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.