南京邮电大学学报（自然科学版）2025，Vol.45Issue(3)：58-66,9.DOI:10.14132/j.cnki.1673-5439.2025.03.007
基于双向门控循环单元网络的心音分段
Heart sound segmentation based on Bi-GRU
摘要
Abstract
In order to improve the positioning precision and accuracy of heart sound segmentation,a heart sound segmentation algorithm based on bi-directional gated recurrent unit(Bi-GRU)network is pro-posed.First,the short-time Fourier transform(STFT)and synchrosqueezing transform(SST)techniques are used to extract the instantaneous frequency features from heart sound signals within short-time win-dows in each time step.Second,the extracted instantaneous frequency features are combined into a se-quence and input into the Bi-GRU network,and the Bi-GRU network explores the contextual dependen-cies of the feature sequence and extract the contextual time-frequency features of the heart sound signals.Finally,the Softmax classifier is used to classify the heart sound signals into four state categories:first heart sound(S1),second heart sound(S2),S1-S2 interval,and S2-S1 interval.The experimental re-sults on the PhysioNet/CinC Challenge 2016 dataset show that the proposed heart sound segmentation al-gorithm achieves an overall accuracy of 93.30％,with an average F1 score of 0.953 8 for S1 and 0.945 0 for S2,and that the proposed algorithm outperforms the baseline heart sound segmentation algorithm LR-HSMM.These demonstrate that the proposed algorithm can effectively segment heart sound signals and provide a basis for feature extraction and analysis of heart sound signals.关键词
心音分段/短时傅里叶变换/同步挤压变换/双向门控循环单元Key words
heart sound segmentation/short-time Fourier transform(STFT)/synchrosqueezing trans-form(SST)/bi-directional gated recurrent unit(Bi-GRU)分类
计算机与自动化引用本文复制引用
卢官明,许梦悦,卢峻禾,戚继荣,赵宇航,王洋..基于双向门控循环单元网络的心音分段[J].南京邮电大学学报（自然科学版）,2025,45(3):58-66,9.基金项目
国家自然科学基金(72074038)、江苏省卫生健康委员会重点项目(K2023036)和南京市卫生科技发展专项资金项目(ZKX22050)资助项目 （72074038）
