The Research And Application Of Human Fall Detection Technology Based On Smartphone

Along with the speed of social aging and the evolution of family structure, the number of "empty nest" are increasing in china. There are about18,240,000old people live in single family. By2030, the population of the elderly are nearly300million, and the proportion of empty families will be reach to90%. The elderly’s health problems are becoming more and more caused by "aging empty nest ". The security of the elderly has become the focus of international attention. Fall is an important factor that affect the elderly’s daily physical and mental health in our country. Fall is the fourth reason in injury death and is the first reason in the above65years old people. Fall could cause a lot of the elderly’s disabilities and affects the quality of the elderly’s daily life. Injuries and medical costs caused by the fall are connect with not only the impact of body when subjected to the fall but also the time length of the aid. Therefore, the occurrence of the elderly’s fall can be real-time detected and promptly issued a distress signal after the fall to minimize serious injury to the body caused by a fall have a great significance.There are many kinds of the human fall detection system’s research technology. The video image analysis and the wearable device detection are the most common methods. The video image analysis system is installed in a room to capture several images of human motion and the image processing algorithm can determine whether a fall occurs. However, the image analysis algorithms are often complex and the cost of system is high. The system only can detect a certain area and would likely to expose the user’s personal privacy due to the acquisition of the image information. The wearable devices fall detection system use the micro sensors carried or worn on the human body to perform fall detection. The system can monitor human activities real-time, and the algorithm can determine whether a fall occurs when the motion parameters of the body changed. The body will have a certain impact on activity because of the sensor device requires the user to wear long, and the maximum distance between the system and the base station is limited are the disadvantages of this system. So the fall detection device can’t be promoted a wide range of use.In recent years, the development of mobile technology and the motion sensors’ miniaturization are greatly promoting the application of fall detection technology which based on mobile phones. Currently, the number of smartphone users are increasing. Smartphone as a part of travel essential tool is indispensable in daily life. Smartphone has call function basically, and itself embedded in a large number of sensors. Such as accelerometers, orientation sensor, digital compass, GPS, gyroscope and distance sensors. Smartphone combine with a fall detection system two important components:fall detection and rescue communications. Use the smartphone as a fall detection system reduce system cost because of allows users to use their existing mobile devices, and GPS built-in cellphone can determine the location of a fall.Use the cellphone built-in sensor to determine fall detection technology has made little progress in domestic so far, and foreign scholars have proposed use the mobile phones embedded in the three-axis acceleration sensor to get the body’s acceleration signal. They use the fall acceleration amplitude and inclination to judge a fall. The alarm can be sent directly through the mobile phone after the fall. However, data collected only by a separate acceleration sensor can’t reflect the body posture change fully. Such as sit down quickly, jogging and some human activities can also produce a higher intensity value characteristic of a similar high acceleration. So fall detection has many false alarms. There existing algorithm use the inclination to determine a fall that only applies to mobile phone fixed on the trunk of users. But it does not meet the mobile phone carried in the usual way in people’s daily life. Based on the above research background, this paper provides a solution to use smartphone as a fall detection system to detect fall events. The smartphone built-in accelerometer and gyroscope sensors obtain real-time information of human motion posture, and use fall detection algorithm to determine the occurrence of a fall event and send intelligent alarm finally. The main work is divided into four parts:falling process analysis and human movement model, fall detection algorithm design, system software and algorithms evaluated.1. Falling process analysis and human movement modelThere are three phases in the process of a fall:weightlessness, collision, and stationary. The body acceleration and angular velocity would have taken place great changes. Three-dimensional Cartesian coordinate system can establish a human torso. Use a torso position as the coordinate origin, and the body’s coordinate system’s three axis-direction are the front of the trunk, the left of trunk and the right of trunk. Any direction in the process of human movement acceleration and angular velocity vector can be broken down into three component on the axis-direction.2. Fall detection algorithm designThe purpose of the fall detection algorithm is to distinguish the process of a fall and daily life activities. Mobile phone is placed in chest pocket of human body in the design of experiment to collect data samples and extract characteristics as the basis. Use the matlab processing experimental data, and we get the acceleration characteristics, the angular velocity, and the similarity characteristics. Use the similarity characteristics to detect a fall in this paper for the first time.The process of algorithm is divided into four partes:(1) Smartphone collect triaxial acceleration information data and angular velocity information data, and save the result data. Calculate the total value of acceleration data and compare the value with acceleration threshold. If the acceleration value is greater than the threshold, the detection will go to the next stage.(2) Calculate the total value of the angular velocity data at the same time and compare the value with angular velocity threshold. If the angular velocity value is greater than the threshold, the detection will go to the next stage.(3) Set the timer to wait for the human body posture stability. Then calculate the similarity values and compare the value with similarity threshold. If the acceleration value is greater than the threshold, the algorithm suspect a fall occur.(4) When algorithm detects a suspected stumble, the system will go to the intelligent alarm mode. User can cancel this alarm in a period time in this mode. If user don’t have any operation, the system will send an emergency alarm information and make a phone call. If any stage in the first three stages characteristic value is less than the threshold, the algorithm will be immediately interrupted and return to the first stage to continue a fall detection.3. The algorithm evaluationTo verify the effectiveness of the fall detection algorithm in this paper, designs fall group and the comparison group of non-fall experiments. And we use the sensitivity, specificity and accuracy assessment indicators to evaluate the performance of the algorithm. The algorithm has a sensitivity of88%, specificity of92%and accuracy of90%, and the experimental results show that the algorithm is an effective fall detection algorithm.4. The system softwareThe fall detection system in this paper is based on the smartphone platform application software development. Select the android operating system for application software development platform. The function module design of software mainly divided into the data management module and alarm processing module. The data management module includes data collection capabilities, data storage, real-time data processing functions. Save the acceleration data and angular velocity data to a SQLite database created by application, and the algorithm processing the data real-time until the algorithm detects a fall.Alarm processing module is intelligent alarm mode processing parts, including start alarm cancel mechanism, start the GPS module to obtain location information, send an emergency message and make a phone call.This paper study technology of using smartphone to detect a fall and mainly monitoring the elderly who living alone in daily life. If detect a fall event or an accident, immediately notify family members and health care workers to improve aid effectiveness and quality of life. Study the elderly fall detection has important social significance, and the fall detection technology research based on smartphone has broad application prospects.In this paper, the research and innovation were done in the following areas:(1) Analysis the human movement and extraction a new feature for fall detection;(2) Designed the fall detection algorithm based on smartphone;(3) The experiment was designed to evaluate the effectiveness of the algorithm;(4) Completed a fall detection system on Android software.