Related Experiment Videos
A cross-sectional evaluation of meditation experience on electroencephalography data by artificial neural network and
Yu-Hao Lee1, Ya-Ju Hsieh, Yung-Jong Shiah
1aDepartment of Electrical Engineering, National Cheng Kung University, Tainan bDepartment of Medical Imaging and Radiological Sciences, Kaohsiung Medical University cGraduate Institute of Counseling Psychology and Rehabilitation Counseling, National Kaohsiung Normal University dDepartment of Medical Imaging, Kaohsiung Medical University Hospital eCenter for Infectious Disease and Cancer Research, Kaohsiung Medical University fInstitute of Medical Science and Technology, National Sun Yat-sen University gGraduate Institute of Medicine, College of Medicine, Kaohsiung Medical University hDepartment of Medical Research, Kaohsiung Medical University Hospital, Kaohsiung, Taiwan iTibetan NyingmapaKathok Buddhist Organization, Sichuan, China jDepartment of Neurology, Kaohsiung Medical University, Kaohsiung, Taiwan.
Abstract:
To quantitate the meditation experience is a subjective and complex issue because it is confounded by many factors such as emotional state, method of meditation, and personal physical condition. In this study, we propose a strategy with a cross-sectional analysis to evaluate the meditation experience with 2 artificial intelligence techniques: artificial neural network and support vector machine. Within this analysis system, 3 features of the electroencephalography alpha spectrum and variant normalizing scaling are manipulated as the evaluating variables for the detection of accuracy. Thereafter, by modulating the sliding window (the period of the analyzed data) and shifting interval of the window (the time interval to shift the analyzed data), the effect of immediate analysis for the 2 methods is compared. This analysis system is performed on 3 meditation groups, categorizing their meditation experiences in 10-year intervals from novice to junior and to senior. After an exhausted calculation and cross-validation across all variables, the high accuracy rate >98% is achievable under the criterion of 0.5-minute sliding window and 2 seconds shifting interval for both methods. In a word, the minimum analyzable data length is 0.5 minute and the minimum recognizable temporal resolution is 2 seconds in the decision of meditative classification. Our proposed classifier of the meditation experience promotes a rapid evaluation system to distinguish meditation experience and a beneficial utilization of artificial techniques for the big-data analysis.