抄録
Automatic real-time sleep staging system has been fabricated for the purpose of chronobiological study. The system is supported by a conventional personal computer and can monitor sleep state for about ten hours continuously. In this system, EEG at C3 lead is fed to a set of IC filters and is divided into seven frequency bands. Output powers of these filters are used as components of feature vectors for the discriminant analysis of sleep stage. This discrimination is performed every 30sec and the final staging is obtained after corrections using contextual informations together with EOG and EMG. The results of the sleep staging with this system showed more than 70% agreement with those of visual inspections based on Rechtschaffen and Kales manual. Furthermore, we found that the principal componenet of the multiple-discriminant analysis can be used as the index of the depth of sleeping and using this index, we can identify REM period automatically with 85% agreement. As conclusion, this system well works for chronobiological study. Some results on REM rhythm will be reported in succeeded paper.
