Skip to main content
. 2022 Jun 2;22(11):4245. doi: 10.3390/s22114245

Figure 1.

Figure 1

Overview of the proposed pipeline for the quantitative assessment of gait from freehand 2D camera recordings. From left to right: S1: Configure the camera and its settings. S2: Collect data with the camera. S3: Extract keypoints using Alphapose. S4: Match the skeleton using PoseFlow. S5: Normalize the skeleton data with one of methods given in Section 2.4. (BoN: box normaliztion; S: shoulder normaliztion; H: hip normalization; LR/RH: left-shoulder right-hip normalization; LS/RH-d: left-shoulder right-hip distance normalization; MS/MH-d: mid-shoulder mid-hip distance normalization; ASH: average shoulder hip normalization). S6: Obtain the normalized keypoint skeleton sequences. S7: Analyze the (distance-based) features derived from skeleton data for classification.