Manipulation Related EEG Brainwave Feature Extraction and Events Recognition for Robotics Learning Applications

摘要

This research is directed to the application of pattern recognition techniques (PCA), for understanding waves patterns resulting from brain Electroencephalography (EEG). The EEG brainwaves are resulting from specific human hand fingers movements, for a defined task. The adopted technique involved four main computational stages. First, EEG dataset collection for a defined grasping task, Luciw et. al. [1]. Secondly, it was a filtration and multi-signals signals conditioning of such multi-dimensional EEG wave sets. Thirdly, dimensionality reduction of the EEG patterns, hence capturing main EEG waves features. Finally, last stage involved using of pattern recognition and classification algorithms for classification of diverse grasping events. Events classification was based on analysis of set of EEG related patterns, then to correlate such patterns with real word experiment fingers movements. The adopted technique is useful in terms of understanding EEG related and hidden patterns, that are useful for a number of robotics direct or indirect learning applications.