Research On Activity Recognition And Identity Authentication Based On Sensors In Smart Phone

The research of human behavior recognition technology describes the user's movement process and behavior mode through image or data information and makes accurate judgment and analysis for what is happening.This technology is closely related to scientific research and is beneficial to the development of pattern recognition,identity authentication and security and warning.With the development of electronic technology,smart phones have become an indispensable part of work and life,in which integrated multi-type sensor hardware is constantly updated,optimized and coordinated with applications to provide more humanized functions.Therefore,the research on the data types and characteristics of smart phone sensors provides more effective ideas for the research in the field of behavior and identity recognition.At present,successful behavior in the field of behavior recognition research can identify the behavior content is simple,the value of application promotion is low.It mainly makes the use of acceleration sensor and gyroscope data;the single type increases the difficulty of data processing.In the field of identity recognition,tamper or imitation technology has been used in recognition methods based on physiological characteristics(such as fingerprint and human face recognition),which reduces the effectiveness and reliability of the method.The behavior characteristics displayed by users are universal and unique,hard to change or be imitated by others,and can be used for identity authentication with higher security level.Aiming at the above problems,this paper uses the smart phone sensor data to improve and study the behavior recognition and identification methods,and the main research results are as follows:1.To improve the low data utilization and single behavior content problems in the existing behavior recognition based on smart mobile sensor data and solve the effects of the mobile storage and normal use when the data acquisition and recognition program run on it,this paper presented a method of human basis behavior recognition based on the smart phone data and a method of human behavior recognition based on the alteration of multi-sensor in smart phone.The former method firstly carries out a multi-level data preprocessing means and dimensionality reduction operations on the original data features and selects the behavioral features with the combination of existing knowledge,and then makes multiple classification of five behaviors by using the decision tree binary classification method,which reduces the amount of training data and improves the accuracy of recognition.The latter method processes the original data into a state that can represent the alteration of the sensor data.Performance comparison is made between Markov chain and Naive Bayesian network when they finished the classification respectively.This method improves the utilization rate of the sensor data and enriches the content of the identifiable behavior.Finally,the effectiveness of the two methods is verified by experimental test.2.To improve the reliability of biometric recognition and apply the results of behavioral recognition to the identification field based on behavioral features at the same time,a gait identification method based on the motion sensor data of smart phones is proposed.The motion sensor uses self-reference frame,and there is an angle between it with the commonly used inertial frame.Firstly,the rotation operation of the self-reference coordinate system is carried out by using the coordinate transformation algorithm to make it coincide with the inertial coordinate system.Then machine learning algorithm is used to identify the change of data characteristics,and to simulate the change of user behavior characteristics when switching.Experiments show that this method can be applied to mobile terminal identification.