Adaboost with SVM-based classifier for the classification of brain motor imagery tasks

  • Authors:

  • Jue Wang;Lin Gao;Haoshi Zhang;Jin Xu

  • Affiliations:

  • Institute of Biomedical Engineering, Key Laboratory of Biomedical Information Engineering of Education Ministry, Xi'an Jiaotong University, People's Republic of China;Institute of Biomedical Engineering, Key Laboratory of Biomedical Information Engineering of Education Ministry, Xi'an Jiaotong University, People's Republic of China;Institute of Biomedical Engineering, Key Laboratory of Biomedical Information Engineering of Education Ministry, Xi'an Jiaotong University, People's Republic of China;Institute of Biomedical Engineering, Key Laboratory of Biomedical Information Engineering of Education Ministry, Xi'an Jiaotong University, People's Republic of China

  • Venue:
  • UAHCI'11 Proceedings of the 6th international conference on Universal access in human-computer interaction: users diversity - Volume Part II
  • Year:
  • 2011

Quantified Score

Hi-index 0.00

Visualization

Abstract

The Adaboost with SVM-based component classifier is generally considered to break the Boosting principle for the difficulty in training of SVM and have imbalance between the diversity and accuracy over basic SVM classifiers. The Adaboost classifier in the paper trains SVM as base classifier with changing kernel function parameter s value, which progressively reduces with the changes of weight value of training sample. To testify the validity of the classifier, the classifier is tested on human subjects to classify the left- and right-hand motor imagery tasks. The average classification accuracy reaches 90.2% on test data, which greatly outperforms SVM classifiers without Adaboost and commonly Fisher Linear Discriminant classifier. The results confirm that the proposed combination of Adaboost with SVM classifier may improve accuracy for classification of motor imagery tasks, and have applications to performance improvement of brain-computer interface (BCI) systems.