Research on Characteristic Extraction of Human Gait

When a human walks, the motion of various components of the body including the head, torso, arms, legs, and feet produce a very characteristic Doppler signature. A new signal model of the walking human, based on the linear frequency modulated (LFM) signal, is proposed. This model contains 12 scatter-points of human, together with kinematic equations of motion of each body part. For each part of human body, an estimated radar cross section was assigned by the radar range equation. Short time Fourier transform (STFT) is used to obtain the human´s spectrogram at first, from which the walking velocity can be estimated. Both a traditional method based on the cadence frequency and a new echo decomposition method are used to obtain the human gait frequency. Several unique features of human targets, like the height of human and so on, can be revealed by utilizing these parameters. Some simulations show that basic information extraction from spectrogram has some ability to identification human targets.