Radar Echoes Human Feature Extraction Algorithm

With the ability to work in all weather conditions, detect target in remote places, and perceive target behind obstacles, radar can be extremely useful sensing technique to observe persons. Radar human detection technique has been the research hotspots in human sensing research field at home and abroad. Radar detects human body by analyzing the features of human body containing in radar echo. Therefore, extraction of human features of radar echo is key factor in radar human detection.The main work of this paper focus on the extraction of human features of radar echo, especially analysis the human features by spectrogram. The main works are as follows:1. Aiming at micro vibrational target model and human multi-scattering centers model, analysis the radar echo model and the theory of human feature extraction.2. Aiming at the problem of extraction of gait frequency, a spectrogram-based method is proposed which is more accurately than directly FFT method. Aiming at human speed estimation problem, multi-spectrograms are analyzed and a spectrogram-based method is proposed to estimate human speed.3. Aiming at the problem of human micro-Doppler feature extraction, the significance of spectrogram features are analyzed and corresponding methods to extract these human features are given. Aiming at the inaccuracy of time-frequency spectrogram, introduce the spectrum zoom method, and improve the micro-Doppler features of human in the spectrogram.4. Aiming at human activities classification, Markov chain computation algorithm is applied to classify human activities such as walk, run.5. Hilbert-Huang transform is applied to human features extraction. The feasibility of this method is firstly analyzed, then apply this method to extract human vibrational and Doppler features. A method based on EMD is introduced to separate Doppler features of different scattering centers.All the features extraction methods are illustrated with simulation results and some are illustrated with experiment measured data. These human feature extraction methods possess extensive application prospect, and are the basis of radar human classification and identification.