The Research Of Gait Recognition Based On Mobile Phone Acceleration Sensor

Biometric recognition systems based on human faces,fingerprints,speech,etc.have already been applied in many fields.Compared with early identification systems such as passwords and graphics,biometric systems can provide more reliable identity.These identification systems all require the user to participate in the operation,and can only perform an identification at a specific stage such as when the device is turned on or unlocked,and cannot be used for continuous identity authentication of the user.The identification system based on gait can identify the user during user walking,so that continuous authentication of the user identity can be achieved,which greatly improves the security of the electronic device.At present,the gait information acquisition technology based on the mobile phone acceleration sensor has a relatively mature application in pedometer,walking and running recognition.Especially the increasingly popular smart wristbands and smart watches have built-in acceleration sensors,and some of them have barcode payment functions and security verification requirements.Therefore,the recognition of gait information by using an acceleration sensor of a mobile phone or a smart wearable device not only has practical significance and practical value,but also meets the needs of information security protection of mobile phone and wearable devices.The contribution of the thesis are as follows:Firstly,using a mobile phone equipped with a triaxial acceleration sensor and software that is easy to develop,a human gait information acquisition software based on a mobile phone is designed.A 50-person gait acceleration database was established to lay a theoretical and technical foundation for the study of human gait recognition.Secondly,an identification method based on the asymmetric feature of human left and right gait is proposed.First of all,preprocessing the gait acceleration signal;second,selecting the first derivative of the Gauss function as the wavelet basis function,detecting the gait mutation point of the human body,extracting the amplitude and time characteristic sequence of the acceleration signal,passing the volunteers performed similarity and difference tests of the left and right gait characteristics,which proved the feasibility of distinguishing the left and right gaits of the human body.Finally,a dynamic time warping algorithm was used to match the extracted amplitude and time feature sequences.The simulation and experimental results show that the method of identification based on the asymmetric features of left and right gait in this paper improves the accuracy of gait recognition and has wide applicability.Thirdly,a novel human gait identification method combining multi-scale entropy and dynamic time warping algorithm is proposed.A multi-scale entropy algorithm is used to process the human body gait acceleration signal in the direction of gravity to obtain the entropy values of the human body gait acceleration signal at various scales and use it as the feature;then the dynamic time warping algorithm is used to calculate the gait characteristics of the volunteers.Distortion distance,simulation results show that the method based on multiscale entropy and dynamic time warping algorithm further improves the accuracy of the identification,providing a new idea for gait recognition.