Research On Human Gait Recognition Based On Millimeter Wave Radar

Human gait recognition is an important research direction of action recognition.It can not only detect human life characteristics and motion status,but also can be used as an effective means of identity recognition.It has important applications in security,criminal investigation and other fields.Traditional human gait recognition is mostly based on wearable sensors or video sensors,which have great limitations in the use of scenes.Research on the extraction and recognition of human micro-motions using radar has been paid more and more attention.Starting from the human walking echo model,this topic studies key technologies such as radar echo-based motion parameter estimation and time-frequency analysis.And through the millimeter-wave radar for actual measurement verification,and then complete the classification and recognition of human movement and gait.Firstly,a walking model based on the Boulic human model is constructed for the walking state of the human body.Considering that the model converts body parts into elliptical rigid bodies,the calculation of the radar echo time delay using the centroid coordinate distance is insufficient in reflecting the micro-Doppler information.A line integral walking model is also proposed.On the basis of this model,the analytical expression of radar echo in the walking state of the human body is derived,and the relationship between the micro-Doppler frequency change of the human walking and the gait parameters is analyzed.Secondly,the time-frequency distribution of radar motion,and three timefrequency analysis methods are introduced.The time-frequency distribution evaluation criteria based on Renyi entropy are used to compare the three methods.In order to better explore the micro-Doppler frequency changes in the time-frequency graph,this paper extracts the envelope of the time-frequency graph and separates the micro-movement parts based on the inverse Radon transform,and echoes is used as the main means to express the micro-Doppler information of human extracts the Doppler frequency change curve of the large and small arms.Based on these envelopes and curves,a digital gait feature extraction method is given.Finally,according to the relationship between the Doppler frequency shift and the gait parameters of the human walking,the digitized gait characteristic values of the human walking are calculated,and the measured data are used for identification to verify the feasibility of the algorithm.In life,people's walking is often interspersed with other actions,so this article also recognizes different actions based on time-frequency diagrams,and proposes a step-by-step human action and gait recognition process,using the advantages of each classifier to effectively identify the state of human behavior.