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

Thank You very much Joan Duprez . I will definitely go through this website. Thank you for giving me your valuable time. 

I also can recommend Mike Cohen's book to understand the underlying mechanics.

If you would prefer touse something already built, EEGLAB is the de facto standard, see the tutorial (e.g. https://sccn.ucsd.edu/wiki/Chapter_08:_Plotting_ERP_images)

Hi Niraj,

You need to read this article by Pfurtscheller and Lopes da Silva 1999 - everything is explained there:

https://www.ncbi.nlm.nih.gov/pubmed/10576479

It's very simple. If you end up having problems I can send you the matlab code.

Good luck!

Birgit

 Thank You very much, Matthew Moore. I will definitely go through this website. Thank you for giving me your valuable time. 

Thank You very much, Birgit Nierula madam. I will definitely go through this paper. Thank you for giving me your valuable time. 

Hi

Iam working currently on similar work

ERD/ERS calculation using EEGLab.

ERP plot/ERSP/Time frequency plots are possible with EEGlab but PSD/ERD is not there directly.

Can someone suggest me on that.

Yes Subasree Ramakrishnan Madam, you can calculate ERD/ERS patterns using the MATLAB. There are two methods to calculate the patterns one is band power (BP) and another is inter-trial variance (IV). If you want to know about it please refer my conference paper " Classification of Motor Imagery Tasks Using Inter Trial Variance In The Brain Computer Interface ".

Thanks.Will go through.

Dear Niraj Bagh ,

I am new to this field.

May I ask you for the MATLAB codes for calcualting ERD/ERS please? :)

Dear Shiva Sedghi

You can calculate ERD/ERS patterns using the MATLAB. There are two methods to calculate the patterns one is band power (BP) and another is inter-trial variance (IV). If you want to know about it please refer my conference paper " Classification of Motor Imagery Tasks Using Inter Trial Variance In The Brain-Computer Interface "

Hello,

I managed to calculate the mu ERD/ERS ratio with the bandpower function from MATLAB, however, my average values are mostly negative and may lower than -1 (-2.90 and -2.40 for example). Should I have more positive values? Is there a range that my rations should have? I only saw values ranging between 0 and -0.06 in other papers, but not as small as mine.

Dear Diana Eliza Gherman,

Thank you for showing your interest in motor imagery based BCI. Actually, you won't get negative values because you are taking the amplitudes of both mu and beta frequency bands. Once you get the amplitudes, you have to take the square of the amplitude in both the bands. The square of amplitude gives power in both the bands. Similarly, you have to calculate powers in both event-related and rest periods. Once you got the powers, you have to subtract the average of the rest power from event-related power. Next, you have to divide the final result by the average of rest power then only you will get ERD/ERS.

The negative values will appear before the planning and execution of the movements and it is known as ERD. On the other hand, the positive values will appear before the end of the movements and it is called ERS. I hope you will understand this concept. Please let me know if you have any queries ?

Hi Niraj,

Thanks for your answer. I tried different methods, but I am new to MATLAB and EEG analysis and although I read a lot of papers, I am quite confused, because each talk about other methods. Firstly I used the bandpower function in a method and the fft function in another, but I do get very different and negative values in both. I would really appreciate if you could answer these questions:

1. What function should I use to get the amplitude?

2. Is it ok if I only look at the mu band?

3. Also, I took the rest from after the event, because there are 3 seconds before the event of "cross fixation" (I work on a BCI dataset), but it is not specified that the participants were resting. Is this wrong?

I would like to mention that I cannot use EEGLAB due to the structure of my dataset. Thank you in advance.

Can you please tell me which BCI data are you using ??

Sure, this: http://gigadb.org/dataset/100542

No need to use MATLAB function, do the following in the MATLAB

1. After preprocessing take the raw EEG signal (C3 and C4 Channels).

2. Band pass filtered to all EEG trials for left hand and right hand MI movements.

3. Calculate power for all trials for LH and RH MI movements.

4. Take the average of all trials separately for LH and RH MI movements.

5. Use the ERD/ERS equation.

6. Plot the final value with respect to the time you will get ERD/ERS patterns.

Yes, you can use the mu band but you can use beta band also. But the variation of patterns you will observe in the mu band is better than the beta band.

Thank you very much. Regarding step 3, how do I calculate the power without a MATLAB function?

Just square it. For example,

x=rand(1, 1152);

y=(x.^2)

Great, I will try this method. Thanks!

I have one more question. Do you think that I need to apply the bandpass filter if I already applied the surface laplacian? I tried the badpass function, but it seems that is not working for matrixes.

Surface laplacian is a spatial filter whereas band-pass filter is temporal filter.

The surface Laplacian can reduce spatial noise and aid in the identification of sources. In BCI research, these two functions of the surface Laplacian correspond to prediction accuracy and signal orthogonality.

Because you have to calculate ERD/ERS in sensorimotor frequency band ( 8-30 Hz) or mu (8-12 Hz) or beta (12-30Hz). In these frequency bands only you will get ERD/ERS. ERD/ERS are subject-specific in nature means some subjects u will get ERD/ERS in 8-12 Hz but for other cases, you may get in 12-20 Hz but generally, it comes between 8-30 Hz. Therefore you have to implement an IIR bandpass filter (Butterworth) to raw EEG signal whose order is around 8 or 10, which will select 8-30 Hz frequency band from raw EEG signal. Search " butter" function in MATLAB that will help for this.

You can calculate ERD/ERS patterns using the MATLAB. There are two methods to calculate the patterns one is band power (BP) and another is inter-trial variance (IV). If you want to know about it please refer my conference paper " Classification of Motor Imagery Tasks Using Inter Trial Variance In The Brain-Computer Interface "

Hello, thank you very much for your contribution, I have a question regarding a response on the blog: 4. Take the average of all trials separately for LH and RH MI movements. 5. Use the ERD / ERS equation. 6. Plot the final value with respect to the time you will get ERD / ERS patterns.

Did you mention those steps, to calculate the averages, do you run time windows to generate moving power averages? or how you calculate the ERD / ERS over time. I really appreciate your answer.

Dear Gerardo Lara, thank you for showing interest in motor imagery-based BCI using ERD/ERS patterns. Yes, you are right, I have used run time windows to generate moving power averages. For Example, Let the time duration of one EEG trial is 9s. Out of 9s, let rest time is 0-3s and movement duration is 3-6s. Then you have to do the following :

1. Calculate power by considering a time window of 0-9s.

2. Repeat for all trials and calculate the ensemble of all trials, Let say, P_event.

3. Next, calculate power by considering a time window of 0-3s for the rest period.

4. Repeat for all trials for the rest period (0-3s), Let say, P_rest.

5. Calculate the average of the rest power, Let say, P_rest_avg.

6. Now use the equation:

ERD/ERS=(P_event)-(P_rest_avg)/(P_rest_avg)*100%

7. Plot the ERD/ERS patterns with respect to time, you will see that the pattern is constant up to 0-3s, and starts decline while the subject perform any movement, and its called ERD. The pattern will start to increase while the subject stops perform the movement and it is referred to as ERS. The ERD/ERS patterns play a vital role in designing an MI-based BCI system.

I hope this concept will helps you. If you have any queries or if you will face any problem in the calculation of ERD/ERS patterns please free to contact me.

Hello Niraj,

Thank you for your suggestions. I am new to MI BCI. I have one doubt. I have a data structure like this.....201x118x168....where 201 is epoch per trial, 118 is the no. of channels, and 168 is the total no. of training trials. The two classes are right hand and feet. The corresponding channels are C3 and Cz right? So if I consider only these two (C3 and Cz) channels, then what is the use of other channels? How can I calculate %ERD/ERS and plot it for the above data? Thank you.