DEVELOPMENT of homeland security systems allowing

1746 IEEE SENSORS JOURNAL, VOL. 10, NO. 11, NOVEMBER 2010 Detecting Hidden Objects on Human Body Using Active Millimeter Wave Sensor Boris Kapilevich, Senior Member, IEEE, and Moshe Einat Abstract The paper describes millimeter wave (MMW) sensors designed for detecting both metallic and nonmetallic objects placed on a human body and hidden under clothes. The sensor is based on the synchronized detection principle and estimating a power of back-scattered signal from hidden objects. Time-gating algorithm combined with preliminary determined threshold level has been implemented in order to reach detection probability of 90% or more for metal and plastic hidden objects at the distance up to 3 m. Index Terms Concealed weapon detection, homeland security, millimeter wave (MMW), sensors. I. INTRODUCTION DEVELOPMENT of homeland security systems allowing remote searching and detecting various dangerous objects hidden under clothing is important for protection against terrorist threats. In order to provide personnel screened for undesirable hidden objects as well as concealed weapons in airports, train stations, embassies, and other secured buildings and locations, conventional searching technologies rely almost entirely on metal detectors to scan a person for concealed weapons and X-ray systems to screen hand-carried items. However, they are not reliable when it is necessary a remote detecting explosives or other nonmetallic weapons. Various types of the sensors and imagers operating in microwave (MW), MMW, and terahertz (THz) bands have been proposed during the last decade. They can be classified into the two main categories passive [1] [4] and active [5] [7]. The passive imagers and detectors are based on analysis of thermal radiation emitted by the targets under screening. Since the emitted power is very low sensitive receivers must be employed in passive sensors. On the other hand, the active imagers and detectors need an additional transmitter illuminating screened targets. Hence, they do not require sensitive receiver but specula reflections may drastically destroy the reconstructed image as well as related information about the targets. The choice of operating frequency is a critical issue for both passive and active sensors. Indeed, MMW can penetrate to a reasonable depth through various materials (better than infrared radiation), and have mm scale resulting in mm scale basic resolution in imaging process (better than MW frequencies). However, MW receivers are basically more sensitive and demon- Manuscript received May 05, 2009; revised December 11, 2009 and February 23, 2010; accepted April 21, 2010. Date of publication June 07, 2010; date of current version September 15, 2010. The associate editor coordinating the review of this paper and approving it for publication was Dr. Dwight Woolard. The authors are with the Department of Electrical and Electronics Engineering, Ariel University Center of Samaria, Ariel, Israel 40700 (e-mail: boriskap@ariel.ac.il; moshee@ariel.ac.il). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/JSEN.2010.2049350 strate better noise performance. Imagers and sensors in THz band [8] [11] have better resolution in comparison with MMW versions but they suffer from low sensitivity. As a result THz detecting systems can operate at short distance, typically, less than 1 m. Examples of active imagers and detectors operating in MW, MMW, and THz bands have been reported in [12] [16]. However, many of them are still not mature. Existing prototypes are unique and expensive. The devices are complicated and must be positioned sometimes in one or two dimensions requiring an expensive positioning system. Another approach is based on the array of mm-wave receivers each of which is serving as an independent pixel. The disadvantage of this arrangement is high cost. This paper presents a low-cost MMW active nonimaging sensor operating in W-band. Experiments demonstrated that both metallic and nonmetallic objects (plastics, ceramics, woods, carton, soap, and cheese, as well as their combinations) were detected at the distance up to 3 m. The W-band sensor is based on synchronized amplitude modulated (AM) detector analyzing the power back-scattered by a target. Since the human body can be considered as a background material for the hidden object, the edge diffraction on boundaries background target creates backscattered peaks during scanning process. Time-gating algorithm combined with the preliminary determined threshold level is suggested to increase the detection probability and to reduce the false alarm rate (FAR). Its implementation has demonstrated the detection probability about 90% or more for metal and plastic hidden objects for a distance up to 3 m. II. MMW SENSOR BASED ON SYNCHRONIZED AM DETECTOR The mm-wave sensor using the synchronized AM detection is shown schematically in Fig. 1. It consists of transmitting and receiving units. The two identical W-band horn-lens antennas with 5 beamwidth and gain about 30 db are used for transmission and reception. The CW Gunn diode generator having the output power 13 dbm at 94 GHz is modulated by frequency 1 KHz resulting AM signal at the generator s output. The signal back-scattered by the target is captured by the antenna and delivered to millimeter wave zero biased Schottky detector with own input sensitivity 800 mv/mw. The detector, utilizing MIC technology and Schottky Barrier Beam Lead Diodes, provides a very economical solution for power detection over the 18 to 330 GHz range. Both high sensitivity and full waveguide bandwidths are achieved simultaneously without external dc bias or adjustments. The detector has a flat frequency response, as its sensitivity shows minor variation over the entire waveguide band [19]. After amplifying and 1530-437X/$26.00 2010 IEEE

KAPILEVICH AND EINAT: DETECTING HIDDEN OBJECTS ON HUMAN BODY USING ACTIVE MILLIMETER WAVE SENSOR 1747 Fig. 3. Calculated R as a function of for different values of MDS. III. LINK BUDGET AND RADAR CROSS-SECTION (RCS) EFFECTS In order to estimate the operating range of this sensor, the standard radar range equation can be used [17] Fig. 1. Simplified block-diagram of the mm-wave sensor based on synchronized detection. Fig. 2. General view of W-band sensor. filtering by BPF with 1 KHz center frequency the signal goes to the input of analog-to-digital converter (ADC). The ADC is integrated with LabView interface which is matched with detecting algorithm written using MatLab code. The sensor was assembled as the fully autonomic unit supplied by 15 V dc. Its general view is depicted in Fig. 2. All the W-band and low-frequency components of the sensor are commercially available at relatively low cost. The overall systems noise performance is characterized by at a distance from the target 3 m. where is the maximum distance to the target; is the effective area of the antenna; is the RCS of the target; and are the transmitted and received powers; and is the gain of the transmitted antenna. Fig. 3 shows the calculated as a function of for different values of minimum detectable signal (MDS) varying between with an assumption that, and. We can see that several meters of operating range of the sensor can be reached with quite moderate level of MDS corresponding to typical sensitivity of the Schottky detector with a post-amplifier. The radar cross section of the detected target is critical parameter determining a detection probability. Indeed, if the RCS of human body with and without of a hidden object are poorly distinguished the detection probability degrades drastically while FAR is increased. The RCS simulations of a human body with hidden objects needs a lot of information about constitutive parameters of a human skin, clothes, the hidden object itself, as well as accurate shaping specifications of a body and target placed on the body. We have simulated RCS of idealized configuration shown in Fig. 4(a) using the CST Microwave Studio [18]. Conductivity of the equivalent human body was and real part of dielectric constant, the material simulating the gun was supposed as a perfect conductor. We assume that the target is illuminated by plane wave in normal direction. Fig. 4(b) shows the scattered E-field for linear polarization without the gun, while Fig. 4(c) corresponds to the same situation with the gun. We can see considerable differences in the far-field RCS (1)

1748 IEEE SENSORS JOURNAL, VOL. 10, NO. 11, NOVEMBER 2010 Fig. 5. Experimental setup used for simulation of human body plastic barrel 20-l with water and plastic box filled with plastic granules. Fig. 4. Simulated RCS of the human body with and without gun. (a) Configuration the body and gun. (b) RCS pattern without gun. (c) RCS pattern with gun. patterns for both situations. However, when real antenna illuminates the target only a part of the reflected energy is backscattered by the hidden object while the rest is reflected by the human body. Depending on distance, beam width and angle of incidence the contribution of both scattered components is varied leading to uncertainty of the detection procedure when they are comparable. IV. MEASUREMENT OF THE BACK-SCATTERED SIGNAL OF SOME TARGETS Accurate theoretical prediction of RCS of real-world targets is difficult to perform due to a variety of their shapes and material combinations. So, experimental study of back-scattered signal should be reasonable for some specific situations. Various targets made of different nonmetallic materials and having different shapes were tested to estimate their back-scattering properties. First of all, we have studied various combinations of a human body plastic-carton. The metallic target (such as a gun) and explosive will be considered below. To simulate the human body, we have employed a 20-l plastic barrel filled with tap water at room temperature and measured video output signal when the AM detecting sensor is placed at the distance 1 m from the target, as shown in Fig. 5. The hidden object was mounted using transparent foam holder in a front of the barrel. The sensor s antennas were oriented in normal direction to the hidden object scanning the target + object in azimuth and elevation around the normal direction. An Az-El programmable positioner was used to record the scattered signatures of a variety of targets given in the Table I. They include different shapes of plastic, carton, Teflon, ceramic, cheese, soap, etc. The parasitic reflection (noise background) from walls of the laboratory did not exceed 5 mv. Since the purpose of experiments is a detection of hidden objects without imaging, we have recorded their signatures as a function of scanning time. Fig. 6 shows the records for the water-filled plastic barrel (20-l) as an example of the human body background. The output signal in volts is plotted as a function of recorded time in azimuth-elevation scanning mode. Similar records were done when the plastic box containing a metal plate was placed in the front of this barrel; see Fig. 7. The output signal in the second case is much higher and it has different pattern compared with the first case. The same scanning process was used to estimate a spatial resolution of the sensor. Two metallic rods with diameter, at the distant 10 cm between them and 50 cm distance from the sensor have been tested; see Fig. 8(a). The proper recorded signature is shown in Fig. 8(b) for single forward and backward scans. We can see that the spatial resolution of 10 cm or sometimes less is achievable for the sensor considered. We have also employed the surrogate of C4 as an explosive material. It is quite lossy media with low real part of dielectric constant. Being hidden on the human body, it absorbs a part of

KAPILEVICH AND EINAT: DETECTING HIDDEN OBJECTS ON HUMAN BODY USING ACTIVE MILLIMETER WAVE SENSOR 1749 TABLE I MEASURED SIGNAL AT THE RECEIVER S OUTPUT AT THE DISTANCE 1m.THE OBJECTS ARE PLACED IN THE FRONT OF CYLINDRICAL PLASTIC BARREL FILLED WITH TAP WATER Fig. 6. Measured reflectance signature of the water-filled plastic barrel (20-l) as a function of time in azimuth-elevation scanning mode. Fig. 7. Measured scattered signature of the water-filled plastic barrel (20-l) with the plastic box containing a metal plate placed in the front as a function of time in azimuth-elevation scanning mode. illuminating RF power. As a result, the reflected signal is less in comparison with the same signal reflected by a human body without C4 ( empty man ). This effect is illustrated by Fig. 9. It should be noticed that the recorded signatures are dependent on orientation of the object leading to change of the reflected amplitude. Fig. 10 shows the person with explosive and gun participating in the experiments. V. COMBINED TIME-GATING/THRESHOLD LEVEL ALGORITHM Doing a lot of experiments with different targets we have noticed that when the antenna s beam crossed the boundary between a human body and the edge of target, a sharp peak response has appeared. This effect can be used for developing algorithm enhancing the detection probability and reducing FAR. Both time-gating characteristic and RF reflected power threshold level must be taken into consideration simultaneously in taking a final decision about true or no detection. Fig. 11 illustrates typical output signal of the sensor after sampling corresponding to scanning area containing 300 sampling points. The person participating in the experiment was located within the range of 50 200 sampling points; see Fig. 11(a). The same record for the gun hidden under shirt is depicted in Fig. 11(b). The distance between tested target and the sensor was 3 m. It should be pointed out that clutters such as parasitic reflections

1750 IEEE SENSORS JOURNAL, VOL. 10, NO. 11, NOVEMBER 2010 Fig. 8. (a) Experimental setup used for estimating spatial resolution and (b) its signature. The distance between metallic rods is 10 cm. from walls and other objects located in a proximity to scanning area may be a source of false alarm. To avoid such undesirable clutter s effect, we have employed only a part of the scanning sector within 50 200 sampling intervals corresponding to actually active scanning sector. Comparing both signatures we can observe the unique peak on the Fig. 11(b) between 170 175 sampling points. The magnitude of this peak exceeds 0.15 V. Assuming that the given peak corresponds to the hidden objects we may enhance the detection procedure by adding the time-gating and threshold magnitude of reflected power level in the processing algorithm (4 sampling intervals and 0.15 V in our case). Based on these features, the following detecting algorithm has been suggested. a) Data is collected from a certain scan and all the samples that are below the selected threshold are ignored as nodetection. b) When a sample is above the threshold, a time gate of samples is activated. If the numbers of samples that are above the threshold are or less, then detection is declared true. c) Else, i.e., the signal stays above the threshold more then samples; it is regarded as not sharp echo and also ignored. The algorithm was operated with the samples in range 50 200 which corresponded to the boundaries of the active checked area. Readings below and above are coming from areas outside of the human target are automatically deleted by the algorithm. Sometimes, there is a parasitic reflection from empty human body satisfying the both predefined criteria Fig. 9. Comparison reflected signals from a human body without and with an explosive. resulting in false alarm. Also, the reflection from human body with hidden object not satisfying these criteria may lead to wrong decision in detecting this object. Therefore, the total percentage of false alarm and true decisions may exceed 100%. Since the decision is on relative measurement basis, it is important to set uniform reference level. This is done by performing calibration of each experimental session under the same conditions where the parameters are quite constant. Still, there may be a drift in the parameters. In order to avoid that, two objects were places at the ends of the target plan: a corner reflector and absorbing material. At the beginning, readings are taken from these objects and regarded as a maximal reflection and noise level. Then, during the experiment, the sensor is aimed periodically to these objects and the reflections are monitored. If a change is noticed it is compensated in the amplifying chain. Therefore, a uniform reference level is obtained. When a static target (a simulating barrel for example) was measured, the obtained readings were repetitive. However,

KAPILEVICH AND EINAT: DETECTING HIDDEN OBJECTS ON HUMAN BODY USING ACTIVE MILLIMETER WAVE SENSOR 1751 TABLE II SUMMARY OF THE MEASURED THRESHOLD MAGNITUDES AND TIME GATES FOR TEN DETECTION TRIALS OF A HUMAN BODY WITH AND WITHOUT HIDDEN GUN Fig. 10. Person with plastic explosive and gun participating in the experiments. TABLE III PERCENTAGES OF TRUE AND FALSE DECISIONS ESTIMATED WITH THE THRESHOLD MAGNITUDES AND TIME GATES CRITERIA Fig. 11. Comparison reflected signals from a human body (a) without a gun and (b) with the hidden gun. when measurements were taken on a real target (a human), the readings were changing in sequent measurements of the same scenario. It was found that when a human is measured, although told to stand still, he moves. There were also minor variations due to uncontrolled movements and a person still moves even when asked to hold his breath. These variations are within the millimeter range and detectable with the sensor. Therefore different readings are obtained each time. In order to verify the above approach, we have done 10 search trials of the human body without the gun and the same 10 trials of the human body with the hidden gun, when the man is asked not to move. Table II summarizes the measured threshold magnitudes and time gate (sampling intervals). Using these data the true false percentage has been estimated on the basis of the following settings: threshold magnitude: 0.13 0.15 V; sampling interval: 1 or 2 points or not specified. The percentages of true and false decisions estimated with the above criteria are given in Table III. One can see that employing only backscattered magnitude without time-gating leads to high levels of false alarm. However, adding timegating with minimizing the threshold magnitude permits to reduce FAR and increase true decisions. For example, in the case of magnitude and time gate,we have obtained " "!. VI. CONCLUSION The millimeter wave sensor designed for detecting both metallic and nonmetallic objects placed on a human body and hidden under clothes has been described. Time-gating/threshold level algorithm has been suggested and realized practically. It has permitted to improve the detection probability up to 90% or more reducing FAR to 10% or less.

1752 IEEE SENSORS JOURNAL, VOL. 10, NO. 11, NOVEMBER 2010 ACKNOWLEDGMENT The authors would like to thank Dr. A. Luzzato for fruitful discussions, B. Litvak, A. Shulzinger, B. Kostjakovsky, and A. Levitsky for their assistance in experiments, and D. Hardon and R. Regavim for their active coordinating efforts. REFERENCES [1] R. Appleby, Passive millimetre-wave imaging and how it differs from terahertz imaging, Phil. Trans. R. Soc. Lond. A, vol. 362, pp. 379 394, 2004. [2] L. Yujiri, M. Shoucri, and P. Moffa, Passive millimeter wave imaging, IEEE Microwave Mag., vol. 4, pp. 39 50, Sep. 2003. [3] V. A. Manasson, L. S. Sadovnik, R. Mino, and S. Rodionov, Novel passive millimeter wave imaging system: Prototype fabrication and testing, Proc. SPIE, vol. 5070, pp. 2 13, Apr. 21, 2000. [4] B. Kapilevich, B. Litvak, M. Einat, and O. Shotman, Passive mm-wave sensor for in-door and outdoor homeland security applications, in Proc. 2007 Int. Conf. Sensor Technol. Appl., pp. 20 23. [5] D. M. Sheen, D. L. McMakin, and T. E. Hall, Three-dimensional millimeter-wave imaging for concealed weapon detection, IEEE Trans. Microw. Theory Tech., vol. 49, no. 9, pp. 1581 1592, Sep. 2001. [6] D. M. Sheen, D. L. McMakin, W. M. Lechelt, and J. W. Griffin, Circularly polarized millimeter-wave imaging for personnel screening, Proc. SPIE, vol. 5789, pp. 117 126, Apr. 21, 2005. [7] B. Kapilevich, M. Einat, B. Litvak, A. Shulsinger, and E. Nehemia, Experimental study of in-door mm-wave imaging resolution limits, presented at the Workshop Nefertiti-2005, Brussels, Belgium, paper #111. [8] R. D. Boykin, A Brief Overview of T-ray (THz) Imaging May 13, 2005. [Online]. Available: http://ric.uthscsa.edu/personalpages/lancaste/di2_projects_2005/t-ray.pdf [9] K. B. Cooper, R. J. Dengler, G. Chattopadhyay, E. Schlecht, J. Gill, A. Skalare, I. Mehdi, and P. H. Siegel, A high-resolution imaging radar at 580 GHz, IEEE Microw. Wireless Comp. Lett., vol. 18, no. 1, pp. 64 66, Jan. 2008. [10] J. C. Dickinson, T. M. Goyette, A. J. Gatesman, C. S. Joseph, Z. G. Root, R. H. Giles, J. Waldman, and W. E. Nixon, Terahertz imaging of subjects with concealed weapons, Proc. SPIE, vol. 6212, pp. 62120Q- 1 62120Q-12, 2006. [11] M. C. Kemp, P. F. Taday, B. E. Cole, J. A. Cluff, A. J. Fitzgerald, and W. R. Tribe, Security applications of terahertz technology, Proc. SPIE, vol. 5070, pp. 44 52, 2003. [12] D. T. Petkie, F. C. DeLucia, C. Casto, P. Helminger, E. L. Jacobs, S. K. Moyer, S. Murrill, C. Halford, S. Griffin, and C. Franck, Active and passive millimeter and sub-millimeter-wave imaging, Proc. SPIE, vol. 5989, pp. 598918-1 598918-8, 2005. [13] G. Tryon, Passive millimeter-wave object detection and people screening, presented at the 2007 SURA Terahertz Appl. Symp., Washington, DC, Jun. 6 8, 2007. [14] R. W. McMillan, N. C. Currie, D. D. Ferris, and M. C. Wicks, Jr., Concealed weapon detection using microwave and millimeter wave sensors, in Proc. Microw. Millimeter Wave Technol. (ICMMT 98), 1998, pp. 1 4. [15] Tadar Millimetre-Wave People Screening System. [Online]. Available: http://www.smithsdetection.com/eng/1371.php [16] Multi-Threat Detection Solutions: Product Informationl Safe Zone Systems. [Online]. Available: http://www.safezonesystems.com/productinfo.html [17] E. A. Wolff and R. Kaul, Microwave Engineering and Systems Applications. New York: Wiley, 1988, p. 52. [18] CST Microwave Studio. [Online]. Available: http://www.cst.com [19] [Online]. Available: http://www.millitech.com/pdfs/specsheets/ IS000026-DXP.pdf Boris Kapilevich (M 94 SM 97) is a Dr.Sc. Technology and Professor of Microwave Engineering at the Ariel University Center of Samaria, Israel. He has authored or coauthored four books, holds 14 patents, and published over 150 papers dedicated both active and passive microwave devices, theory of guided waves, electromagnetics, measurements and material characterization, and mm-wave imaging, as well as some other related topics. Moshe Einat received the B.Sc., M.Sc., and Ph.D. degrees from Tel-Aviv University, Tel-Aviv, Israel. He is specialized in high-power microwave electron tubes integrated to ferroelectric cold electron guns. He works on various experimental projects in the area of microwave, millimeter and THz waves, both from the generation as well as the application point-of-view. He is a member of the engineering faculty at the Ariel University Center of Samaria, Israel.