Remote Pulse Rate Estimation Method Based on Principal Signal Extraction and Wavelet Transform

DOI HANDLE

Bibliographic Information

Other Title
  • 主要信号抽出とウェーブレット変換による非接触脈拍数推定法

Search this article

Abstract

In this paper, we propose a remote pulse rate estimation method from video under fluorescent light. Our method focuses on the following two assumptions. One is that the pulse rate is approximately equivalent at any points on the skin. Another is that the pulse rate does not change so quickly while keeping us at rest. First of all, we track face to extract skin region from video. Next, we divide the skin region into blocks and convert the time sequence of color images within the region to color signals. After that, we detrend each color signal and convert them to chrominance signals. Then, we extract principal signal from chrominance signals by pruning of signal with different peak intervals and using singular value decomposition, under the assumption of the pulse rate equality over any surface of the skin. Finally, we use wavelet transformation to estimate the pulse rate under the assumption of the stability of the pulse rate while resting. As a result, our method achieves improving estimation accuracy. The mean absolute error is 0.762bpm, the mean absolute percentage error is 1.04% , the standard deviation of error is 0.899bpm, the correlation coefficient is 0.987 and, in 90% of data, the error is less than 2bpm.

Journal

Details 詳細情報について

  • CRID
    1390853650463334528
  • NII Article ID
    120007119706
  • NII Book ID
    AA12746425
  • DOI
    10.15002/00023869
  • HANDLE
    10114/00023869
  • ISSN
    24321192
  • Text Lang
    ja
  • Data Source
    • JaLC
    • IRDB
    • CiNii Articles
  • Abstract License Flag
    Allowed

Report a problem

Back to top