A Coherent Through-the-Wall MIMO Phased Array Imaging Radar Based on Time-Duplexed Switching

Transcription

1 A Coherent Through-the-Wall MIMO Phased Array Imaging Radar Based on Time-Duplexed Swithing Qinghao Chen a, Kevin Chetty a, Paul Brennan b, Lai Bun Lok b, Matthew Rithie, Karl Woodbridge b a Dept. of Crime and Seurity Siene, University College London, London, U.K.; b Dept. of Eletronis and Eletrial, University College London, London, U.K.; Defense Siene and Tehnology Lab, Salisbury, U.K. ABSTRACT Through-the-Wall (TW) radar sensors are gaining inreasing interest for seurity, surveillane and searh & resue appliations. Additionally, the integration of Multiple-Input, Multiple-Output (MIMO) tehniques with phased array radar is allowing higher performane at lower ost. In this paper we present a 4-by-4 TW MIMO phased array imaging radar operating at.4 GHz with 00 MHz bandwidth. To ahieve high imaging resolution in a ost-effetive manner, the 4 Tx and 4 Rx elements are used to synthesize a uniform linear array (ULA) of 16 virtual elements. Furthermore, the transmitter is based on a single-hannel 4-element time-multiplexed swithed array. In transmission, the radar utilizes frequeny modulated ontinuous wave (FMCW) waveforms that undergo de-ramping on reeive to allow digitization at relatively low sampling rates, whih then simplifies the imaging proess. This arhiteture has been designed for the short-range TW senarios envisaged, and permits suffiient time to swith between antenna elements. The paper first outlines the system harateristis before desribing the key signal proessing and imaging algorithms whih are based on traditional Fast Fourier Transform (FFT) proessing. These tehniques are implemented in LabVIEW software. Finally, we report results from an experimental ampaign that investigated the imaging apabilities of the system and demonstrated the detetion of personnel targets. Moreover, we show that multiple targets within a room with greater than approximately 1 meter separation an be distinguished from one another. Keywords: Multiple Input Multiple Output Radar, Phase Array, Radar Imaging, Through Wall, Time-Duplexed radar. 1. INTRODUCTION The ability to detet the presene of individuals behind visually obsuring strutures is of signifiant benefit to polie, seurity and emergeny agenies in many situations. Signifiant amount of researh has attempted to apply radar detetion and imaging tehniques to senarios that onsist of an obsuring wall barrier. Through-Wall (TW) radars have been employed in short range detetion appliations whih are generally used to detet the motion, loation and signs of life indiators of human targets inside buildings. To detet a moving target, Continuous Wave (CW) radars an be utilized to detet only the Doppler information related to target s movement. CW radar an be regarded as only a motion detetor without providing any range information. However, the miro-doppler signature an still be utilized to differentiate the target s motions with the help of reognition algorithms. Single hannel FMCW radar has been used to detet the range of the target behind the wall but not any angle information, whih an only provide the 1-D information. Therefore, the detetion apability is limited by the beam-width of the antenna. To detet -D or even 3-D information of the target behind the wall, Peabody et al. [1] designed a UWB FMCW MIMO radar ranging from GHz to 4GHz and a 8 transmit 13 reeive phased antenna array to generate the radar image behind the wall. The hirp generation of this radar is built via a ramp generator and the voltage ontrol osillator and the transmit peak power is W. In addition, the range gating method is implemented as a band pass filter. Dilsavor et al. [] built a VNA-based syntheti aperture radar by moving the single transmit and single reeive antenna and post proess these data by applying different weights orresponding to their antenna positions. However, this VNA-based radar laks the ability to detet moving targets behind the wall as they need to move their one hannel Tx and Rx antenna to form the syntheti aperture. Another ombination of the SAR proessing and phased array MIMO Radar Sensor Tehnology XXI, edited by Kenneth I. Ranney, Armin Doerry, Pro. of SPIE Vol , SPIE CCC ode: X/17/$18 doi: /1.646 Pro. of SPIE Vol

2 radar is to mount the antenna arrays and radar on the truk [3]. The truk moves in the azimuth diretion and the array is in the elevation diretion to obtain the 3-D imaging apabilities. The most similar work to us in the radar design is Huang et al. [4], who designed a time-duplexed phased array MIMO system for maritime navigation. However, their system was not designed to operate with parameters ommensurate for through-wall stand-off detetion senarios. The array design between the system is also different. The work from Peabody et al. is also similar to our work in terms of employing time-duplexed MIMO tehniques however the radar design is different: we use a DDS to generate the 00MHz linear hirp whilst they utilized a Voltage Controlled Osillator (VCO) to generate the Ultra-Wide-Band (UWB) hirps of GHz bandwidth. In this paper, we present a TW FMCW Phased-Array MIMO radar entered at.4ghz with 00MHz bandwidth and the onventional path antenna array. To build a low-ost MIMO radar, we make use of swithing tehnique to form multiple virtual array elements [5]. Next, we apply FFT based array proessing to the MIMO data. Moreover, we demonstrate the imaging apability of the system using targets moving behind a wall. To ahieve this result, we also employ a moving target indiation (MTI) filter to mitigate the effets of the wall. In setion, we desribe the signal model of the swithed phased array MIMO radar and signal proessing methods to obtain the radar images. In setion 3, the design and implementation of the radar system and the antenna array are explained. Next in setion 4, the imaging results from an experimental trial are illustrated, inluding single and multiple moving targets. Finally in setion 5, we onlude the paper by summarizing the main advanements from the researh..1 Model of Phased Array Radar. SIGNAL MODEL AND PROCESSING METHOD The through-the-wall (TW) MIMO phased array radar desribed in this paper onsists of a four element transmitter (TX) array and a four element reeiver (RX) array. To minimize diret signal interferene (DSI) the arrays are separated by 1m, and offset by a height of 1.4m. Both the TX and RX arrays have equal element spaings of.4λ and 0.6λ respetively. This onfiguration synthesizes a 16 element virtual uniform linear array (ULA) with an equal element spaing of 0.3λ, and was hosen to avoid grating lobes by a large margin in the MIMO array proessing. Figure 1 illustrates the real and virtual synthesized elements in the antenna arrays. Consider N t elements in the TX array and N r elements in the RX array where t=4 and r=4. Their physial oordinates are denoted as x i TX, i = 1,,3,4, and the oordinates of the RX are denoted as x j RX, j = 1,,3,4. Assuming that the far field ondition is met, then the round trip distane P ij between the i th element in TX array to the j th element in RX array via a point satterer at point p is approximated as twie of the distane between the virtual element x ij = (x i TX + x j RX )/ and the point satter p [6, 7]: P ij TX i RX j p x p x (1) p xij Spaing between TX and RX array 0.6m 1 H H.4À 0.3X 0.6X 0.88m 0.036m 0.07m 0.864m m 016 TX ARRAY Virtual ARRAY RX ARRAY Figure 1. Plaement of the Phased array, inluding the TX Array, RX Array and the Virtual Array. Pro. of SPIE Vol

3 . Signal Model of the MIMO FMCW Radar The signal transmitted by the MIMO radar is a linear frequeny ontinuous wave s(t), with the arrier frequeny f, bandwidth B and sweep time T. For the nth hirp the signal s n (t) an be represented as the following: s ( t) A n o exp( j f ( t ( n 1) T)) exp( j( t ( n 1) T) ) u( t ( n 1) T), () where A 0 denotes the amplitude of the transmitted signal, μ = B/T is the hirp rate and u(t) is a retangular funtion defined as: 1,0 t T u( t). (3) 0, others Suppose the far field ondition is met, the return delay between virtual element x ij and a point satterer at the range R and the angle θ (orresponding to the boresight) an be represented as: where is the speed of the light. Then the n th reeived hirp signal an be expressed as: r R xij sin ij, (4) ij, n( 0 r n ij t t) A K s ( t ) z( ), (5) where the amplitude of the reeived signal is A, the refletion oeffiient of the target is K r and z(t)is the noise signal. Finally, after de-ramp proessing and low pass filtering, the beat signal an be expressed as [8]: b( t) s n t n 1 ) ( t) rij, n ( t) Kr A0 exp( j f ij ) exp( jij ) exp( jij T. (6) From Eq.(6), it an be seen that the beat signal onsists of three terms that inorporate both the range and angle information of the target. Firstly, the exp(jπμτ ij t) term represents the frequeny of the beat signal and indiates the range of the target as the follows f ij j4r ( t) exp( t). (7) It is worth noting that, due to the far field assumption, the return delay measured by all virtual elements τ ij are approximately equal to R/. Additionally, the range resolution is proportional to the inverse of the hirp duration T. Seondly, the seond and third phase terms exp(jπf τ ij )+exp(-jπμτ ij ) in Eq.(6) indiates the bearing of the target. In the appliation of TW imaging, within the detetable range, the square phase term exp(-jπμτ ij ) an be approximated as the onstant due to the relatively small distane differene among the array elements and the square funtion as well. Therefore, the phase term of the de-ramped signal an be expressed as: R x f ( ij sin ) ij 4f xij sin 4x ij sin C C, (8) where C is the onstant. The 4π/λ then validates the two-way path delay and suggests that designing the TX and RX arrays suh that the inter-element spaing in the virtual array is 0.3λ,would generate the minimum grating lobes around +-74 degree with the sanning angle from -45 to 45 degree. However, in the pratial TW appliations, as the path antenna would have the 3dB beam-width around 100degree, these grating lobes would be suppressed..3 Signal Proessing Method for -D Imaging The frequeny spetrum U(f) and the range information U(R) in Eq.(6) and Eq.(7), an be estimated by performing the Fast Fourier Transform (FFT) or any other mathing pursuit algorithms, suh as MUSIC on the beat signal b(t). Moreover, the steering vetor of orresponding to the i th element in TX and j th element in RX, denoted as a ij an be denoted as: Pro. of SPIE Vol

4 xij sin aij ( ) exp( j4 ). (9) In this way, the azimuth profile, or the angular profile P(θ) an be estimated as the following: Nt Nr P( ) a ( ) U( f ). (10) ij i1 j1 It is worth noting that FFT, MUSIC and any method related to the mathing pursuit an be utilized to obtain the range and the azimuth profiles. As illustrated in setion., the range resolution is related to the frequeny resolution, therefore, the range resolution is limited by the bandwidth of the hirp, as the following equation shows: 1 R T B. (11) The angular resolution is then dependent on both the aperture of the MIMO virtual array and the given angle θ aording to the following equation:, (1) A e d( NtNr 1)os where d is the inter-element spaing of the virtual array and A e is the virtual aperture of the antenna array. Furthermore, the azimuth resolution at the range R and the given angle θ an be estimated as:.4 Swithing Tehnique for Expanding the MIMO Channels R R azimuth R. (13) d( NtNr 1)os The radar onsists of a diret digital synthesizer (DDS) whih is onneted diretly to a four-way swith eah of whih feeds a radiating element Nt in the TX array. This enables MIMO signal transmissions through time-duplexing. Here, the sanning time and hirp duration are limited by the motion of the target of interest. Firstly, in terms of the swithing shemes, the target is restrited to a movement of less than a quarter of the wavelength whilst the system is sanning. That is to say, the maximum veloity of the target v is defined as: In this way, the upper bound of the hirp sweep time T u1 is limited by: v T ( Nt 1) / 4. (14) T u 1 4v max ( Nt 1). (15) The sweep period is also limited by the aeleration of the target to ensure ohereny. It is defined that the hange of the radian frequeny f d is over the hirp period should be less than the frequeny resolution, 1/T. The Doppler frequeny f d of the target is: Thus the hange of the Doppler frequeny within the hirp period an be expressed as: This gives the seond upper bound of the hirp period, T u as the following: v f d, (16) dv T a 1 f d T. (17) dt T Pro. of SPIE Vol

5 T u. (18) amax 3. HARDWARE IMPLEMENTATION In this setion, firstly, the phased array antenna is illustrated and seondly, the system arhiteture of the proposed swithed MIMO FMCW radar is shown. In general, the phased array radar onsists of 4-element TX antenna array and a 4-element RX antenna array. The radar is entered at.5ghz and the bandwidth is 00MHz, with the hirp time of 1ms. 3.1 Design of the Phased Array Antenna A path antenna was hosen due to its ease of fabriation and with the aeptable 90 degree azimuth beam-width. However, due to the 3% bandwidth limitation of the normal path antenna, we utilized the orrugation tehnique along the width of the path antenna to inrease the bandwidth. The following Figure shows the S11 refletion oeffiient plot with -7.18dB S11 between.4ghz and.6ghz, therefore the path antenna designed is suitable for the requirement of the radar. As the signal transmitted is in the range of.4ghz and.6ghz, the element spaing in the TX is set as the.4 times of the wavelength at the frequeny of.5ghz, whih is 0.88 meter and the element spaing in the RX is set to 0.6 of the wavelength, whih is 0.07 meter. M1 0.0 Ref: Att: 0.0 db 0 db.4ghz RBW: 10 khz Trae: Clear /Write TG Att: 0 db Suppr: Off -7.18dB M3.6GHz -1.07dB M1 S11 (intórp) Mag 10.0 M Center:.5 GHz Span: 1 GHz Figure. S11 measurement from the Vetor Network Analyzer of the single path antenna; measurement obtained here are -7.18dB at.4ghz and -1.07dB at.6ghz. The bandwidth is approximately 10MHz, with the -8dB S11 measurement. 3. System Arhiteture of the MIMO FMCW Radar The system arhiteture of the MIMO FMCW radar is illustrated in Figure 3. First, the VCO and PLL are utilized to generate a Continuous Wave waveform at GHz and this signal will be fed to the DDS for generating the linear hirp signal, with 1.04ms hirp period and 00MHz bandwidth, sweeping from.4ghz to.6ghz. After this, the bandpass filter and the amplifier are utilized to filter out undesired signals and amplify the desired signal. A splitter is then used to separate the signal into two hannels, one is used for the referene signal to the mixer, prepared for the de-ramp proessing, while the other is onneted to the four-way swith to expand the swithing array. The return signal is olleted from RX array and fed into the bandpass filter with the amplifier, and finally to the mixer. After that, the ative filter is utilized to lean and amplify the de-ramped signal related to the range information. Finally the de-ramped signal is digitized by the ADC and passed to the PC by USB able for the imaging proessing. Pro. of SPIE Vol

6 0m 3ppr^*. 1 m (approx.) X4 Time -Multiplexed Swithed Tx Array -E=.4A {zaaanl Fathesonu==/orwideróarrASiTh -EEE- --r7-40 m (approx.) ummy lement fsee separae ^) Mathed lengths X4 swith integratedon PCB MOW hng ba kfe unting.wate.4-.6ghz.4-.6ghz erctl DDS - GHz Clok frequeny GHz Synthesizer Figure 3. System arhiteture of the proposed MIMO Radar 4. EXPERIMENTAL RESULTS AND ANALYSIS In this setion, we present the experimental results of both single input single output (SISO) and MIMO set up in the following senario. In setion 4.1, the experiment set-up and in setion 4., the results of both the SISO and MIMO proessing are shown and analyzed. 4.1 Experiment Set-Up The experiment is set up in the luster room and the orridor in University College London. As shown in Figure 4, the TX and RX arrays are 1.5m away from the 10m depth plasti wall perpendiularly. The TX and RX arrays are tha0.4m away from eah other. In addition, there are five test positions ranging from P1 to P5 in the orridor in ase of the lutters in the luster room. It is worth noting that the orridor is approximately.m long. For eah position in the experiment, a target is walking in the orridor as shown in the double arrows, from one end of the wall to the other end perpendiularly to the wall as well. A A A A A m 04.m ",10m lm 0.8m 1.5m W W l. v 0.4m TX Offie Figure 4. Experiment Set Up; The radar is plaed 1.5m away from the 10m depth wall and the Tx and Rx part of the radar is 0.4m. The two arrows next to the position indiator represent the walking route of the target at 5 positions. The walking route of the target at eah position is from one end of the wall to the other end. The detailed geometry an be shown in this figure. Pro. of SPIE Vol

7 4. Results First we present the range-time results without using the MTI filters as shown in Figure 5 (a) in the Single Input Single Output (SISO) mode. It is worth noting that this result is proessed by only one TX hannel and RX hannel from the same experiment data as shown in the MIMO results. In Figure 5 (a), we an learly see three stati lines in the rangetime results, the first of whih is the DSI near 0m. The strongest line at approximately 1.5m is attributed to the first wall between the orridor and the offie. There is a weaker stati line around 3.85m whih we regarded as the response from the seond wall. To detet the moving target behind the wall, we applied a Finite Impulse Response (FIR) digital filter on the slow time de-ramp signal and then performed the FFT based array proessing. The relevant moving target result for Figure 5 (a) is shown in Figure 5 (b) and (), where Figure 5 (b) learly shows the walking pattern among the 5 different positions. The final part of the MTI range-time result between 35 and 45 seonds indiates the result of a human target walking aross the orridor to get bak to the offie. Figure 5 () is the zoomed-in figure showing the detailed signatures from the positions P and P3. It ould be shown that in Figure 5 (b) that the power of the range-time results in P and P3 are stronger than other positions and P1 exhibits the lowest power. We hypothesize that this ours beause P and P3 are loser to the boresight of the antenna arrays. In addition, Figure 5 () displays some walking pattern signatures from 10s to 15s in the range of 4m to 6m whih is out of the physial limits of the orridor. We explain this in terms of the double boune effet: as P3 is lose to the boresight of the antenna array, the double bouned signal might exhibit higher powers. Next, the MIMO -D images of the 5 positions are shown in Figure 6 (a) to (e) respetively. Here, the plots are normalized and shown with -5dB to 0dB dynami range. It an be seen from the 5 figures that different positions exhibit different angle estimation results. It is worth noting that the signal lutter ratio in Figure 6 (a) is lower as there is less power refleted from the P1. This orresponds to the SISO result shown in Figure 5 (b). Finally, as shown in Figure 7 we present the MIMO -D image of two targets walking in the opposite diretion. In this senario, the right target exhibits less power than the left target as it s further away from the radar. (a) (b) Pro. of SPIE Vol

8 () Figure 5. Single Input and Single Output Range-Time Plot of senario 1: (a) result without wall mitigation tehnique (b) result with moving target indiator filter applied () Zoomed-in Results of figure 5 (b), for targets walking in P and P3. (a) (b) Pro. of SPIE Vol

9 () (d) Figure 6. Through-Wall Imaging Results Among Five Different Positions in the hall: (a) P1, (b) P, () P3, (d) P4, (e) P5. (e) Pro. of SPIE Vol

10 (a) (b) Figure 7. MIMO Imaging Results of Two Moving Targets with the right target approahing to the wall and the left target leaving away from the wall. 5. CONCLUSION In this paper, we demonstrated a time-duplexed MIMO phased array FMCW radar design and the imaging results in a through-wall senario. We fabriated the Tx and Rx array using the path antenna arrays and the FMCW radar is entered at.5ghz and with bandwidth of 00MHz. It suessfully demonstrated the imaging apability of the moving target in a through-wall senario. Future work would fous on investigating imaging apabilities among various wall types and through-wall imaging apabilities in the stand-off distanes. REFERENCES [1] J. E. Peabody, G. L. Charvat, J. Goodwin, and M. Tobias, "Through-wall imaging radar," Linoln Lab. J, vol. 19, pp. 6-7, 01. [] R. Dilsavor, W. Ailes, P. Rush, F. Ahmad, W. Keihel, G. Titi, et al., "Experiments on wideband through-the-wall radar imaging," in Defense and Seurity, 005, pp Pro. of SPIE Vol

11 [3] G. B. Barrie, "Through-wall syntheti aperture radar (TWSAR) 3D imaging: algorithm design," DTIC Doument004. [4] Y. Huang, P. V. Brennan, D. Patrik, I. Weller, P. Roberts, and K. Hughes, "FMCW based MIMO imaging radar for maritime navigation," Progress In Eletromagnetis Researh, vol. 115, pp , 011. [5] J. Li, P. Stoia, and X. Zheng, "Signal synthesis and reeiver design for MIMO radar imaging," IEEE Transations on Signal Proessing, vol. 56, pp , 008. [6] Y. Qu, G. Liao, S.-Q. Zhu, X.-Y. Liu, and H. Jiang, "Performane analysis of beamforming for MIMO radar," Progress In Eletromagnetis Researh, vol. 84, pp , 008. [7] J. H. Ender and J. Klare, "System arhitetures and algorithms for radar imaging by MIMO-SAR," in Radar Conferene, 009 IEEE, 009, pp [8] P. Laomme, Air and spaeborne radar systems: An introdution: William Andrew, 001. Pro. of SPIE Vol