Phase compensation transform for human detection with LFMCW radar
Introduction
Human target detection is a focus topic due to its wide application in military combat situations, surveillance, and search and rescue operations. Detecting human targets is challenging for radar given the following reasons: the low radar cross section and the micro-Doppler caused by the varying velocity of human motion [1], [2].
The development of a signal processing algorithm to improve the detection signal-to-noise ratio (SNR) is a competitive approach to cost effectively and maximize the potential of radar. Long-time integration is a traditional method that is mainly divided into two classes, namely, incoherent integration [3], [4], [5], [6], [7], [8] and coherent integration [9], [10], [11], [12], [13], [14], [15], [16], [17], [18], [19], [20]. Incoherent integration methods integrate the magnitude of a target, whereas coherent integration methods combine magnitude and phase simultaneously to obtain a high signal processing gain.
The dynamic programming algorithm (DPA) [3], [4] and the Hough transform (HT) [5], [6], [7], [8] are well-known incoherent integration methods. The DPA integrates measurements along possible target trajectories and returns trajectories whose measurement sum or merit function exceeds the threshold. The HT method integrates the target slots that exceed the first low decision threshold to suppress false alarms. HT-based long-time integration methods transform radar raw data into range-Doppler space and improve the detection performance of targets with constant velocities or accelerations. The integration loss of the DPA and HT is larger than those of the coherent integration methods, and the performance degrades when the SNR of the target is extremely low.
For coherent integration, the classical moving target detection (MTD) method has been widely applied to integrate moving targets with unknown velocity [10], [11]. The signal processing gain of MTD is limited by a target's resident time in a range or a Doppler resolution cell. The Keystone transform method [12], [13], [14], [15], [16] compensates the cross range unit effect blindly and does not destroy the pulse phase modulation of high-speed objects during long-time integration. Radon-Fourier transform [17], [18], [19], [20] realizes long-time coherent integration via a joint search along range and velocity directions based on the coupling relationship between radial velocity, range walk, and the Doppler frequency of moving targets.
These integration methods are designed for targets with linear velocities, such as constant velocity or acceleration. However, the Doppler frequency of a walking human target varies similarly to a pendulum and spreads in the Doppler domain [1], [2], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], thereby limits detection ability when long-time coherent integration methods are used.
Numerous studies have focused on extracting the features of human motion to detect human targets as well as detecting or classifying human targets with extracted features. The micro-Doppler effect caused by the periodic rotational motion of a human body is investigated in [21], [22], [23], [24], and the Doppler spread after coherent integration caused by the micro-Doppler effect is demonstrated. The human walking parameter is estimated in [25], [26], [27] by combining radar measurements and the Boulic model in virtual reality. In [28], six features are extracted from the Doppler spectrogram based on the micro-Doppler signatures of human activity, and a support vector machine is trained to classify human activities.
Nonlinear approximation modeling of the true phase history for human targets is derived in [2], and an optimized nonlinear-phase (ONLP) detector for the parameter estimation of human targets through maximum likelihood estimation is analyzed. The ONLP detector exhibits better detection performance than the MTD method. An enhanced ONLP detector using a multichannel radar system is developed in [1] to improve radar detection performance for human targets by exploiting knowledge on human gait.
In this paper, a coherent integration algorithm called phase compensation transform (PCT) is proposed to improve detection performance for human targets with a linear frequency modulated continuous wave (LFMCW) radar system. In the PCT algorithm, the beat signal of a human target is derived based on the Boulic model for an LFMCW radar. Then, the nonlinear phase characteristic of the beat signal caused by micro-Doppler from human motion is analyzed. A phase compensation signal is constructed with the obtained characteristic and utilized to compensate the nonlinear phase in the range cell during a coherent integration interval (CPI). Next, the compensated range cell is fast Fourier transformed (FFT), and the power of human echo is integrated. The PCT algorithm exhibits a higher integration efficiency and better detection performance for human targets than the MTD and ONLP methods. In addition, the proposed PCT algorithm can estimate human motion parameters such as average radial velocity, gait frequency, and torso initial phase.
This paper is organized as follows. In Section 2, the beat signal of a human target based on the Boulic model is derived for an LFMCW radar. Then, a time-frequency analysis of the beat signal from the human target is conducted to demonstrate the Doppler spread of human motion. In Section 3, the beat signal of the human target is simplified, and the nonlinear phase characteristic caused by micro-Doppler is analyzed. Subsequently, the phase compensation signal is constructed. The PCT algorithm is presented in detail in Section 4, and the simulation results and data sampled from the LFMCW radar system developed in the laboratory are presented in Section 5 to evaluate the performance of the proposed PCT method.
Section snippets
Beat signal modeling for human target with LFMCW radar
In this section, the radar radial motion displacement of a human body is modeled using the Boulic model. Next, the beat signal of a human target is derived for an LFMCW radar. Finally, a time-frequency analysis is conducted to demonstrate the Doppler spread of human motion [31].
Phase compensation signal construction
Based on the time-frequency analysis of the human beat signal, the instantaneous Doppler of a walking human is a superposition of the average radial Doppler and the micro-Doppler from each body part. The phase of spectrum SBH, m(fB, m) in the human target's range cell exhibits a periodic sine act during a CPI. If a compensation signal with a phase that conjugates to that of SBH, m(fB, m) is constructed and compensates the nonlinear phase of SBH, m(fB, m), the power of the human echo with the
Phase compensation transform algorithm
Based on the phase compensation signal H, the nonlinear phase of the human beat signal is counteracted by iteratively searching for an optimal motion parameter in the fG-θFB plane, and the spread power in the Doppler domain is concentrated by FFT.
In the nth range cell, SBH(n) is denoted as the input of PCT, it haswhere SBH, m(n) is the output of the nth range cell in the mth PRI during a CPI in (21).
The phase compensation transform procedure for SBH(n)
Performance evaluation
The performance evaluation is divided into two sections. The simulation results are given in Section 5.1 to demonstrate the integration performance of the PCT and MTD methods, the human motion parameter estimation performance of the PCT and ONLP methods, and the detection performance of the PCT, MTD, and ONLP methods. The sampled data from the LFMCW radar developed in the laboratory are provided in Section 5.2 to show the integration performance of the PCT and MTD methods.
Conclusion
The proposed PCT method compensates the nonlinear phase cause by micro-Doppler from human walking motion during a CPI. The compensated result is obtained through FFT, and the Doppler spread caused by periodic micro-Doppler is integrated sufficiently; thus, human target power is concentrated. For human target detection, the PCT algorithm exhibits better integration efficiency than the traditional MTD method, higher estimation precision of motion parameters than the ONLP method, and better
CRediT authorship contribution statement
Renli Zhang: Conceptualization, Investigation, Funding acquisition. Zhili Zhang: Data curation, Writing - original draft. Xin Zhang: Software. Yue Tang: Writing - review & editing. Weixing Sheng: Project administration.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgment
This work was supported in part by the National Natural Science Foundation of China under grant 61971224 and the Natural Science Foundation of Jiangsu under grant BK20180457.
References (33)
- et al.
Kinematic model-based human detectors for multi-channel radar[J]
IEEE Transact Aerosp. Electron. Syst.
(2012) - et al.
A nonlinear-phase model-based human detector for radar[J]
IEEE Transact Aerosp. Electron. Syst.
(2011) - et al.
A novel dynamic programming algorithm for track-before-detect in radar systems[J]
IEEE Trans. Signal Process.
(2013) - et al.
New analytical approach to detection threshold of a dynamic programming track-before-detect algorithm[J]
IET Radar Sonar Navigation
(2013) - et al.
Search radar detection and track with the Hough transform. I: system concept[J]
IEEE Transact Aerosp. Electron. Syst.
(1994) - et al.
A multi-dimensional hough transform-based track-before-detect technique for detecting weak targets in strong clutter backgrounds[J]
IEEE Transact Aerosp. Electron. Syst.
(2011) - et al.
Search radar detection and track with the Hough transform. II: detection statistics[J]
IEEE Transact Aerosp. Electron. Syst.
(1994) - et al.
Search radar detection and track with the Hough transform. III: Detection performance with binary integration[J]
IEEE Transact Aerosp. Electron. Syst.
(1994) - et al.
Long-time coherent integration detection of weak manoeuvring target via integration algorithm, improved axis rotation discrete chirp-Fourier transform[J]
IET Radar Sonar Navigation
(2015) Introduction to Radar System [M]
(2002)