The Design And Implementation Of The Elderly Fall Monitoring Terminal Software Based On Mobile Phone

With the continuous improvement of the degree of social population aging, the frequent falls of the elderly are being more and more obvious. The biggest problem in the elderly fall is not timely relief, but no one can help, unable to contact the guardian and other reasons lead to the occurrence of disability and even death of elderly. Therefore, it is very important to develop a fall monitoring system for real-time monitoring of falling and sending out help information. At present, the main research trend in the field of fall monitoring is based on the wearable fall monitoring system. Fall detection system based on mobile phone is a kind of special wearable system, mobile phone belongs to daily activities, easy to carry and not to forget, comfortable to use, does not affect the user's daily life. Therefore, it is more suitable for elderly to use. Development of smart phone in domestic is relatively late, the fall detection technology based on smart phones is relatively little, if a falling monitoring application software can be developed based on smart phone, elderly safety will get greater protection.This content mainly study fall monitoring software design and implementation based on Android smartphone, monitor human's behavior in real time by using mobile, when human body falls, fall position is accessed and text message is send automatically for help. Firstly, human activity is classified, fall and activity of daily behavior is distinguished, human body coordinates is built, the specific location and sampling frequency of mobile phone wear is determined, acceleration and angular velocity data of human body are obtained by using the acceleration and gyro sensors built in the mobile phone. Secondly, noise signal of the captured data is filtered out by using Kalman filtering, six characteristics of human is extracted to distinguish fall and activity of daily behavior, they are:posture of body angle, minimum free fall, time of weightlessness, energy expenditure, inclination of human body, the similarity. Again, two fall detection algorithm is designed. Threshold of acceleration and angular velocity, amplitude and frequency, similarity arc compared in the first level of algorithm, the suspected fall event is determination Preliminary; six characteristics of the classification is learn by using support vector machine algorithm in the second level of algorithm, the classifier is used to make the final decision on whether fall happened. Weka data mining platform is used to analysis four kinds of commonly used pattern recognition algorithm in the selection of pattern recognition algorithm, support vector machine algorithm is determined in this paper, method of grid search and cross validation is used to find the optimal value of penalty factor and kernel function parameter in the algorithm. Then, the software is divided into 4 modules, data, fall recognition, alarm, parameter setting, respectively. The SQLite is used as database to storage data, the open source libsvm library is used for fall recognition, BaiduMap API is combined with GPS to access location information, system interface is called to send text messages and call the function. Finally, the fall experiment analysis and software test are carried out.Fall experimental results show that the sensitivity of the algorithm is 83%, the specificity is 82.25%, accuracy is 82.5%, AUC is 0.9187, every indicators show that the design of fall detection algorithm is very effective. Through the test of the software, the software function is complete and effective, good stability with strong compatibility.