计算机工程与应用 ›› 2018, Vol. 54 ›› Issue (17): 135-140.DOI: 10.3778/j.issn.1002-8331.1705-0135

• 模式识别与人工智能 • 上一篇    下一篇

基于心冲击信号的睡姿识别

张艺超,袁贞明,孙晓燕   

  1. 杭州师范大学 信息科学与工程学院,杭州 311121
  • 出版日期:2018-09-01 发布日期:2018-08-30

Sleep position recognition based on ballistocardiogram signal

ZHANG Yichao, YUAN Zhenming, SUN Xiaoyan   

  1. College of Information Science and Engineering, Hangzhou Normal University, Hangzhou 311121, China
  • Online:2018-09-01 Published:2018-08-30

摘要: 研究证明,睡眠质量与睡姿有着密切关系,不良的睡姿甚至会加剧多种疾病的潜在风险。为了更精准地进行睡眠健康监控,提出了一种基于心冲击(BCG)信号的睡姿模式识别算法,使用非接触、无干扰的压电薄膜传感器采集BCG信号,在腰腹部采集仰卧、俯卧、左侧卧和右侧卧4种睡姿信号,经小波变换降噪等预处理后提取基于J波的特征值,设计并比较基于神经网络和KNN的睡姿识别分类器。实验结果表明,神经网络睡眠识别算法的平均正确识别率为93%,KNN算法为84%,因此基于BCG信号的神经网络睡姿识别算法可以广泛用于睡眠监测应用。

关键词: 心冲击信号, 小波变换, 特征提取, 神经网络, 睡姿识别

Abstract: Studies show that sleep position can affect one’s sleep quality, and inappropriate position may potentially increase risks for multiple diseases. In this paper, a Ballistocardiogram(BCG) signal based sleeping position pattern recognition algorithm is presented. BCG signals from four different sleeping positions(supine, prone, left lateral and right lateral) are collected with a piezoelectric film sensor in a non-contact way. Eight characteristics are extracted from the preprocessed BCG signals after wavelet-based denoising, and the sleep position pattern is recognized using BP neural network and KNN. The experimental results show that recognition accuracy of the neural network algorithm and KNN are 93% and 84%, respectively. Therefore, BCG-based sleep position recognition with neural network has a potential to be widely used in sleep monitoring applications.

Key words: ballistocardiogram, wavelet transform, feature extraction, neural network, sleeping position recognition