Assessing the impact of signal normalization: preliminary results on epileptic seizure detection.

Signal normalization is an essential part of patient independent algorithms, for example to correct for variations in signal amplitude from different parts of the body, prior to applying a fixed threshold for event detection. Multiple methods for providing the required normalization are available. This paper presents a systematic investigation into the effects of five different methods using epileptic seizure detection from the EEG as an illustration case. It is found that, whilst normalization is essential, four of the considered methods actually decrease the ability to detect seizures, counteracting the algorithm aim. For optimal detection performance the effects of the signal normalization illustrated here should be incorporated into future algorithm designs.