The Elderly Fall State Detection System Based On MEMS Sensor Research And Design

Nowadays our country has entered the aging stage, the average life expectancy for people with the improvement of quality of life and gradual growth, growth in the number of "empty nest" families are also a lot of, to solve these problems, the elderly health and safety issue has become one of today's society's priorities. According to the problems related to data according to the survey, each year because of the fall cause health problems to hospital has more than 50%, falling injury caused 80%of all need hospital treatment, the global each year, more than ten thousand people lost their lives because of the fall can not get timely treatment. The above data can see, if not timely rescue after the fall, the injury will deteriorate further, and even affect the life risk. In this paper, according to the present at this stage the elderly health problems, solve the problem after a fall can get timely treatment study a fall detection method based on wearable devices, and use SIM908 was done after the fall in wireless positioning and remote alarm, make fall injury of old people get timely treatment.This article first daily actions to the human body acceleration signal is analyzed, the eigenvalue able to distinguish between different motion state is obtained. Then, summarized the fall detection algorithm on the existing now and study, from the aspects of practicability, accuracy, real-time, comparison, extraction based on wearable sensors using decision tree classification method.System user the acceleration sensor based on MEMS technology ADXL345 add speed signal acquisition human movement process, and will get the signal acquisition to STM32F103ZET6 microprocessors, through the I2C bus transmission by arithmetic mean filter algorithm to remove noise by an accidental factors, and then the microprocessor using decision tree classification algorithm to determine analysis, for the fall, if the test result is to use GPS positioning module, then fall information to guardian or medical rescue center through GPRS remote alarm.This system adopts the algorithm based on decision tree ID3 classification accuracy rate fall down test results in addition to forward was 90%, the rest of the fall detection accuracy is over 92.8%, and improve the existing algorithm is characteristic of a single, real-time performance and disadvantages such as leakage by guardian privacy and no strict requirement for gear wear position, acceleration sensor and the body into at any Angle. Also through the communication and alarm module debugging, realized in the ward inspection for fall, communication module to complete the alarm operation.