Toward study of features associated with natural sleep posture using a depth sensor
Xuhong Liu, Shahram Payandeh · 2016
Estimating the sleeping posture of a person under natural resting environment is a very complex problem. In this study, a new non-invasive framework is proposed for exploring features defining sleep postures of a person. Three dimensional depth scans as well as the cross-sectional scans of the static sleeping body are developed using the measured depth data. These features can then be used in order to estimate the changes in patterns associated with the head, limbs and torso position in depth images. Scans of the sleeping postures are taken at various cross sectional directions. 1D and 2D Fast Fourier Transformation are explored in order to provide complementary feature of the scans. We also generate grids on the horizontal sleeping plane. In the depth image, each grid has a quadrangle configuration which can further be divided into two triangular meshes. We calculated a surface normal vector to each of the mesh. Given a 2D scan strip perpendicular for the sleeping plane, variations of the normal along such scan strips can be used as other local features. These selected features can be further integrated as a part of natural sleep posture estimation methods.