Research And Implementation Of Smartphone-based Fall Detection For The Elderly

With significant population ageing and the changing of family structure,the elderly living alone is becoming increasingly commonplace.Among the profound consequences,falls are the most worrying events especially when happened but without timely help as they are the leading cause of death,injury and hospital admissions among the elderly population.Thus real-time detection and timely request rescue has become an urgent problem in elderly care.The detection of falls has attracted the attention of researchers.There are also many relevant detecting methods.However;existing researches and solutions have many inconveniences,like requiring fixed placed location and orientation,or multiple devices or low accuracy.To solve these problems,for the hardware part,this paper uses a smart phone as a detection device taking its advantages of powerful process ability,convenient communication interface and rich peripheral sensors that include accelerometer,gyroscope and barometer.This makes no other specific devices needed for the detection.For the algorithm part,a detection method combining threshold judgment and pattern recognition is proposed to monitor the total acceleration level during the fall and altitude variation before and after the fall.Experimental result shows that the detection can be accomplished timely and correctly no matter which pocket and what direction the smart phone is placed.And it comes with an average of 91.67%sensitivity and 88.50%specificity.As a conclusion,this result also proves the effectiveness of the proposed approach.Specifically,this paper mainly involves the following works:1.This paper built a platform to collect the sensing data and store the data on network and locally,which provides a research platform for the algorithm study.2.Through the analysis of the variation and statistic characteristics of the sensing data during the fall,this paper extracted the features of data sets and proposed a detection method combining threshold judgment and pattern recognition.3.According to the comparison of the commonly used classification methods in pattern recognition,the paper figured out a feasible classification method for fall betevior--Decision Tree Model.4.The fall detection system was implemented based on the Android smart phone.Through the evaluation of different functions-communication,locating and fall detection,the performance of the fall detection system was verified.