Research On Recognition Method Of Bad Behavior Based On Smart Phone

With the continuous improvement of the processing power of smart phone and the development of micro-low power sensor,sensor-based behavior recognition has become a research focus.This thesis focuses on the users of smart phone and studies the bad behavior recognition method in mobile phone use,making users paying attention to bad behaviors like phubbing while enjoying the intelligent experience brought by smart phone.The main work of this thesis includes:First of all,by collecting the acceleration data of three different mobile phones position(held on the chest,put in the pant pockets and jacket pockets)in five scenarios,such as sitting,walking,going upstairs,going downstairs and taking a bus,the median filter,high-pass filter and windows are used to remove the noise,separate the gravity acceleration signal from the original data,and separate the original acceleration signal respectively.Then,the mean value,variance,root-mean-square root and correlation coefficient of the extracted signals of x,y and z axis are extracted respectively.Finally,compare the performance of integrated classifier with decision tree(J48),random forest(RF),k-nearest neighbor(KNN),naive bayes(NB)and support vector machine(SVM)classification.The experiment verifies the effectiveness of the recognition method based on the integrated classifier during the use of smart phone.Secondly,the recognition method of the bad behavior of the head.For the phenomenon of phubbing,this thesis designs and implements a method for calculating the neck bending angle based on the fusion image processing and sensor technology of smart phone.First,collect facial images and three-axis acceleration signals through a smart phone front camera and a built-in sensor.Second,face recognition is based on the Adaboost method of extended Haar features.Then,through Adaboost method for face recognition based on extended Haar features,pupil location based on gray-scale integration method,edge detection algorithm based on Canny operator mouth position,then get the face angle,mobile Angle were obtained through acceleration signal,tilt angle of mobile phone is obtained by the three axis acceleration signal.Finally,according to the angle of the face and tilt angle of the phone,the neck bending angle is calculated to identify the bow posture.Finally,a rule-based classification method is adopted to recognize the behavior.A large number of experiments show that the algorithm proposed in this thesis has a good recognition effect.Thirdly,based on the above research,we designed and realized a bad behavior recognition and warning system.The system has app and web,and includes data collection,bad behavior recognition,bad behavior warning,historical data query,health guide,health advisory,personal information and other functions.