Hierarchical Self-organizing Maps of NIRS and EEG Signals for Recognition of Brain States

摘要

Recent advances in temporal data mining of brain activity with NIRS and EEG signals allow us to recognize brain states in higher resolution. However, brain states are not always distinct from each other and often differ in temporal granularity. This paper revisits Dennett's three levels of stance, the DIKW model for the design of two self-organizing maps (SOMs), which contributes to recognition of a hierarchy of brain states with finer granularities. The experimental results show that two brain states at different levels can be accurately identified by applying different training data for each level of SOM.