QR decomposition based recursive least square adaptation of autoregressive EEG features

Electroencephalography (EEG) has potential medical, military and industrial applications. To date, no method is said to be a standardized EEG estimator. The aim of this study is to realize QR decomposition based recursive least square estimator for EEG feature extraction. The features are modeled as adaptive autoregressive model. Linear discriminant analysis is performed to classify the extracted features for a dual class experiment. For validation, right- and left-hand movement imaginations based EEG experiments are conducted. Further validation, carried out by a comparative study with other adaptive (least mean squares and recursive least squares) algorithms, demonstrates the effectiveness of the proposed method.