STATE CLASSIFICATION OF AUTONOMIC NERVOUS ACTIVITY BASED ON RELATIVE HEART RATE SPECTRA
山下, 翔平YAMASHITA, Shohei
562015-03-24 , 法政大学大学院理工学・工学研究科
This paper proposes an efficient algorithm to classify the autonomic nervous activity based on the heart rate variability (HRV). Data has been provided by MIT-Harvard division of health sciences and technology. 14 volunteers’data ware collected at 8 distinct states by drug administration. Binary decision scheme with three layered perceptron (3LP) was shown to be effective for the classification. Careful data cleaning and classification indices for each binary decision lead us to the accurate classification of sensitivity 0.897 and specificity 0.979. The pattern classification with Support Vector Machine (SVM) showed comparable performance with (3LP). The method employs relative power spectra hence useful for individualized health monitoring.