An Enhanced Posture Prediction-Bayesian Network Algorithm for Sleep Posture Recognition in Wireless Body Area Networks

Algorithm 1

Posture prediction-Bayesian network (PP-BN).

Algorithm

Input  : Video V{f₁, f₂… f_n}generate set of frames from video sequence, Buff₁F, set of input videos,

      Buff₂SL(F), sleep postures match with the general human pose dataset extracted from accelerometer sensor.

Output: Sleep Posture Prediction (Supine, Prone, Right Lateral Recumbent, and Left Lateral Recumbent).

Start  : Read the input video:

     Buff1 = Read (video.avi)

     Ff1, f2….fn          //frame extraction

     # Where Buff₁ is the storage space to store the video frames.

Iterate : a=1: x ;b=1:y

     F_k (a,b)Articulate the frames to 64 x 64 pixel resolution

     H_a,b = f(a,b)

      Calculate the threshold values

          # H is the NULL pulse point

          # G is the 2D- Gaussian Kernel Coefficient.

          # Synchronize the Heartbeat rate with normalized sleep postures

       # Object detection based on P-HOG object detector

        # Update the buffer with noiseless sleep posture frames

        # Estimate the upcoming postures based on the predecessor

end Iterate

       # Use Bayesian Networks to predict the successive postures from the previous positions

Stop