A Comparative Analysis of Correlation and Correntropy in Graph-Based Brain Computer Interfaces
Luisa Fernanda Suárez Uribe, Carlos Alberto Stefano Filho, Vanessa Brischi Olivatto, Levy Boccato, Gabriela Castellano, Romis Attux, Vinícius Alves de Oliveira, Diogo Coutinho Soriano

DOI: 10.14209/sbrt.2017.213
Evento: XXXV Simpósio Brasileiro de Telecomunicações e Processamento de Sinais (SBrT2017)
Keywords: Brain-computer interfaces information-theoretic learning correntropy graph measures motor imagery
Abstract
This work presents a comparative analysis of correlation and correntropy in the context of graph-based braincomputer interfaces using motor imagery. These two statistical entities are used in the construction of the graphs, from which features are extracted. The results indicate that correntropy has a more consistent performance over the different graph measures, hence deserving to be considered as a relevant option by researchers of the field.

Download