how to add events to EEG data acquired through EMOTIV EEG.?

[brianbaloch]

Dear all, please help me. I am using EMOTIV EEG to determine person's cognitive state. Either person is in stress or happy etc. After acquiring EEG signal through EMOTIV, How I add events(markers) of stress, not stress . happy & not happy. I am using EEGLAB and BCILAB. Please help me how I add events/markers to continuous EEG signal.

[brianbaloch]

Please guyz help me. I have continuous EEG data which is obtained through Emotiv epoch. how do I add events to it?. Actually data is continuous for monitoring stress level of user of any application. So it would be regression problem. Should I extract epochs ? and convert it to classification problem. If so please tell me how do I add events to it.?

[DMRoberts]

You can relabel events with basic EEGLAB tools, go through the EEGLAB tutorial to learn about it. Through the same tutorial you can learn how to add new event markers to data, but for timing precision these are usually sent during data collection.

[cboulay]

@brianbaloch , the way you asked the question, it is unclear if you already know what the user's state is and you simply want to add that known information to the data, or if you don't know the state and you want help decoding the state from the EEG. The former is easy and @DMRoberts provides some tips in his answer. The latter is a very difficult problem and probably requires a PhD or two to figure it out.

[brianbaloch]

Dear @DMRoberts please attach the link of EEGLAB tutorial for adding events.so that I follow it. Would it be labelling the events manually or automatically.?

[brianbaloch]

@cboulay . Dear I am asking the question that I have to monitor continuously stress level of a user who is using any device. I guess it is regression problem. But I need your valuable comments on it. My second query is that How do I add events to Continuous EEG data using EEGLAB? manually or Automatically.?

[brianbaloch]

@cboulay , Dear , I don't know the state of user. I want to decode state from the EEG.

[cboulay]

For the EEGLAB Event tutorial, see here.

For decoding stress level from EEG, you're in over your head. This is not something you can do reliably without a couple experts in the field helping you, and even then I'd be dubious. Your question is sort of like, "I have a lawnmower and a smartphone. How do I build a self-driving car?"

Just to give you a hint of what's required:

1. Read the academic literature to try to find robust EEG correlates of stress
   • I just did a quick search and didn't find anything obvious, but it might exist.
2. Design a task that can reliably elicit a stress response.
3. Write a program to present this task to the participant and make sure it inserts markers into your recorded data stream whenever the task should evoke no-stress and different markers for when it should evoke stress.
4. Collect EEG and behavioural data from a participant performing this task.
5. Preprocess your data to remove artifacts and other noise signals.
6. Separate your data into no-stress and yes-stress segments based on the markers in the data.
7. Extract from your segmented and cleaned data different features including those you identified in step 1.
8. Use machine-learning to build a classifier that discriminates between no-stress and yes-stress. You can also try to build a regression model if you think your task can reliably evoke different stress levels. Your markers will have to indicate the stress level. You'll need more trials for this than for a binary classifier.
9. Repeat this process for a bunch of subjects (8-12) and see if the features the classifier selects for discriminating stress are consistent across participants.
   • If yes, congratulations you should probably publish this, but...
     • Probably all you are discriminating is something in the task itself, not stress. Go back to step 2 and design a better task.

In case you can't tell, I don't think you're going to get this to work reliably. If all you want to do is measure stress, regardless of modality, then you should probably look into galvanic skin response and blood tests. If for some reason you need to measure stress with EEG and do it in a way that doesn't require (at least) 10 minutes of re-calibration every time you put the device on a person (even the same person), then I think you should be prepared to lose a few years of your life and possibly never achieve it. If you go that route then you should at least go to grad school so you can get some credit for your work.

I won't offer you any more advice than this.

[brianbaloch]

Dear I am PhD student. I have EEG device. As this platform is related to EEG. So I was seeking help. Thanks alot for your timely and valuable help.