for classification emotion using EEG signal, I have transferred 1D EEG to 2D image and then used CNN to classify emotion. As CNN works better on the image. But it did not generate more accurate results as expected.
a. For 1D signal (like EEG) processing which one is better? Directly classify signal or firstly signal to image than classify?
b. For 1D signal classification which (type of) deep learning approach normally works well?
Dear Md. Rabiul Islam ,
Look the link, maybe useful.
Article A NOVEL APPROACH TO CLASSIFY AND CONVERT 1D SIGNAL TO 2D GRA...
Conference Paper Classification of Motor Imagery EEG Signals with multi-input...
Regards,
Shafagat
Dear Md. Rabiul Islam ,
convolutional neural networks (CNN) can work both with images (2D/3D input) and with signals (1-D input).
It would be nice to compare the performance achieved using both input types.
I would recommend you to integrate the features extracted by the CNN with some hand-crafted statistical features (e.g. mean, standard deviation). You can do so by adding directly such features to the first fully connected layer.
Actually, CNN have performed very well on EEG data, similarly to recurrent neural networks (RNN) and long short-term memory (LSTM) networks, thus it is a good idea to use CNN.
Dear Md. Rabiul Islam
These articles might be helpful, have a look:
1. https://www.frontiersin.org/articles/10.3389/fnhum.2020.00338/full
2. https://www.frontiersin.org/articles/10.3389/fnhum.2020.00365/full
3. https://res.mdpi.com/d_attachment/sensors/sensors-20-03491/article_deploy/sensors-20-03491.pdf
Kind Regards
Qamar Ul Islam
- Badges
- Science topic
- Similar topics
- Data Mining
- Classification