论文部分内容阅读
目的构建一个用于区分室颤节律和非室颤节律的多参数融合BP神经网络。方法从心电信号中提取出18个特征值,分别标记心电信号的形态分布、高斯性、幅度谱、变异性、复杂度等各方面特征;以这些特征值作为输入向量,构建一个多参数融合BP神经网络进行训练,得到一个用于区分室颤节律与非室颤节律的分类器。结果与结论将构建的BP神经网络分别基于VFDB数据库和CUDB数据库进行实验,辨识准确率分别高达98.61%和95.37%;相较于现有方法,辨识性能显著提高。
Objective To construct a multi-parameter fusion BP neural network for distinguishing ventricular fibrillation rhythm and non-ventricular fibrillation rhythm. Methods Eighteen eigenvalues were extracted from the ECG signals, and the morphological distribution, Gaussianity, amplitude spectrum, variability and complexity of the ECG signals were marked respectively. With these eigenvalues as input vectors, a multi-parameter Fusion BP neural network for training, to be used to distinguish between ventricular fibrillation rhythm and non-ventricular fibrillation rhythm classifier. Results and Conclusion The constructed BP neural network was tested based on the VFDB database and the CUDB database respectively, and the recognition accuracy was 98.61% and 95.37% respectively. Compared with the existing methods, the recognition performance was significantly improved.