心电信号在个人身份识别中的稳定性研究

Study on the Stability of ECG Signal in Personal Identification

摘要: 心电信号作为一种重要的人体生理数据，将其用于身份识别的研究已经较为成熟。但是，人类个体的心电信号在较长时间跨度情况下是否能够保持良好的特征稳定性，将对心电信号能否真正用于实际身份识别具有重要影响。本文采用卷积神经网络(CNN)算法构建识别模型，在模型识别准确率稳定保持在97%以上的前提下，对健康青年群体首次、一月后、三月后、六月后等不同时间跨度心电信号在身份识别中的稳定性进行了研究。实验将首次采集心电数据作为训练数据，然后使用该模型验证志愿者一月后、三月后、六月后采集的心电数据识别准确率。结果显示，一月后的心电信号平均识别准确率可达98.02%，实验证明了健康青年人的心电信号在较长时间跨度下能够保持一定的稳定性。

Abstract: ECG signal is an important human physiological data and the research on using it for identification has been relatively mature. However, whether the ECG signals of human individuals can maintain good feature stability over longer time spans will have an important impact on whether ECG signals can actually be used for practical identification. In this paper, we use convolutional neural network (CNN) algorithm to construct recognition models. Under the premise that the accuracy of model recognition can be consistently maintained above 97%, the stability of ECG signals in identification was studied for different time spans in healthy youth groups at first, after one month, after three months and after six months. The experiment will collect ECG data as training data for the first time, and the model was then used to validate the accuracy of identification of ECG data collected from volunteers after one month, three months and six months. The results showed that the average recognition accuracy of the ECG signal after one month was up to 98.02%. It was demonstrated that the ECG signals of healthy young people can maintain some stability over a longer span of time.