Abstract: In order to reduce the falls of the elderly, the kinematics data from 54 elderly people were collected in the laboratory. Taking the thoracic spine, knee, scapula and pelvis as the research objects, and the average displacement of joint points in sagittal plane, coronal plane and cross-section as the feature dimension, the prediction model was constructed, and the SVM algorithm was applied to indentify and predict elderly people who are likely to fall. By comparing the data, the minimum dimension that can achieve higher prediction accuracy was obtained. The results show that the prediction accuracy of the proposed model is 87.5% when the parameters of SVM are optimized by particle swarm optimization(PSO)and genetic algorithm(GA), and that the same prediction accuracy can be achieved by establishing three dimensions through pelvic position.

Key words: elderly, average displacement, support vector machine(SVM), fall, prediction model