Data-driven brain computer interface in real environments

Since brain signals measured in real environments should vary due to the variation in brain states and to the various kinds of noises existing in the environments. To develop practical brain computer interface (BCI) systems that cope with such inevitable variability, one possible scenario is to prepare a brain signal database that covers various situations occurring in the target environment. In this study, we present two studies of such data-driven BCI systems. The first one is an NIRS-based decoding system of natural behaviors, which is able to decode natural movements within a daily-life environment called BMI house (main contributor: Takeshi Ogawa, ATR). The second one is a data-driven subject transfer decoding system, which performs EEG-based decoding of spatial attention by a new user by fully utilizing a database of EEG signals taken from many users (main contributor: Hiroshi Morioka, Kyoto University/ATR).