Research And Implement On Human Activity Classification Based On Radar Signatures Of Micro-motion

Due to the wide demand for physical security,disaster recuse and surveillance,the study on the radar feature of human detection and tracking is in a high cluttered environment.The application of human activity recognition and classification based on micro-motion feature has become an area of interest in recent years.Firstly,some simple time-frequency analyses are used in the micro-Doppler signal processing,they suffer from the low time-frequency resolution or the existence of the cross-terms,the above problems would influence the feature extraction for the multi-component signal.Secondly,the research on human activity recognition is short of the analysis and consideration interference motion during the experiment,which leads to the poor antijamming performance and practicability.As for above problems,this paper introduces the study and analysis on the human activity recognition and classification used continue wave radar.The major job is improving the auto-terms concentration and eliminating interference.Through the time-frequency analysis,feature extraction and activity classification of human microDoppler signal,the human activity classification system is implemented and has good accuracy.The main contribution of this paper can be summarized as follows:(1)Analysis and extraction of micro-Doppler signal caused by human motion.We propose the multi-window adaptive S-method(AS-method)distribution used in the time-frequency analysis for radar signal.Firstly,we use the multiple orthogonal Hermite functions to improve the time-frequency resolution of micro-Doppler signal.And then,we use the adaptive S-method to choose the window with different length to suppressing the oscillating component caused by cross-terms.This method can make good compromise in the auto-terms concentration and cross-terms suppressing,which contributes to the multi-component signal separation.At last,the effective micro signal is extracted by threshold segmentation and envelope extraction.(2)Micro-Doppler feature extraction and activity classification.Based on the time-frequency distribution of human micro-motion,we analysis the characteristic of micro-Doppler for different human activities,including the feature selection,extraction and classification.During the feature extraction,we propose the peak matching algorithm,which uses the dispersion of the envelope and time interval to extract valid peaks and get peak pairs,this method can eliminate disturbing signal.And then,combining the characteristic of different activities,we put forward efficient features and apply Support Vector Machine based on decision tree to classify human activity,they can provide ideal classification accuracy with a simple classified algorithm.(3)Based on the above information,we use the software radio equipment to build the human activity classification platform.Some experiments are carried out to validate the performance of classification accuracy and anti-interference.