Development of a Non-Invasive Brain-Computer Interface for Neurorehabilitation

• Autor(a) principal: Pinto, Rui Dinis Ribeiro da Silva
• Data de Publicação: 2015
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10362/28549
• topic: Brain-Computer Interface; Emotiv; Neurorehabilitation; Signal Processing; Feature Extraction and Classification; Stroke; Domínio/Área Científica::Engenharia e Tecnologia::Outras Engenharias e Tecnologias

Neurological disorders, in particular Stroke, have an impact on many individuals worldwide. These individuals are often left with residual motor control in their upper limbs. Although conventional therapy can aid in recovery, it is not always accessible, and the procedures are dull for the patient. Novel methods of therapy are being developed, including Brain-Computer Interfaces (BCIs). Although BCI research has been flourishing in the past few years, most rehabilitation applications are not yet suitable for clinical practice.This is due to the fact that BCI reliability and validation has not yet been achieved, and few clinical trials have been done with BCIs. Another crucial factor, is that modern BCIs are often comprised of inconvenient hardware and software. This is a major factor of aversion from both patients and clinicians. This Master Dissertation introduces the EmotivBCI: an easy to use platform for Electroencephalogram acquisition, processing and classification of sensorimotor rhythms with respect to motor action and motor imagery. The acquisition of EEG is done through 8 channels of the Emotiv Epoc wireless headset. Signals are pre-processed, and the 2 best combinations of channel/frequency pairs that exhibit the greatest spectral variation between the rest and action conditions are extracted for different time frames. These features are then used to build a feature matrix with 2 sets of attributes and 2 class labels. Finally the resulting feature matrix is used to train 3 different classifiers, in which the best is selected. The EmotivBCI enables users to keep record of their performances, and provides additional features to further examine training sessions. To assess the performance of the EmotivBCI, two studies were conducted with healthy individuals. The first study compares classification accuracies between two different training paradigms. The second study evaluates the progress in performance of a group of individuals after several training sessions.