MR24-01 FMCW Radar for the Detection of Moving Targets (Persons) Inras GmbH Altenbergerstraße 69 4040 Linz, Austria Email: office@inras.at Phone: +43 732 2468 6384 Linz, September 2015
1 Measurement Setup In this measurement report a 24-GHz MIMO frontend ADF TX2RX8 D03 and the Radarbook with a USB 3.0 module are used to conduct measurements with the aim to detect moving persons on a football field. A photo of the utilized radar system is shown in Fig. 1. The USB 3.0 module enables measurements with data rates greater than 1 GBit/s and is therefore ideally suited to collect the raw measurement data even if fast chirp transmit waveforms with high repetition rates are used. Scenarios with a duration of approximately 30 s are recorded and stored as Matlab files. The processing of the measured data sets is done offline with Matlab. Figure 1: Photo of the ADF TX2RX8 D03 MIMO frontend. The radar system is placed on a football field and mounted in a high of 75 cm above the ground. A picture showing the view from the point of the radar system is shown in Fig. 2. The goal of the measurement report is to analyze the performance of the system for detecting persons, which move in front of the radar. In a first attempt simple processing algorithms are applied in order to derive limits for the maximum operating distance of the radar system in this type of arrangement. Figure 2: Photo of the measurement scene from the point of the radar system. Inras GmbH 2 AnHa
2 FMCW Radar Configuration The radar system was operated in a range-doppler mode. A sketch of the transmit waveform is shown in Fig. 3. The waveform consists of N Chirp consecutive upchirps with a duration of T RampUp and a repetition period T Int. After a packet of N Chirp upchirps the next packet is initiated after the time T Wait. The time interval T Int is generated in the FPGA in order to ensure a precise timing relation, which is required for range-doppler processing. Figure 3: Transmit waveform with N Chirp adjacent upchirps. The parameters of the FMCW system are summarized in Tab. 1. During conducting measure- Parameter Descirption Values f Strt Start frequency 24.00 GHz f Stop Stop frequency 24.25 GHz T Upchirp Upchirp duration 256 µs T Int Chirp repetition interval 512 µs T Wait Wait time 200 ms N Chirp Number of chirps 200 N Number of samples per chipr 2560 f s Sampling frequency 10 MHz Table 1: Parameters of FMCW radar system. ments the first transmit antenna is activated and N samples are recorded during the upchirp for all receive channels. The utilized waveform enables different modes of processing. In range-doppler mode the adjacent chirps can be used to estimate the distance and the velocity of moving targets. In addition, the range-doppler processing can be combined with digital beamforming techniques in order to estimate the positions of the targets in a two-dimensional plane. Inras GmbH 3 AnHa
3 Measurement Results: Range Profile To show the performance of the measurement system and to analyse the reflections from the sourrounding measurements without a defined target (dominante reflection from a corner cube) have been carried out. In Fig. 4 the measured range profiles for a single receive channel are shown. The left plot shows the range profile for a single measurement and the right profile for 200 averaged measurements. Both measurements show strong reflection up to distances of 120 m. To verify that the reflections are caused by real objects and not by disturbances indoor measurements on the ceiling with the same settings have been carried out. In Fig. 5 the results of the indoor measurements Figure 4: Left: Range profile for a single measurement; Right: Range profile for the average of 200 adjacent chirps. are plotted for comparison. The range profiles do not show reflections at higher range bins and the noise floor of the radar system constant in the observed interval. In addition, the averaged range profile shows a lower noise floor without any ghost targets. If range-doppler processing is Figure 5: Left: Indoor range profile for a single measurement; Right: Indoor range profile for the average of 200 adjacent chirps. applied the chirps are processed corherently. Therefore, it is important to verify that no internal disturbances exist in the processed data. Inras GmbH 4 AnHa
4 Measurement Results: Range-Doppler Processing As shown in Fig. 4, the range profile contains reflection from the surrounding. These reflections can be strong in amplitude and therefore they tend to mask the reflections form the desired objects. In this application we want to detect and in a next step track persons moving in front of the radar. The reflections from persons are in general much smaller in amplitude than the reflections from the environment. In Fig. 6 the range profiles for the measurements of a single receive channel are plotted over time. For generating the measurements one person was moving towards the radar. The second person started the measurements and moved away from the radar after 5 s. At a time instance of 20 s the person turned around again. The left plot shows that the desired signatures Figure 6: Left: Range profile for a single channel over time; Right: Range profile for a single channel over time with subtraction of the reflections from the stationary targets. of the moving persons are embedded in unwanted reflections form the environment. For instance a corner cube was positioned at a distance of 13 m. A very simple method to remove non-moving targets is to subtract the mean of the N Chirp adjacent chirps from the measurements before calculating the range profile. The results are shown in the right plot of Fig. 6. If the mean is subtracted, only the moving targets are visible in the range profiles. Range-Doppler processing allows the calculation of the distance and the velocity simultaneously. Hence, the calculated range-doppler map enables to distinguish between stationary and moving targets. In addition, targets with different velocities can be separated even if they are located at the same range bin. For instance in Fig. 7 the range-doppler map is shown for the time instance t = 12 s. To calculate the range-doppler map only the first 64 chirps have been used. The map shows two targets, one heading towards the radar and the second at a distance of 20 m with positive velocity moving away from the radar system. Because of the moving arms and legs the reflections are spread over multiple velocity bins. Apart from the ability to separate targets regarding their velocity, the range-doppler processing also enables a significant signal processing gain due to the coherent processing of the adjacent chirps. Hence, the maximum range can be increased. In the left plot of Fig. 8 the range-doppler map for a person in a distance of 45 m is shown. In the right plot the range profil for a velocity of 2 m/s Inras GmbH 5 AnHa
Figure 7: Range-Doppler map for time t = 12 s with two targets. The target with the negative velocity is moving towards the radar and the target with the positive velocity is moving away from the system. is shown. The spectral SNR is approximately 25 db, which means that a reliable detection can be carried out. In addition, the observed SNR indicates that the maximum operating distance can be extended to higher distances. Figure 8: Left: Range-Doppler map for a target in distance of 45 m to the radar system; Right: Range profile for the velocity bin containing the maximum value. Finally, a very simple detection algorithm has been implemented. The detection was carried out in the range-doppler map of a single receive channel. After detecting the target in the range- Doppler map, the angle of incidence is calculated by digital beamforming for the estimated range and velocity bin. Thereafter, the position of the detected target in the x-y-plane is calculated. In the Fig. 9 the estimated positions of the detected targets are marked. The red positions indicate targets moving away from the radar and the blue targets correspond to targets moving towards the radar system. Inras GmbH 6 AnHa
Figure 9: Target positions for two persons moving in front of the radar. The red positions mark targets moving away from the radar and the blue positions correspond to targets moving towards the radar system. The 24-GHz FMCW radar system with the implemented range-doppler processing can be used to detect moving persons. In the presented measurement report movements up to 50 m have been investigated. The measurement results reveal that a reliable detection up to this distance is possible. The observed spectral SNRs indicate that the operating range can be extended to distances up to 75 m for the presented measurement setup. In addition, digital beamforming techniques could be used to synthesize multiple beams before performing the range-doppler processing. In this event, the additional processing gain can be used to further improve the performance. This is done on the expense that multiple beams have to be processed. Inras GmbH 7 AnHa