A Proposal for Markerless Gait Analysis Based on 3D Points Cloud

This paper proposes a solution for markerless gait analysis based on point cloud data, which are three-dimensional data obtained by sensors such as LiDAR, that can capture body movements in real-time and in various environments. Gait analysis aims to evaluate the differences in symmetry between the left and right sides of the body, which can indicate diseases such as Parkinson’s and Alzheimer’s. The point cloud is a sparse matrix with N points in 3D space and the intensity of reflected light, which does not have an ordering between the points and the minimum distance between them is responsible for the density of the cloud. In the proposed method the point cloud is transformed into voxels, which are cubic units that facilitate data processing, and each set of points in the voxel is used for person detection. The points of the person are used to segment the body parts, and each part is used to estimate the joint position. The joint position is used to construct a rigid body model, where a person is represented by their segments. The rotations and translations of the segment over time are used to build the gait analysis graph. By comparing the rotation of each segment over time, it is possible to calculate the difference between the left and right sides and detect at what moment of the movement the difference is large enough to indicate the onset of pathology.