Aiming at problem that falling down movement of the elderly is very complex and falling down label is difficult to achieve

a stumbling and falling system based on Tri-training semi-supervised algorithm is proposed.The system uses the 3D acceleration sensor to collect movement accelerating data

and extraction of feature is done on the data and partial sample is labeled. Tri-training algorithm is used to train the classifier in the next step.Trained classifier is served to recognize stumbling and falling. The sensor of data acquisition is designed as a wearable device in particular and the server uses Java to write a program for data processing and analyzing. The test results show that the proposed method can effectively improve the accuracy of the recognition by applying a large number of unlabeled data. The experimental results demonstrate that the system can meet the needs of the elderly in their daily lives

and some unexpected falls are able to give timely detection and alarm.