Skip to main content
. Author manuscript; available in PMC: 2018 Mar 5.
Published in final edited form as: IEEE J Biomed Health Inform. 2017 Mar 2;22(2):525–536. doi: 10.1109/JBHI.2017.2676878

Fig. 4.

Fig. 4

Overview of the proposed approach: (a) Tracking of facial landmarks and head pose, (b) per-frame encoding through Stacked Denoising Autoencoders, (c) extraction of per-frame dynamics (amplitude, velocity, and acceleration), and (d) per-video encoding through Improved Fisher Vector coding or Compact Dynamic Feature Set.