文章摘要
沈建飞,陈益强,谷洋.基于时频信息融合网络的非干扰呼吸检测方法[J].高技术通讯(中文),2020,30(10):998~1009
基于时频信息融合网络的非干扰呼吸检测方法
  
DOI:doi:10.3772/j.issn.1002-0470.2020.10.002
中文关键词: 非接触检测; 雷达检测; 生理检测; 呼吸检测
英文关键词: non-contact detection, radar detection, vital sign detection, respiratory rate detection
基金项目:
作者单位
沈建飞  
陈益强  
谷洋  
摘要点击次数: 287
全文下载次数: 218
中文摘要:
      为了能够克服复杂环境中的噪声影响,实现基于多普勒雷达的鲁棒呼吸信号检测,本文提出了一种基于时频信息融合网络的非干扰呼吸信号检测方法。该方法利用多普勒雷达采集用户的胸腔运动信息,提取双通道混频信号(I和Q通道),从时域和频域2个维度,构建时频信息融合网络进行呼吸频率的识别。针对时域信号,使用长短期记忆(LSTM)网络提取信号的有效周期信息;针对频域信息,使用选带傅里叶变换(ZoomFFT)实现细粒度的频域特征计算,并结合卷积神经网络(CNN)提取频域特征的有效信息;最后,融合2个层面的信息,通过Lowess平滑方法,实现对用户呼吸频率的精准检测。实验表明,该方法比其他常用信号处理方法识别平均误差、标准差更小,对不同距离、不同朝向的呼吸都可以进行有效识别。
英文摘要:
      To overcome the noise influence in respiratory detection based on Doppler radar, a non-contact respiratory detection method is proposed based on the time and frequency fusion network. This method collects information of the chest movements with Doppler radar, and analyses the mixing signals I and Q from two domains. For the timedomain, the period information of the signal is extracted with the long short-term memory (LSTM) network. For the frequency domain, the zoom fast Fourier transform (ZoomFFT) is used to implement the fine-grained frequency domain features, and the CNN network is used to extract the effective information from the frequency domain features. The last part converges two-dimension information to achieve accurate detection of the user’s respiratory rate with the Lowess smoothing method. Experiment results show that the proposed method can achieve less average error and root mean square error than other common signal processing methods, and it can effectively identify the breathing rate from different distances or different orientations.
查看全文   查看/发表评论  下载PDF阅读器
关闭

分享按钮