近年来,毫米波雷达信号在医疗监测领域的应用日益广泛,实现雷达信号到心电信号的精准映射已成为满足日常持续性非接触心电监测需求的关键挑战.详细介绍了毫米波雷达信号处理流程,探索了雷达信号与心电信号的细粒度映射关系,引入基于卷积块注意力机制模块(convolutional block attention module,CBAM)的卷积自编码器(convolutional autoencoder,CAE)与双向长短期记忆(bidirectional long short-term memory,BiLSTM)组合的CAE-BiLSTM深度学习网络,实现了雷达信号到心电图的非线性转换.实验结果表明,所提方法在形态学精度上的中位数为0.92,特征峰预测误差低于50 ms,显著增强了雷达信号与心电信号的映射关系,为非接触式心电信号的生成提供了新思路.
Non-contact ECG reconstruction algorithm based on millimeter wave radar
With the wide application of millimeter-wave radar signals in medical monitoring,accurately mapping these signals to ECG signals has become a key challenge in meeting the needs for daily continuous non-contact ECG monitor-ing.The signal processing flow of millimeter-wave radar was introduced in detail,the fine-grained mapping relation-ship between radar signals and ECG signals was explored,and the nonlinear transformation from radar signals to elec-trocardiograms was achieved through the introduction of the CAE-BiLSTM deep learning network,which was a hybrid of a convolutional autoencoder(CAE)and bi-directional long short-term memory(BiLSTM),incorporating the convo-lutional block attention module(CBAM).The results show that the median morphological accuracy of the proposed method is 0.92,and the feature peak prediction error is less than 50 ms.The proposed approach significantly enhances the mapping relationship between radar and ECG signals and offers a new idea for generating non-contact ECG signals.
NETL
NSTL
万方数据
