facilitate and accelerate the natural incorporation of the state-of-the-art techniques from the different areas of knowledge. The researchers firstly discuss an extensive literature review and classification of BCI technologies based on Wolpaw’s architecture. As the number of disabled people grows it is vital that researchers develop devices that assist in making the everyday life of such people more bearable. However when developing technologically advanced human computer interfaces researchers have often ignored the importance of the look and feel of the device. In this paper the researchers assess the importance of the Theory of Social Weight when considering BCIs .
Key words: Brain Computer Interfaces, Social weight, health
Millions of dollars are spent by the government of different nations trying to reincorporate into the society thousands of people with severe disabilities. The following figures from the United
States illustrate the magnitude of the problem.
There are approximately 200000 cases of spinal cord injuries – growing at a rate of 11000 new cases per year. Of these new cases 56% are in the
16 – 30 age group and the recovery rate is 0.7%[1].
Our research concerns the use of Brain Computer
Interfaces as devices to bring an alterative channel of control and communication between people severely disabled and the “outside” world. BCIs are intended to enable both the severely motor disabled as well as the healthy people to operate electrical devices and applications through conscious mental activity [2].
Brain Computer Interface research is multidisciplinary involving scientists, computer professionals, communication specialists and usability experts among others. However, much more work is required on the effect of human factors on usability [3]. A BF++ framework for the implementation of a wide range of Cognitive
Biofeedback Systems (CBF) has been developed by [4]. This BF++ system has been described as the most efficient, the most portable across different platforms, the most generic and the one that allows the realization of the cheapest final systems for BCIs in particular. As BCIs become more viable options for users to interact with the physical world it is timely to consider the social aspects behind the design of these devices.
The authors of this paper discuss the applicability of the Theory of Social Weight [5] to
Brain Computer interfaces for studying the effects of human factors on BCIs. The Social Weight of an item of technology is defined to be the measure of the degradation of social interaction that occurs between the user and other people caused by the use of that item of technology [5]. It is important to understand that the nature of a BCI is to adapt to humans instead of following the traditional method of forcing human beings to adapt to the computer interface. Social weight is an important concept as humans are inherently social beings. [5]
The structure of the paper is organized in three main sections. The first section outlines the problem formulation and is followed by the architecture of a BCI according to Wolpaw [6]. It provides an overview of the different methods used to improve the functionality of a BCI after the first
BCI was created twenty eight (28) years ago by
Jacques J. Vidal et al in 1973 [7]. The nextsection outlines the Theory of Social Weight, its significance in wearable computing and discusses its applicability to BCI research. The last section gives our conclusions as well as pointing the way to further research
We have been involved in devising two frameworks as a result of extensive literature

reviews of the work already done on Brain
Computer Interfaces research [8] with the objective of performing, on the next stage of the study, an evaluation of the suitability of a BCI as a wearable computing networked device. The initial framework was based on Wolpaw’s architecture
[6] but further work was undertaken in collaboration with Neil Squire Institute in Canada on the classification and categorization of BCI technologies under the framework proposed by
Mason and Birch in [9] .
The categorization of the
BCI work in this paper is exclusive to Wolpaw’s proposed architecture [6] but a comparison of the problems encountered when classifying the various
BCI systems under the two different architectures is currently being analyzed and will form the subject of a future paper.
Before considering the applicability of Social
Weight Theory to Brain Computer Interfaces it is necessary to provide an overview of the architecture of a BCI system. It is explained and identified in accordance to the different BCI system phases set out below:
1.
Signal Production
2.
Signal Processing Component
3.
Rendering Component
Table 1 shows the linkages among these 3 phases of the BCI and the interaction with computers.
Phases of the BCI
Signal
Production
Explanation Link with
Computer
Processes
N/A
Signal
Processing
Component
Rendering
Component
Electrophysiologica l activity produced by the user
Acquisition, amplification, digitalization, filtering and feature extraction of the signal
Translation of the signals to actions and/or commands.
E.g. Computer
Screen, switching off a device.
Filtering of the signal
(or
Preprocessi ng, e.g.
Artifact
Removal),
Categorizat ion
(Feature
Extraction)
Translating
Process
Table 1: BCI Phases
The Signal Production is the electrophysiological activity produced by the user. The Signal
Processing Component involves the acquisition, amplification, digitalization, filtering and feature extraction of the signal. The rendering component includes the translation of the signals to actions and/or commands.
The official architecture of a BCI is still under discussion, but according to Wolpaw et al. [6] a working BCI is composed of the following parts:
1.
Signal Acquisition
2.
Signal Processing: a.
Preprocessing or Artifact removal
(Includes the cancellation of unwanted data, or noise, as EOG (Electro
EMG Oculoogram)
(Electromyographic). b.
Feature Extraction c.
Translation Algorithm and
3.
Output Device and Output
4.
An Operating Protocol
Figure 1: Procedural Architecture of a BCI
3.1 BCI Classification
As described above, there are many characteristics that differentiate one BCI from another. Some of these characteristics can sometimes play a crucial role in the choice of one BCI system over others.
This will depend on the type of application or problem that requires solving.
One important BCI project is “The Thought
Translation Device” (TTD) by Birbaumer in 1999
[10], which was created to help as a communication device to severely handicapped people by choosing letters on a screen. This device was tested with totally paralyzed, artificially respirated and artificially fed patients, achieving 80-95% success. The individuals needed to be trained for several weeks and months to self-

regulate their slow cortical potentials, as means of communication with the TTD [10].
In 1988 the character recognition BCI system by Farwell and Donchin from the University of
Illinois was an important BCI application breakthrough [11]. A short review of this project is presented below. A mental prosthesis is used for assessing the speed of a P300-based BCI [11]. This system uses P3 evoked potential (one type pf EEG signal) to write a word in a computer. The P3 component of the evoked potential is a response that occurs 300-450ms after a task-relevant stimulus. Since the P3 component is an inherent response, it requires little training for a subject to use a brain-computer interface (BCI) with this component as the control signal [3]. To select the letters, the user has to count the number of times that a row or column containing the letter flashes
(at 10 Hz). The system differentiates the letter from the response amplitude from that row and column that is reliably larger from the rest. (2.3 characters/min) (Accuracy 95%).
Figure 2: Display used in the Mental Prosthesis
[12].
From the few projects involving invasive techniques, one the most sound ones for its innovative and effective use of epidural electrodes is called “The Brain Response Interface” created by Sutter in 1992 from the Wadsworth Center in
USA. The Brain Response Interface [13] requires the implant of epidural electrodes in the user’s brain. The interface utilizes visual evoked potentials (VEP) as input for a word processor (10-
12 words/min), with an accuracy of 90%.
The Implanted BCI created by Kennedy et al
[14] controls a cursor by measuring the rate of neural firing. Neural wave shapes are converted to pulses and the quantity of the pulses decides the action to follow, where the first and second pulses translate to moving up or down, right or left; and three pulses translates to a click on the computer mouse.
Scientists at the Ludwig Bolzmann Institute of
Medical Informatics and Neuroinformattics, Graz,
Austria developed the ERS/ERD Cursor Control
[15] This system reads the sensorimotor cortex and monitors event-related synchronization and desynchronization (ERS/ERD). The user has to control a cursor horizontally on the screen.
(Accuracy 89-100% from 5 students and 60-51% from the other two).
Scientists from the Middendorf Scientific
Services Inc., USA developed the Steady Visual
Evoked Potentials ( SVEP) BCI Flight simulator.
Steady-state visual evoked potentials (SSVEPs) reflect power changes at the stimulus driving frequency and have been used to assess brain activity reflecting cognitive processing [16]. This
BCI system uses Steady Visual Evoked Potentials and operant conditioning methods to train people to be able to control the direction of plane left or right. Scientists at the Laboratory of Nervous
System Disorders, Wadsworth Center, New York
State Department of Health and State University of
New York, Albany 12201, USA developed the .
Mu
Rhythm Cursor Control. [17] The users can move the cursor up and down by imagined movement with an accuracy 90%.for vertical cursor control.
Researchers at the Faculty of Science at the
University of Technology Sydney, in Australia developed the Mind Switch Environmental Control
Unit (ECU). This is activated by a burst in alpha brainwaves when a person closes their eyes ([18].
It is used to control a simple on/off switch, but can be easily adapted to operate a communication device. It has proved to be more than 90 per cent reliable in its latest trials with severely disabled people such as quadriplegics and stroke, polio or cancer-based sufferers [18]. The trials took place in the persons’ homes and participants used the
ECU to control their television:
 switching on and off
 changing channels,
 altering the volume.
Most turned on the switch in one to three seconds more than 90 per cent of the time [18]. The above examples illustrate the potential of Brain Computer
Interfaces but it is important that these devices are adaptable to the human wearing them. Our aim is to examine BCIs taking Social Weight Theory into account and is outlined in the next section.
4 Social Weight Theory And BCIS
In order to be able to perform daily activities effectively, the BCI has to be a transparent device that allows the user to be action focused, with the

freedom required to do this. The researchers contend that implementing wireless technologies with BCIs would help with the process of achieving such interface transparency. The BCI could start to be considered as an alternative input to the control devices on a daily basis. Ideally the
BCI could be adapted to be a wearable computer and, as such, will need to address the issue of
Social Weight [5]. As defined in Section 1, the
Social Weight of an item of technology is defined to be the measure of the degradation of social interaction that occurs between the user and other people caused by the use of the item of technology.
The components of Social weight are set out in the table below:
Component Meaning
Cognitive load
(CL)
Human processor resources required by the task
Physical Presence Form factor, weight, fashion relevancy
Technology apprehension
Apprehension of other people
Social Weight = g(CL + PP +TA)
Table 2: Components of Social Weight [19]
4.1 Cognitive Load of a BCI
Thomas demonstrated the cognitive load of reading a display on various devices as measured against a model human processor and loss of eye contact [19]. Wearable devices such as Head
Mounted Display Units have a high social weight as they may partially or totally occlude an eye.
Reading from a PDA is classified as having a large
Cognitive Load on the human processor and large loss of eye contact. By contrast, a wrist PC has a medium cognitive load on a human processor and involves only medium loss of eye contact. A pager monitor was classified as having a small/medium cognitive load on a human processor and no loss of eye contact. The BCI on the other hand could be considered as having a high cognitive load on a human processor but would score low on the loss of eye contact as the BCI is worn on the head and does not require loss of eye contact . However it is noted that with the Mind Switch[19 ] eye contact is lost as the switch is in fact operated by closing the eyes.
On the other hand the cognitive load on the human processor required by the task of a Brain
Computer Interface has to be measured for every
BCI system and application, as sometimes the use of this interface can vary considerably from one system to another. This will depend of the type of
BCI utilized and the task that is being performed.
For instance the cognitive load of the “Finger
Switch“ by Mason and Birch [9] that uses low frequency bands and is driven by motor cortex recordings would score higher than the one by
Sutter [13] that utilizes Steady Visual Evoked
Potentials (SVEPs) where the user is not required to fully concentrate on the task, as the different frequencies in colour at which the user is staring are matched in the visual cortex.
4.2 Physical Presence of a BCI The Form Factor
Form factor refers to the physical size and shape of the device – in this case the BCI. The difficulty in attaching the electrodes would add to its form factor. The physical presence of a BCI is also problematic –weight is low but its fashion relevancy would count against it.
Figure 3: Brain Computer Interface Cap
As shown it the photo above, wearers of a BCI cap immediately attract the attention of other people but not necessarily for the right reasons. BCI artifacts should be made to look good or in fashion. A person wearing a BCI would feel vulnerable as part of an excluded or socially rejected group. Given that many wearers of BCIs are already disabled the BCI could further damage their self esteem.
One of our authors participated in many sessions using a BCI whilst working at the Neil Squires
Institute. The wearer of the BCI found it difficult

to use and difficult place on her head. The BCI requires precision in positioning the electrodes in the exact location on the scalp. An assistant was required to put the necessary gel in each of the electrodes. The user finds this task a detriment to using a BCI as the process normally can take from thirty to forty minutes. A needle is used for application of the gel and the residue in the subject's hair after the BCI was removed was very distressing. There are also issues with having thick hair, as the gel in some cases doesn’t make contact with the scalp. Observers find the look of the typical BCI unappealing and some researchers have attempted to disguise the BCI as a baseball cap (see next section).
As humans are inherently social it is important to design devices
 that people will actually use – if they feel inhibited they will limit their use of the device
 that do not limit how and when users access the device data {Thomas}.
In future research more precise results of the
Social Weight of a Brain Computer Interface may be obtained by measuring a specific BCI system and application, as sometimes the use of this interface can vary considerably from one system to another. For instance, the social weight of using a
BCI when navigating an aeroplane will score lower than the BCI system utilized to talk to somebody else [Spelling Device]. In the first case, the pilot is expected to use and be “seen“ with many types of devices to control and monitor the navigation of the plane, whereas in the second case you are not expected to be seen wearing a special device to talk or the device should not overwhelm the observers. For this reason Craig et al have developed the baseball cap BCI [19] for their BCI.
As baseball caps are so common, people do not give wearers a second glance .
Figure 4: The New Mind Switch:
The Wearable Computer eSuit [5] also attempts to make the computing interfaces as unobtrusive as possible. The following devices are used as examples: a 2 line * 64 character watch display, a pager motor in a shoulder pad and LED lights on a suit cuff.
The trajectory the BCI area has been continously limited, as shown in this paper, by technology constraints. Some of these constraints could be potentially solved by the use of interdisciplinary collaboration, advanced computer model techniques, improved hardware for online signal processing and communication tools. The convergence of digital and biological networks provides a more realistic platform for issues and potential problems to solve, for instance effective communication for the paralysed, breaking the frontiers of space and time, giving different forms of expression to humankind and effective learning techniques. This could lead to the futuristic conversion and transmission of information carried in bits and bytes to spikes for the stimulation of the corresponding neurons leading to a more effective human-to-human communication and interaction.
Our literature review and interaction with BCIs indicates that most of the research has been done in the areas of alternative communication channels for people with physical disabilities. The researchers believe that personalizing BCIs by correlating EEG findings and additional techniques like fMRI, EMG and MEG would make it possible to identify some user’s key signal characteristics when performing specific daily activities of a person; for instance the maneuverability of industrial machines samples, or more simply on the control of house appliances.
6.1 Wireless BCI
Much work needs to be done on BCI to get a BCI that is a wearable computer interface but it is now in the realms of possibility. The development of the first BCIs seen as a wearable computer, hardware standardized and application specific (for example by using web services), would work as a catalyst for finding many more applications that use the input of a BCI in their systems.

The researchers intend to prototype a wireless
BCI . When considering the suitability of a BCI as a Wearable computer, some of the first things that come to our mind are speed, flexibility and usability among others. One of the visible obstacles that it is necessary to overcome is actually to be able to wear as transparent as possible a BCI. This implies making the interface unnoticed, but also wearable computer interfaces affords and requires mobility, and the freedom from the need to be connected by wire to an electrical outlet, or communications line.
6.2 Personalized fMRI EEG BCI
As scientists are now working on understanding the brain are now able to capture detailed images of brain activity the moment a thought occurs using magnetic resonance imaging (MRI) [15].
The researchers' hypothesis is that with this method the functioning, accuracy and speed of a
Brain Computer Interface could be improved by personalizing the BCI with data obtained with the use of studies in functional neuroimaging (fMRI) and magneto encephalography (MEG) of the BCI user. By cross relating this information with the real time data obtained from the EEG signaling, new and more specific combinations of points showing EEG activity are expected to be found and pinpointed. By obtaining a more specific and personalized combination of data, a BCI could be previously trained utilizing adaptive intelligent computer systems techniques with the objective of achieving a faster and more accurate EEG pattern.
It is expected to obtain more accurate results by recreating and controlling the environment, for instance by the use of virtual reality or augmented reality as a matter of having a virtual constant in the environment to interact with the interface when performing the fMRI and MEG studies, and then utilize the same environment when using the proposed previously trained EEG BCI.
REFERENCES
[1]Black, M. Biologically Embedded Computing, http://www.cs.brown.edu/people/black/
[2] Mikko Sams and Jukka Heikkonen Brain
Computer Interface
http://www.lce.hut.fi/research/bci/
[3]Bayliss, J. D., The use of the P3 evoked potential component for control in a virtual apartment , IEEE Transactions on Neural Systems and Rehabilitation Engineering, v.11, no.2, June
2003.
[4]Luigi Bianchi, Fabio Babiloni , Febo Cincotti,
Serenella Salinari, Maria Grazia Marciani. An
Object Oriented Approach To Biofeedback
Applications For Disabled People
,
Bled,
Slovenia, 2003
[5] Aaron Toney, Barrie Mulley, Bruce H.
Thomas, and Wayne Piekarski , Minimal Social
Weight User Interactions for Wearable Computers in Business Suits ISWC2002, 6th International
Symposium on Wearable Computers
October 7-10, 2002 - Seattle, Wa, USA
[6] Wolpaw, J.R., et al., Brain-computer interfaces for communication and control.
Clin
Neurophysiol, 2002. 113 (6): p. 767-91.
[7].Vidal, J.J., Toward Direct Brain-Computer
Communication.
Annual Review of Biophysics and Bioengineering, 1973. 2 : p. 157-180.
[8].Felix Navarro, K. Wearable Wireless Brain
Computer Interfaces in Augmented Reality
Environments ITCC, 2004 p643
[9].Mason, S.G. and G.E. Birch, A
Generalframework for brain-computer interface design.
IEEE Trans.Neural Syst.Rehabil.Eng,
2003. 11 (1): p. 70-85.
[10].Birbaumer, N., et al., The Thought
Translation Device – communication using selfcontrol of slow cortical potentials.
IEEE
Tansactions on Rehabilitation Engineering, 2000: p. 190-193.
[11]Farwell, L.A. and E. Donchin, Talking off the top of your head: toward a mental proshesis utilizing event-related brain potentials.
Electroencephalogr Clin Neurophysiol, 1988: p.
510-523.
[12] Donchin, E., K.M. Spencer, and R.
Wijesinghe, The mental prosthesis: assessing the speed of a P300-based brain-computer interface.
IEEE Trans Rehabil Eng, 2000. 8 (2): p. 174-9.
[13]Sutter, E.E., The brain response interface: communication through visually-induced electrical brain responses.
Journal of Microcomputer
Applications, 1992. 15 (1): p. 31-45.
[14]Kennedy, P.R., et al., Direct control of a computer from the human central nervous system.
IEEE Transactions on Rehabilitation Engineering,
2000. 8 (2): p. 198-202.
[15]Babiloni, F., et al. Mahalanobis distancebased classifiers are able to recognize EEG patterns by using few EEG electrodes . in The 23rd
Annual International Conference of the IEEE
Engineering in Medicine and Biology Society (Cat.
No.01CH37272). IEEE. Part vol.1, 2001.
Piscataway, NJ, USA.Oct.
2001. Istanbul, Turkey.
. IEEE.
[16] Middendorf Scientific Services, Virtual
Simulation and Training, http://www.virtualsimulation.com/associates/midd endorf.htm

[17] Perlstein WM, Cole MA, Larson M, Kelly K,
Seignourel P, Keil A. Steady-state visual evoked potentials reveal frontally-mediated working memory activity in humans.
Neurosci Lett. 2003 May 22;342(3):191-5. http://www.ncbi.nlm.nih.gov/entrez/
[18] Craig, A. Good news for disabled - UTS
Mind Switch a winner at innovation awards http://www.uts.edu.au/new/releases/2003/March/2
7.html2004
[19] Thomas B Social Weight and pervasive
Computing, Proceedings of the 2 nd Monash
CoolCampus Workshop on Mobile and Pervasive
Computing in the University of the Future, 2004.
14.Mann, S. Wearable Computing as means for
Personal Empowerment . in International
Conference on Wearable Computing ICWC-98.
1998. Fairfax VA.