Research Of Human Motion Recognition Algorithm And Application Based On Multi-sensor

As society changes in family structure and the number of aging population, the "empty nest" become a common phenomenon that the number of elderly people who live alone increases and old people’s lives have undergone great changes from depending on their children to take care of previously to automatic model now.Therefore, automatic monitoring of the elderly daily life is important for the elderly and benefit to social stability.Monitoring of human movement generally brings large amount of calculation and a lot of energy consumption, These make it difficult to move to portable terminals and systems require frequent replacement of the battery or charging system problems, This paper came up with a human motion recognition algorithm based on the two-tier decision tree classifier,and based on the two-tier decision tree algorithm, we design a complete fall detection system, the realization of the elderly fall detection identification.The main work of this paper is concluded as follows:1. Analysis and summary of the existing theory and research method based on sensor for human motion recognition. And then the data preprocessing methods, feature extraction and selection of technology, classification and recognition of the recognizer and short distance wireless communication technology are detailed analysis.2. This thesis proposed a human action recognition algorithm based on the two-tier decision tree, according to the characteristics of human motion and the motion characteristics of the data. In order to reduce the system energy consumption and the complexity of the algorithm, the proposed method reduce the extraction and processing of acceleration data. The human motion recognition algorithm use Kalman filter and the length of the semi-2s overlapping windows to process the data. The result shows that the recognition rate of the system is improved and the complexity of the algorithm is reduced.3. This paper collected eight motions data of 20 experimenters by using MPU6050 sensor module and wireless communication module Bluetooth to train and validate the accuracy of the recognizer. The results show that the accuracy of recognition based on the two level decision tree is 98.44% and 94.16% respectively. By compared with other classifier, the performance of all aspects are better.4. Design a complete set of the elderly fall detection alarm system. The specific fall detection algorithm is based on the two-tier decision tree classifier. And the experimental data show that the accurate recognition rate of the fall detection system of is up to 95%. It also shows the fall detection system has certain market value.