Research On Non-invasive Gait Identity Recognition Using Muiti-source Data

With the development of intelligent mobile terminals,intelligent terminal security issues become increasingly prominent.The gait-based user identification method utilizes the built-in sensor of the wearable device to realize the identification of the user identity,has the device independentity,is not restricted by the scene and has the non-interference to the user,is an effective non-interference user identity method.The traditional noninterference gait recognition method has some problems in the field of risk control of wearable intelligent terminal equipment.Most of the existing schemes are to identify and verify the gait by means of acceleration sensors,gyroscopes and other sensors,Thus the risk of mobile phone control.As the identification of the method set a lot of restrictions,resulting in the promotion and use of the technology has some difficulties.For example:the sensor device to be fixed in the ankle,knee,waist and other positions at the same time need to put the phone in a specific location and to ensure that a specific orientation,the user needs to do a specific action.Second,the use of gait for identification and validation of technology applied to the field of risk control needs a complete and reliable system architecture.Therefore,this paper proposes a non-interfering identification and verification method,which is independent of location and behavior.It realizes the identification of non-interfering user identities by extracting the valid features and constructing the multi-data fusion model through the feature migration of user behavior and device location.This is the core of the establishment of a complete system architecture and prototype system.The implementation of the architecture method improves the accuracy and robustness of user identification and improves the human-computer interaction experience.The experimental results show that the proposed system architecture is also conducive to the improvement of the overall accuracy of the system and the method has-a high recognition rate and very low FPR(false positive)characteristics,so that in the case of non-interfering users to improve the APP and intelligence Mobile phones and other intelligent terminal equipment security.The main contributions of this paper are as follows:(1)An effective feature extraction method for user noninterference gait recognition is proposed.In order to eliminate the influence of user behavior and device location in gait recognition process,in addition to extracting user gait features,user behavior features and device location feature are extracted separately.At the same time,for user behavior due to individual differences and device location differences Model and position model drift problem,a feature migration method based on behavioral location correlation is proposed.The source domain and target domain of different user behavior and device location are transformed into the same reconstructed Hill Porter space by using migration component analysis,To eliminate the difference of features,to improve the non-interference gait identity robustness.(2)A non-interfering gait identification method for multivariate data fusion is proposed.In order to solve the influence of user behavior and wearable device location on gait identification,the acceleration sensor data are used to extract the valid feature for the user behavior and the wearable device respectively.The random behavior of the random forest(RF)algorithm is used to construct the user behavior and(MD-UIM),which is a multi-model data fusion based on the multi-data fusion model.The experimental results show that the five combinations of user behavior and intelligent device position can be effectively identified.The accuracy of the test is 5%higher than that of the traditional method,and the validity of the effective feature extraction method is verified.(3)Design and implement a non-interfering gait recognition prototype system based on multivariate data fusion.Based on the proposed method,a non-interfering gait identification prototype system based on multi-data fusion is designed and implemented.The system is composed of intelligent mobile phone acceleration sensor data acquisition module,multi-data fusion non-interference different identification module and cloud management module,which can realize the user identification under the premise of non-interference user.