EEG on-line analysis for autonomous adaptive interface
| Autor: | Yukinori Kakazu, Hiroshi Yokoi, Yuko Ishiwaka |
|---|---|
| Rok vydání: | 2002 |
| Předmět: |
Engineering
Artificial neural network medicine.diagnostic_test business.industry Interface (computing) Biomedical signal Pattern recognition General Medicine Electroencephalography Machine learning computer.software_genre Task (project management) Line (geometry) medicine Reinforcement learning Artificial intelligence business computer |
| Zdroj: | International Congress Series. 1232:271-275 |
| ISSN: | 0531-5131 |
| DOI: | 10.1016/s0531-5131(01)00721-x |
| Popis: | Our purpose was to develop an interface that closely adapts to individuals such as bedridden people. By quantifying the frustration as the human manipulates the machines from the biomedical signal, and making it to be the teaching signals of machine learning (ML), we aimed at the development of the system in which the machine adapts to the human. In this paper, we extract the characteristic vector whether the examinee feels comfort or discomfort from an electroencephalogram (EEG). The artificial neural network (ANN) is employed to extract the characteristic vector. For the machine learning, reinforcement learning is used and their rewards are extracted signals from physiological signals. As a basic experiment for extracting comfort and discomfort from the physiological signals, the EEG measurement experiment is carried out under an unpleasant task. The input signals to the ANN, for the characteristic vector extraction are examined. Finally, several results of EEG measuring under the discomfort situation are shown in which the frustration accumulated for the examinee. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|
Full Text from ScienceDirect