Principal Component and Independent Component Analysis of Brain-Generated Biopotential Measurements

• Activity Number: 256
• Type: Contributed
• Date/Time: Monday, August 1, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): George Freitas von Borries*+ and Murilo Coutinho Silva and Loyane Christina Soares Rocha and Ricardo Freitas von Borries
• Companies: Universidade de Brasília and Universidade de Brasília and Universidade de Brasília and The University of Texas at El Paso
• Keywords: Principal Component Analysis ; Independent Component Analysis ; Electroencephalography

Abstract

Recent techniques in digital signal processing allow detection of different kinds of mental activity from brain-generated biopotentials. Electroencephalograph (EEG) methods, as opposed to invasive measurements, have the advantage of requiring only surface biopotentials to be acquired, and are therefore more suitable for regular-use HMI equipment in applications as classification of thought patterns for machine control by persons with motor disabilities. This paper uses Principal Component Analysis (PCA) and Independent Component Analysis (ICA) of EEG signals to investigate data reduction source activated areas of brain when submitted to a sequence of visual stimuli. The data was collected in the Biopotentials Imaging Lab (BIML) at UTEP, using a 128-electrode acquisition system. EEG data reduction and source identification allow one to understand brain activity during different tasks and to reduce the number of dimensions in classification procedures. By applying this technique, we expect to improve classification of different tasks with respect to usual classification techniques applied to raw data and to better understand brain activation by visual stimuli.