A 24 GHz FM-CW Radar System for Detecting Closed Multiple Targets and Its Applications in Actual Scenes

Transcription

1 2016 by the authors; liensee RonPub, Lübek, Germany. This artile is an open aess artile distributed under the terms and onditions of the Creative Commons Attribution liense ( Open Aess Open Journal of Internet of Things (OJIOT) Volume 2, Issue 1, ISSN A 24 GHz FM-CW Radar System for Deteting Closed Multiple Targets and Its Appliations in Atual Senes Kazuhiro Yamaguhi A, Mitumasa Saito B, Takuya Akiyama A, Tomohiro Kobayashi A, Naoki Ginoza A, Hideaki Matsue A A Tokyo University of Siene, Suwa, Toyohira, Chino, Nagano, Japan, , yamaguhi@rs.tus.a.jp, GH14601@ed.tus.a.jp, GH15607@ed.tus.a.jp, G112033@ed.tus.a.jp, matsue@rs.suwa.tus.a.jp B CQ-S net In., Torigoe 7-8, Kanagawa-ku, Yokohama-shi, Kanagawa, Japan, , saitoh@kpe.biglobe.ne.jp ABSTRACT This paper develops a 24 GHz band FM-CW radar system to detet losed multiple targets in a small displaement environment, and its performane is analyzed by omputer simulation. The FM-CW radar system uses a differential detetion method for removing any signals from bakground objets and uses a tunable FIR filtering in signal proessing for deteting multiple targets. The differential detetion method enables the orret detetion of both the distane and small displaement at the same time for eah target at the FM-CW radar aording to the reeived signals. The basi performane of the FM-CW radar system is analyzed by omputer simulation, and the distane and small displaement of a single target are measured in field experiments. The omputer simulations are arried out for evaluating the proposed detetion method with tunable FIR filtering for the FM-CW radar and for analyzing the performane aording to the parameters in a losed multiple targets environment. The results of simulation show that our 24 GHz band FM-CW radar with the proposed detetion method an effetively detet both the distane and the small displaement for eah target in multiple moving targets environments. Moreover, we develop an IoT-based appliation for monitoring several targets at the same time in atual senes. TYPE OF PAPER AND KEYWORDS Regular researh paper: 24 GHz, FM-CW radar, distane measuring, small displaement measuring, multiple targets detetion, tunable FIR filtering 1 INTRODUCTION Reently, rapidly evolving wireless ommuniation tehnologies enable us to obtain various piee of information from a lot of manufaturers in the wide range of the tehnology field. The IoT (Internet of things) is a key topi in network field for developing systems and servies, and it enables various physial devies to ollet various piees of information suh as sensor information. Radar has been used as a sensing appliation to detet in wide range of areas suh as on the ground, on the sea, in the air, and in spae. Radar systems an detet various piees of information aording to the appliation area by transmitting and reeiving using radio waves. Twenty-four GHz band radar systems are based on 1

2 Open Journal of Internet of Things (OJIOT), Volume 2, Issue 1, 2016 Figure 1: Blok diagram of FM-CW radar system the ARIB standard T73 [2] in Japan as sensors for deteting or measuring mobile objets for a speified low power radio station. And the 24 GHz band radar system an be applied to various fields suh as seurity and medial imaging in indoor and outdoor environments. Various radar systems have been reported [11, 16, 6, 19]. Pulsed radar systems an measure the period between the transmitted and reeived signals. Pulsed radars an detet the distane in the far field. However, the targets in the near field annot be deteted orretly by a pulsed radar. Doppler radar systems an measure the frequeny differene between the transmitted and refleted signals. A Doppler radar an detet the moving veloity of the target, but it annot detet the distane of the target. FM-CW (Frequeny-Modulated Continuous- Wave) radar systems [12, 3, 13, 17, 1] are the most widely used for deteting the distane of the target objet in the near field and the small displaement of the target. In MIMO (Multiple-Input Multiple-Output) radar systems [9, 5], several radars are used at the same time for measuring the target s information. In these radar systems, eah radar has both advantages and disadvantages. Therefore, the types of radars are determined aording to the neessary information and the situation. In this paper, we foused on an FM-CW radar system to detet human breathing in the near field. In a previous study, we reported the design, performane analysis, and appliations when using a 24 GHz band radar system to detet both the distane from the radar and the small displaement aused by human breathing [21]. The radar system ould orretly detet both the distane to the human from the radar and the small displaement aused by the human breathing. However, it was diffiultto detet the distanes and displaements for multiple humans at the same time. To detet the distanes and displaements in a losed multiple target environment, we proposed a detetion method for signal proessing with a tunable FIR (Finite Impulse Response) filter [4] in the FM-CW radar system [20]. Furthermore, a performane analysis for the FM-CW radar system was performed using omputer simulations. In this paper, the proposed detetion method onsiders both the differential detetion and the tunable FIR filtering to detet both the distanes and small displaements in single and losed multiple target environments. The results of the omputer simulations show that the proposed FM-CW radar system with FIR filtering ould detet targets orretly when the distane between targets was greater than 0.5 m. Moreover, appliations based on the IoT for deteting human movements in atual senes were desribed by using the developed FM-CW radar system. The IoT-based radar system ould ollet information on the breathing of several people by using network transmission. The results show that the FM-CW radar system ould detet human breathing in atual senes, and the IoTbased radar system would monitor the human breathing remotely. The rest of this paper is organized as follows. In Setion 2, we desribe the priniple of the FM-CW radar system and its basi performane in a single target environment. Setion 3 presents the proposed detetion method in a single target environment, and Setion 4 presents the proposed detetion method in a multiple target environment. In Setion 5, we disuss the results of the omputer simulation and field experiments. Finally, Setion 6 onludes this paper. 2 FM-CW RADAR SYSTEM In this setion, a basi system model of the FM-CW radar and basi performane for measuring a distane and a small displaement of a target in a omputer simulation are desribed. 2.1 Priniple of FM-CW Radar Figure 1 shows the blok diagram of an FM-CW radar system [15]. FM-CW radar is a type of radar that 2

3 K. Yamaguhi et al.: A 24 GHz FM-CW Radar System for Deteting Closed Multiple Targets and Its Appliations in Atual Senes Figure 2: Sawtooth frequeny modulation transmits a ontinuous arrier modulated by a periodi funtion, suh as a sawtooth wave, to provide the range data. In Figure 1, VCO means voltage-ontrolled osillator, D/A means digital-to-analog onversation, BPF means bandpass filter, and A/D means analog-todigital onversation. In the FM-CW radar system, the frequeny modulated signal at the VCO is transmitted from transmitter Tx; then, signals refleted from the targets are reeived at reeiver Rx. The transmitted and reeived signals are multiplied by a mixer, and beat signals are generated by multiplying the two signals. The beat signals pass through a low pass filter, and an output signal is then obtained. In this proess, the frequeny of the input signal is varied with time at the VCO. The modulation waveform has a linear sawtooth pattern [14] (shown in Figure 2). This figure illustrates the frequeny-time relation in the FM-CW radar; the red line denotes the transmitted signal, and the blue line denotes the reeived signal whih has some delay time. Here, f 0 denotes the enter frequeny, f w denotes the bandwidth of sweep frequeny, and t w denotes the sweep time. The transmitting signal V T (f, x) is represented as V T (f, x) = Ae j 2πf x, (1) where f denotes the frequeny at a ertain time, x denotes the distane from the transmitter whose position is x = 0, A denotes the amplitude value of the transmitted signal, and denotes the speed of light. The refleted signal V R (f, x) is represented as V R (f, x) = K Aα k γ k e jϕ k e j 2πf (2d k x), (2) where γ k and ϕ k are the refletive oeffiients for the amplitude and phase of the kth target, respetively. α k denotes the amplitude oeffiient for transmission loss from the kth target, and d k is the distane between the transmitter and the kth target. Here, at the reeiver whose position is x = 0, Equation (2) is rewritten as V R (f, 0) = K Aα k γ k e jϕ k e j 2πf (2d k). (3) The beat signal is generated by multiplying the transmitted signal in Equation (1) by the reeived signal in Equation (3) at the position x = 0. After passing through an LPF, the output signal V out (f, 0) is generated by V out (f, 0) = K A 2 α k γ k e jϕ k e j 4πfd k. (4) By using signal proessing, the distane and displaement for the target are given from the generated output signal in Equation (4). By using Fourier transform, the distane spetrum of the output signal P (x) is alulated as follows. P (x) = = f0+ fw 2 f 0 fw 2 f0+ fw 2 f 0 fw 2 V out e K K = A 2 α k γ k e jϕ k 4πf j df A 2 α k γ k e jϕ k e j 4πfd k e f0+ fw 2 f 0 fw 2 [ K = A 2 α k γ k e jϕ k e j 4πf 0(d k x) 4πfx j e j 4πf(d k x) df df { } ] 2πfw (d k x) f w sin. (5) In this equation, the funtion of sin (x) denotes sin (x) = sin x x. (6) The amplitude value of the distane spetrum P (x) in 3

4 Open Journal of Internet of Things (OJIOT), Volume 2, Issue 1, 2016 Equation (5) is given as P (x) = A 2 K α k γ k e jϕ k e j 4πf 0(d k x) { } 2πfw (d k x) f w sin [ A 2 f w K α k γ k { } ] sin 2πfw (d k x), (7) Table 1: Parameters in omputer simulations Parameters Values Center frequeny f (GHz) Bandwidth of sweep frequeny f w 50, 100, 200, 400 (MHz) Sweep time t w 1024 (µs) Sampling time for sweep 0.1, 1, 10 (µs) Number of FFT points 4096 Window funtion hamming and we have equality if and only if the phase omponents φ k + 4πf0(d k x) for all of k are equal. Here, we assumed that the number of targets is 1. The distane spetrum in Equation (5) is rewritten as [ P (x) = A 2 α 1 γ 1 e jϕ1 e j 4πf 0 (d 1 x) { } ] 2πfw (d 1 x) f w sin, (8) and the amplitude value of the distane spetrum is given as { } P (x) = A 2 2πfw (d 1 x) α 1 γ 1 f w sin. (9) This equation indiates that the distane of the target is generated by the amplitude value of the distane spetrum. The phase value of the distane spetrum P (x) is represented as P (x) = ϕ 1 + 4πf 0 (d 1 x) = θ 1 (x). (10) Here, beause θ 1 (x) satisfies π θ 1 (x) π, the displaement of the target is ( π ϕ 1) 4πf 0 d 1 (π ϕ 1) 4πf 0. (11) If the phase value satisfies φ 1 = 0, Equation (11) is rewritten as 3.11 [mm] d [mm] with f 0 = [GHz]. That is, a small displaement of the target within ±3.11 [mm] is generated by the phase value of the distane spetrum. On the other hand, the maximum distane for measuring d max is f w f = [Hz], t w /t s d max = [m], (12) 4 f Figure 3: Range resolution for distane spetrum aording to bandwidth of sweep frequeny f w Figure 4: Error value for distane spetrum aording to sampling time for sweep where t w denotes the sweep time, t s denotes the interval time for sampling, and f denotes the frequeny resolution of the distane spetrum. For example, in the ase when t w = 1024 µs and t s = 1 µs, the maximum distane d max is 384 [m]. 2.2 Analysis of Basi Performane First, we desribe the basi performane of the 24 GHz band FM-CW radar. The parameters for the omputer simulation are listed in Table 1 and were determined in aordane with ARIB standard T73 [2]. The enter 4

5 K. Yamaguhi et al.: A 24 GHz FM-CW Radar System for Deteting Closed Multiple Targets and Its Appliations in Atual Senes (a) 3D view. (b) 2D view. Figure 5: Distane spetrum for measuring moving target Figure 6: Measured displaement frequeny was GHz, and the bandwidths were 50, 100, 200, and 400 MHz. Note that the 400 MHz bandwidth was only used for the omputer simulation beause of radio law standards in Japan. The sweep time was 1024 µs, sampling times for the sweep were 0.1, 1, and 10 µs, number of FFT points was 4096, and the hamming windows were used as the window funtion in signal proessing. We assumed that a stati target was loated at 10 meters from the transmitter and reeiver, and the distane spetrums were outputted with various parameters. Figure 3 shows the amplitude value for the distane spetrum versus the measured distane with various sweep bandwidths. The results show that the sweep bandwidth affeted the range resolutions of the measured distane, and a widely swept bandwidth ould improve the range resolution. In the ase when t s = 1 µs, the range resolutions of the measured distane with f w = 50, 100, 200, and 400 MHz were ±5, ±1.5, ±1, and ±0.5 m, respetively. Figure 4 shows the amplitude value for the distane spetrum versus the measured distane with various sampling time. The results show that the sampling interval affeted the error of the measured distane, and a short sampling interval ould redue the error value for the distane. In the ase when f w = 200 MHz, the error value of the measured distane with t s = 10 µs was about 0.5 m. Figure 5 shows the results for measuring a slowly moving target when f w = 200 MHz and t s = 1 µs. The target moved slowly from a distane of 10 m to 20 m with intervals of 0.5 m. Figure 5(a) shows the amplitude value versus the measured distane versus the target distane (shown in 3-dimensional view), Figure 5(b) shows the measured distane versus the target distane (shown in 2-dimensional view). The olor in (b) orresponds to the strength of the amplitude value in (a). From these figures, it was onfirmed that the distane ould be measured orretly aording to the positions of the moving target. Figure 6 shows the results of measuring a target with small displaement, and the measured displaement versus the target displaement was outputted. The objet was loated at 10m from the reeiver, and the objet moved from -5mm to 5mm with intervals of 0.1mm. A small displaement ould be measured by the phase value of the distane spetrum, and the measured displaement orresponds to the target displaement. Note that the measured displaement denotes the relative displaement, and it does not orrespond to the absolute distane between the reeiver and the target objet. A small displaement was orretly measured within ± 3.11 mm with the parameters of the FM-CW radar system in this paper although a displaement of more than ±3.11 mm let to unertainty. 3 PROPOSED TECHNIQUES FOR SINGLE TARGET DETECTION To detet a target s information aurately, suh as distane and small displaement, we propose an algorithm for eliminating the signals refleted from the bakground objets. This setion desribes the proposed algorithm and shows the results of omputer simulations and field experiments. 5

6 Open Journal of Internet of Things (OJIOT), Volume 2, Issue 1, 2016 (a) Without differential detetion (b) With differential detetion (Proposed) Figure 7: Distane spetrum for measuring moving target distane with / without differential detetion in environments in whih multiple objets are loated. 3.1 Detetion for Single Targets As mentioned in the above setion, the FM-CW radar system ould measure the distane and the small displaement at the same time for a single target. However, being able to reeive only the refleted signal on the target at the reeiver was a speial ase. Generally, the reeiver reeives signals refleted from many objets inluding the desired target. Therefore, when there were several objets, signal proessing for deteting the distane spetrum from only the desired target was required. The proposed method removed the signals from the other objets by using the differential detetion of distane spetrum. Figure 7 shows the distane spetrum when the target objet moved from a distane of 10m to 20m, and the other objets were loated at a distane of 15m and 20m. The transmitted signal was refleted on the target and the other objets, and the reeiver reeived several refleted signals. Therefore, the distane spetrum of the other objets was also generated by the FM-CW radar system in Figure 7(a), and the distane spetrum of the target ould not be deteted aurately. In partiular, when the refletion oeffiient of the target was lower than that of the other objets, the distane spetrum of the other objets had a higher Table 2: Parameters in field experiments Parameters Center frequeny f 0 Bandwidth of sweep frequeny f w Sweep time t w Sampling time for sweep t s Output power on transmission Antenna gain Range of distane Range of relative displaement Values (GHz) 200 (MHz) 1024 (µs) 1 (µs) 7 (mw) 11 (dbi) (m) ±3.11 (mm) amplitude value than that of the target. In the proposed differential detetion, the distane spetrum of the other objets P 0 was first generated beforehand (shown in Figure 7(a)). The distane spetrum P 0 was generated when there were no people for deteting the distane and the small displaement. Then, the distane spetrum of both the target and the other targets P, whih inludes P 0, was subtrated by P 0. By using the differential detetion, the distane spetrum of the target and the other targets was removed from the distane spetrum of the other targets P 0. Therefore, only the distane spetrum of the desired target P P 0 was deteted. Figure 7(b) shows the distane spetrum by using the proposed differential detetion method, and the distane spetrum of the target was orretly measured. Compared with the measured distane spetrums in Figure 7(a) and (b), it was learly onfirmed that the proposed method ould detet the target distane by using the differential detetion. The proposed differential detetion ould effetively detet the distane of a moving or stati target from multiple refletions of the bakground stati objets. 3.2 Field Experiments and Appliations In this setion, we desribe setup onditions and results of field experiments, and demonstrates the IoT-based FM-CW radar appliation Setup Conditions for Field Experiments To evaluate the effetiveness of the proposed method for deteting the target distane and displaement, we developed a FM-CW radar system and arried out the experiments with the radar system in an atual environment. Table 2 lists the parameters. The developed FM-CW radar system obtained a ertifiate of onformity with regard to the tehnial regulations 6

7 K. Yamaguhi et al.: A 24 GHz FM-CW Radar System for Deteting Closed Multiple Targets and Its Appliations in Atual Senes Results of Field Experiments (a) Without differential detetion (b) With differential detetion (Proposed) Figure 8: Distane spetrum for measuring moving target distane with / without differential detetion. FM-CW radar Distane Small displaement Figure 9: Setting example of FM-CW Radar for deteting human breathing Figure 10: Displaement for measuring movement of human breathing in Artile 38-6, paragraph 1 of the radio law [18] in Japan, and the developed FM-CW radar system satisfied the ARIB standard T73 in Japan [2]. Figure 8 shows the distane spetrum of a moving target. A person walked away from the FM-CW radar and then ame lose between 2m to 10m. In Figure 8(a), several distane spetrums of the person and the bakground objets were outputted. The distane spetrum of the moving person was not learly deteted in Figure 8(a) beause of the signals refleted from bakground objets. To detet the distane spetrum of the moving person by the differential detetion method, the distane spetrum without the person was measured beforehand. By generating the distane spetrum of the bakground objets beforehand, the distane spetrum of the moving person was orretly deteted in Figure 8(b) with the proposed differential detetion. Therefore, the FM-CW radar system ould measure the movement of the target person effetively. Figure 9 shows the setup of the FM-CW radar system for deteting human breathing in atual senes. The FM- CW radar satisfied the safety guidelines, and the details of the safety guidelines are desribed in the Appendix. Figure 10 shows the results of measuring the small displaement for human breathing. The person s hest movement was measured within the range of relatively small displaement. In Figure 10, the period of breathing was deteted to be about 4s, and the breathing movement was deteted within ±2mm Examples of Appliations In this setion, we show an example of appliation of the 24 GHz band FM-CW radar system in a hospital. Figure 11(a) shows an example of the results for deteting the human breathing of a single target in a room in the hospital. The distane spetrum in this example was measured as follows. 1. Measuring the distane spetrum without any people. 2. A person omes to the bed. The radar reeives signals from the person s body. 3. The person lies asleep on the bed. The radar detets the person s breathing movement. By generating the distane spetrum of the bakground objets without the person, only the distane spetrum of the person was deteted. When the person omes within the range of the radar, the radar system ould detet the signals refleted from the person, and the distane spetrums of the person s body were deteted. One the person had laid down on the bed, the radar system ould detet the small displaement due to the person s breathing. By using the differential detetion method, 7

8 Open Journal of Internet of Things (OJIOT), Volume 2, Issue 1, 2016 Figure 12: Setup with two targets (a) (b) Figure 11: Examples of appliations of the FM-CW radar system the distane and small displaement of the moving objet were aurately deteted. By onneting the developed FM-CW radars in several rooms in the hospital, deteting several targets ould be ahieved through aessing the network. An example of monitoring several targets is shown in Figure 11(b). The developed FM-CW radars were onneted to the network system, and the deteted distane and small displaement data were sent to a server in the nurse station. In the nurse station, the status of several targets were olleted and monitored, and alert information indiating, for example, falling out of bed and stopping breathing ould be monitored. Although the trial experiments were just getting started, the remote monitoring of human breathing ould help in hospital senes. 4 PROPOSED TECHNIQUES FOR MULTIPLE TARGET DETECTION In this setion, we improved the FM-CW radar system to detet multiple targets by using signal proessing. 4.1 Multiple Target Detetion Problems As mentioned in the above setion, the FM-CW radar system ould detet both the distane and displaement Figure 13: Simulated results of detetion for losed multiple targets orretly in single target environments. However, the FM-CW radar ould not detet both the distane and displaement orretly in losed multiple targets environments. To detet losed multiple targets, we propose a detetion method that uses the signal proessing with a tunable FIR filter [4]. 4.2 FIR Filtering The proedures of the proposed detetion method for multiple losed targets are as follows. We assumed that the FM-CW radar was loated at a short distane from the beds as shown in Figure 12. Radio waves were emitted from the FM-CW radar, and the two targets laid down on their beds. The distanes from the radar to eah target were different, so the peaks of the distane spetrum were deteted at different frequenies. In this ase, the deteted result shown on the bottom left in 8

9 K. Yamaguhi et al.: A 24 GHz FM-CW Radar System for Deteting Closed Multiple Targets and Its Appliations in Atual Senes Figure 14: Blok diagram of proposed multiple target detetion in signal proessing Figure 13 was the sum of these two targets, and it was diffiult to detet both the distane and phase values for eah user orretly. To solve this problem, signal proessing with the tunable FIR filter was used in the proposed detetion method. Figure 14 shows the blok diagram of the proposed detetion method. First, FFT was arried out on the A/D onverted signal to obtain the distane spetrum. After FFT, the peaks of the amplitude value of the distane spetrum were alulated for eah user. In the peak position detetion proessing, the two frequeny values relating to the targets were obtained; then, FIR filters were designed aording to the peak positions of the desired targets. The tunable FIR filter had a enter frequeny orresponding to the alulated frequeny in the peak position detetion, and had the linear property for the phase value. By using the designed FIR filter, bandpass filtering was arried out for the distane spetrum. After passing through the bandpass filter, the amplitude and phase value for eah target were deteted orretly. 4.3 Computer Simulations To ebaluate the effetiveness of the proposed FIR filtering method, omputer simulations were arried out. This setion desribes the setup onditions and the results of the omputer simulations Setup Conditions The parameters for the omputer simulations are listed in Table 3 and are based on the ARIB standard T73 [2]. The enter frequeny was GHz, and the frequeny bandwidths were 100, 200, 400, and 800 MHz. Note that the 400 and 800 MHz bandwidths were only used for the Table 3: Parameters in omputer simulations 100, 200, 400, 800 Parameters Values Center frequeny f (GHz) Bandwidth of sweep frequeny f w (MHz) Sweep time t w 256, 512, 1024, 2048 (µs) Bandwidth for FIR filter B 200, 400, 800, 1600, 3200 (Hz) Number of FFT points 4096 Window funtion hamming Table 4: Setup for targets Parameters Target 1 Target 2 Distane 1.0 (m) 2.0 (m) Amplitude of Displaement 2.5 (mm) 2.0 (mm) Period of displaement 10.0 (s) 7.5 (s) omputer simulation beause of the radio law standards in Japan. The sweep times were 256, 512, 1024, and 2048 µs, sampling time of the sweep was 0.1 µs, number of FFT points was 4096, and the hamming windows were used as the window funtion in signal proessing. The bandwidths for FIR filter were 200, 400, 800, 1600, and 3200 Hz. Table 4 lists positions, amplitudes, and phases of target 1 and target 2. In the following setions, we desribe the evaluation values for analyzing the results of multiple target detetion, and the performane analysis aording to the bandwidth for the sweep frequeny, sweep time, bandwidth for FIR filter and position of target 2 were desribed Evaluation Values for Amplitude and Phase Values Figure 15 shows the evaluation values for deteting the amplitude and phase values of the distane spetrum of multiple targets. As evaluation values, we defined the degree of amplitude separation for the amplitude value and NMSE (Normalized Mean Square Error) for the phase value. The degree of amplitude separation is represented by S k = 10 log P min P k = 10 log P (f min) P (f k ) [db], (13) where k denotes the number of targets, S k denotes the degree of amplitude separation, P min denotes the minimal value of the amplitude whose frequeny is f min, and P k denotes the peak value for the k-th target with a frequeny of f k. 9

10 Open Journal of Internet of Things (OJIOT), Volume 2, Issue 1, 2016 (a) Amplitude (a) Amplitude (b) Phase (b) Phase Figure 15: Definitions of evaluation values for omputer simulation Figure 16: Evaluation values versus bandwidth of sweep frequeny The NMSE for the phase value is represented by NMSE = N d i d i 2 i=1 d i 2, (14) where N denotes the number of sampling points, and d i and d i denote the sampled signals of the deteted and setup values for a target, respetively Performane for Bandwidth of Sweep Frequeny Figure 16 shows the evaluation values versus the bandwidth of sweep frequeny f w for (a) the amplitude value and (b) the phase value. As shown in (a), the degrees of amplitude separation were about -2.5 db and -19 db when f w = 100 MHz and f w = 800 MHz, respetively. Beause the frequeny resolution of the distane spetrum inreased when inreasing bandwidth of sweep frequeny, separating the targets to determine the amplitude beame easy. As shown in (b), the NMSE was about 2 % when the bandwidth of sweep frequeny f w = 100 MHz, and the NMSE also improved when inreasing bandwidth of sweep frequeny. Although there were a few disadvantages relating to the NMSE of target 2, the multiple target detetions ould orretly be ahieved beause a degree of amplitude separation of -1 db and an 10

11 K. Yamaguhi et al.: A 24 GHz FM-CW Radar System for Deteting Closed Multiple Targets and Its Appliations in Atual Senes (a) Amplitude (a) Amplitude (b) Phase (b) Phase Figure 17: Evaluation values versus sweep time NMSE of 6 % were high enough for deteting multiple targets in pratial use Performane for Sweep Time Figure 17 shows the evaluation values versus the sweep time t w for (a) the amplitude value and (b) the phase value. As shown in (a), the degree of amplitude separation for target 1 kept at about -4 db, whih was not in aordane with the sweep time. The degree of separation for target 2 was -1 db while the sweep time t w = µs. In the ase when t w = 2014µs, the degree of amplitude separation was almost zero db. That is, it was diffiult to detet the peak value of the amplitude for target 2. Moreover, ompared to the results Figure 18: Evaluation values versus bandwidth of FIR filter in (b), the worst NMSE for target 2 was about 4 % when t w = 2048 µs. Beause the resolution of the distane spetrum dereased with inreasing sweep time, the sweep time should be less than 1024 µs Performane for Bandwidth of FIR Filter Figure 18 shows the evaluation values versus the bandwidth of FIR filter B for (a) the amplitude value and (b) the phase value. As shown in (a), the degree of amplitude separation for target 1 kept at -2 db, whih was not in aordane with the bandwidth of FIR filter B. When bandwidth of FIR filter B was more than 1600 Hz, the degree was more than -1 db. Therefore, it was 11

12 Open Journal of Internet of Things (OJIOT), Volume 2, Issue 1, 2016 (a) Amplitude Figure 19: target 2 (b) Phase Evaluation values versus positions of diffiult to detet the peak value. As shown in (b), the NMSE for target 1 was less than 3 %, but the NMSE for target 2 was more than 5 % when B = 1600 and 3200 Hz respetively. Beause the wide bandwidth of the FIR filter was enough not to ut off the signal of the other target ompletely, the bandwidth of the FIR filter should be less than 1600 Hz Performane for Distane Between Targets Finally, we show the results for deteting multiple targets when target 2 omes lose to target 1 from a distane of 2 m to 1.5 m. Figure 19 shows the evaluation values versus the position of target 2 from the radar for (a) the amplitude value and (b) the phase value. The position of target 1 was 1 m from the radar. Compared to the results, the degree of amplitude separation and NMSE beame dereased as target 2 approahes target 1. As shown in (a), when the bandwidth of sweep frequeny was more than 400 MHz, the degree of amplitude separation had good properties, whih was enough to detet the peaks. However, in the ase when f w = 200 MHz, the degree for target 2 was about -0.3 db, and it was diffiult to detet the peaks. As shown in (b), when f w was less than 400 MHz, the NMSE ould be kept at less than 5 %. However, in the ase when f w = 800 MHz, the NMSE dereased. It was diffiult to detet the peaks of amplitude for targets as the distane between targets dereased. Therefore, the enter frequeny of the FIR filter was not enough to ut off the other target s signal, and the imperfet FIR filter affeted the NMSE values. Although there were a few diffiulties in determining the parameters of the proposed deteting method for FM-CW radar, the proposed deteting method ould be effetive for deteting the distane and the small displaement at the same time in multiple target environments. 5 DISCUSSION As for related work, MIT group reported a radar system for deteting human breathing [1]. The system ould also detet movement, and the person breathing in the room ould be deteted aurately. However, in Japan, the bandwidth for sweep frequeny was muh smaller than that of the MIT s radar system beause of the ARIB standard T73 [2]. As shown in Figure 3, the resolution of the distane spetrum depended on the bandwidth. Therefore, the smaller bandwidth affeted the deteted distane and small displaement. Moreover, in a losed multiple target environment, some distane between the radars and eah target is required for distinguishing the distane spetrum of eah target. Therefore, in this paper, we used a tunable FIR filtering method in signal proessing to detet losed multiple targets at the same time. In this paper, the omputer simulated analysis was only desribed for deteting multiple targets at the same time. In the simulation, the huge omputational ost of designing an FIR filter aording to the position of the targets is required. Furthermore, in the field experiments, the proposed multiple target detetion algorithm must be implemented on the hardware. As future work, we plan to analyze the performane taking into aount the omputational ost versus the filtering results to develop the FM-CW radar system with FIR filtering. 12

13 K. Yamaguhi et al.: A 24 GHz FM-CW Radar System for Deteting Closed Multiple Targets and Its Appliations in Atual Senes Moreover, for pratial use of the IoT-based radar system, we have to analysis the performane of both the radar and network transmission. The number of the developed IoT-based radars is low, so the network performane requirements, suh as throughput and endto-end delay time, should be analyzed as future work. 6 CONCLUSION In this paper, we developed a 24 GHz band FM- CW radar system for deteting the distane and small displaement of a target. The basi performane of a FM-CW radar was evaluated by using omputer simulation, and the distane and small displaement of a single target were measured in field experiments. To detet both the distane and small displaement of multiple targets, we proposed a FM-CW radar system with a tunable FIR filter. The proposed deteting method generates a tunable FIR filter whose enter frequeny orresponds to the peak positions of the distane spetrum of eah target. The tunable FIR filtered signal an be used to detet the distane and displaement for eah target orretly. In omputer simulations, the performane of the FM-CW radar system in a losed multiple moving target environment was analyzed in aordane with the bandwidth of sweep frequeny, sweep time, bandwidth of FIR filter, and distane between targets. As a result, the 24 GHz band FM-CW radar with the proposed detetion method ould effetively detet both the distane and the small displaement for eah target in multiple moving targets environments. And it was onfirmed that the proposed detetion method an detet both the distane and small displaement orretly when the distane between targets was greater than 0.5 m. Moreover, we developed an IoT-based appliation for monitoring several targets at the same time in atual senes. As future work, we will try to implement hardware for the proposed FIR filtering and arry out the field experiments. ACKNOWLEDGEMENTS Part of this work was supported by the Ashita wo Ninau Kanagawa Venture Projet of Kanagawa, Japan. The authors thank Prof. Toshio Nojima at Hokkaido University in Japan for his valuable advie and for analyzing the safety aspets of the developed FM-CW radar system aording to the safety guidelines. REFERENCES [1] F. Adib, H. Mao, Z. Kabela, D. Katabi, and R. C. Miller, Smart homes that monitor breathing and heart rate, in Proeedings of the 33rd Annual ACM Conferene on Human Fators in Computing Systems, ser. CHI 15. New York, NY, USA: ACM, April 2015, pp [2] ARIB STD-T73 Rev. 1.2, Sensors for Deteting or Measuring Mobile Objets for Speified Low Power Radio Station, english/html/overview/do/1-std-t73v1 2.pdf, Assoiation of Radio Industries and Businesses Std., November 2005 (in Japanese). [3] W. Butler, P. Poitevin, and J. Bjomholt, Benefits of wide area intrusion detetion systems using fmw radar, in Seurity Tehnology, st Annual IEEE International Carnahan Conferene on, Otober 2007, pp [4] A. E. Cetin, O. N. Gerek, and Y. Yardimi, Equiripple fir filter design by the fft algorithm, IEEE Signal Proessing Magazine, vol. 14, no. 2, pp , Marh [5] E. Fishler, A. Haimovih, R. Blum, D. Chizhik, L. Cimini, and R. Valenzuela, Mimo radar: An idea whose time has ome, in Radar Conferene, Proeedings of the IEEE, April 2004, pp [6] S. Fujimori, T. Uebo, and T. Iritani, Short-range high-resolution radar utilizing standing wave for measuring of distane and veloity of a moving target, Eletronis and Communiations in Japan Part I - Communiations, vol. 89, no. 5, pp , January [7] ICNIRP, ICNIRP statment on the guidelines for limiting exposure to time-varing eletri, mangeti, and eletromagneti fields (up to 300 ghz), Health Physis, vol. 97, no. 3, pp , [8] IEEE Std C , IEEE Standard for Safety Levels with Respet to Human Exposure to Radio Frequeny Eletromagneti Fields, 3 khz to 300 GHz, IEEE Std., April [9] P. S. Jian Li, MIMO Radar Signal Proessing. John Wiley & Sons, In., Otober [10] Safety guidelines for use of radio waves, dwn/guide38.pdf, Ministry of Internal Affairs and Communiations, June 1990 (in Japanese). [11] S. Miyake and Y. Makino, Appliation of millimeter-wave heating to materials proessing, 13

14 Open Journal of Internet of Things (OJIOT), Volume 2, Issue 1, 2016 IEICE transations on eletronis, vol. 86, no. 12, pp , Deember [12] T. Saito, T. Ninomiya, O. Isaji, T. Watanabe, H. Suzuki, and N. Okubo, Automotive fmw radar with heterodyne reeiver, IEICE transations on ommuniations, vol. 79, no. 12, pp , Deember [13] W. K. Saunders, Post-war developments in ontinuous-wave and frequeny-modulated radar, IRE Transations on Aerospae and Navigational Eletronis, vol. ANE-8, no. 1, pp. 7 19, Marh [14] W. Sediono and A. Lestari, 2d image reonstrution of radar indera, in Mehatronis (ICOM), th International Conferene On, May 2011, pp [15] M. Skolnik, Radar Handbook, Third Edition. MGraw-Hill Eduation, February [16] M. Skolnik, Introdution to Radar Systems. MGraw Hill, Deember [17] A. G. Stove, Linear fmw radar tehniques, IEE Proeedings F - Radar and Signal Proessing, vol. 139, no. 5, pp , Otober [18] Ministry of Internal Affairs and Communiations, Radio law (law no. 131), go.jp/main sosiki/joho tsusin/eng/resoures/laws/ 2003RL.pdf, May [19] T. Uebo, Y. Okubo, and T. Iritani, Standing wave radar apable of measuring distanes sown to zero meters, IEICE Transations on Communiations, vol. 88, no. 6, pp , June [20] K. Yamaguhi, M. Saito, T. Akiyama, T. Kobayashi, and H. Matsue, A 24 ghz band fm-w radar system for deteting losed multiple targets with small displaement, in Ubiquitous and Future Networks (ICUFN), 2015 Seventh International Conferene on, July 2015, pp [21] K. Yamaguhi, M. Saito, K. Miyasaka, and H. Matsue, Design and performane of a 24 ghz band fm-w radar system and its appliation, in Wireless and Mobile, 2014 IEEE Asia Paifi Conferene on, August 2014, pp APPENDIX In general, devies using eletromagneti waves must satisfy guidelines on human exposure to eletromagneti fields, see, for example, IEEE C95.1 in USA [8] and ICNIRP in Europe [7]. MIC have instituted guidelines in Japan [10]. The developed 24 GHz band FM-CW radar in this paper has the following properties. The power of the transmitter is 7 mw, transmitting antenna gain is 11 dbi, effetive radiated power is 88 mw, radiation angle of the transmitting wave is about 50 degrees, and distane between the transmitter and the person is 2.5 m. Aording to the radar equation, the eletri field strength E and the power density P on the human body is alulated as E = = 0.65 [V/m], 2.5 P = E2 z 0 = π = [mw/m 2 ]. Aording to the guideline in [10], these parameters must satisfy E 61.4 [V/m], P 1 [mw/m 2 ]. Therefore, the developed 24 GHz band FM-CW radar system in this paper suffiiently satisfies the onditions in the guideline. 14

15 K. Yamaguhi et al.: A 24 GHz FM-CW Radar System for Deteting Closed Multiple Targets and Its Appliations in Atual Senes AUTHOR BIOGRAPHIES Kazuhiro Yamaguhi was reeived the M.S. and Ph.D. degrees in information siene from Hokkaido University, Hokkaido, Japan, in 2009 and He is urrently an Assistant Professor at the Tokyo University of Siene, Suwa, in Japan. His researh interests inlude Internet of Things, wireless ommuniation, signal and image proessing, and holography. Mitsumasa Saito was reeived the B.S. degree from University of Eletro-ommuniations, Tokyo, Japan, in From 1978 to 2006, he has been engaged in researh and development of seurity and multimedia devies, eletroni display for onsumer eletronis at SONY Corporation in Japan. In 2009, he established ompany of CQ-S net In. Kanagawa, Japan. His researh interests inlude Internet of Things, radar tehnology, and signal and image proessing. Takuya Akiyama was reeived the B.E. and M.S. degrees from Tokyo University of Siene, Suwa, Japan, in 2014 and His researh interests inlude digital signal proessing and wireless ommuniation. Tomohiro Kobayashi was reeived the B.E. degree from Tokyo University of Siene, Suwa, Japan, in He is urrently working toward the M.S. degree. His researh interests inlude digital signal proessing and wireless ommuniation. Naoki Ginoza was reeived the B.E. degree from Tokyo University of Siene, Suwa, Japan, in He is urrently working toward the M.S. degree. His researh interests inlude digital signal proessing and wireless ommuniation. Hideaki Matsue was reeived the B.S. and Ph.D. degree from University of Eletroommuniations and Tokyo Institute of Tehnology, Tokyo, Japan, in 1978 and 1993, respetively. From 1978 to 2004, he has been engaged in researh and development of digital mirowave radio-relay system, network arhiteture for personal ommuniation at NTT Eletrial Communiation Laboratories in Japan. He is urrently a Professor at the Tokyo University of Siene, Suwa, in Japan. His researh interests inlude Internet of Things, network arhiteture, and wireless ommuniation. 15