Closed scheuchzer closed 10 years ago
Filtering raw signal at the beginning of featureExtraction improves cross validation results for Patient 2, but only the probability of seizure, the probability of early seizure diminishes slightly. However, the computation time increases dramatically. Around 2 hours were needed to finish Patient 2.
Tried low pass and notch filter and no improvement. Closing.
The filtering could be done right at the beginning of featureExtraction which affects all features or only in the getFrequencyFeature method. Which filters should be used? Band-pass (1Hz/60Hz)?