The Study And Implementation Of The The Fall Detection System Based On Sensor

Falls have brought great physical and psychological harm to the elderly, the research of the body fall detection have become a hot spot in pattern recognition domain in recent years in the domestic and overseas. In this paper, the body fall detection methods have been studied in order to conduct fall detection effectively and timely. Fall detection application software on the Android mobile and body fall threshold detection hardware and software prototype system have been developed. The main work is as follows:(1) Study machine learning algorithms of human body fall detection. In view of the fall detection binary classification problems, firstly build a mathematical model of human spatial coordinates, by selecting sensitive time and frequency domain characteristics to fall such as acceleration amplitude, energy, and poor standards to constitute a multi-dimensional feature space, and then use the singular value decomposition(SVD) to reduce the dimensionality and reconstruct the feature space in order to optimize the feature space, finally use Support Vector Machine(SVM) to classify fall behavior based on fall characteristics.(2) Study threshold algorithm of human body fall detection.In order to overcome the excessive consumption and low real-time performance of machine learning method, further fall detection threshold method was studied. According to the amplitude range’s analysis of experimental data in the magnitude area, magnitude acceleration and inclination angle,the multiple thresholds and posture rules have been built to conduct human fall detection.(3)Develop body fall detection application software on the Android mobile. Use the good interactive mode, portability and location ability of the Android platform to develop the real-time mobile application software of human fall detection, the alarming information will be sent with the user’s position obtained when falling is detected.(4) Develop body fall threshold detection hardware and software prototype system. The systerm hardware module is composed with micro acceleration sensor, Fujitsu microprocessor and micro wireless communication module. Software module is completed with embedded C language to inplement acceleration data acquisition, the I/0 control, A/D conversion, data storage, UART asynchronous serial communication, and alarming SMS will be sent with GPRS.The results of simulation and experiments show that the singular value decomposition method and SVM classifier can achieve good recognition rate.Besides, in the actual wear experiments, both the fall detection software based on the Android platform and hardware prototype system can effectively distinguish between falls and daily behaviours, and the fall detection algorithm has been verified that is suitable for the elderly.