Howe-Patterson
EEG kept as waveform
- Applies FIR filtering, downsampling, and rolling normalization to the EEG channel.
- The model still receives EEG as a time-series waveform.
Why This EEG Example Matters
same raw EEG
Bragi-Marinosson
EEG turned into features
- Uses raw EEG windows to extract wavelet-packet and Hjorth descriptors.
- The model receives feature vectors, not the original EEG trace.
Takeaway
Performance differences come from signal choice and preprocessing, not only from the final classifier.
One example
EEG From The Same Record
RAW EEG
HOWE-PATTERSON
BRAGI-MARINOSSON
Turns the same EEG window into derived descriptors before classification.
raw EEG window
wavelet packet
Hjorth stats
feature vector
The visual difference is the point: Howe still sees a waveform, while Bragi replaces the waveform with extracted EEG descriptors.