Design Of Body Inertial Information Collection System And Research On Fall Recognition Algorithm

At present,our country's population aging issue is becoming more prominent.Because of physical decline,the aging people have the high probability of fall,which will bring great physical and psychological hurt to the elderly,seriously interfere with the normal living of the elderly.Thus,the study on fall recognition is has great significance.Based on body fall process analysis,this paper has designed and realized a human body inertial information collection system.And by the human body inertial system,the collection of the body inertial data like acceleration and angular velocity was completed.And this paper has completed the design and the simulation of fall recognition algorithms respectively based on thresholding method,especially completed the design and the simulation and comparison analysis based on Hidden Markov Model.The innovation point of this paper lies in that two characteristic variables that resultant acceleration and resultant angular velocity were taken to build Hidden Markov model,then SVM was used to make classification and recognition for fall and activities of daily living,which improved the sensibility and the specificity of fall recognition compared to the algorithm that only one single characteristic variable like acceleration was taken.The main work of the paper is:(1)In the base of analyzing the difference between the fall and activities of daily living,select the kinematics characteristic variables of acceleration,angular velocity and attitude angle as characteristic parameters for fall recognition,complete the general framework of the body inertial information collection system,respectively complete the fall recognition process design based on threshold value method and HMM.(2)Complete and realize the design of the body inertial information collection system,including the design of sensor node module,base station module and upper computer software platform.Realize the function of collected data storage,wave form display,data processing,parameter configuration and other function of upper computer software data.(3)Design and simulate a fall recognition algorithm base on thresholding method.Experiment was designed and the inertial data of fall and activities of daily living was acquired.By SVM sorting technique,classified the fall and activities of daily living,computed the thresholding value of acceleration and attitude angle,recognized fall based on thresholding values,and discussed the recognition results.(4)Design and simulate a fall recognition algorithm base on thresholding method.By using the time series of body acceleration information and the angular velocity information which is a short time before fall collision,train the characteristics of acceleration and the angular velocity,respectively build HHM.During the recognition,by calculating the posterior probability which is the result of observation series matched HHM,use SVM to sort,then recognized fall from activities of daily living.At last,compare the fall recognition method where HMM was built by only one characteristic variable,analyze the result of the two fall recognition methods primarily.