基于希尔伯特振动分解和卷积神经网络的融合特征心电识别算法

在信息高速发展的当代社会,5G技术的问世将极大地助力社会经济和信息化发展,而隐私安全和信息安全愈发得到重视,因此公众会对身份的识别技术提出了更高要求。然而,传统基于密码、ID卡以及新型的基于人脸和指纹的识别方法存在易丢失、遗忘和窃取或易于伪造和获取复制等问题而存在极大的安全隐患。为提高身份识别的可靠性和准确率,提出了基于希尔伯特振动分解和卷积神经网络的融合特征心电图信号识别算法。首先采用基于重叠组收缩阈值算法和平移不变的消噪算法对含噪心电信号去噪,其次利用盲源分割技术将心电信号分割成固定时长的心电片段,再次采用基于希尔伯特振动分解的时频分析方法获得心电信号的时频表示图,通过提出的心电残差卷积神经网络对时频表示图实现特征提取和降维,最后通过Softmax分类器实现分类和识别。以Physionet数据库的ECG-ID数据集验证提出算法的性能,采用10折交叉验证法得到平均识别率为99.08%。结果表明,提出的心电识别算法具有高效的识别性能和良好的应用前景。

Fusion Feature ECG Identification Algorithm based on Hilbert Vibration Decomposition and Convolutional Neural Network

In the contemporary society where information is developing rapidly,the advent of 5G technology will greatly contribute to the development of social economy and informatization,and privacy security and information security are increasingly valued,so the public will put forward higher requirements for identification technology.However,traditional identification methods based on passwords,ID cards,and new face and fingerprint-based methods have the potential to be lost,forgotten and stolen,or easy to forge and obtain copies,which poses great security risks.In order to improve the reliability and accuracy of identity recognition,a fusion feature ECG signal recognition algorithm based on Hilbert vibration decomposition and convolutional neural network is proposed.Firstly,a noisy algorithm based on overlapping group shrinkage threshold algorithm and constant translation is used to denoise the noisy ECG signal.Then,the blind source segmentation technology is used to segment the ECG signal into ECG segments of fixed duration.Then,a time-frequency analysis method based on Hilbert vibration decomposition is used to obtain the timefrequency representation of the ECG signal,and the proposed ECG residual convolutional neural network is used to extract features and reduce dimensionality.Finally,Softmax classifier is used to realize classification and identification.The ECG-ID dataset of the Physionet database is used to verify the performance of the proposed algorithm,and the average recognition rate is 99.08%using the 10-fold cross-validation method.The experiment results indicate that the proposed ECG recognition algorithm has efficient recognition performance and good application prospects.