An Efficient Approach for Patterns of Oriented Motion Flow Facial Expression Recognition from Depth Video

Abstract

Patterns of directed motion flow (POMF)  from optical flow data is a novel feature illustration method  that we have a tendency to propose in this paper to  recognize the correct facial expression from facial  video.The POMF encodes the directional flow data with  increased native texture small patterns and computes  completely distinct directional motion data.It demonstrates  its ability to recognize facial data by capturing the spatial  and temporal changes caused by facial movements through  optical flow and allowing it to examine both domestic and  foreign structures.Finally, the hidden Markoff model  (HMM) is trained on the expression model using the  POMF bar graph.The objective sequences are generated by  using the K-means agglomeration method to create a  codebook in order to instruct through the HMM. Over  RGB and depth camera-based video, the projected  technique's performance has been evaluated. The results of  the experiments show that the proposed POMF descriptor  is more effective than other promising approaches at  extracting facial information and has a higher classification  rate.