Work

ロバストな脳コンピュータインタフェース実現のためのマルチバンド解析と特徴抽出

Year: 2018
Type: dissertation
Abstract: The development of robust BCI technology for real life applications is one of the most challenging tasks due to noise effects, number of adequate command generations, appropriate methodologies as well as real time usages.This study proposes multiband analysis and feature extraction methods for designing robust brain-computer interfaces (BCIs) including Steady-State Visual Evoked Potential (SSVEP)-and motor imagery (MI)-based BCIs.The SSVEP-based BCI is also implemented with supervised and unsupervised way.For designing both unsupervised and supervised SSVEP-based BCI, canonical correlation analysis (CCA) is the most popular and widely used method.In SSVEP-BCI, alpha band components at around 10 Hz can be more dominant than the SSVEP.In the standard CCA-based unsupervised approach, the canonical correlation values between SSVEPs and the sine-cosine waves tends to decrease as flickering frequency increases.This can lead in decreased accuracy of SSVEP detection in higher frequencies.To address this problem, the SSVEP signals are divided into two subbands.The first subband contains all the target frequencies of SSVEPs.The second one includes the SSVEP signal corresponding to a desired number of higher order stimulus frequencies.The full dimension of artificial reference signals (RS) is used with first subband, whereas a reduced dimension of artificial RS is employed with second subband to compute canonical correlation.The weighted sum of the obtained correlation values is used to recognize the frequency of an SSVEP.To further improve the BCI performance, the existing unsupervised methods and proposed BsCCA are extended with filter bank fashion.
Language: English
Open Access status: green
OpenAlex ID: W7208065318
Cites: 134
Cited by:
Related to: 9