Research On Human Movement Information Recognition Based On Channel State Information

At present,the research of human behavior recognition plays a pivotal role in the field of human-computer interaction.With the rapid development of wireless networks,Wi Fi signals have been widely popularized and applied in public life.As an emerging research hotspot,the use of Wi Fi signals to collect mobile information of the human body is attracting attention in many fields such as smart homes and intrusion detection.The use of wireless networks to collect human body movement information has many advantages such as simple deployment,low cost,and extensive application scenarios.At the same time,the channel state information in the Wi Fi signal can be used to analyze the gait information of the human body,thereby providing a strong data basis for the identity authentication technology.The gait of each person is the same as the fingerprint iris of each person,which is unique to each person.It is not easy to be imitated and is only related to each person's height,weight and walking habits Due to the multipath effect,different people are walking.Time will have different effects on the amplitude and phase of the channel state information.Therefore,this thesis proposes a human gait recognition approach based on channel state information.This approach first preprocesses the collected CSI data.After preprocessing,a walking interval cutting algorithm with adaptive threshold is used to segment the walking interval and the stationary interval in the data.Then the data in the walking interval is transformed into a carrier energy map,which contains rich human walking characteristics.Finally,a convolutional neural network is used to achieve the recognition of various gaits.In the verification stage of the experiment,when there are 10 testers,the accuracy of this approach is over 91%,showing good recognition performance.At present,when researchers are studying human body movement information,there are few studies on the direction of human body movement.The study of the direction of human movement is also very important.It can play an important role in many application scenarios.For example,when taking care of the elderly,their different directions of movement represent different things to do,which can be done in time through the different directions of movement of the elderly.Prejudge their behavior.Therefore,this thesis proposes a approach of human body movement direction recognition based on channel state information.First,phase calibration and preprocessing are performed on the CSI data,and then the Doppler velocity,the energy characteristics of each frequency band and the characteristics of wavelet packet coefficients after wavelet packet decomposition are used as eigenvalues.Finally,support vector machines are used to classify the direction parallel to the line of sight and the direction perpendicular to the line of sight.In the experimental verification stage,the recognition accuracy of the two directions was over 93%,which achieved the expected results.The human body movement information recognition approach based on channel state information proposed in this thesis can show high recognition accuracy in gait recognition and movement direction recognition.It has research significance and application value in intrusion detection and elderly care.