Research Article

Application of Human Posture Recognition Based on the Convolutional Neural Network in Physical Training Guidance

Table 1

PCKh comparison results on the MPII dataset.

MethodHeadShoulderElbowWristButtockKneeAnklePCKh

Iterative error feedback95.591.581.672.382.673.066.379.5
Simple stacked hourglass network model98.196.091.186.990.187.283.490.1
Network model based on improved stacked hourglass98.396.291.788.090.587.884.990.7
Joint subset partition and labeling94.090.083.377.282.575.668.581.6
Convolutional pose machines97.694.988.583.888.182.579.187.8
Model of this study97.796.090.686.089.786.382.689.7