Human posture estimation and correction based on the CPM and the Pearson correlation coefficient
Xin Liu, HongYuJie Xiao, Jia Cheng · 2021
This paper proposes a mechanism for estimating and correcting fitness posture based on deep learning. First, use the Convolutional Pose Machine (CPM) to estimate the human body posture from the collected human motion images. Then 14 key bone points of the human body can be obtained after correction, and the pixel positions of these 14 bone points are used as the criterion for judging the motion posture. Subsequently, citing the Pearson correlation coefficient as a correlation basis, compared Pearson correlation coefficients of each area of the standard posture with the posture to be measured, then get the weighted sum of the comparison results of each area which is used as a criterion for judging whether the human body posture is standard. Experimental results show that this mechanism has the accuracy of high degree.