循环谱分析在心律失常分类中的应用研究

发布时间：2018-05-27 14:10

本文选题：心律失常分类 + 循环谱　；参考：《计算机科学与探索》2017年11期

【摘要】：心电信号心律失常分类性能主要取决于有效的特征提取和分类器设计。针对传统心律失常分类研究中,多数研究直接利用时域或者频域特征实现心律失常分类,对于多类别的分类性能仍有待提高。鉴于此,选用循环谱分析方法实现心律失常多分类任务。假设信号处于非平稳状态,建立更符合心电信号实际状态的模型去捕捉心电信号中的隐含周期实现心律失常分类。在提取形态特征和时频域小波系数特征之外,利用循环谱技术提取了谱相关系数特征用于后续多分类任务。除此之外,比较了人工神经网络、传统支持向量机和超限学习机分类器在该实验环境下的分类性能,通过多组对比实验,结果表明,利用循环谱技术结合超限学习机分类器进行心律失常分类,可以区分10类心律失常并在MIT-BIH心律失常数据库上实现了98.13%的平均分类准确率。
[Abstract]:The performance of ECG arrhythmia classification mainly depends on the effective feature extraction and classifier design. In the traditional classification of arrhythmia, most of the studies directly use the time-domain or frequency-domain features to achieve the classification of arrhythmias, but the classification performance of multi-category still needs to be improved. In view of this, circulatory spectrum analysis was used to achieve multi-classification task of arrhythmia. Assuming that the signal is in a non-stationary state, a model which is more consistent with the actual state of the ECG signal is established to capture the hidden period in the ECG signal to realize the classification of arrhythmia. In addition to morphological features and wavelet coefficients in time-frequency domain, spectral correlation coefficients are extracted by cyclic spectrum technique for subsequent multi-classification tasks. In addition, the classification performance of artificial neural network, traditional support vector machine and over-limit learning machine classifier in this experimental environment is compared. Using circulatory spectrum technology and learning machine classifier to classify arrhythmias, 10 kinds of arrhythmias can be distinguished and the average classification accuracy is 98.13% on MIT-BIH arrhythmia database.
【作者单位】：天津大学电子信息工程学院;
【基金】：国家自然科学基金,No.61271069~~
【分类号】：R541.7;TP18

【相似文献】