CART Decision Tree Based Human State Estimation Algorithm and Research

Human state estimation is an important technical tool in the field of elderly mobility assistance. In order to further improve the accuracy of human state recognition, a CART decision tree based human state classification algorithm was proposed for the auxiliary support system. Firstly, the human inertial sensor posture data was pre-processed, the human body link model was used to calculate the human body center of gravity for posture detection of sitting and standing motion process, extract features and training samples, then the dimensionality of the posture data was reduced by PCA method to get the best subset of data with good reflection of human body posture, the motion process includes four consecutive states: sitting, rising, standing and falling, and the state transition stages were subdivided into another four categories, and finally the CART decision tree algorithm was used to train the samples to generate CART rule trees to classify the human posture. The experimental results show that the classification accuracy of the human posture detection system under the constructed CART decision tree is up to 98%, which is better than the SVM classification results, and the estimation is correct in the vicinity of the state transition, and the accuracy of the state estimation is reasonable.