Identification of Observations of Correct or Incorrect Actions using Second Order Statistical Features of Event Related Potentials

摘要

The identification of correct or incorrect actions is a very significant task in the field of the brain-computer interface systems. In this paper, observations of correct or incorrect actions are identified by means of event related potentials (ERPs) that represent the brain activity as a response to an external stimulus or event. ERP signals from 47 electrodes, located on various positions on the scalp, were acquired from sixteen volunteers. The volunteers observed correct or incorrect actions of other subjects, who performed a special designed task. The recorded signals were analysed and five second order statistical features were calculated from each one. The most prominent features were selected using a statistical ranking procedure forming a set of 32 feature vectors, which were fed to a Support Vector Machines (SVM) classifier. The performance of the classifier was assessed by means of the leave-one-out cross validation procedure resulting in classification accuracy 84.4%. The obtained results indicate that the analysis of ERP-signals that are collected during the observation of the actions of other persons could be used to understand the specific cognitive processes that are responsible for processing the observed actions.