Multimedia Tools and Applications

, Volume 74, Issue 19, pp 8655–8667 | Cite as

Tactile-force brain-computer interface paradigm

Somatosensory multimedia neurotechnology application
Article

Abstract

This study explores the extent to which a neurotechnology multimedia application utilizing tactile-force stimulus delivered to the hand holding a force-feedback joystick can serve as a platform for a brain-computer interface (BCI). We present a successful application of an extended multimedia paradigm beyond the classic vision and auditory based approaches. The four pressure directions are used to evoke tactile brain potential responses, thus defining a tactile-force brain computer interface (tfBCI). We present brainwave electroencephalogram (EEG) signal processing and classification procedures leading to successful online interfacing results. Experiment results with seven advanced and five naive users performing online BCI experiments provide a validation of the hand location tfBCI paradigm, while the feasibility of the concept is substantiated by noteworthy information-transfer rates.

Keywords

Tactile media interface Brain-computer interface Brain somatosensory evoked response Brainwave signal processing and classification 

1 Introduction

A state-of-the-art BCI is usually based on mental visual and motor imagery paradigms, which require substantial user training and good eyesight of the user [17]. Alternative solutions have recently been proposed to make use of spatial auditory [12] or tactile (somatosensory) modalities [1, 4, 9, 10, 15] to enhance brain-computer interface comfort. The concept proposed and reported in this paper is based on a previous pilot study conducted by the authors [6], and further extended experiments [5]. It expands on the brain somatosensory (tactile) channel previously reported by our research group [8, 9, 13] to allow for tactile-force based stimulus application. The rationale behind the use of the tactile-force somatosensory channel is that it is usually far less loaded and more intuitive to learn compared with auditory or even visual modality interfacing applications. Additionally a very recent report [4] has confirmed the superiority of the tactile BCI in comparison with visual and auditory modalities, as tested with a locked-in syndrome (LIS) user [11].

Another report [15] proposed utilizing as the tactile BCI a Braille code stimulator with 100 ms static force push stimulus delivered to each of six fingers to generate a somatosensory evoked potential (SEP) response and the following P300 attentional modulation. The P300 response is a positive electroencephalogram SEP deflection starting at around 300 ms and lasting for 200−300 ms after an expected stimulus in an oddball (random) series of distractors [17]. In this paper, examples of averaged P300 responses are depicted in the form of color coded diagrams in Fig. 4 and using time series plots with purple lines with standard errors in Fig. 5.

The P300 brain response is considered to be the most reliable and easy to capture from EEG in the majority of human users. Thus, the P300 is commonly used in BCI applications [14, 17].

This paper extends our previous report [6] on the novel successful application of the tactile-force BCI by adding a test with a group of naive BCI users (who had never participated in any BCI experiments before). The results of the naive user study are presented in order to validate the proposed tactile-force brain computer interface (tfBCI) usability even with beginners. This confirms that there is no necessity for the extensive training which often limits BCI applicability in the case of imagery paradigms [17]. We present very encouraging results obtained with seven advanced and five naive healthy users of whom the majority scored far above chance levels of accuracy in online BCI experiments.

The remainder of the paper is organized as follows. The next section introduces the materials and methods used in the tactile-force BCI study. It also outlines the experiments conducted. The results obtained in EEG online experiments with seven advanced and five naive healthy BCI users are then discussed. Finally, conclusions are formulated and directions for future research are outlined.

2 Materials and methods

The experiments in this study involved seven advanced (six males and one female; mean age 24.71 years, with a standard deviation (SD) of 7.5 years) and five naive (BCI beginners; three males and two females; mean age 22.4 years, SD 0.55 years) healthy users. All the experiments were performed at the Life Science Center of TARA, The University of Tsukuba, Japan. The online EEG BCI experiments were conducted in accordance with The World Medical Association Declaration of Helsinki - Ethical Principles for Medical Research Involving Human users. The psychophysical and EEG recordings for BCI paradigm experiment procedures were approved by the Ethics Committee of the Faculty of Engineering, Information and Systems at University of Tsukuba, Tsukuba, Japan. The participants agreed voluntarily to take part in the experiments. The details of the tactile-force stimulus creation, psychophysical and EEG experiment protocols are described in the following subsections.

2.1 Tactile-force stimulus

The tactile stimuli were delivered as movements generated by a portable computer in a MAX 6 [2] environment, as depicted in Fig. 1, in the form of a visual interface with instructions to the user. Each tactile stimulus was generated via a Force Feedback Joystick Driver for Java [16]. The stimuli were delivered to the user’s right palm via the FLIGHT FORCE joystick by Logitech.
Fig. 1

The visual instruction screen presented to the users during the psychophysical experiment developed in MAX 6 [2] is presented in panel (a) above. The force-feedback (or tactile-force) joystick FLIGHT FORCE by Logitech used in experiments reported in this paper is depicted in panel (b) above. The tactile-force stimulus was delivered to the user’s dominant hand. Four pressure stimuli executed by the force—feedback joystick (depicted with red arrows above), defined as North, East, South and West, were executed randomly from a computer causing the joystick’s handle to move and push the user’s hand automatically

There were four stimulus tactile-force direction patterns delivered in random order to the user’s hand. The directions were labeled North, East, West, and South, as depicted in Fig. 1. For example, the North direction stimulus interaction caused the joystick to generate a forward tactile-force pressure on the user’s hand holding it. Similarly the South, East, and West stimulus directions caused, respectively, backward, right, and left tactile-force pressure on the user’s hand. The joystick returned to the center position (no pressure) following each stimulus after a presentation time of 100 ms (see Tables 1 and 2 for a summary of details of the experiment conditions).
Table 1

Tactile-force psychophysical experiment protocol conditions and details

Condition

Detail

Users

7 advanced (6 males and 1 female; mean age 24.71

years with SD 7.5 years) and 5 naive

(3 males and 2 females; mean age 22.4 years with

SD 0.55 years)

Tactile stimulus length

100 ms

Inter-stimulus interval (ISI)

900 ms

Stimulus generation

FLIGHT FORCE joystick by Logitech

Number of trials for each user

20 (including in total 80 targets + 240 non–targets)

Table 2

Conditions and details of the tfBCI EEG experiment

Condition

Detail

 

7 advanced (6 males and 1 female; mean age 24.71

years, SD 7.5 years) and 5 naive users

(3 males and 2 females; mean age 22.4 years,

SD 0.55 years)

Tactile stimulus length

100 ms

Inter-stimulus interval (ISI)

300 ms and 400 ms for advanced and naive users, respectively

EEG recording system

gUSBamp active wet EEG electrodes system

Number of the EEG channels

16

EEG electrode positions

Cz, CPz, P3, P4, C3, C4, CP5, CP6, P1, P2, POz, C1, C2, FC1, FC2, and FCz

Reference electrode

Left earlobe

Ground electrode

On the forehead (FPz electrode position)

Stimulus generation

FLIGHT FORCE joystick by Logitech

Number of averaged trials

10 (one target + three non-targets in each single trial)

During both the psychophysical and EEG experiments, the user held the joystick handle using the dominant hand (right in the case of all the users participating in this study) and responded (button press in the psychophysical experiment and mental confirmation/counting in the EEG experiment) only to the instructed direction. The instructions about which directions to attend to were presented visually using the same MAX 6 environment program that created the stimulus and communicated it via the Java driver to the joystick as depicted in Fig. 1. The experiments were designed in an oddball presentation paradigm [14, 17] in order to generate clear brainwave responses in BCI experiments (to generate so-called “aha” or P300 responses).

2.2 Tactile-force psychophysical experiment protocol

A psychophysical experiment was conducted to investigate the influence of tactile-force stimulus direction on the user’s behavioral response times and accuracy. The behavioral responses were collected using a trigger button on the joystick handle and the MAX 6 program. The user was instructed which stimulus to attend to in each session by an arrow on the computer display pointing in the direction of a target, as depicted in Fig. 1. In each psychophysical experiment, the user was presented with 80 target and 240 non-target directions as stimuli, in order to collect enough responses for subsequent statistical analysis.

Each trial was composed of 100 ms tactile-force pressures delivered to the user’s hand in a randomized order with an inter-stimulus interval (ISI) of 900 ms, thus composing a stimulus-onset asynchrony (SOA) based experiment of 1000 ms. Every random trial thus contained a single target and three non-targets. A single session was composed of the twenty trials for each tactile-force target. The choice of the relatively long ISI was justified by a slow behavioral response in comparison to the EEG evoked potential, as described in the next section. The details of the tactile-force psychophysical experiment protocol are summarized in Table 1.

The behavioral response times were registered with the same MAX 6 program, also used for the stimulus generation and instruction presentation, as depicted in Fig. 1. The goal of the psychophysical experiment was to investigate the behavioral response times and target recognition accuracies in order to test cognitive loads (task difficulties) generated by the four varying tactile-force stimuli. The results of the experiment are discussed in Section 3.1.

2.3 EEG tfBCI experiment protocol

In the BCI experiments, EEG signals were captured with a portable EEG amplifier system g.USBamp by g.tec Medical Instruments GmbH, Austria. Sixteen active wet EEG electrodes were used to capture brain waves with event related potentials (ERP) with attentional modulation elucidated by the so-called “aha” or P300 response [17]. The EEG electrodes were attached to the head locations Cz, CPz, P3, P4, C3, C4, CP5, CP6, P1, P2, POz, C1, C2, FC1, FC2, and FCz, as in the 10/10 intentional system [3]. A reference electrode was attached to the left mastoid and a ground electrode to the forehead at FPz position. No electromagnetic or electromyographic (EMG) interference was observed from the moving joystick. Details of the EEG experiment protocol are summarized in Table 2.

The EEG signals were recorded and preprocessed by a BCI2000-based application [14], using a stepwise linear discriminant analysis (SWLDA) classifier [7] with features drawn from the 0−800 ms ERP interval. The sampling rate was set to 256 Hz, the high-pass filter to 0.1 Hz, and the low-pass filter to 40 Hz. The ISI was 300 ms for advanced and 400 ms for naive users. The longer ISI for naive users was chosen in order to reduce the difficulty of the experiment for the beginners. Each tactile-force stimulus duration was set to 100 ms for the both experiment user groups.

Instructions to the user about which tactile-force stimulus direction to attend to were presented visually, as in the previous psychophysical experiments, using the MAX 6 program (see Fig. 1). Each target was presented 10 times in a random series with the remaining 30 non-targets in a single intended direction classification step. A procedure of ten single ERP responses averaging was used in order to enhance the P300 response in noisy EEG [7, 14], which is a common practice for enhancing the BCI classification accuracy, yet lowering significantly the interface speed.

3 Results

This section presents and discusses results that we obtained in the psychophysical and in the online tfBCI experiments. The very encouraging results obtained in the psychophysical and tfBCI paradigm experiments support the proposed concept of the tactile-force modality.

3.1 Tactile-force psychophysical experiment results

The psychophysical experiment accuracy results are summarized in Table 3, depicted in the form of a confusion matrix in Fig. 2, and as response time distributions plots in Fig. 3, where the median response times and the interquartile ranges are depicted for each direction (see also Fig. 1 for the directions).
Table 3

Advanced and naive user psychophysical experiment accuracy results (note, this is not a case of binary accuracy, but one with a theoretical chance level of 25 %) in the tactile-force interface task. The accuracy results are not significantly different within or between the user groups

Advanced users

North

East

South

West

Average

#1

95.0 %

95.0 %

100.0 %

100.0 %

97.5 %

#2

95.2 %

100.0 %

100.0 %

100.0 %

98.8 %

#3

85.0 %

95.0 %

90.0 %

95.0 %

91.3 %

#4

100.0 %

100.0 %

100.0 %

100.0 %

100.0 %

#5

100.0 %

100.0 %

100.0 %

100.0 %

100.0 %

#6

95.2 %

100.0 %

100.0 %

100.0 %

98.8 %

#7

100.0 %

100.0 %

100.0 %

95.0 %

97.5 %

Overall advanced users

    

97.7 %

Naive users

North

East

South

West

Average

#1

100.0 %

100.0 %

90.0 %

95.0 %

96.3 %

#2

95.0 %

90.0 %

95.0 %

100.0 %

95.0 %

#3

90.0 %

95.0 %

90.0 %

100.0 %

93.8 %

#4

100.0 %

90.0 %

75.0 %

80.0 %

86.3 %

#5

100.0 %

90.0 %

100.0 %

100.0 %

97.5 %

Overall naive users

    

93.8 %

Fig. 2

Tactile-force interface psychophysical experiment results in the form of confusion matrices of the grand mean average user accuracies of the advanced (a) and naive (b) user groups, respectively. The rows in the above matrices denote the instructed targets and the columns denote the user responses. A diagonal of the matrix represents the correct response, while the off-diagonal values indicate user errors. The last column summarizes missed responses in experiments. Numerical percentage values represent response accuracies. In the experiments conducted, the mean errors were marginal (below three percent). There were also no systematic errors observed (common mistakes between pairs of patterns), which further validates the tactile-force stimulus design

Fig. 3

Distribution plots of the tactile-force psychophysical experiment response time distributions of the four joystick directions North, East, South and West by advanced (a) and naive (b) user groups. The differences between the medians were not significant (as tested with a pairwise Wilcoxon rank sum statistical test) with either of the two user groups. Significant differences (p ≪ 0.001) were observed between the two user groups, as tested with the same Wilcoxon method. The red lines depict response time interquartile ranges of the response time distributions. The outer shapes of each of the plots (colored bar widths) depict the response distributions (number of responses within each time bin)

These results confirm the stimulus related cognitive load similarity, since the behavioral responses for all the directions were basically the same within the experiment user groups (as resulted with non-significant median differences from a pairwise Wilcoxon rank sum test). Significant median response differences (p ≪ 0.001 as tested with the same test) were observed between the groups (faster response times in the case of the advanced users), which confirmed better performances by the trained advanced users (accustomed to the experiment paradigm). This finding validated the design of the following tfBCI EEG experiment, since the four tactile-force patterns resulted in similar cognitive loads, within the users groups, as confirmed by the same accuracies in Table 3 and Fig. 2, as well as by response times depicted in Fig. 3.

3.2 Online EEG Tactile-Force BCI Experiment Results

The results of the online tfBCI paradigm EEG experiment conducted with the seven advanced and five naive users are presented in Fig. 4 in the form of matrices depicting ERP latencies with P300 response together with area under the curve (AUC) feature separability analyses. We also present averaged topographic plots of the evoked responses at the highest and lowest ERP separability latencies in the target vs. non-target averaging scenario. The highest average differences were found at 445 ms and 432 ms (as calculated by AUC) for advanced and naive users respectively (non-significant differences), which perfectly represented the P300 response peaks as can also be seen in Fig. 5, where target and non-target responses are visualized separately for each electrode. Figure 5 also presents a very interesting post-P300 attentional modulation in the case of the advanced users, which with the majority of electrodes chosen for our experiments had extended positive ERP modulation beyond the classic P300 peak in a range exceeding the 300−600 ms range up to 1000 ms. Such attentional modulation is an effect of longer training or the experiment paradigm familiarity of the advanced user group.
Fig. 4

Grand mean ERP and AUC scores leading to the final classification results of the participating advanced (a) and naive (b) users. The top panels in both the figures above represent head topographic plots of the target versus non-target AUC scores. (AUC is a measure commonly used in machine learning intraclass discriminative analysis, and AUC > 0.5 usually confirms feature separability.) The top left panel, in each of the above figures, represents a latency of the largest difference as obtained from the data displayed in the bottom panel of the figure. The top right panel depicts the smallest AUC latency. The topographic plots also show the electrode positions. The fact that all the electrodes received similar AUC values (red) supports the initial electrode placement in the tfBCI EEG experiments conducted. The second panel from the top represents averaged EEG responses to the target stimuli (P300 response in the range of 300−600 ms). The third panel from the top represents averaged EEG responses to the non-target stimuli (no P300 response). Finally, the bottom panel depicts the AUC of target versus non-target responses (P300 response latencies can again be easily identified here by the red color-coded values)

Fig. 5

Greater details (compared with results summarized in Fig. 4) of the grand mean ERP’s in the form of time series plots depicting differences with standard errors of the advanced (a) and naive (b) users. The purple lines represent averaged brainwave responses to targets and the blue lines represent responses to the ignored non-targets. The P300 (“aha”) responses are clearly depicted in the range of 300−600 ms. The P300 response modulations are longer in the case of the advanced users depicted in panel (a). The more advanced BCI users are usually able to utilize so-called “cognitive effort” related responses which further extend the available brainwave ERP latencies for the subsequent feature extraction and classification

The online tfBCI accuracies (as obtained with the SWLDA classifier) of the all seven advanced and five naive users are summarized in Table 4. All the users from both the groups participating scored well above the chance level of 25 %. Three out of the seven advanced users reached 100 % accuracy based on the 10 ERP responses averaging, which is a very good outcome for the proposed tfBCI prototype. Based on the accuracies obtained, we calculated the information transfer rate (ITR) scores in order to allow a simple comparison of the proposed tfBCI paradigm with other published approaches. The mean ITR scores obtained were in the range of 0.05 bit/min to 4.72 bit/min for the advanced and of 0.04 bit/min to 1.83 bit/min for the naive users (see Table 4). The maximum results toped 7.50 bit/min and 2.37 bit/min for advanced and naive users, respectively.
Table 4

Ten trials (brain responses) averaging-based classification BCI accuracy (note, this is not a binary P300 classification result but a spelling result with a theoretical chance level of 25 %) in tactile-force task using the classic SWLDA classifier [7] for the advanced (3.75 selections/minute) and naive (3.00 selections/minute) users. (See (1) and (2) for ITR calculation details)

Advanced

BCI accuracy of the session

Mean

Mean

Maximum

users

#1

#2

#3

#4

Accuracy

ITR

ITR

#1

100 %

25 %

75 %

75 %

68.8 %

2.29 bit/min

7.50 bit/min

#2

75 %

75 %

50 %

50 %

62.5 %

1.69 bit/min

2.97 bit/min

#3

100 %

100 %

50 %

100 %

87.5 %

4.72 bit/min

7.50 bit/min

#4

50 %

0 %

25 %

50 %

31.3 %

0.05 bit/min

0.78 bit/min

#5

25 %

50 %

25 %

50 %

37.5 %

0.21 bit/min

0.78 bit/min

#6

100 %

50 %

75 %

25 %

62.5 %

1.69 bit/min

7.50 bit/min

#7

25 %

0 %

50 %

50 %

31.3 %

0.05 bit/min

0.78 bit/min

Average

 

54.5%

1.53 bit/min

3.97 bit/min

Naive

BCI accuracy of the session

Mean

Mean

Maximum

users

#1

#2

#3

#4

Accuracy

ITR

ITR

#1

75 %

0 %

0 %

50 %

31.3 %

0.04 bit/min

2.37 bit/min

#2

75 %

50 %

50 %

75 %

62.5 %

1.35 bit/min

2.37 bit/min

#3

75 %

25 %

25 %

25 %

37.5 %

0.17 bit/min

2.37 bit/min

#4

25 %

0 %

50 %

50 %

31.3 %

0.04 bit/min

0.62 bit/min

#5

75 %

75 %

50 %

75 %

68.8 %

1.83 bit/min

2.37 bit/min

Average

 

46.3 %

0.67 bit/min

2.02 bit/min

The ITR was calculated as follows,
$$ ITR = V \cdot R, $$
(1)
where V is the classification speed in selections/minute (3.75 selections/minute and 3.00 selections/minute for advanced and naive users’ in the experiment setting in the study presented). R stands for the number of bits/selection calculated as,
$$ R=\log_{2}N + P \cdot \log_{2}P + (1 - P) \cdot \log_{2}\left(\frac{1 - P}{N - 1}\right), $$
(2)
where N represents the number of classes (four in this study). P is the classification accuracy (see Table 4). The results are considered to be good outcomes in comparison with state-of-the-art BCI’s [17].
We also calculated offline the BCI accuracies and resulting ITR’s in a single trial (single brain response) processing scenario in order to show the possible strength of the tfBCI. The offline brainwave analysis results are summarized in Table 5. The averaged accuracies were lower compared to averaged trials presented in Table 4, yet the maximum scores obtained by the majority of the users allowed the attaining of 75 bit/min and 60 bit/min by advanced and naive users, respectively. Such high scores demonstrate the possible strength of the proposed tfBCI paradigm.
Table 5

Mean results of single trial classification (single brain response-based classification) BCI accuracy obtained in offline processing of experiment results summarized in Table 4, using the same SWLDA classifier [7] for the advanced (allowing for 37.5 selections/minute) and naive (allowing 30.0 selections/minute) users. (See (1) and (2) for ITR calculation details.) This demonstrates future possibilities for tfBCI, once better signal processing and machine learning are applied in online experiments

Advanced users

Mean single trial based accuracy

Mean ITR

Maximum ITR

#1

31.3 %

2.31 bit/min

7.78 bit/min

#2

37.5 %

2.06 bit/min

7.78 bit/min

#3

48.7 %

7.03 bit/min

29.72 bit/min

#4

50.0 %

7.78 bit/min

75.00 bit/min

#5

12.5 %

0.00 bit/min

0.00 bit/min

#6

50.0 %

7.78 bit/min

29.83 bit/min

#7

37.5 %

2.06 bit/min

29.72 bit/min

Average

38.2%

4.15 bit/min

25.69 bit/min

Naive users

Mean single trial based accuracy

Mean ITR

Maximum ITR

#1

50.0 %

6.23 bit/min

23.77 bit/min

#2

43.8 %

3.61 bit/min

6.23 bit/min

#3

43.8 %

3.61 bit/min

6.23 bit/min

#4

31.3 %

0.44 bit/min

6.23 bit/min

#5

50.0 %

6.23 bit/min

60.00 bit/min

Average

43.8 %

4.02 bit/min

20.49 bit/min

4 Conclusions

Psychophysical and EEG experiments with a novel sensory modality interface have been presented with the results obtained, which confirm the feasibility of the practical application of the novel four-commands tactile-force BCI. The results of experiments with seven advanced and five naive healthy users confirm our hypothesis of the tfBCI application validity.

The EEG experiment with the tfBCI paradigm has confirmed that tactile-force stimuli can be used easily (yet enhanced with prior BCI training, as shown by a comparison between advanced and naive users) and successfully, with ITR scores ranging up to 7.50 bit/min in the online interfacing case using the SWLDA classifier based on the ten trials averaging setting. The offline processing example shows the possible strength of the proposed tfBCI paradigm, with the ITR score reaching 75 bit/min.

The results presented offer a step forward in the development of a novel multimedia tactile modality neurotechnology application. The current paradigm obviously still needs improvement and modifications for implementation online with a shorter ISI and lower averaging rate necessary to improve the EEG features separability. These requirements determine the major lines of study for future research. However, even in its current form, the proposed tfBCI can be regarded as a practical solution for LIS patients (locked into their own bodies despite often intact cognitive functioning), who cannot use vision or auditory based interfaces due to sensory or other disabilities.

We plan to continue this line of tactile-force BCI research in order to further optimize the feature extraction, signal processing and machine learning (classification) methods. Next we will test the paradigm with LIS patients in need of multimedia tactile-based BCI communication neurotechnology applications.

Notes

Acknowledgements

This research was supported in part by the Strategic Information and Communications R&D Promotion Program, no. 121803027, of The Ministry of Internal Affairs and Communications in Japan.

Author Contributions

Programmed the tactile-force stimulus generation and delivery interface: SK, TMR. Performed the EEG experiments: SK. Analyzed the data: SK, TMR. Conceived the concept of the tactile-force BCI: TMR. Wrote the paper: TMR, SK.

References

  1. 1.
    Brouwer AM, Van Erp JBF (2010) A tactile P300 brain-computer interface. Front Neurosci 4 (19). doi:10.3389/fnins.2010.00019. http://www.frontiersin.org/neuroprosthetics/10.3389/fnins.2010.00019/abstract
  2. 2.
    Max 6 (2012). http://cycling74.com/
  3. 3.
    Jurcak V, Tsuzuki D, Dan I (2007) 10/20, 10/10, and 10/5 systems revisited: Their validity as relative head-surface-based positioning systems. NeuroImage 34(4):1600–1611. doi:10.1016/j.neuroimage.2006.09.024. http://www.sciencedirect.com/science/article/pii/S1053811906009724 CrossRefGoogle Scholar
  4. 4.
    Kaufmann T, Holz EM, Kuebler A (2013) Comparison of tactile, auditory and visual modality for brain-computer interface use: A case study with a patient in the locked-in state. Front Neurosci 7:129. doi:10.3389/fnins.2013.00129. http://www.frontiersin.org/neuroprosthetics/10.3389/fnins.2013.00129/abstract CrossRefMATHGoogle Scholar
  5. 5.
    Kono S (2014) Tactile-force brain-computer interface paradigm. Bachelor degree thesis, School of Informatics -University of Tsukuba, TsukubaGoogle Scholar
  6. 6.
    Kono S, Aminaka D, Makino S, Rutkowski TM (2013) EEG signal processing and classification for the novel tactile-force brain-computer interface paradigm. In: International conference on Signal-Image Technology Internet-Based Systems (SITIS) 2013, pp 812–817. doi:10.1109/SITIS.2013.132, arXiv:1310.1593
  7. 7.
    Krusienski DJ, Sellers EW, Cabestaing F, Bayoudh S, McFarland DJ, Vaughan TM, Wolpaw JR (2006) A comparison of classification techniques for the P300 speller. J Neural Eng 3(4):299. http://stacks.iop.org/1741-2552/3/i=4/a=007 CrossRefGoogle Scholar
  8. 8.
    Mori H, Matsumoto Y, Kryssanov V, Cooper E, Ogawa H, Makino S, Struzik ZR, Rutkowski TM (2013) Multi-command tactile brain computer interface: a feasibility study. In: Oakley I, Brewster S (eds) Haptic and Audio Interaction Design 2013 (HAID 2013), Lecture Notes in Computer Science, vol 7989. Springer-Verlag, Berlin Heidelberg, pp 50–59. arXiv:1305.4319 Google Scholar
  9. 9.
    Mori H, Matsumoto Y, Struzik ZR, Mori K, Makino S, Mandic D, Rutkowski TM (2013) Multi-command tactile and auditory brain computer interface based on head position stimulation. In: Proceedings of the Fifth International Brain-Computer Interface Meeting 2013, p. Article ID: 095. Graz University of Technology House, Publishing, Asilomar Conference Center, Pacific Grove. doi:10.3217/978-4-83452-381-5/095, http://castor.tugraz.at/doku/BCIMeeting2013/095.pdf
  10. 10.
    Müller-Putz G, Scherer R, Neuper C, Pfurtscheller G (2006) Steady-state somatosensory evoked potentials: suitable brain signals for brain-computer interfaces?. IEEE Trans Neural Syst Rehab Eng 14(1):30–37. doi:10.1109/TNSRE.2005.863842 CrossRefGoogle Scholar
  11. 11.
    Plum F, Posner JB (1966) The diagnosis of stupor and coma. FA Davis, PhiladelphiaGoogle Scholar
  12. 12.
    Rutkowski TM, Cichocki A, Mandic DP (2009) Spatial auditory paradigms for brain computer/machine interfacing. In: Proceedings of the international workshop on the principles and applications of spatial hearing 2009 (IWPASH 2009). Miyagi-Zao Royal Hotel, Sendai, p 5Google Scholar
  13. 13.
    Rutkowski TM, Mori H (2014) Tactile and bone–conduction auditory brain computer interface for vision and hearing impaired users. J Neurosci Methods. doi:10.1016/j.jneumeth.2014.04.010
  14. 14.
    Schalk G, Mellinger J (2010) A practical guide to brain–computer interfacing with BCI2000. Springer-Verlag, London LimitedCrossRefGoogle Scholar
  15. 15.
    van der Waal M, Severens M, Geuze J, Desain P (2012) Introducing the tactile speller: an ERP-based brain–computer interface for communication. J Neural Eng 9(4):045,002. doi:10.1088/1741-2560/9/4/045002, http://stacks.iop.org/1741-2552/9/i=4/a=045002 CrossRefMATHGoogle Scholar
  16. 16.
    Wischenbart M (2010) ForceFeedback joystick driver for Java. http://boat.lachsfeld.at/ffjoystick4java/
  17. 17.
    Wolpaw J, Wolpaw EW (2012) Brain-computer interfaces: principles and practice. Oxford University PressGoogle Scholar

Copyright information

© Springer Science+Business Media New York 2014

Authors and Affiliations

  1. 1.Life Science Center of TARAUniversity of TsukubaTsukuba-shiJapan
  2. 2.RIKEN Brain Science InstituteWako-shiJapan

Personalised recommendations