Signal decomposition by multi-scale PCA and its applications to long-term EEG signal classification

Data coming from a real-world complex system are usually contaminated by certain levels of noise or some irrelevant components, which do not contribute to improve signal classification accuracy. Also in signal de-noising, the performance of any statistical method used to recover the original signals may be impacted by the noise. In this paper, we propose the multi-scale principal component analysis (PCA) method, which combines discrete wavelet transform and PCA for de-noising and decomposing complex biomedical signals in both spatial and temporal domains for signal classification. We also develop a new classification method, called Empirical Classification (EC), based on the characteristics of data we analyzed. These methods were applied to a publicly available EEG database for the purpose of epilepsy diagnosis and epileptic seizure detection. Our study shows that signal decomposition by the multi-scale PCA method coupled with the EC method, leads to a highly promising classification accuracy in classifying epileptic EEG signals. Our methods are also applicable for classifying biomedical images.