基于ECG的身份识别技术

发布时间：2018-01-13 09:16

本文关键词：基于ECG的身份识别技术　出处：《浙江大学》2014年硕士论文　论文类型：学位论文

【摘要】：身份识别是安全领域非常重要的一个组成。ECG心电信号由于其普遍性、唯一性、稳定性和可测性,逐渐在身份识别领域得到了越来越广泛的应用。 ECG身份识别的核心在于心电信号特征的提取、分类、识别。但是,心电信号本身十分微弱,又是通过体表的导联采集,常常混入各种噪声干扰。这些噪声干扰影响了波形的识读,给特征提取的准确性带来了挑战。因此,对心电信号进行预处理,滤波去除噪声是一个很重要的课题。同时,身份识别不同于疾病识别,对于特征的合理筛选也是非常关键的工作。通过分析一些已有的预处理方法,本文提出了三个方面来改进。第一,构造心电信号模型,添加噪声后滤波,通过滤波后的信噪比和相关系数,选择最适合的小波函数。第二,改进传统闽值滤噪法。第三,提取波形特征时,选取不受心率影响的特征,避免心率变化给身份识别带来的不利影响。最后,本文分别用RBF神经网络、支持向量机和朴素贝叶斯三种分类方法测试身份识别的效果,都取得了满意的识别率,用实验证明了该身份识别技术的可行性。
[Abstract]:Identity recognition is a very important component in the field of security. ECG ECG signal has been more and more widely used in the field of identity recognition because of its universality, uniqueness, stability and testability. The core of ECG identification is the extraction, classification and recognition of ECG signals. However, ECG signals are very weak and collected by the lead of body surface. Often mixed with a variety of noise interference, these noise interference affects the waveform reading, and brings a challenge to the accuracy of feature extraction. Therefore, the ECG signal is preprocessed. Filtering and removing noise is a very important task. At the same time, identity recognition is different from disease identification, and it is also a key task for proper feature screening. Through the analysis of some existing pretreatment methods, this paper proposes three aspects to improve. First, construct ECG signal model, filter after adding noise, filter the signal-to-noise ratio and correlation coefficient after filtering. The most suitable wavelet function is selected. Secondly, the traditional threshold method is improved. Thirdly, when extracting waveform features, the features that are not affected by heart rate are selected to avoid the adverse effect of heart rate change on identity recognition. Finally, this paper uses RBF neural network, support vector machine and naive Bayesian classification methods to test the effect of identity recognition, and achieved a satisfactory recognition rate. The feasibility of the identification technology is proved by experiments.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN911.7

【参考文献】