人工神经网络在HRV分析中的应用研究

摘要

心率变异性(Heart Rate Variability,HRV)反映了心脏神经活动的紧张性和均衡性,是一种检测自主神经性活动的非侵入性指标.近几十年来大量研究已充分肯定了自主神经活动与多种疾病有关系,特别是与某些心血管疾病的死亡率,尤其是猝死率有关.为此,在应用人工神经网络进行HRV分析的基础上,采用误差反向传播网络(Back Propagation,BP)及其改进算法,实现了对HRV信号的初步识别.对不同算法神经网络进行了参数设置尝试和训练测试.测试结果表明,隐层节点数为10及学习速率为0.5时,采用附加动量法(动量学习率为0.3)即可保证整个神经网络训练及检测识别的正确率为93.96％,且稳定性较好.应用人工神经网络算法分析HRV,为心电信号智能分析提供了新的研究领域和应用空间.%HRV reflects the tension and balance of cardiac nerve activity,which is a non-invasive index to detect autonomic nervous ac-tivity. In recent years,a lot of research has fully affirmed the autonomic nervous activity correlated a variety of diseases,especially with some cardiovascular disease mortality,even the sudden death rate. Therefore,on the basis of application of the artificial neural network to HRV analysis,the error Back Propagation ( BP) and its improved algorithm is utilized,in order to realize the preliminary identification of the HRV signals. The diverse neural network is conducted in parameters setting and training test. The test results show that when the num-ber of hidden nodes is 10 and the learning rate is 0. 5,the additional momentum (momentum learning rate is 0. 3) is used to guarantee the correct rate of 93. 96％ and better stability in the neural network training and recognition. The artificial neural network algorithm is ap-plied to the HRV analysis,which expands the new research field and space for the intelligent analysis of the ECG signal.