Research On Human Daily Activity Recognition System Based On Ensemble Learning

Sports are very important for the health of human body, and exercise is an important means to prevent disease. Therefore, the system of scientific and effective human body activities analysis has the vital significance to the national health care. However, the lack of sports scientific guidance and feedback, the movement effect discounted sports injury worse. With the advent of the accelerometer, gyroscope and other portable sensor technology, the health system based on wearable body sensor network has received extensive attention in recent years. It can be said, wearable computing and human motion analysis has come into our lives, and it will become an important technical means for the future of human health and disease prevention.The accurate measurement of human physical activities recognition by multi sensor fusion was realized and it focuses on the integrated learning machine learning method in multi-sensor for human motion analysis theoretical innovation and application, so as to provide scientific guidance for human health. The following works were done in this study:1. A wearable network for real-time acquisition of sensor signals in human motion is set up. The main sensing device includes three axis acceleration and three axis gyro digital sensor which are worn on different body position of human body. In addition, in order to improve the energy utilization rate of wearable health monitoring system, the method of sensor information collection and data transmission is improved in this paper based on Bluetooth 4 technology. And the effectiveness of the proposed method is verified by the experimental results of CC2541 and nRF51822.2. The theoretical basis of integrated learning and the principle of random forest algorithm are analyzed, and the two methods are integrated in this paper to propose an integrated learning algorithm based on the relevant information of the sensor. The effectiveness of the proposed method is verified by using the open human body motion information database. The experimental results show that the proposed algorithm has a good recognition performance.3. The recognition performance of the traditional static recognition algorithm will become worse due to the difference between individuals and the change of individual behavior. In order to solve this problem, this paper proposed the optimization algorithm based on incremental learning and the related experiment were carried out on different individuals. The experimental results show that this algorithm effectively acquires good effects.In general, combined with wearable technology, signal processing and wireless communication technology, a wearable health monitoring system based on the integrated learning of human daily activities recognition was developed. Finally, it proves the reliability and validity of the system by the corresponding software and hardware experiment.