I am new to EEG Signal processing. I downloaded DEAP dataset (preprocessed) in .mat format. Once we load them to workspace (a metrix with channel numbers and data), how can we proceed up to a feature space? I read several papers but methodology is bit confusing. specially I have following sub questions.
1. Are we framing the stream? if so what is the time duration for a chunk?
2. How we load features of a single frame to matlab?
I found EEG lab but I would like to start this from scratch to understand the process.
http://www.eecs.qmul.ac.uk/mmv/datasets/deap/
Can you post a small part of the data so that we can answer you accurately?
Thank you George.
Size of data is 40 x 40 x 8064 (40 trials x 40 channels x data).
A part of data
>> y = data(1:5,1:5,1)
y =
0.9482 0.1247 -2.2165 1.0057 5.0959
10.2602 9.4919 7.1287 10.4775 4.0194
1.0130 0.0332 0.7228 9.7638 8.4035
-7.6584 -9.4793 -5.0810 -5.7302 -7.7049
-1.8111 -2.0578 -4.0229 -3.1591 -5.5728
the data is adequately documented at the source:
http://www.eecs.qmul.ac.uk/mmv/datasets/deap/readme.html
ie, 60 seconds post-stim and 3 seconds pre-stim, so 63 seconds total, downsampled to 128Hz. what about this is unclear? (I don't know what you mean by "features" - this data is just microvolts from scalp electrodes...)
Hi
It 's dependent on which type of features you need.
e.g. PSD, HOC, Coherence, etc
Oh Sorry I got a bit confused with the question. If that is the case then I suggest to extract a single feature try.
a = data(1:no. of trials,1:no. of channels,data)
I have uploaded the signed End User License Agreement. But have not yet received the dataset from http://www.eecs.qmul.ac.uk/mmv/datasets/deap/ this. I need the dataset immediately. Please help.
For me it seems that it could be useful for you to learn more about the EEG itself "to understand it from the scratch". What is measured by EEG and what might it tell us about the underlying brain function. There is no need for more unguided extraction of "features" from EEG data that is not driven by hypothesis. Computational power is increasing rapidly and "feature extraction" methods are being used all over, most of the time noone is really knowing what is happening to the data. Without a clear goal many works on EEG time series are not being interpreted any more. This is not a gain for EEG-research but will lead to more confusion.
For a first glance file:///C:/Users/Thalamus/Downloads/9783540879183-c1.pdf
and then start going through the literature until you get to sophisticated EEG analyzes using EEGLab et al...
Hi Senaka,
what is the question that you want to answer? If you explain your aim it would be easier to suggest literature and methods that are suitable for you.
Best,
Julian
Thank you all for comments.
Hi, Neumann,
Simply what I want is to extract features from available EEG data and detect the emotion category of the subject. I am planning to compare existing classification techniques and identify the best technique to classify emotions.
I have experience with voice-based features. But not with EEG. Can you suggest some good features to start with?
You can have a look at event related potentials. You can obtain them by averaging over trials. However, I suggest that you start learning about EEG analysis first. There are plenty of tutorials. I recommend the FieldTrip toolbox:
http://fieldtrip.fcdonders.nl/tutorial
The tutorials are clear and the toolbox is very versatile.
Best,
Julian
This source is very good:
http://fieldtrip.fcdonders.nl/references_to_review_papers_and_teaching_material
Hi, I have uploaded the signed End User License Agreement. But have not yet received the username and password. Could you share this data with me? thanks so much!
- Badges
- Science topic