Doppler-based detection and tracking of humans in indoor environments
Introduction
Through-wall radar imaging is a topic of current research interest in connection with law enforcement, urban area operations and search and rescue missions. Developments to date include wideband radar using impulsive [1], [2], [3], [4] or stepped frequency [5], [6] waveforms to obtain down-range resolution down to a few centimeters. Since the attenuation of the signal through walls increases rapidly as a function of frequency [7], [8], frequencies below 5 GHz are usually employed. One particular aspect of through-wall imaging is the monitoring of human activities. For human movement detection and tracking, Doppler sensing is a natural choice since stationary indoor clutters can be suppressed to highlight human movements. Further, low-cost, low-power Doppler sensors can be readily implemented. Finally, human movements give rise to very unique micro-Doppler features [9], [10], [11], which may be exploited for recognition [12], [13]. An early example of a Doppler-based system is the motion detector radar with a continuous wave phase detector to detect moving targets [14]. However, this system does not provide any location information. The Radar Flashlight system that was later developed in [15], [16] utilizes an X-band FMCW radar system to detect the respiratory motion of still humans behind walls. In [17], a pulse-Doppler waveform is used to compute the range of humans through walls. By using a real array with multiple elements, the bearing information in azimuth and elevation could also in principle be obtained, although at the expense of higher cost. Use of synthetic aperture concept is more challenging due to the short time required to collect coherent data from the human movement.
In this paper, we discuss the application of the Doppler concept to detect and track humans in complex indoor environments. First, we focus on the micro-Doppler characteristics of humans. In Section 2, human gait data are represented in the joint time–frequency space using the short-time Fourier transform (STFT). The Doppler spectrograms of both simulated and measured data are studied to identify various movements. The reassigned joint time–frequency transform (RJTF) [18], [19], [20], [21] is also explored for its superior signal localization properties to aid the interpretation of the micro-Doppler features.
Next, we focus on the use of Doppler sensors to track multiple human subjects indoors. A low-complexity Doppler radar concept [22], [23], [24], [25], [26], [27] is first reviewed in Section 3. The concept exploits the Doppler separation among movers to track multiple humans in three dimensions using only three antenna elements and two carrier frequencies. In Section 4, we study ways to improve the multiple-human tracking performance. We show that by incorporating a multiple element array, it is possible to combine Doppler processing with spatial beamforming to resolve different targets in the Doppler and bearing space [28]. Additional frequency diversity is also explored to determine the range of targets.
Finally, some of the challenges to Doppler-based human tracking are discussed. In Section 5, we study the effect of walls on the signal amplitude decay and phase distortion. These wall effects can cause significant degradation in the signal-to-noise ratio as well as error in the bearing estimation of humans through walls. This is studied using both simulations based on the finite-difference time-domain (FDTD) simulation as well as measurements on real walls. Conclusions are given in Section 6.
Section snippets
Micro-Doppler characteristics of human movements
Contrary to rigid targets commonly encountered in many radar applications, a human is a complex non-rigid body under movement. As a result, many interesting features of human movement such as the swinging arms and legs can be observed using a Doppler radar. These features are termed micro-Dopplers and they have been studied in a number of publications [9], [10], [11], [29], [30] through the use of joint time–frequency representations.
We first illustrate the micro-Dopplers from human gait using
Doppler-based human tracking
While the individual human Doppler was shown to be quite interesting, it is also desirable to simultaneously gather location information on multiple humans indoors. For this purpose, a very low-complexity radar concept that exploits the Doppler separation between multiple movers was investigated at the University of Texas [22], [23], [24], [25], [26], [27]. The concept combines the Doppler discrimination offered by human movements with the direction of arrival (DOA) information that can be
Doppler tracking combined with spatial beamforming and frequency diversity
In this section, we study ways to improve the tracking performance of multiple humans with overlapping Dopplers. By combining Doppler processing with spatial beamforming in a multi-element (greater than 2) array, it becomes possible to resolve targets along two dimensions, Doppler and DOA. It is further possible to resolve targets along the range dimension by using multiple frequencies. The improvement in the performance towards multiple human tracking is first studied using Monte Carlo
Wall effects
One of the major challenges in the tracking of humans is the accuracy of the DOA measurements under through-wall operations. To better understand the wall propagation phenomenology, we present in this section some measurement and simulation results in connection with the operation of the Doppler-based radar testbed.
The two important wall effects on RF signal propagation are the attenuation of the signal amplitude through the wall and the phase distortion of the signal due to multipath and
Conclusion
In this paper, we have presented the principles of Doppler processing to detect and track human movers in indoor environments. The main advantage of Doppler sensing is that stationary clutters can be suppressed. The potential application of micro-Doppler features for the identification of human movements and the presence of carried objects was discussed. The RJTF was investigated as an alternate representation of the Doppler spectrogram that provides improved signal localization compared to the
Acknowledgments
This work was supported by the Office of Naval Research, the Texas Higher Education Coordinating Board under the Texas Advanced Technology Program, the National Science Foundation Major Research Instrumentation Program and NSF Grant CBET-0730924.
References (40)
- G. Franceschetti, J. Tatoian, D. Giri, G. Gibbs, Timed arrays and their application to impulse SAR for through-the-wall...
- S. Nag, M.A. Barnes, T. Payment, G. Holladay, Ultrawideband through-wall radar for detecting the motion of people in...
- A.M. Attiya, A.M. Bayram, A. Safaai-Jazi, S.M. Raid, UWB applications for through wall detection, in: IEEE Antennas and...
- Y. Yang, A.E. Fathy, See-through-wall imaging using ultra-wideband short-pulse radar system, in: IEEE Antennas and...
- A.R. Hunt, A wideband imaging radar for through-the-wall surveillance, in: SPIE Proceedings of the Sensors, and...
- et al.
Design and implementation of near field wideband synthetic aperture beamformer
IEEE Trans. Aerosp. Electron. Systems
(2004) - L.M. Frazier, Radar surveillance through solid materials, in: SPIE Proceedings of the Command, Control, Communications,...
- et al.
Prediction and measurement of wall insertion loss
IEEE Trans. Antennas Propagat.
(1999) - J.L. Geisheimer, E.F. Greneker, W.S. Marshall, High-resolution Doppler model of the human gait, in: SPIE Proceedings of...
- et al.
Time Frequency Transforms for Radar Imaging and Signal Analysis
(2002)
Human walking estimation with radar
IEE Proc. Radar Sonar Navig.
Analysis of time varying signals with small BT values
IEEE Trans. Acoust. Speech Signal Process.
Improving the readability of time frequency and time scale representations by the reassignment method
IEEE Trans. Signal Process.
Algorithms for computing the time-corrected instantaneous frequency (reassigned) spectrogram, with applications
J. Acoust. Soc. Am.
Cited by (104)
Track detection of high-velocity resident space objects in Low Earth Orbit
2023, Advances in Space ResearchCitation Excerpt :Since then, track identification within spectrogram images has been used in a plethora of applications. These range from the monitoring of a patient’s heart rate studies by Lee et al. (2011) to the tracking of a vehicle by Dutta et al. (2006) and Roy et al. (2011) or human motion such as Ram et al. (2008), Harmanny et al. (2014) and Tahmoush and Silvious (2009). Morrissey et al. (2006) used sonograms to identify marine mammals through their calls.
Received power-based vital signs monitoring
2021, Contactless Vital Signs MonitoringThrough-the-wall radar detection using machine learning
2020, Results in Applied MathematicsCitation Excerpt :One particular approach is the use of radar frequency imaging to retrieve the properties of an unknown object from its back-scattered fields. Previous work has used Doppler-type radar to detect and analyze humans where there is no direct line of sight, whether it be studying human motion with standard Doppler radar [1], noiseforms [2], or micro-Doppler radar, which looks for smaller scale movements such as arm movement [3] and heartbeats [4]. In [5], a Support Vector Machine approach was used to discriminate between child and adult in a through-the-wall setting.
Gait-based Human identification using acoustic sensor and deep neural network
2018, Future Generation Computer SystemsCitation Excerpt :The aforementioned problems seriously impede the applications of the video/image techniques in a wide range. Numerous efforts have been devoted to investigating promising alternatives for human identification to enable ubiquitous applications of human identification under all conditions with a fast processing rate [11,12]. For example, a wide range of alternatives, such as radar, wearable sensor and floor sensor based techniques have been demonstrated to be effective in gait-based human identification [13–15].
A linear sampling method for through-the-wall radar detection
2017, Journal of Computational PhysicsDevelop Indoor Object Detection and Localization System Based Image Processing
2024, AIP Conference Proceedings