期刊
BIOMEDICAL SIGNAL PROCESSING AND CONTROL
卷 77, 期 -, 页码 -出版社
ELSEVIER SCI LTD
DOI: 10.1016/j.bspc.2022.103822
关键词
Generative Adversarial Network (GAN); Deep Learning; Heart Rate Variability (HRV); Autonomic Nervous System (ANS); Dataset augmentation; Classification; Recurrence Plot analysis; Chinese Chi meditation; Kundalini yoga
This study proposes an alternative approach to analyzing the effects of Chinese Chi meditation and Kundalini yoga on cardiac dynamics using 'HRVGAN'. The approach augments the dataset of HRV signals and analyzes them directly without the need for signal conversion. The results indicate that meditation significantly affects cardiac dynamics, and the choice of meditation technique leads to different changes. The proposed HRV-GAN model achieves high sensitivity and accuracy, making it beneficial for automated analysis of HRV signals with limited data samples.
The alternative approach highlighted in our study ameliorates the process of analyzing the effect on cardiac dynamics due to the practice of Chinese Chi meditation and Kundalini yoga by the implementation of 'HRVGAN'. It works by augmenting the dataset of Heart Rate Variability (HRV) signals by 31 times with the aid of the Generative Adversarial Network (GAN) and after that analyzing them without the need for any other equivalent conversion of the signal. The study highlights the distinct effects of practicing different types of meditation on HRV by augmenting the initial dataset entirely in an automated deep learning approach. The visualized results from the architecture are verified statistically. The dataset chosen for the study comprised just 24 meditative HRV samples in total, which have been successfully augmented to 744 samples for classification. The results indicated that meditation affects cardiac dynamics significantly, and diversion varies with the choice of meditation technique. The results obtained from the proposed semi-supervised deep learning model-'HRV-GAN' was verified statistically by plotting required histograms, Recurrence Plots (RP), and Recurrence Quantification Analysis (RQA) analysis, which were in agreement with each other. HRV-GAN has achieved a sensitivity of 96.667% and an accuracy of 95.556 %. Our innovation can benefit the medical and biomedical community for research involving automated analysis of HRV signals because 'HRV-GAN' can perform appreciably even with fewer data samples for training a deep learning model.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据