The Research Of Identification Algorithm Based On Gait Features In WiFi Environment

With the continuous expansion of the application field and scope of information technology,indoor target identification technology has gradually become an indispensable part of people's daily life.Most of the existing system realize identify recognition through computer vision to recognize human fingerprint,iris,face and other biological features.However,in the complex indoor environment,theses camera-based system fail to work under darkness and None-Line-of-Sight(NLo S)environment.On the other hand,wide-deployed indoor coverage of wireless signals can not only be used for target identification,but also privacy protection,which attracted the attentions of many researchers.In recent years,schemes based on human gait features provide a new idea for identity recognition.Personalized home services can be developed through gait features extraction to determine whether the current user is a legitimate target(homeowner,etc.).However,two problems need to be addressed among existing gait feature-based identity recognition system: Firstly,due to the immaturity of gait detection methods,the results of identity recognition are greatly affected by the extracted inaccurate features.Secondly,most of the existing systems demand users walking along a predefined route,which leads to the limitations of the application scenarios.To tackle the above problems,this paper studies the related algorithms of indoor identity recognition system,and validate the performance of our proposed system in a typical indoor environment:Firstly,based on the statistical characteristics of electromagnetic field,the relationship between power of the received electric field and human motion is analyzed theoretically,and the indoor multipath propagation equivalent model is constructed.At the same time,starting from the essence of the signal,the influence of human motion on the angular spectrum is studied in depth.By analyzing the statistical characteristics of angular spectrum and the autocorrelation function of electric field power,the velocity estimation model is derived from the autocorrelation function of electric field power.Secondly,according to the theoretical model of speed estimation,a peak detection algorithm based on Bayesian estimation is designed to extract target moving speed from the autocorrelation function.Meanwhile,the gait characteristic parameters of target motion are further estimated from speed information,including acceleration,gait cycle and stride length information.Based on the above gait features,a multi-target classification algorithm is designed to realize identity recognition.Finally,in order to validate the performance of our proposed algorithms,including signal preprocessing,peak detection,gait parameter estimation,motion detection,identity recognition and other functional modules,extensive experiments are conducted in several typical indoor environments.The results show that our proposed algorithms achieve an average correct rate of identity recognition more than 92%,which is far beyond the needs of indoor identity recognition service.