Presentation 2010-07-27
A Study on Estimating a Vowel Read Silently from Human Electrocorticogram
Shigeyuki IKEDA, Tomohiro SHIBATA, Naohiro TSUYUGUCHI, Naoki NAKANO, Yoshinobu HARA, Rieko OKADA, Kazushi IKEDA, Amami KATO,
PDF Download Page PDF download Page Link
Abstract(in Japanese) (See Japanese page)
Abstract(in English) In recent years, Brain Machine Interface (BMI), which retrieves cortical brain activity by invasive or noninvasive method of measurement and connects external computer has been studied intensively. The ultimate goal of this study is to develop a BCI for verbal communication without utterance. As a step toward the goal, we conducted a silent reading task in which two epilepsy patients with Electrocorticogram (ECoG) electrodes were asked to silently read a vowel, and investigated whether the vowel read was identified by discriminating vowels from the ECoG with Support Vector Machine (SVM). Using data of all electrodes, we confirmed that over-fitting arises due to few all trials of task. Therefore, we estimated discriminating performance of each electrode, then made the color-map of these performance. From the result, we selected electrodes of high performance, then estimated discriminating performance of combination of selected electrodes with test data. As a result, we obtained 53-63% of discriminating performance, which is much higher than the chance level (33%).
Keyword(in Japanese) (See Japanese page)
Keyword(in English) ECoG / BMI / Silent reading / SVM
Paper # NC2010-37
Date of Issue

Conference Information
Committee NC
Conference Date 2010/7/20(1days)
Place (in Japanese) (See Japanese page)
Place (in English)
Topics (in Japanese) (See Japanese page)
Topics (in English)
Chair
Vice Chair
Secretary
Assistant

Paper Information
Registration To Neurocomputing (NC)
Language JPN
Title (in Japanese) (See Japanese page)
Sub Title (in Japanese) (See Japanese page)
Title (in English) A Study on Estimating a Vowel Read Silently from Human Electrocorticogram
Sub Title (in English)
Keyword(1) ECoG
Keyword(2) BMI
Keyword(3) Silent reading
Keyword(4) SVM
1st Author's Name Shigeyuki IKEDA
1st Author's Affiliation Graduate School of Information Science, NAIST()
2nd Author's Name Tomohiro SHIBATA
2nd Author's Affiliation Graduate School of Information Science, NAIST
3rd Author's Name Naohiro TSUYUGUCHI
3rd Author's Affiliation Graduate School of Medicine, Osaka City University
4th Author's Name Naoki NAKANO
4th Author's Affiliation Kinki University Hospital Neurosurgery
5th Author's Name Yoshinobu HARA
5th Author's Affiliation Graduate School of Biological Science, NAIST
6th Author's Name Rieko OKADA
6th Author's Affiliation Graduate School of Medicine Kinki University
7th Author's Name Kazushi IKEDA
7th Author's Affiliation Graduate School of Information Science, NAIST
8th Author's Name Amami KATO
8th Author's Affiliation Kinki University Hospital Neurosurgery:CREST, JST
Date 2010-07-27
Paper # NC2010-37
Volume (vol) vol.110
Number (no) 149
Page pp.pp.-
#Pages 6
Date of Issue