利用分析和外观模型的混合心电图人体识别方法

发布时间：2018-04-06 07:10

本文选题：心电图　切入点：人体识别　出处：《计算机工程》2017年03期

【摘要】：为提高心电图(ECG)生物识别对噪声和变化信号的敏感度,优化区分心跳特征,提出一种混合的ECG人体识别方法。联合心跳分析和外观属性,使用Fisher线性判别分析方法从这些属性中获取主要属性,在Fisher空间计算心跳特征,使用时域特征计算ECG特征。在一系列心跳推导时域特征后,通过标准化和缩放ECG信号获取形态学特征。采用最大化类间个体散列与类内个体散列的比率选择最具区分性的特征。实验结果表明,该方法在健康个体组成的大型数据库上的识别准确率为99.24%。当人体处于心率失调的情况下等错误率(EER)为0.76%,处于混合状态时EER为1.31%,验证了该方法具有较好的鲁棒性。
[Abstract]:In order to improve the sensitivity of electrocardiogram (ECG) biometrics to noise and change signals, and to optimize the identification of heartbeat characteristics, a hybrid ECG human body recognition method was proposed.Combined with heartbeat analysis and appearance attributes, the main attributes are obtained from these attributes by Fisher linear discriminant analysis, heartbeat features are calculated in Fisher space, and ECG features are calculated using time-domain features.After deriving a series of heartbeat time domain features, morphological features are obtained by standardizing and scaling ECG signals.The most distinguishing features are selected by using the ratio of maximization of inter-class individual hash to intra-class individual hash.The experimental results show that the recognition accuracy of this method is 99.24on a large database composed of healthy individuals.When the human body is in the condition of heart rate disturbance, the equal error rate (er) is 0.76, and the EER is 1.31 when the human body is in the mixed state, which proves that the method has good robustness.
【作者单位】：四川师范大学数学与软件科学学院;
【基金】：国家自然科学基金青年项目“复杂医疗保健数据的统计推断和过程控制”(11301364) 四川省教育厅科研重点项目(13sa0137)
【分类号】：R540.41;TP391.41

【相似文献】