Research And Implementation Of Context Awareness Based Accident Prevention During Mobile Phone Usage

Context awareness is originated from the pervasive computing field.Context awareness was first proposed by Schilit et al.[1] in 1994.Since context awareness was proposed,a large number of researches and applications based on context-aware technologies have appeared.In recent years,more and more researchers dedicate to combine context-aware technologies with hot technologies in other fields to solve the key issues which are closely related with people's daily lives,such as driverlessness,and so on.At the same time,with the development of social networks,mobile internet and other technologies,people are increasingly dependent on mobile phone[2].However,excessive reliance on smart phones is causing a considerable part of the population to face potential personal accidents caused by the use of mobile phones.Based on the above considerations,this paper uses contextaware technology to research accident prevention methods while using mobile phone and implement a corresponding system.The research about accident prevention methods while using mobile phone,which is based on context awareness,is a high-level issue.This paper divides the research on accident prevention methods while using mobile phone into three sub-problems: human behavior perception,spatial position perception and interactive perception.The human behavior perception sub-problem refers to the type of behavior which is perceived by mobile phone holders,is different from problem of traditional human behavior identification.This paper focuses on the type of behavior by mobile phone holders in different mobile phone holding modes.The spatial location perception sub-problem refers to sensing whether the mobile phone holder is in a dangerous area where accidental injury may occur.The interactive perception sub-problem refers to sensing the interaction ways between the mobile phone holder and the mobile phone to determine whether the mobile phone holder is in the process of using the mobile phone.Research on accident prevention methods while using mobile phone will be based on the output of multiple sub-issues.For human behavior perception sub-problems,experiments were conducted based on random forest and neural network.After comparing the experimental results,a random forest model was chosen to solve the sub-problems of human behavior perception.The data used to solve the human behavior perception sub-problems are collected from builtin sensors of smart phone.This article uses the Nexus 5X smart phone as the data source.Each row contains eight attributes including three-axis angular acceleration,three-axis acceleration,longitude,latitude and so on.A total of 400,000 behavior data including behaviors such as walking,biking and driving a car are collected under different mobile phone holding modes(mobile phone position,screen orientation,head orientation,etc.).These data include 32 sorting tags.All the behavioral data were denoised in the data collection stage,which guaranteeing the quality and high availability of the data.Based on the collected behavioral data,the random forest model is trained,optimized,analyzed and evaluated.The final model has a classification accuracy rate of about 96% for 32 labels,showing good classification effect and high availability.For the spatial position perception sub-problem,this paper conducts experiments based on the spatial position trajectory of the mobile phone holder and the road network information.On the one hand,it concerns about the absolute distance between the current location of the mobile phone holder and a road section in the road network.On the other hand,it focuses on the matching degree between the spatial position trajectory of the mobile phone holder over a period of time and a road segment in the road network.Using the weighted output of two result to judge whether the mobile phone holder is located in areas which may have unexpected injury.For the interactive mode perception sub-problem,this paper uses the API provided by the smart phone operating system to monitor the operation of the mobile phone holder's touch screen and other operations.Then using the result of monitoring to judge whether or not the mobile phone holder is interacting with the mobile phone and in what manner.Based on the output of the three sub-problems,this paper constructs an accidental injury recognition model during the use of mobile phones,and finally applies this model to realize an accidental injury prevention system which can be used in real-time to warn mobile phone holders while their behavior may cause accidental injury.