The Development Of Activity Monitoring System Based On Phone Accelerometer

With the social development and the improvement of living standards, people’s life style has changed enormously. Many disease, such as obesity, diabetes, hypertension and insomnia, have happened more and more common. It has become a serious threat to human health. Thus monitoring and identification of the body’s daily activities has a vital significance to solve the above problems. Human activity recognition based on triaxial acceleration sensor is a emerging research direction in the field of pattern recognition. First, the acceleration signal of human motion is collected by using triaxial accelerometer. Then, we will preprocess the acceleration signal and extract features from the raw acceleration data. Finally, the human activity is classified using the extracted features, in order to reach the purpose of monitoring human activity.The intelligent phone with built-in accelerometer is used in the human activity recognition. And then an activity monitoring prototype system is built to monitor the daily behavior of human.(1) In order to identify the five kinds of daily actions, namely walk, upstairs, downstairs, running and stand, the body acceleration data collected by phone accelerometer is processed by an FIR low-pass filtering algorithm using the function of hamming window according to the specific application environment. The SVM classifier is usedto classify the extracted features. The best solutionwill be choosedto realize the recognition of five actions by comparing three different kinds of time domain feature sets and frequency domain feature sets.(2) A prototype system is implemented based on jQuery mobile framework and struts2 framework,by using the programming techniques including Java, JavaScript and Ajax. It is a Browser/Server style system. By realizing function of thecollection of acceleration signal, uploading and preprocessingthe acceleration signal, features extraction,the extracted featuresclassification, the prototype systemimplementsthe monitor of the daily behavior of human based on mobile accelerometer.(3) The performance of the prototype system is tested by comparing two feature sets and three wear positions for three different models of intelligent phone, the time domain feature set includes standard deviation, skewness, correlation coefficient, the first quartile and frequency domain feature set includes 64 pre-set FFT coefficients. The experimental results show that the prototype system is feasible and effective.