EEG analysis for BCI applications


Wilmer Castro, Universidad de Nariño. San Juan de Pasto - Colombia, 2019

Since its first application, electroencephalography has been a technique used primarily to understand the behavior of the brain. Studying the electroencephalographic (EEG) signals has tried to decipher the intentions of a person and the actions that can be exercised on certain devices just by imagining it. This concept is what is known as a brain-computer interface or BCI, systems that are currently used with great expectation in people with disabilities or with a high degree of disability, given that it is presented as a new form of communication, making individuals with partial paralysis or total of their limbs can perform tasks such as: write on a monitor, move in a wheelchair, handle prosthesis among others.

The basic function of a BCI system is to measure the electrical signals coming from the brain, process them, extract characteristics and allow the user to interact with the environment through a physical device. The extraction of these characteristics leads to a deep analysis of the EEG signal in order to find characteristic patterns, that is, those related to the user's intentions and that will be used as output commands. This is where the efficiency of a BCI system is still restricted, because the EEG signals at the moment of being acquired are affected by noise and artifacts, besides presenting a low spatial resolution, avoiding a correct characterization of the physiological phenomena and deriving in strenuous hours of training by the user before accurately performing the intentions you want to perform.

In this work an alternative methodology is proposed that allows to analyze and process EEG signals for application purposes in BCI systems. For this, a comparative process of different characterization methods, feature selection and dimension reduction was developed, in order to determine a model of EEG signal classification balanced between efficiency and computational cost.




ą
wilmer castro,
3 jul. 2019 20:28
Comments