Data Science publications for EEG analysis like EEGNet: A Compact Convolutional Network for EEG-based Brain-Computer Interfaces use the raw EEG data (see: github vlawhern/arl-eegmodels ) rather than gif or png files of the data to predict the data.

Why does it make more sense to work with the raw data instead of the images of the data?

My asumption would be that detecting special patterns in EEGs (like edges, slopes etc.) would also work with the same or similar neural network topologies like those uesd for image recognition.


Raw data is already noisy, why would you corrupt it even more converting it to images with limited resolution, and possibly lossly compression formats such as .jpg?

In theory, it could work. But using the raw data makes more sense, and should require fewer parameters and samples. Your network would need to figure out that left to right is advancing in time, bottom to top is magnitude, and that each channel is in a different level in the image. Not worth it.

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.