Hello,
I understand that you want to extract 64 features from EEG data within the frequency range of 4-45 Hz in EEGLab. For this, you need to consider how features are typically extracted and what might be limiting them to only 42 features. You can write custom MATLAB scripts to extract features. EEGLab functions can be called within your scripts for preprocessing and feature extraction. For extracting all the features, please follow the below given measures:
- Ensure Proper Filtering: Make sure your data is properly filtered to the range of 4-45 Hz. You can use the EEGLab GUI or script commands to apply the appropriate band-pass filter.
- Feature Extraction Methods: There are various methods to extract features from EEG data like Power Spectral Density (PSD) which helps in extracting power in different frequency bands. Time-domain features can also be used such as mean, variance, skewness, kurtosis, etc. Frequency-domain features can be extracted using Fourier Transform or Wavelet Transform. Time-Frequency representations can also help such as Short-Time Fourier Transform (STFT) or Wavelet Transform.
- Increasing Feature Count: To increase the feature count, please follow this example MATLAB code:
% Load EEGLab and your dataset
eeglab;
EEG = pop_loadset('your_dataset.set');
The EEG data is filtered between 4 and 45 Hz.
% Filter the data to the desired frequency range
EEG = pop_eegfiltnew(EEG, 4, 45);
Divide Frequency Bands Further: Instead of using broad bands (e.g., theta, alpha, beta, gamma), you can subdivide these into narrower bands. For example, instead of having a single alpha band (8-12 Hz), you can divide it into smaller bands like 8-10 Hz and 10-12 Hz.
% Define frequency bands (example of more granular bands)
freq_bands = [4 8; 8 12; 12 16; 16 20; 20 24; 24 28; 28 32; 32 36; 36 40; 40 45];
Use Multiple Channels: If you have multiple EEG channels, features can be extracted from each channel separately and combined.
% Initialize feature matrix
num_channels = size(EEG.data, 1);
num_features = length(freq_bands);
features = zeros(num_channels, num_features);
Combine Multiple Methods: Use a combination of different feature extraction methods (e.g., combine time-domain and frequency-domain features). Band power is computed for each band and each channel.
% Loop through each channel and extract features
for ch = 1:num_channels
for f = 1:num_features
% Extract band power (example feature)
band = freq_bands(f, :);
psd = bandpower(EEG.data(ch, :), EEG.srate, band);
features(ch, f) = psd;
end
end
% Reshape or aggregate features if necessary
final_features = reshape(features, 1, []);
The above example script will generate a set of features based on band power for each frequency band and each channel. If you have 64 channels and 10 frequency bands, this would give you 640 features. You can adjust the number of frequency bands and the feature extraction method as needed to achieve your desired number of features.
I hope this helps!
