Performance of Nested vs. Non-nested SVM Cross-validation Methods in Visual BCI: Validation Study

Mohammed Abdulaal, Alex Casson, Patrick Gaydecki

    Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

    1376 Downloads (Pure)


    Brain-Computer Interface (BCI) is a technology that utilizes brainwaves to link the brain with external machines for either medical analysis, or to improve quality of life such as control and communication for people affected with paralysis. The performance of BCI systems depends on classification accuracy, which influences the Information Transfer Rate. This motivates researchers to improve their classification accuracy as best possible. A bias problem in reporting accuracies by using non-nested cross-validation methods was thought to increase accuracy. The aim of this paper was to validate and quantify such a concept by using a low-cost commercial EEG recorder to classify visually evoking face vs scrambled pictures, and report high accuracy using non-nested cross validation. The algorithm employed Independent Component Analysis followed by feature extraction with sample covariance matrices. The data were then
    classified using Support Vector Machines. The accuracy was tested with nested and non-nested cross-validation methods; accuracies obtained were 63% and 76%, respectively.
    Original languageEnglish
    Title of host publication2018 26rd European Signal Processing Conference (EUSIPCO)
    Publication statusPublished - 2018

    Research Beacons, Institutes and Platforms

    • Manchester Institute for Collaborative Research on Ageing


    Dive into the research topics of 'Performance of Nested vs. Non-nested SVM Cross-validation Methods in Visual BCI: Validation Study'. Together they form a unique fingerprint.

    Cite this