Research On Gait Identification Technology Based On Inertial Sensor In Changing Environment

In this paper,gait identification is a kind of identification technology,which usually uses the gait data collected by acceleration sensor.Gait identification technology can be used in various fields.Because the human gait is affected by the tester's own factors and external environment,gait data collected in the different time and different environment is different.This results in the low recognition rate of gait identification.The purpose of this paper was to solve the problems of gait identification from changing environment.In this paper,the changing environment was summed up as the change of wearers' movement,the change of phone's placement position and the change of environment.Aimed at these problems,we researched the following methods in this paper.This paper put forward a scheme of "classification before identification".Two recognition models were designed for this scheme.In the first stage of this paper,a triple-group network authentication model was designed.By reducing the distance between similar data and increasing the distance between different groups of data,the "smaller distance inside the group and larger distance among the groups" was realized.The accuracy of authentication for "illegal users" was 0.9.In the second stage of individual identification,firstly,we used the method of class activation heat map to visualize the model "black box".On the basis of visualization,we could make the recognition rate higher by intercepting the data.Then,aimed at the problems from changing environment,we used the methods based on feature transfer to solved the problem of low accuracy.In this paper,through the scheme of two stages,first group identification and then individual identification,recognition rate of gait identification was improved.And the problems caused by changing environment were solved.Finally,the person identification in changing environment was realized.