Human posture recognition based on projection histogram and Support Vector Machine
Abderrazak Iazzi, Mohammed Rziza, Rachid Oulad Haj Thami · 2018
In this paper, we propose a human posture recognition based on the human shape. We represent human shape by using the projection histogram based on the bounding box, where we divide it horizontally and vertically in several lines oriented by several angles, and the intersection between them provides local features as a shape descriptor. By using the Support-Vector Machine (SVM) classifier, we map each histogram to one type of postures including lying, standing, bending and siting posture. We compare our method with two shape descriptors such as Shape Context(SC) and Ellipse-based projection histogram. To show the performance of our method, we based on two datasets and the results present that our method achieves a high accuracy in human posture recognition.