The Study Of The Human Fall Detection Algorithm Based On Sensor

Fall is one of the main factors affect human body health, in recent years, many domestic and foreign scholars and research institutions in the related fields of fall detection make a study. This paper analyzes the movement characteristics of fall behavior and studies the effective fall detection algorithm, then developes the fall detection application based on Android operating system. The specific researches of this paper include the following aspects:(1) Data analysis and feature extraction of fall behavior. The smart device is fixed in right wrist of the human body and data comes from the motion sensor which were build in the smart device; In order to obtain valid data, the original data is preprocessed by the third moving average filter. Then the sensitive characteristics to fall behavior is studied, the motion amplitude, tilt level, rotation degree and other characteristics are extracted during the human movements, and the influence of those characteristics to fall behavior is analyzed.(2) Fall detection algorithm based on PSO-SVM. The Support Vector Machine classifier which has a good effect of binary classification is used to detect fall behavior, then the Particle Swarm Optimization is used to obtain the best parameters of SVM. Then the test data is analyszed according to the constructed SVM classification model. The simulation results have shown that the algorithm has a good classification performance. But the complexity of this method is high, the real-time performance is poor.(3) Fall detection algorithm based on DTW and D-S evidence theory. The Dynamic Time Warping is adopted to preliminary judge the human behavior respectively based on motion amplitude, tilt level, rotation degree characteristics. In order to obtain the best effect of classification, the D-S evidence theory is adopted to fusion the preliminary judgement of three characteristics, then making a final decision of human movement pattern, thus improving the accuracy of fall detection. The simulation results have verified the effectiveness of the algorithm.(4) Human fall detection system based on Android platform. The fall detection system is developed based on DTW and D-S evidence theory in the Android operating system. This system has three modules including initial setup, fall detection, warning and help. The performance of this system is analyzed in different brands of mobile phone. The experimental results have shown that this system can operate stably with high-speed on the smart devices,and the accuracy of this system is almost same with simulation results. This system has good detection performance.