Research On Fall Monitoring System Based On Intelligent Mobile Computing

With the intensification of Chinese aging process,the problem of fall as the most serious accidental injury leading to the death of the old people is more and more serious.How to use information technology to reduce the risk of accidental fall and the fall injury has been one of the most important problems to be solved urgently in the community.In this thesis,we focus on three problems of fall risk assessment and early warning,fall detection and prediction of fall injury.And we design the prototype of fall monitoring system based on intelligent mobile computing.The specific research work of ours is as follows:1.On the risk assessment of fall,a matched case-control method is used to extract risk factors for fall risk assessment,and we get ten factors such as fall history,aid or support,medical diagnosis,mental state,gait,vertigo,drug use,poor eyesight,excretion and hypertension as the main risk factors.At the same time,logistic regression is applied to build the risk assessment model.And the DBSCAN(Density Based Spatial Clustering of Applications with Noise)algorithm is used to cluster the historical information of falling,which is used to assess and predict the fall intensity in the area that users are located.2.In this thesis,we use the sliding window algorithm to collect the triaxial acceleration time series and extract time-domain characteristics and frequency domain characteristics of three axis acceleration.And we use the convolutional neural network(CNN)to classify the fall data combined with the long and short term memory network(LSTM).With the comparison of the decision tree and k-Nearest-Neighbor algorithm and support vector machine and the naive bayesian algorithm,CNN-LSTM is selected in fall detection,which the recognition accuracy is 98%,98% and 98% for forward fall,backward fall and lateral fall.The dynamic time warping is used in this thesis to recognize the direction of fall,which is combined with the overturned angle of phone.And the accuracy of the dynamic time is 93.33%?90%?86.66% for forward fall,backward fall and lateral fall.At the same time,we design the experiment of thin film sensor to detect the force of different parts of the body under different fall directions,which is aimed to recognize the injuries of fall lately.4.In this thesis,we use smart phone to collect three axis acceleration sensor signal,and we designed a prototype of fall monitoring system based on intelligent mobile computing.With the help of location,network link and other functions of smart phone,a complete flow of fall warning,detection and rescue is implemented on the prototype,which helps users get help in time when they fall.