Brain Computer Interfaces - Science topic

I do such work in my consulting company, and have been using BCIs for years..

Keep in mind that a “true“ BCI, in which a small group of neurons (or even a single neuron) can be used as an actuator (sy to activate an external swirch or drive a display, etc.) requires implanted microelectronics and is therefore medically invasive and requires the involvement of a neurosurgeon. Moreover, the microelectodes tend to kill the neurons close to them (via immune system or cytokine consequence). So avoid such approaches unless you are willing to get FDA clearance, pay the neurosurgeon $10,000+ per day, etc.

,The best alternative approaches use either scalp-derived EEG (raw waves - unreliable) or quantitative EG (qEEG), which is probably the best of all worlds. QEEG metrics are quite sensitive to consequences, it is Non-invasiv as it reads from the scalp, and users can be trained to operate a device or move a robot are by learning how to make subtle changes in brain state. a

My company supplies (American Brain Forensics) has designed such such interfaces on a per custom basis for a number of different applications, including for military/government research, or for more mundane applications.

if interested in finding out more, contact me at walambos@mac.com, or call 813.235.2470 and leave a message, I’m a computational neuroscientist (and licensed neuropsychologist) with a consulting Barmecidal to my business and would me happy to discuss it!

best,

William A. Lambos, Ph.D.

‘Neuroscientist, Data Scientist

‘CEO, Computational Neuroscience, Imc.

(A D/B/A of American Brain Forensics)

If you accept my candidacy professor Junhua Li, I'm your guy! Or if you prefer, let my institution fund a scientific visit to your prestigious laboratory.

Regards,

Issam

Based on the information you provided, it seems that you have obtained electrode information for each electrode in the UNM dataset for Parkinson's disease. However, the information provided does not seem to include the spatial coordinates of the electrodes.

The other numbers in the electrode information may represent various electrode properties or measurements, but without additional information or context, it is difficult to say for certain what they represent. For example, the first number in the third set of brackets may represent the electrode impedance or resistance, while the numbers in the other brackets may represent other electrical properties or measurements.

To obtain the spatial coordinates of the electrodes, you may need to consult additional documentation or metadata associated with the UNM dataset. This metadata may include information such as the electrode locations or the electrode montage used to acquire the data.

Alternatively, you may be able to infer the electrode locations from the electrode labels themselves. The electrode labels typically follow a standard naming convention, such as the 10-20 system, which specifies the approximate location of each electrode on the scalp. By using this information, you may be able to estimate the spatial coordinates of the electrodes.

Once you have obtained the spatial coordinates of the electrodes, you can use them to generate an MNE object in Python. The MNE Python library provides several functions for reading and importing electrode coordinates, such as mne.channels.read_custom_montage() and mne.channels.make_standard_montage(). By using these functions, you can create an MNE object that includes the electrode locations and can be used for further analysis.

Those are more analytic than BCI problems. More and more research in BCIs (and classification of EEG i general) is focused on getting models to work with raw, unprocessed data. In my opinion, way too little attention is paid to model evaluation and to final intended purpose. I think you should pay attention to when overfitting (and evaluation with n-fold crossvalidation) is beneficial (when creating a speller for a disabled person, for example meant to be used by that specific person and noone else) or detrimental (when a model is meant to be used by anyone in a plug-and-play fasion).

Abir Hosen Ashik Here are a few prospective Brain Computer Interface (BCI) study topics that a Master's student in Computer Science and Engineering (CSE) can explore for their research project in 2023, with an emphasis on machine learning and deep learning-based custom models:

1. Development of a low-cost BCI device for recording brain signals and classifying brain activity utilizing electroencephalography (EEG) and machine learning methods to increase the accuracy of BCI systems.

2. Deep learning algorithms for deciphering brain signals in BCI and their use to operate robotic arms or wheelchairs are being investigated.

3. The use of Generative Adversarial Networks (GANs) in BCI for building more robust and accurate models of brain signals, particularly for use in controlling prosthetic limbs, is being investigated.

4. The application of BCI in non-invasive brain stimulation for the treatment of neurological illnesses such as depression, Parkinson's disease, and chronic pain is being researched.

5. Investigate the use of BCI to improve the performance of virtual and augmented reality systems.

6. Deep learning techniques are being used to develop a BCI system for speech production for those with communication difficulties.

7. The application of BCI in monitoring and managing the brain activity of persons suffering from chronic insomnia and other sleep disorders is being researched.

Please keep in mind that the area of BCI is continually expanding, with new research subjects and applications being produced on a regular basis. It is critical to keep current and study recent articles in the subject to have a comprehensive picture of the current situation.

Hi,

There are several ways to do this.

1. EEGLAB with MATLAB

2. LabVIEW

Use Instrument I/O > Instr Asst > NI DAQmx

Best regards

Elias

Besides on the comments that are going to be posted, I would consider that feedback and response times are important metrics in the implementation of an effective BCI. You do not want to have something that reacts almost real-time with the use of a system that can only provide slow action, remember that the brain activity in EEG is measuring the activation of a certain group of neurons, located at a specific location based on specific stimuli and it takes an important amount of time to achieve that with untrained subjects.

In such sense, the feedback method has to be selected carefully based on the previous literature reviews. The response has to be corrected from any errors based also in the measuring methods. EEG is most common for low-cost research but I highly doubt you will end up doing MEG, Spikes or ECoG without the participation of a big laboratory.

Hope this information could be useful...

J. Rafiee, thanks a lot for the information. I will give it another try and if it is necessary, I will describe my technical issues to those web communities.

You may be attracted by the idea of development a simple Brain-Computer Interface application for controlling a real-life device, even a smartphone, or a simulation involving a mechatronic system, such as a multifunctional robot aimed for helping people with disabilities.

Eric Andrews Thank you for your informative response. Initially, we are attempting to reproduce an emotion classifier, moreso replicating existing models. Eventually, we intend to use the emotion classifier to generalize to new participants with a generative music algorithm.

Hence, the emotion classifier may not need to be trained on new participants if the bootstrapping approach you mentioned works. However, eventually, we intend to recruit participants that can feed through the model, and then be sent to a generative algorithm. A good summary of this attached (Ehrlich, 2019):

The only difference would be our design does not use an LDA model for emotion classification, as I believe Ehrlich's design does. We also need access to DEAP pre-processed data. We only have it within one of our classifier datasets. Do you have any idea how the preprocessed MATLAB data can be accessed?

Hi I have been playing with BCI for music (using a Neurosky Mindwave headset until we can get the Emotiv one) and used BrainwaveOSC (it is open source on GitHub) to turn the EEG data into OSC. I found a MaxMSP patch to turn the OSC into MIDI, but not yet got it working. Would love to hear more about your work!

The first thing that comes to mind then would be playing with class weights. Simply assign a greater weight to the minority class so that the penalty of a misclassification is comparable to that of the majority one.

If you are using Keras it should be straightforward: store the class-weight mapping in a dictionary and feed to the model when calling .fit() with the class_weight argument.

In PyTorch you can do the same by passing the weights to the loss function via the weight argument.

If you haven't tried, give it a shot and good luck!

Hi Victor!

I think the number of patients depends on what you want to demonstrate. If you are looking for a preliminary study, with few patients you can sense the magnitude of the effect of your intervention. After this preliminary study, you can calculate better the number of patients you need to robustly demonstrate an effect of neurofeedback.

Moreover, you must keep in mind the statistical analysis is different if you have few or more patients.

Best regards,

Jorge de Francisco

yes you can use reinforcement and other techniques in the extraction process and even you can doing a hybrid classification in the extraction feature process

Lutsenko E. V. Diagnostics and forecasting of professional and creative abilities by the method of ACK-analysis of electroencephalograms in the "Eidos" system / E. V. Lutsenko, A. N. Lebedev / / Polythematic network electronic scientific journal of the Kuban State Agrarian University (Scientific Journal of KubGAU) [Electronic resource]. - Krasnodar: KubGAU, 2003. – №01(001). P. 59-61 – - IDA [article ID]: 0010301009. - Access mode: http://ej.kubagro.ru/2003/01/pdf/09.pdf, 0.188 cu. p. l.

EEG forecast the success of psychomotor test while reducing the level of wakefulness: an analysis of the results of the study / T. N. Shchukin, V. B. Dorokhov, A. N. Lebedev, E. V. Lutsenko // Polythematic network electronic scientific journal of Kuban state agrarian University (Scientific journal of Kubsau) [Electronic resource]. - Krasnodar: KubGAU, 2004. – №04(006). P. 290-306. - IDA [article ID]: 0060404022 – - Access mode: http://ej.kubagro.ru/2004/04/pdf/22.pdf, 1,062 cu. p. l.

EEG forecast the success of psychomotor test while reducing the level of wakefulness: a description of the experiment / T. N. Shchukin, V. B. Dorokhov, A. N. Lebedev, E. V. Lutsenko // Polythematic network electronic scientific journal of Kuban state agrarian University (Scientific journal of Kubsau) [Electronic resource]. - Krasnodar: KubGAU, 2004. – №04(006). P. 277-289. - IDA [article ID]: 0060404021 – - Access mode: http://ej.kubagro.ru/2004/04/pdf/21.pdf, 0.812 cu. p. l.

EEG forecast the success of psychomotor test while reducing the level of wakefulness: statement of the problem / T. N. Shchukin, V. B. Dorokhov, A. N. Lebedev, E. V. Lutsenko // Polythematic network electronic scientific journal of Kuban state agrarian University (Scientific journal of Kubsau) [Electronic resource]. - Krasnodar: KubGAU, 2004. – №04(006). P. 268-276. - IDA [article ID]: 0060404020 – - Access mode: http://ej.kubagro.ru/2004/04/pdf/20.pdf, 0.562 cu. p. l.

Hi Ivana,

This paper is an excellent starting point http://www.columbia.edu/cu/biology/pdf-files/faculty/Yuste/yuste%20et%20al.nature2017.pdf.

Additionally, this year there has been a lot of work from IEEE trying to push for a standard which deals with many ethical aspects of the technology: https://standards.ieee.org/content/dam/ieee-standards/standards/web/documents/presentations/ieee-neurotech-for-bmi-standards-roadmap.pdf.

You should check the work of Ricardo Chavarriaga which has produced a lot of awesome research around these topics.

Hope that helps.

Rodrigo

Tricky question. Depends on your Epoc+ Version and dongle version. Now Emotiv requires the subscription. So, headset version =<1.1 still works with most of open-source workarounds.

Thank you sir. Shivabalan Kathavarayan Ramu .

Following

Following

Try:

or

It may seem weird but this is what we did. If you have access to a P300 Speller (BCI2000 or OpenVibe) configure the standard 6x6 matrix to use 250 ms ISI. This will give you four stimulations per second. On OpenVibe you can use the standard p300 sample that comes with the installation bundle. You just put somebody to wear you cap or headset and instruct them to pay attention to all the letters. You get the signals from O1 and O2. If you extract all the epochs and average all of them you must see the four bumps. Now what you can do is to go step by step and reducing that value, lower and lower. You can repeat this all over until you have two good differentiable frequencies by counting the number of bumps on the averaged segments. 60 ms and 80 ms work good. You can bound gradually if you LCD screen refreshing works. Keep in mind that if you run this on Windows, it is not a RTOS, so you can have sometimes weird behavior with the timing. You need a faster computer btw.

Second option: pick an Arduino UNO and LDR light sensor with a couple of resistances. You can write a very easy code to get the light values continuously and transmit them to a serial connected computer. You put the sensor very close to the monitor, and program some UI application to generate the flashing at whatever frequency you want. If everything is OK you will see exact same frequency on the signal obtained from the Arduino and check if the LCD refreshing is OK (https://maker.pro/arduino/tutorial/how-to-use-an-ldr-sensor-with-arduino)

Finally, the attached paper proposed a very detailed procedure to verify exactly all the frequencies from the stimulation and to check if the synchronization is OK.

Hope it helps and good luck !

Will you consider to pilot test the effects of stimulating the C3/C4 vs. the SMA using a repeated-measures design by randomizing the order of test stimulations? Test stimulations may be scheduled to be two or three days apart to minimize the learning effect.

Of course, what constitutes intrinsic or extraneous cognitive load depends on what needs to be learned. For example, if the goal of learning is to comprehend concepts incorporated in some text, using jargon may constitute an extraneous cognitive load.

Sorry I have no medical, computing or scientific research expertise. My work is in practical learning and teaching: see ‘Inside Teaching: How to Make a Difference for Every Learner and Teacher’ (Routledge, 2017).

Hi Sai,

The amplitude of SSVEP signals can be modulated by subjects' attention. So as the subject's attention fluctuates during the experiment, you will see SSVEP amplitude change.

Though I believe that the single trial variability mostly come from, except measurement noise and various artifacts, changes in background activity which can be influenced by your alertness, fatigue, mind wandering etc.

I have a similar question. Can you use an off the shelf EEG cap with your existing Nihon Kohden EEG equipment? Thanks

GINis a another great repository - this one in particular has massively parallel data recorded in macaque during motor tasks.

https://doid.gin.g-node.org/f83565d148510fede8a277f660e1a419/

The history of brain–computer interfaces (BCIs) starts with Hans Berger's discovery of the electrical activity of the human brain and the development of electroencephalography (EEG). In 1924 Berger was the first to record human brain activity by means of EEG.

SRM Institute of Science and Technology having Brain Computer Interface facility in India here is its http://www.srmuniv.ac.in/content/brain-computer-interface-lab link. Few other Institutes are there which you can visit for same purpose like wise National Brain Research Center (NBRC), Center for Neurosciences IISC Banglore and National Institute of Mental Health and Neurosciences Banglore.

Not yet, but we can right now define normal from abnormal brain response

We can also define ill brain tissue responsible of excessive discharge for example for epilepsy and other neurological disease

Please refer to this open access article and it's wide references. It may help you a bit.

HI Eduard

my personal opinion is that the development of closed-loop responsive/adaptive brain stimulation based on intracranial electrophysiological recordings will be the most exciting and most clinically useful implementation of BCI in the near future.

Using implanted electrodes a neurostimulation device can capture pathological brain activity (e.g. seizure activity/pathological oscillations in Parkinson's disease/Tourette Syndrome or similar) and adjust its stimulation to the therapeutic demand in real-time. The conducted studies are preliminary but the field is expanding rapidly right now.

Have a look here:

Best

Julian

I believe so, I have to use their software to see all the outputs from the EEG. If you buy it monthly its only $99/month

emotion classification based on brain signal all the best

In fMRI, the signal is modelled with double gamma functions, not waves, which is far more complex. To even increase complexity, the BOLD signal is lagged, and to make things worse, it depends on the brain region and also depends on the individual. Therefore, the signal is spatially blurred and temporarily too. So, you may expect that brain function is much more complex than satellites and TV signal (which is artificially produced). The BOLD signal also encompasses a lot of noise and the non-digitalised signal offers more paths to un-noise the signal (e.g. using smoothing, or trending; what would you get if you smooth a 0s and 1s field map? too contrasted, maybe).

When we do machine learning with neural data, the most difficult part is this one: how to convert the signal into useful features? And most of the proposed methods fail here. For example, using brain data to decode categories of emotions. They just don’t work, even the commercially available ones, not because of the mathematical part, but because of the data preprocessing. Therefore, all contributions are welcome… but I am not really getting why digitalizing the signal would be helpful to understand/model brain functioning. May you elaborate on this? You define a threshold and convert the data into 0s and 1s and…?

Here you may find one strategy that I used in the past to convert the signal into features:

In this case, I averaged the second and third value after stimulus onset (but not digitalized it). I am trying now different strategies that include time stamping in order to model the hemodynamic response (because fMRI’s temporal resolution is not optimal, this is a hard hurdle, because data is scattered).

BTW, the BOLD signal is highly nonlinear! Cheers!

This question is old but it is still unanswered, so here I go...

The Toeplitz matrix is just the beginning of the algorithm, and these are just matrices where all the values from the same diagonal have the same constant value. In this case, as you will have a 1 in the first column of the matrix, in the k-th row, every other column will have to have a one on that only diagonal, zero elsewhere. Think of it as the identity matrix "pushed" downwards.

They used it to represent the time-locked delay of the ERP, i.e. how many sample points do you need to wait since the onset of the stimulus until you actually start seeing the ERP. Because you multiply this matrix with A, which is the ERP signal, the result is the A matrix, representing the ERP, pushed also downwards and delayed k units.

I like the idea of "evoked potential algebra" that the authors use in the paper ( )

Page 138 of Lotte's Brain Computer Interfaces 1 book offers a brief explanation of the algorithm that perhaps may help you as well (optimization).

You can find an implementation in OpenVibe's source code and also in the MNE-Python package.

The project looks interesting, may we know details of the project such as the goals or the desired outcomes? Can you also expand on why use brain computer interface for speech emotion recognition? 

Thank You Dibakar Pal, but Link is not working.

understood.

thanks again.

Hi Niraj,

I am also completely new to the field and I can't help you directly  but you should take a look at Mike Cohen's website (http://mikexcohen.com/lectures.html). It has very good lectures on time-frequency based analyses and all the computations are done in Matlab (with downloadable scripts).

Hope it helps

It depends on the tasks.

Hi Yubo,

in addition to the suggestions of Samer and Sam, can it happen that you focuse your search in the occipital channels? (having in mind that visual info in processed in that lobe)...

ICA algorithms typically take sample*channel matrices. So epoching itself will not have any effect on the results of the algorithm because the data is re-concatenated in most packages/programs (e.g. EEGLAB)  before submission to the algorithm. 

Removing data samples in the form of entire epochs will have an effect on the results of ICA algorithms. Especially if those epochs contain large, unique movement artifacts. ICA algorithms will typically place unique movement artifacts in single components.

It would probably be best to submit all epochs to the algorithm unless there are some egregious movement artifacts in the recording. No EEG will be able to be recovered during periods of large movement artifact.

Was this helpful? I've attached a book chapter for reference. Section 6 contains the relevant discussion. 

Brain Computer Interfaces typically acquires signals generated from the brain (Neurons) and then extracting some information from those signals. Those are low frequency signals around 0-80Hz. In BCIs no signals are applied onto the brain for getting information.

But there are some other technologies/works that apply signals on brain for different tasks. e.g. "Deep Brain Stimulation" 

Thank you Benjamin and Vladimir for your answers! I am astonished that the considerable progress in neural implants is possible with extracellular recording... I had a similar answer in the facebook neuroscience group, although with slightly different measures (40 - 50 micrometers for large neurocyte cell bodies). And I had information on the spacing of the Utah-array to be about 400 micrometers. From this I concluded that a single electrode might be about 50 micrometers or comparable to a neurocyte cell body and produced the image attached to this post. (Precise measures are not that much important here... And, yes, I am producing these images myself.) I am planning to produce some more specific images displaying neurocytes in the retina and cochlea (and a very similar tip) and composing this to a little news feature on my website (http://news.scivit.com). This should be available in a few weeks. Many thanks again and comments helping improve my visualizations are always welcome,

Thomas

Thank you Dirk for your reply, I know there are many types of brain tumors but my goal is to know if there is a distribution  pattern of  blood perfusion  that characterizes some tumors, I not study a specific brain tumor.

The reason could be the that the covariance matrix does not have full rank anymore, e.g., its inverse turns out complex. And a reason for that could be common average electrode spatial filtering. Did you check for this?

Hi Edward,

I have some experience with the U-probe. It's really durable and perfect for repeatedly penetration. I'd say I love this probe if it's not that expensive. One thing I should point out is, due to the low impedance, the signal recorded from U-probe is not as good as regular single electrode (still all right for isolation). 

I have no idea about the V-probe. Presumably they should be similar except the shank. Hope someone else could share their experiences.

-Ji Dai 

Perhaps there is one good explanation for a focus on midline electrodes: a frequent enrollment of homologous cortices, making for a bilateral signal whose combined field is maximal medially. But other than that, it is generally a mistake to focus only on some electrodes, midline or otherwise. This is true both under the aim of interpreting one signal (its spatial organization is important: even as one can get good measurement of some properties at peak electrodes, the investigation should not stop there), and under the aim of interpreting several simultaneous signals. If your question was tongue-in-cheek, then I share your stupefaction that so many investigators would go through the trouble of preparing, collecting and analyzing 64, 128 or more electrodes, only to redact a report that shows very few of them. A lot of EEG misconceptions are hidden in this limited release of spatial information, such as inappropriately identified rhythmic activity, e.g. an alpha rhythm interpreted as mu just because it has residual power at electrode CZ.

Also take a look at the Proceedings of ACM ASSETS Conference.

Thanks!

I do not know how this question found its way to me.My expertise is topics on psychology and neuroscience.  This program obviously has more work required. 

Sorry. Bob Webb

Hello Kaspar Schindler

please check pdfs

Thanks, Vsevolod! But the question is: how does our brain behave when faced with unforeseen situations, those not determined by innate motor skills? For a long time, little attention was paid to this problem. It was for the most part studied with regard to theories of the executive functions: the functions that allow an individual to design, plan objectives, carry out projects aimed at one purpose, monitor (and if necessary anticipate) his/her own actions to adapt to environmental changes. Numerous sub-skills are included in this sphere which are coordinated between themselves: for example, the inhibition of a response at an inappropriate moment (or its deferral to a more appropriate moment); the implementation of a plan of sequences of action useful for a purpose; the representation of the task that includes relevant information (memorised or perceived immediately) associated with the desired result. The associative cortex of the frontal lobe supervises these functions at three different levels of operation: a) the dorsomedial, assigned with functions of working memory necessary for the selection and maintenance in the memory of the aims of the behaviour; b) the mesial, assigned with the integration of the emotional and motivational aspects necessary for the continuation of the action and c) the orbital, assigned chiefly with the inhibition of behaviour and instinctual impulses. For the effective completion of the movement they collaborate with the frontal lobe, the cerebellum, the basal ganglia and the spinal cord for an effective execution of the movement.
 In this representation, the space for new sequences of action is reserved for lower levels of the motor control, where the procedures for the correction of the errors proceed from a superior to an inferior order and the individual movements are combined in routine and subroutine sequences that lead to the goal directed action. This hierarchical model renders the motor functions subordinate to the higher brain activities and drastically limits the understanding of new sequences of action. 

Catching a flying object, hitting a moving target, pronouncing words with great fluidity or playing a piano, call for such fast times that exclude the possibility of corrections by means of sensory feedback: an action may have success or failure, but due to the time constraints it cannot be correct. In what way, therefore, can the brain achieve such rapid muscle action sequences, to the point of overcoming the temporal limits of the neural circuits?

Hi Hessel,

The most important part is that you want independent recording to avoid spurious synchrony.

Moreover I do not recommend to have asymmetrical setup. While the GND location on the body is not of great interest (but the convention is earlobe/forehead), just place the GND electrode at the same location for both subject.

Louis

Hi Mr. Barthel,

Thank you for your answer. Unfortunately the project is already finished. I quantified the volumes manually or using 3D object counter per slice. But I will try your suggestion later.

Regards,

Alexander

Dear Rayyan! Please try to evaluate some papers in the attachment ro my letter. I hope you will find your own way...

Vladimir

I think the mean amplitude would be the jest one (if I cirrectly underdtand what you mean in your question).

Pankaj,

The 2011 review paper from Neuropsychologia is probably the most complete review of the ERP and prospective memory literature. There are some newer studies that have appeared in the last couple of years; however, these have not really introduced new component but rather refined our understanding of components identified in previous research. I have attached a PDF of the Neuropsychologia paper, I hope that you find it useful in your research. Cheers, Rob

Together with Urban Alehagen, we used a combination of selenium and Q10, but perhaps only the selenium component was responsible for the preventive effect?

 Alehagen, U., & Aaseth, J. (2015). Selenium and coenzyme Q10 interrelationship in cardiovascular diseases–A clinician's point of view. Journal of Trace Elements in Medicine and Biology, 31, 157-162.

Here's a recent paper evaluating OpenBCI performance relative to other commercial grade equipment. By Jeremy Frey, presented at the BCI Society 2016 meeting this week in Pacific Grove, CA (Asilomar Conference Center).

Dear Akshay Raj Gollahalli 

Goals of the organizers

The goal of the "BCI Competition III" is to validate signal processing and classification methods for Brain-Computer Interfaces (BCIs). Compared to the past BCI Competitions, new challanging problems are addressed that are highly relevant for practical BCI systems, such as

session-to-session transfer (data set I)

small training sets, maybe to be solved by subject-to-subject transfer (data set IVa),

non-stationarity problems (data set IIIb, data set IVc),

multi-class problems (data set IIIa, data set V, data set II,),

classification of continuous EEG without trial structure (data set IVb, data set V).

Also this BCI Competition includes for the first time ECoG data (data set I) and one data set for which preprocessed features are provided (data set V) for competitors that like to focus on the classification task rather than to dive into the depth of EEG analysis.

The organizers are aware of the fact that by such a competition it is impossible to validate BCI systems as a whole. But nevertheless we envision interesting contributions to ultimately improve the full BCI.

Goals for the participants

For each data set specific goals are given in the respective description. Technically speaking, each data set consists of single-trials of spontaneous brain activity, one part labeled (training data) and another part unlabeled (test data), and a performance measure. The goal is to infer labels (or their probabilities) for the test set from training data that maximize the performance measure for the true (but to the competitors unknown) test labels. Results will be announced at the Third International BCI Meeting in Rensselaerville, June 14-19, and on this web site. For each data set, the competition winner gets a chance to publish the algorithm in an article devoted to the competition that will appear in IEEE Transactions on Neural Systems and Rehabilitation Engineering.

Hi Yves,

Maybe this paper where ANN is compared among other models (MARS, PPR, SVR, RF) is useful to you.

Cheers,

Manuel

Unfortunately I can not tell you much about the quality of the data sent, which I didn't test (if this is what you mean), but there are more info about the packages in the documentations on the developer website. 

I am working on BCI based SSVEP speller. I am acquiring SSVEP data with the help of flickering block (with a particular frequency) on the computer screen. CCA is the most common method to detect the frequency of the signal.  What can be the other possible classification methods or feature extraction transforms I can apply to detect the accurate frequency of the obtained signal?  and What are the features which can be used for the classification of biomedical  signals?

Dear Atef!  suggest you get acquainted with the detailed work in this direction. Please read the first one informative article in the application.

Niftireg is also very good for the kind of registration you want to do.

Hi Yves,

you might want to check out the (open-source) implementation in the FieldTrip Matlab toolbox:

In FieldTrip, Hjorth's method is implemented within the function ft_scalpcurrentdensity, see: http://www.fieldtriptoolbox.org/reference/ft_scalpcurrentdensity

Yves, this paper might be useful to you:

Eliasmith et al (2012) "A Large-Scale Model of the Functioning Brain"

If your BCI system is used to control an external device (e.g., a robot), accuracy may be more important. If your BCI system is used for communication, such as a P300-based  BCI speller, ITR may be the better criterion.

I agree that it may be frustrating when you don't get an answer to your problem, but you have to realize that this is a free and open-source software, meaning people are not making money out of it. So people who answer on the forum or write the documentation have to have the knowledge, motivation and time to do so, as they are doing it for free, purely out of good will...

There are a variety of ICA algorithms -- some more computationally demanding than others, which can be found here:

Hope this helps,

-Michael