Analysis of the Radar Doppler Signature of a Moving Human Traian Dogaru Calvin Le Getachew Kirose U.S. Army Research Laboratory RF Signal Processing and Modeling Branch
Outline Use Doppler radar to detect and classify moving humans, particularly in concealed environments (e.g., behind vegetation, walls, etc) We analyze the Doppler spectrum of a moving human in the frequency or time-frequency domains All data based on computer models simulating a pulse-doppler radar We relate patterns in the Doppler spectrograms to EM scattering phenomenology We investigate two discrimination problems: Detecting whether the human carries a weapon Discriminating a moving human from a moving dog Applications: Sensing through the wall radar Foliage penetration radar
Computing the Doppler Signature of the Walking Human Narrowband response for each frame Human body mesh animation EM simulation Computing radar return Frame #40 Frame #1 FDTD Extracting I-Q data vs. slow time sampling Plotting spectrogram x Re-arranging data in 2-D array Frequency FFT #1 FFT #2 FFT #N FFT Overlapping short-time FFT windows Time
Sample Animation of the Walking Human The human body is modeled as a uniform dielectric with r = 50, = 1 S/m Previous studies showed that this simplified model produces practically the same radar signature as the complete body model
Doppler Spectrogram of a Walking Man Motion cycle period = 2 sec Center frequency = 1 GHz Bandwidth = 80 MHz CPI = 0.6 sec PRF = 40 Hz The black line represents the velocity of a point on the man s chest. The spectrogram clearly tracks the forward motion of the torso. The zigzag pattern in the spectrogram displays certain smearing, due to limited resolution.
Changing the Aspect Angle 30 azimuth 90 azimuth The spectrogram pattern varies strongly with the azimuth angle (or walking direction) f c = 1 GHz, BW = 80 MHz, CPI = 0.3 sec, PRF = 40 Hz, V-V polarization
Operating at Lower Frequencies Motion cycle period = 2 sec Center frequency = 300 MHz Bandwidth = 40 MHz CPI = 0.6 sec PRF = 40 Hz FDTD simulation The pattern changes dramatically at UHF frequencies. A detailed analysis showed that the EM phenomenology is different that the previous cases.
Animation of the Walking Human with AK47
RCS of a Standing Human with or without a Rifle V-V polarization V-H polarization The RCS is similar in co-polarization (with or without the rifle), but stronger for the human with rifle in cross-polarization RCS computed at 1 GHz
RCS of a Standing Human with or without a Rifle V-H to V-V ratio The V-H to V-V ratio is larger for the human with rifle, at almost any azimuth angle between about -60 to 60 azimuth (beyond that, the rifle is masked by the human) The result holds for practically any rifle tilt angle This ratio can be used for discriminating a human with a rifle
Doppler Spectrum of a Walking Human with or without a Rifle V-V polarization V-H polarization The Doppler spectrum shows the same polarization differences as the RCS (stronger copolarization response for the human without rifle and stronger cross-polarization response for the human with rifle) 0 azimuth (walking straight at the radar)
Doppler Spectrum of a Walking Human with or without a Rifle V-H to V-V ratio 0 azimuth 30 azimuth The V-H to V-V ratio is much larger for the human with rifle, around the average velocity; the opposite is true at the ends of the Doppler spectrum (because the fast arm motion occurs only for the human without rifle) This results is very strong at 0 azimuth, but also holds for any azimuth angle between about -60 to 60 azimuth We threshold out the data below -40 db (V-V) and -55 db (V-H)
Animation of the Walking Dog
Spectrogram of a Walking Dog vs. a Walking Man There are clear differences between the Doppler spectrograms of a walking dog and a walking human, especially at the upper end of the spectrum 0 azimuth (walking straight at the radar) f c = 1 GHz, BW = 80 MHz, CPI = 0.3 sec, PRF = 40 Hz, V-V polarization Walking dog Walking human
Animation of the Crawling Man
Spectrogram of a Walking Dog vs. a Crawling Man One can notice more similarities between the spectrograms of a walking dog and a crawling human Walking dog Crawling human
Doppler Spectrum of a Walking Dog, a Walking Man and a Crawling Man Detail Notice much larger spectral components above 1 m/s for the walking human The walking dog and crawling man have similar spectra However, the walking dog spectrum has larger shape factor (slower roll-off), due to faster leg motion
Conclusions Analysis of the Doppler spectrum may help in human target detection, discrimination and classification, including biometric features We demonstrated the use of sophisticated tools for mesh animation/conversion and radar signature modeling and analysis Realistic human motion animation models are key in this analysis We showed how the EM scattering phenomenology dictates the patterns noticeable in the Doppler spectrograms Discriminating a human carrying a rifle can be achieved by exploiting the difference between the cross- and co-polarized radar returns A human in walking or crawling motion can be discriminated from a walking dog by looking at certain metrics in the Doppler spectrum The most important next step is to validate these techniques using measured radar data