Research On Human Fall Prediction System Based On BP Neural Network

As the problem of aging population in our country becomes more and more serious,the dependency ratio increases,and the empty nest phenomenon of the elderly becomes more and more common.Under the conditions of saving manpower and material resources,it is more and more important to protect the health of the elderly more efficiently.Falling is the most common sudden accident in the life of the elderly,and it can cause serious physical and mental damage to the elderly.Therefore,the research of fall protection equipment for the elderly is of great significance.This paper proposes a method for predicting human fall based on BP neural network,and implements an embedded system that can predict human fall as the research content.The waist acceleration and angular velocity data of the human body in different states are collected through experiments,the feature vectors extracted from the experimental data are used as data samples to train the BP neural network,and the weights and bias parameters of the BP neural network after the training are extracted.An embedded system is designed,and the forward propagation of the BP neural network is realized on the embedded system according to the extracted neural network parameters,and the physical construction of the system is completed.The real-time system is verified by simulation and fall experiments.The reliability of sex and fall prediction functions.The main research contents are as follows:1.A data acquisition device is designed.The data acquisition device is worn on the waist of the human body to simulate the daily activities and falling process of the elderly.Through experiments,the data on the fall process of the human body and the waist acceleration and angular velocity of daily life are collected.Analyze and study the human body's fall process and the law of waist acceleration and angular velocity changes in daily life,and extract characteristic data from it.Specify the feature vector format input to the BP neural network,calculate each group of feature vectors and add labels to each group of feature vectors.2.Draw up the initial structure of the BP neural network,create a BP neural network model on the MATLAB platform,train and test the designed BP neural network model using feature vectors and labels as training samples,and find the most reasonable through continuous experimentation and comparison BP neural network structure.After the training of the BP neural network is completed,the weights and biases of each neuron of the neural network are extracted.3.Used MATLAB to calculate the second-order polynomial piecewise fitting table of the hidden layer activation function(Sigmoid)and the output layer activation function(Softmax).The FPGA-based hardware method is selected to implement the BP network,and the Verilog HDL language is used to design the digital circuit of the trained BP neural network forward propagation.The simulation results prove that the hardware BP network can meet the real-time requirements and accuracy requirements.4.Design the overall structure of the embedded system and the internal structure of the FPGA chip,and design the system's low-power operation mechanism at the software level.The hardware circuit and PCB circuit board of the system were designed using Altium Designer software,and the physical production of the system was completed.The compiled program is programmed into FPGA and MCU,and the reliability of the system function is verified through experiments.This research verified the theory of human fall trend recognition based on BP neural network through computer simulations and experiments,and designs a fall prediction system.The forward propagation of BP neural network is realized on the FPGA chip of the system,and the BP neural network is completed.The hardware implementation of the network,experiments and simulation results show that this method can accurately identify the human body's falling trend,and can send out an early warning signal before the impact of a fall.The human fall prediction embedded system designed in this paper has important reference significance for the research in the field of fall protection for the elderly.