A Method of Learner's Sitting Posture Recognition Based on Depth Image
Xing Zeng, Sun Bei, Wang Enlong, Wusheng Luo, Taocheng Liu · 2017
Real-time detection of learner's sitting posture not only helps prevent myopia in time but also promotes the improvement of learning efficiency.However, most of the current sitting detection methods have the shortcomings of low detection variety and recognition rate, et al.Based on this, a sitting posture detection method based on Cartesian plane projection is proposed.The sitting depth images are projected into three Cartesian planes respectively.The background removal, interpolation scaling and normalization are performed for each projection map.The projection feature is obtained and the PCA is used to reduce the dimension of the feature.Finally, the projection feature and the front view HOG feature are fused to generate the new posture feature vector.In the experiment we collected 20 people, each person 14 kinds of sitting posture to form test database and the use of random forest to classify the extracted sitting posture characteristics.The experimental results show that this method can effectively detect the learner's sitting posture and it is superior to the existing method in recognition accuracy and recognition speed.