You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

78 lines
3.7 KiB

  1. \section{System Architecture}
  2. \label{sec:design}
  3. We designed a thermal-box to collect the data. It has four Grideye sensors on the
  4. corners of a 10 cm square and a Lepton 3 at the central. Figure~\ref{fig:method} shows
  5. the system of our method. It consists four parts. The first part is to fuse multiple
  6. data from Grideye sensors into a low-resolution image, since the resolution of a
  7. single Grideye sensor too low to make a decision. The
  8. second part, we train the SRCNN model with fused Grideye image
  9. as low-resolution and downscaled Lepton 3 image as high-resolution image.
  10. The third part, we use the
  11. Super-resolution image to train a neural network model for recognizing current pose
  12. is lay on back or lay on side. The last part, to reduce the noise and effect cause by
  13. the residual heat on bed after turning over. We
  14. remove the noise by median filter, and determine the current pose according to
  15. the trend of the possibility from recognition network.
  16. \begin{figure}[tbp]
  17. \begin{center}
  18. \includegraphics[width=0.9\linewidth]{figures/method.pdf}
  19. \caption{Illustration of Proposed Method.}
  20. \label{fig:method}
  21. \end{center}
  22. \end{figure}
  23. \subsection{Grideye Data Fusion}
  24. On the thermal-box, there are four Grideye sensors. At the beginning, we let
  25. the thermal-box faces to an empty bed and records the background temperature.
  26. All the following frames will subtract this background temperature. After that,
  27. we resize four $8 \times 8$ Grideye images to $64 \times 64$ by bilinear
  28. interpolation and then merge them dependence on the distance between thermal-box and
  29. bed, distance between sensors and the FOV of Grideye sensor. In our case, $D_B$ is
  30. 150 cm, and $D_s$ is 10 cm.
  31. \begin{enumerate}
  32. \item $D_b$ is the distance between bed and thermal-box.
  33. \item $D_s$ is the width of sensor square also the distance between adjacent sensors.
  34. \item $F$ is the FOV of Grideye sensor which is about 60 degree.
  35. \item $Overlap = 64 - 64 \times (\frac{D_s}{2 \times D_b \times tan(\frac{F}{2})})$
  36. \end{enumerate}
  37. \subsection{Turning Over Determination}
  38. We train a SRCNN model by the fused Grideye image and downscaled Lepton 3 image,
  39. and use it to enhance all following Grideye frames to SR frames. We labeled some SR frames
  40. into two categories, lay on back and lay on side. Since the input
  41. data is very small, we use a neural network consist one 2D convolution layer, one
  42. 2D max pooling, one flatten and one densely-connected layer. The possibility of
  43. output has a very large various just after turn over because the model cannot
  44. distinguish the residual heat on bed and the person as Figure~\ref{fig:residual_heat} shown. This
  45. situation will slowly disappear after one or two minutes.
  46. To determination the pose, first we use a median filter with a window size of five
  47. to filter out the noise. Then, find the curve hull line of the upper bound and
  48. lower bound of the data. Finally, calculate the middle line of upper bound and
  49. lower bound, and regrad it as the trend of the pose changing. Figure~\ref{fig:trend}
  50. shows the filitered data and these lines.
  51. We divide every data into 10 second time windows. If the middle line of the time window
  52. is at the top one fifth, or the trend is going up, it is a lay on back.
  53. Otherwise, it is a lay on side. If there are three
  54. continuously same poses, and different from the last turning over, it will be count as
  55. another turning over.
  56. \begin{figure}[tbp]
  57. \centering
  58. \minipage{0.25\columnwidth}
  59. \includegraphics[width=\linewidth]{figures/Lepton_residual_heat.png}
  60. \caption{Residual heat on bed.}
  61. \label{fig:residual_heat}
  62. \endminipage
  63. \minipage{0.55\columnwidth}
  64. \includegraphics[width=\linewidth]{figures/MinMax_2.pdf}
  65. \caption{Trend of pose.}
  66. \label{fig:trend}
  67. \endminipage
  68. \end{figure}