Study On Pedestrian's Micro-motion Separation Based On Micro-Doppler Analysis

Since proposed,the micro-Doppler(m-D)effect has been widely used in target detection,feature extraction,classification,recognition,parametric inversion and so on.In addition to the bulk motion of the mass center of human body,different part of body exists distinct micro movement relative to the center of mass,such as periodic swinging of limbs,vibration of skin and chest caused by breath and heartbeat.Considering the m-D effect,more useful information can be extracted from the radar echoes.Aside from the inherent characteristics including backscattering intensity and body size,it also reflects the micro motion of body,which would improve the capability of target recognition.However,due to the nonstationary characteristic of m-D signal,the aliasing of m-D signatures of different parts in time-frequency(TF)domain poses a great challenge to the feature extraction process of various body parts.Therefore,effective separation of pedestrian's micro-motion parts based on m-D characteristic is a great challenge.Hence,the study on separation algorithm of pedestrian's micro-motion parts based on m-D Analysis has important theoretical significance and application value,which not only provides technology accumulation for refined recognition and classification to a certain extent,but also can apply to refined recognition and classification of human gait,which has important application value in battlefield surveillance,the warning of terrorist attack,evidence of emergency,medical assistance,precision medicine and so on.Based on the practical demand of human target recognition and classification,the study on separation algorithm of pedestrian's micromotion parts based on m-D analysis is carried out.This dissertation focuses on establishing pedestrian radar echo model and separation algorithm of pedestrian's micro-motion parts.The main contents are summarized as followed:1.Establish the motion model of typical micro-motion including vibration,rotation and conical motion,and m-D frequency induced by typical micro-motion is derived.For the nonstationary characteristic,joint TF analysis is applied to analyze the micro-motion characteristic of m-D signatures,which lay a foundation for the following research of pedestrian's m-D characteristic analysis and micro-motion separation.2.Aiming at the challenge of pedestrian's complex movement mechanism,combined with the kinetic characteristic of body parts movement and radar cross section(RCS)analysis based on burn area in medicine,the radar echo model of pedestrian movement is established.Based on the hypothesis that the human body can be modeled as 12 attributed scattering centers and there is no occlusion between body parts,the echo of pedestrian is obtained by summing up the echoes of all parts.The motion trail of different part is simplified as a sinusoid curve for sine-like characteristic manifested by micro motion of different part.By analyzing the backscattering characteristics,burn area in medicine is introduced to calculate RCS of human body,so that radar echo model suitable for pedestrian characteristic analysis is established.Joint TF analysis method is applied to analyze pedestrian m-D characteristic in detail.3.Aiming at the problem of pedestrian m-D signatures occlusion and separation difficulty,a separation algorithm based on principal component analysis(PCA)-Kmeans method is proposed.Firstly,a one-dimensional echo is converted to a two-dimension input matrix after preprocessing,and mean normalization process is applied to the input matrix.Then,orthogonal principal components acquired after PCA process form a eigen subspace with a lower dimension,and the echo is decomposed into orthogonal principal components.Lastly,based on the frequency probability density distribution of principal components,K-means method is applied to cluster principal components,and each cluster represents echo of different parts.4.Considering the fact that there is remained occlusions in clusters after PCA-Kmeans separation algorithm,a separation algorithm based on inverse Radon transform is proposed for further separation of pedestrian m-D signatures.Firstly,inverse Radon transform is applied to estimate the parameters of the strongest m-D component based on signal's TF representation.Then,the strongest m-D component is extracted from the original mixed signal.Lastly,the energy ratio of the strongest m-D component and the original energy is calculated and compared.If the energy ratio is lower than threshold,the algorithm terminates and the strongest m-D component will be discarded.Otherwise,signal without the strongest m-D component will undergo next separation process.Components obtained by this iterative algorithm corresponds to different m-D signature of different part of pedestrian.