Anthropometric ratios for lower-body detection based on deep learning and traditional methods

Lower body detection can be used in many applications such as detection of walking patterns, monitoring exercises and knee rehabilitation. Nevertheless, it might be challenging to detect the lower body, especially in various lighting condition and occlusion. This paper presents a novel lower body detection framework using proposed anthropometric ratio comparing deep learning and traditional detection methods. According to the result, the proposed framework as anthropometric-convolutional neural networks (A-CNNs) maintains high accuracy (more than 90%) while the anthropometric-traditional (A-Traditional) techniques for lower body detection achieves a satisfactory performance about 74 % of accuracy. The proposed framework helps to successfully detect the lower body’s accurate boundary under various illumination and occlusion situations for lower limb monitoring.