Aiming at the problems that local optima are easily trapped in and low estimation accuracy is achieved in the pose estimation of human non-rigid body motion, an accurate human motion pose estimation method based on improved machine learning was proposed. Three-dimensional coordinates and multi-directional key points of the human body were adopted as core features to characterize human motion postures, and high-level estimation and recognition of human postures were realized. The weights and biases of each layer of the convolutional neural network in machine learning were optimized via the particle swarm optimization algorithm, and an immune mechanism was introduced to prevent the model from falling into local optima.With the acquired optimal weights and biases, the convolutional neural network was modified. The features of human motion postures were fed into the improved network, and accurate estimation results were obtained. The experimental results show that the features extracted by the proposed method are representative and reliable, and the key points of human motion pose estimation can comprehensively cover the pose image. The accuracy of human motion pose estimation is high, and the practical application effect is good.
CaiMin-min, HuangJi-feng, LinXiao, et al.Motion frame acquisition method based on human pose estimation and clustering[J]. Journal of Graphics, 2022, 43(1): 44-52.
ZhuLi-ping, TangLiang, ZhuKai-jie, et al. Multi-objective human pose estimation in Motion Scenario [J]. Computer Engineering and Design, 2023,44(7):2156-2162.
LvHeng, YangHong-yu. A 3D human pose estimation method based on spatiotemporal motion information interaction modeling[J].Journal of Computer Graphics, 2024, 45(1): 159-168.
MaXiao, YanYu-dong. Human pose estimation in basketball scene based on multi-scale temporal and spatial characteristics[J]. Journal of Central South University for Nationalities (Natural Science Edition), 2023, 42(1): 95-102.
[11]
CuiH, DahnounN.Real-time short-range human posture estimation using mmwave radars and neural networks[J].IEEE Sensors Journal, 2022, 22(1):535-543.
ChenYang, LiuQin-ming, LiangYao-xu. Research on machine learning strategies based on small sample unbalanced device data[J]. Journal of University of Shanghai for Science and Technology, 2022, 44(4):407-416.
LiuYi-song, GaoHan-lu, CaiKai-xiang.3D human motion prediction based on spatio-temporal graph convolutional network and non-autoregressive model[J]. Applied Research of Computers,2024, 41(3): 956-960.
ZhaoWei, LiYi. Human pose estimation optimization and animation generation based on Kinect[J].Journal of Computer Applications, 2022, 42(9): 2830-2837.
WuDa-fei, YangGuang-yong, FanKang-sheng, et al. Improved particle swarm optimization algorithm with multi-strategy fusion[J].Application Research of Computers, 2022, 39(11): 3358-3364.
LiZhou-zi, FengYue, LinZhuo-sheng, et al.Face recognition optimization of SVM model based on improved PSO algorithm[J]. Computer Simulation, 2022, 39(4): 241-247.
GuoXin, LiLi-jun. Trajectory planning of robotic arm based on artificial immunity-improved particle swarm optimization algorithm[J].Journal of Mechanical Transmission, 2024, 48(5): 33-40.