These are the codes that I used to preprocess the DEAP EEG dataset (https://www.eecs.qmul.ac.uk/mmv/datasets/deap/index.html).
The preprocessing steps are (Step1_Preprocessing.m):
1. Resampling (250 Hz).
2. Rereference using common average (pop_reref).
3. Bandstop filter @ 50 Hz (CleanLine).
4. Independent Component Analysis (runica - infomax extended).
5. Dipole fitting (dipfit).
6. ICLabel.
After that, I manually reviewed all components and rejected those suspicious (Step2_ReviewICA.m).
Finally I selected the data corresponding to the emotional states of interest (Step3_Selection.m):
1. Surprise.
2. Neutral.
3. Fear.
4. Disgust
With the data preprocessed and selected, I created nonoverlapping windows of 5 seconds of EEG (Step4_Windows.m).
These windows were merged with the similar windows from the SEED VII dataset (https://github.com/dcarolpz/Preprocessing_SEEDVII_EEG) to train a Neural Network to classify emotions based on EEG data (Step5_NeuralNetwork.m).
by: Diego Caro López, 30-Mar-2026.
Queries: dgcarolp@hotmail.com // A00833057@exatec.tec.mx