The selection of proper discriminative cognitive tasks - A necessary prerequisite in high-quality BCI applications

• Main Authors: Dobrea, Monica-Claudia, Dobrea, Dan Marius
• Format: Conference Proceeding
• Language: English
• Subjects: Application software; Artificial neural networks; autoregressive model; Bayes quadratic classifier; Brain computer interfaces; Brain modeling; brain-computer interface application; EEG-discriminative mental tasks; Electroencephalography; Information technology; Multilayer perceptrons; Psychology; Signal processing; Tongue
• ISSN: 2325-5315; 2325-5331
• DOI: 10.1109/ISABEL.2009.5373706

While in brain computer interface (BCI) field the research is focused basically on finding improved processing methods leading to both high classification rates and high bit transfer rates, in this paper the same BCIs performances are addressed but, this time, with the emphasis set on the subject-specific discriminative cognitive tasks selection process. In this respect, a set of twelve electroencephalographic (EEG)-discriminative mental tasks was proposed to be studied in conjunction with four different subjects. For each subject, a particular set of four mental tasks was selected. The classification performances corresponding to these particular sets of tasks were obtained using some standard processing methods (i.e., the autoregressive model of the EEG signals and a multilayer perceptron classifier trained with back-propagation algorithm). The superior classification rates achieved for the selected sets compared to other set of mental tasks commonly used in the 4-class BCI studies (i.e. the set proposed by Keirn and Aunon) promote the idea of subject-oriented mental tasks selection process as a necessary preliminary step in any high-quality BCI application.