Research On User Behavior Recognition Based On Smart Phone

At present,user behavior recognition is everywhere in our daily life,and it is applied in many fields,such as medicine,sports,human-machine interaction and identity recognition.But before that,this kind of research is basically using some related professional wearable sensors or external sensors.This series of sensors are most expensive,not conducive to the expansion of the market,and bring inconvenience to our research.With the rapid development of mobile smart phone,most smartphones have absorbed a variety of different sensors and related technologies.This paper is complying with this hot situation,then puts forward the research of user behavior recognition based on smart phones.After comparing Chinese and foreign literature synthetically,this paper can not only improve the accuracy of five basic human behaviors recognition,such as standing,walking,running,upstairs and downstairs,but alsoidentifymore complex behaviors.In order to collect the sensor data of these sports behavior,this paper developed a special sensor data acquisition APP based on Android smart phone,which not only can collect all the sensor data of the ordinary smart phone,but also record the related motion trajectory.This paper will also deal with the human activity recognition and postural conversion data sets based on smart phone,filter by median filter and third-order Butterworth low pass filter,and then remove noise at 20 Hz angle frequency filter.By calculating the average value of the acceleration and the value of the gyroscope,the standard deviation,the mean absolute difference,the maximum value,the minimum value and so on,the 561 dimension feature vector isformed.Thesefeature vectors are used to represent the user's motion behaviors.Then we use neural network to identify and classify,and then compare it with traditional classification.The experimental result shows that the overall recognition rate of user behavior recognition based on smart phone is more than 91%,and somebasic actions reached99%.In this paper,neural network is used to solve the problem of low recognition rate of the previous methods,and it also proves the feasibility of the research plan.The results of this paper are as follows:1.We design and develop a behavior recognition APP based on Android mobile2.We preprocess the collected raw data,and use neural network to classify and identify the data.3.We study the classification results of user behavior based on neural network, then we compare them with traditional classifiers,the recognition accuracy rate has improved.