JP6279931B2 - Radar device, guidance device, and radar signal processing method - Google Patents

Description

  Embodiments described herein relate generally to a radar device, a guidance device, and a radar signal processing method.

  For example, in a pulse radar device used for a guidance device, a wide pulse bandwidth is required to achieve high distance resolution. However, increasing the bandwidth of one pulse places a burden on the processing circuit. For this reason, there has been proposed a radar device called a synthetic band radar that divides a bandwidth into a plurality of frequency bands and assigns the frequency signals of the divided bands to a plurality of pulses for transmission / reception.

  In the synthetic band radar, a plurality of frequency signals assigned to a plurality of pulses and transmitted / received are subjected to inverse Fourier transform, and then a waveform is synthesized. At this time, since each frequency signal has a different transmission / reception time, the phase difference caused by the movement of the target between them must be corrected. For this purpose, the relative movement speed between the target and the radar is required, and the synthetic band radar apparatus requires high speed detection accuracy.

  Here, in a normal radar apparatus, a plurality of pulses are Fourier transformed to generate a Doppler spectrum, and a velocity is detected by detecting a Doppler peak. However, since the synthetic band radar device transmits and receives pulses having different frequencies, it is difficult to Fourier-transform all the pulses as they are.

  Therefore, as a speed detection method often used, for example, a spectrum is generated by Fourier-transforming pulses having the same frequency (hereinafter referred to as frequency step), and the obtained multiple spectra are added by amplitude or power (non-coherent). Integration) and detecting the peak from the addition result to calculate the speed (see Non-Patent Document 2). Another speed detection method is to generate a spectrum by performing Fourier transform for each pulse of the same frequency step, detect each peak from the obtained multiple spectra, calculate the speed, and calculate the obtained multiple speeds. There is a way to average.

  In the former method, although the Doppler frequency is different for each frequency step, it is assumed that they are almost the same and non-coherent addition of a plurality of spectra is performed, so that a floor is likely to occur in speed detection accuracy. On the other hand, since the latter method detects a peak from each spectrum, at a low SNR (Signal-to-Noise Ratio), erroneous detection of a peak frequently occurs and accuracy deterioration becomes remarkable. Therefore, at a low SNR where the floor is not a problem, the former method can detect the speed with high accuracy.

  On the other hand, for example, in order to improve the confidentiality of radar specifications, there is a method in which the pulse repetition interval (PRI) is changed variously within one transmission period (Coherent Processing Interval: CPI). When this method is applied to the synthesis band, it can be considered that the PRI is changed for each frequency step. Even in that case, high speed detection performance is required as in the above method.

  Note that the Fourier transform method includes a fast Fourier transform (FFT) in which the degree of freedom of the sample sequence configuration is low but the processing is fast, and a discrete Fourier transform (Discrete Fourier Transform) in which the degree of freedom is high but the speed is low. DFT). Since a radar device operating in real time requires high-speed processing, it is desirable to use FFT rather than DFT for Fourier transform.

  The most famous method of FFT is butterfly operation, and the number of samples needs to be a power of two. This is a kind of the Cooley-Turkey method, but essentially does not need to have a base of 2. The FFT method further includes a factorization method, a Rader method, and the like, and it is known that an approximately fast Fourier transform is possible if an integer is obtained by appropriately combining these methods.

  Note that FFT is a speed-up of DFT, and FFT other than butterfly computation is often described as DFT. However, in the following description, DFT describes a complex kernel and puts it in a sample string. Let us show the basic method of multiplying and adding one. In addition, the Cooley-Turkey method including butterfly computation, the factorization method, the Rader method, and all the methods that are speeded up by combining these methods are referred to as FFT.

Japanese Patent Laid-Open No. 2003-167049 Special Table 2000-507356

Donald R. Wehner, “High-Resolution Radar,” ch.5, Artech House Radar Library Series (1994) Inaba: Multi-target separation method using multi-frequency stepped ICW radar, IEICE technical report SANE2005-1

  By the way, in the synthetic band radar device, the Fourier transform of the pulse trains in which the PRI is different for each frequency step is equivalent to the Fourier transform of the sample trains having different sampling intervals. Therefore, if the number of pulses in each frequency step is the same, even if a signal having exactly the same frequency is subjected to Fourier transform, the relative position in the spectrum where the peak appears changes for each frequency step. Therefore, even if non-coherent integration for adding the bins having the same spectrum is applied, the peak positions are different, so that the peaks are not added. This is because the bin interval frequency changes corresponding to the PRI.

  In the Fourier transform, instead of simply performing FFT on the sample sequence, it is possible to specify the frequency and perform DFT to calculate a specific frequency component. In this case, since the same frequency component can be calculated in all frequency steps, using DFT can solve the problem that the PRI differs for each frequency step. However, DFT is computationally intensive and has many uses that are not practical.

  If the SNR is high, the speed may be detected individually for each frequency step. However, as described above, this method has a significant deterioration in speed detection accuracy at a low SNR. At low SNR, detection is desired after increasing the detection gain by non-coherent addition, but this becomes difficult if the PRI is changed for each frequency step.

  The present embodiment has been made in view of the above problems, and provides a radar apparatus, a guidance apparatus, and a radar signal processing method capable of obtaining high speed detection accuracy even at a low SNR in a variable PRI in which a PRI handles a pulse train that differs for each frequency step. For the purpose.

In order to solve the above problem, the radar apparatus according to the present embodiment includes a transmission unit, a reception unit, and a signal processing unit. The transmission unit generates a plurality of pulses at each of a plurality of steps repeated within one transmission period (CPI) at a unique pulse repetition interval (PRI) different from each other , and a plurality of pulses at each step Are sequentially transmitted as radar waves. The receiving unit receives a reflected wave of the radar wave. The signal processing unit calculates a Doppler spectrum by performing Fast Fourier Transform (FFT) on a reception signal corresponding to the plurality of pulses obtained at the step, and obtains a target Doppler spectrum from the calculation result of the Doppler spectrum. Estimate speed. At this time, the signal processing unit performs the FFT after adding the number of zeros determined for each step around the value of the plurality of pulses in each of the plurality of steps, and adds the PRI, the FFT frame length, and the addition. The number of zeros to be determined is determined so that the time corresponding to the FFT frame length after adding the zeros is substantially the same in all steps.

  In this way, even when the PRI is different for each step, it is possible to substantially align the peak positions in the spectrum of the same target while generating the spectrum by FFT with a relatively small number of samples. Become. As a result, Doppler detection using non-coherent integration is possible, and the target speed can be detected with high accuracy even at a low SNR.

It is a block diagram which shows the structure of the radar apparatus which concerns on 1st Embodiment. FIG. 3 is a timing diagram illustrating transmission pulses generated for each frequency step in the first embodiment. FIG. 6 is a waveform diagram illustrating an example of a Doppler spectrum obtained for each transmission pulse in the first embodiment when the PRI is common to all frequency steps and FFT is performed with the same number of samples. (A) is a scatter diagram showing the relationship between each frequency step number, PRI length, and bin interval frequency when the PRI is changed for each frequency step and FFT is performed with the same number of samples, and (b) is a scatter diagram showing (a). It is a wave form diagram which shows the Doppler spectrum for every transmission pulse at the time of setting to the relationship shown. In 1st Embodiment, it is a timing diagram which shows each relationship of PRI length in each frequency step, FFT frame length, the total number of samples at the time of FFT, and the number of zero added at the time of FFT. In the relationship shown in FIG. 5, it is a conceptual diagram which shows the frequency bin space | interval in each frequency step, and the number of spectrum bins of each frequency step. (A) is a scatter diagram showing the relationship between the frequency step number and the number of FFT frame samples in the first embodiment, and (b) is the relationship between the frequency step number, the PRI length, and the bin interval frequency in (a). FIG. FIG. 8 is a scatter diagram showing the relationship between the frequency step number, the PRI length, and the bin interval frequency when the example shown in FIG. 7 is rounded at a time interval corresponding to the sampling rate. (A) is a waveform diagram in which the Doppler spectrum of all frequency steps calculated by the PRI and the number of samples in the FFT frame obtained by the rounding process shown in FIG. 8 is overwritten, and (b) is a waveform diagram showing all frequency steps shown in (a). FIG. 4C is a waveform diagram showing a result of non-coherent integration of the Doppler spectrum of FIG. 1A with an amplitude, and FIG. FIG. 3 is a block diagram illustrating a specific configuration example of a speed estimation unit in the first embodiment. It is a wave form diagram which shows the example of a simulation of the Doppler spectrum which concerns on 2nd Embodiment and raised the target speed and turned back 5 times in the case of Low-PRF. In 2nd Embodiment, it is a conceptual diagram which shows the relationship between the Doppler spectrum period for every frequency step, and the Doppler frequency corresponding to a target speed estimated value. In 2nd Embodiment, it is a conceptual diagram which shows a mode that the number of a spectrum component carries out the cyclic shift with respect to the bin number in the movement destination bin of target prediction speed. FIG. 11 is a waveform diagram showing a result obtained by shifting the Doppler spectrum shown in FIG. 11 by shifting the bin corresponding to the predicted value to the center of each spectrum for each frequency step, and shifting again so that the shifted bin numbers are aligned at all frequency steps. It is. It is a block diagram which shows the specific structural example of the speed estimation part in the case of performing aliasing correction in the radar apparatus which concerns on 2nd Embodiment. (A), (b) is a block diagram which shows the specific structure of the folding correction | amendment part shown in FIG. 15, respectively. (A) is a waveform diagram showing an example in which a peak is applied to the end of the Doppler spectrum in the synthetic band radar, and (b) is a waveform diagram showing an example in which no peak is applied to the end of the Doppler spectrum of (a). It is a block diagram which shows the specific structural example of the speed estimation part in the case of performing folding correction | amendment by 2 systems of processing in the radar apparatus which concerns on 3rd Embodiment. In the radar apparatus which concerns on 4th Embodiment, it is a wave form diagram which shows a mode that the same spectrum was repeatedly arranged from the absolute frequency to the frequency | count of folding including the Doppler frequency corresponding to a prescription | regulation maximum relative speed for every frequency step. It is a block diagram which shows the specific structural example in the case of identifying the Doppler frequency of the correct frequency | count of return in the speed estimation part of the radar apparatus which concerns on 4th Embodiment. FIG. 15 is a waveform diagram illustrating a state in which noise is removed from the Doppler spectrum illustrated in FIG. 14 in order to describe a radar apparatus according to a fifth embodiment. In the radar apparatus according to the fifth embodiment, assuming that the carrier frequency of each frequency step increases in ascending order in the order of the number, the PRI length, the FFT frame length, the total number of samples at the time of FFT, and the FFT It is a timing diagram which shows the relationship of each number of zero added at the time. In the relationship shown in FIG. 22, it is a conceptual diagram which shows the frequency bin space | interval in each frequency step, and the number of spectrum bins of each frequency step. In the relationship shown in FIG. 22, it is a scatter diagram which shows PRI length and bin space | interval frequency for every frequency step. FIG. 22 shows the result of calculating the Doppler spectrum at each frequency step in the relationship shown in FIG. 22, where (a) shows a waveform when no noise is present, and (b) shows a waveform when there is noise. It is a block diagram which shows a structure in case the radar apparatus which concerns on this embodiment is a synthetic band radar. It is a block diagram which shows a structure in case the radar apparatus which concerns on this embodiment is a frequency agility radar. It is a block diagram which shows the structure at the time of using the radar apparatus of this embodiment for a guidance device. It is a flowchart which shows the flow of the signal processing used as the basis in the radar apparatus of this embodiment.

  Hereinafter, embodiments will be described with reference to the drawings. In the following description and illustrations, only parts that are essentially necessary for the configuration of the embodiment are shown, and parts that are not related to the operation of the embodiment are omitted.

(First embodiment)
FIG. 1 is a block diagram showing the configuration of the radar apparatus according to the first embodiment. The radar apparatus shown in FIG. 1 includes a synthesizer 1, a radar wave transmitter 2, a circulator 3, an antenna 4, a radar wave receiver 5, and a signal processor 6. The signal processing unit 6 includes a zero addition unit 61, an FFT 62, and a speed estimation unit 63.

  The synthesizer 1 generates a radar wave frequency signal. The radar wave transmission unit 2 pulsates the frequency signal generated by the synthesizer 1 with the PRI determined for each step, arranges the pulse train of each step in units of PRI, and outputs it as a radar wave for transmission. At this time, the radar wave transmission unit 2 performs modulation processing such as frequency chirping on the radar wave as necessary. The radar wave generated in this way is transmitted from the antenna 4 through the circulator 3 in a predetermined direction.

  The transmitted radar wave is reflected by various objects such as clutter and obstructions in addition to the target. These reflected waves returning to the radar apparatus are captured by the antenna 4 and sent to the radar wave receiving unit 5 via the circulator 3 as a radar wave reception signal. The radar wave receiving unit 5 performs processing such as filtering, baseband conversion, analog-digital (A / D) conversion, etc., on the radar wave reception signal based on the frequency signal generated by the synthesizer 1, thereby providing a radar wave. In the case where modulation processing is applied to the signal, demodulation processing such as pulse compression is performed to convert it into a digital baseband signal. At this time, a signal (hereinafter referred to as a pulse representative value) to be subjected to speed detection processing is extracted for each frequency step and sent to the signal processing unit 6.

  In the radar apparatus shown in FIG. 1, one antenna 4 is shared for transmission and reception through the circulator 3, but this configuration is not essential for the present embodiment. For example, a separate antenna and a directional coupler are used. Other configurations such as sharing by may be used. Further, for the sake of illustration, the display is shown as a one-element antenna, but an array antenna may be used. The detailed configuration of the antenna has nothing to do with this embodiment.

  The zero adding unit 61 constituting the signal processing unit 6 receives notification of PRI information at each step of the radar wave from the radar wave transmitting unit 2. Then, a sample sequence consisting of pulse representative values is sequentially input from the radar wave receiving unit 5 for each step, and the number of zeros determined for each PRI in each step is added to the sample sequence consisting of pulse representative values, for example, behind. Then, an FFT frame is formed for each step. The FFT unit 62 performs FFT on the sample sequence of the FFT frame formed by the zero addition unit 61 for each step to obtain a Doppler spectrum. The speed estimation unit 63 detects the target speed from the Doppler spectrum of each step obtained by the FFT unit 62.

  In the radar apparatus having the above configuration, the processing operation as a synthetic band radar will be described in detail below. In the synthetic band radar, each step is a frequency step having a different frequency for each step.

FIG. 2 is a timing chart showing transmission pulses generated for each frequency step in the radar apparatus shown in FIG. 1 in order to explain the synthetic band radar. In FIG. 2, the horizontal axis indicates time, the vertical axis indicates frequency, and the rectangular bar indicates a pulse having a predetermined pulse length and a predetermined bandwidth. The time interval between the individual pulses is PRI, and one carrier frequency, that is, one frequency step, transmits / receives N _p pulses, moves to the next frequency, and returns N _p pulses. Repeat the procedure of sending and receiving. The number of frequency steps is N _f , and is repeated N _f times in total.

  In this embodiment, PRI is a unique value set for each frequency step. First, in order to explain the operation of a normal synthetic band radar, an example of a Doppler spectrum when PRI is common to each step is shown in FIG. 3 shows. In FIG. 3, the number of pulses in one frequency step is set to 10, the number of frequency steps is set to 10, and one pulse representative value is extracted from each pulse for each frequency step under the condition without Doppler ambiguity. The Doppler spectrum obtained by FFT is shown. In the FFT, in order to smooth the spectrum waveform, a large number of zeros were added after the pulse representative value to interpolate between frequency bins.

  Since the carrier frequency is different for each frequency step, the original Doppler frequency is slightly different for each frequency step, but in the state where the number of pulses is small and there is no Doppler ambiguity, each frequency difference is very small. There is a peak at the same frequency. It is because of noise that there is a slight height difference or shift.

  Next, a case where the PRI is simply changed for each frequency step and FFT is performed with the same number of samples as it is will be described with reference to FIG.

  FIG. 4A shows the PRI length of each frequency step and the bin interval frequency of the spectrum, and FIG. 4B shows the Doppler spectrum when all frequency steps are FFTed with the same number of samples. As a procedure for determining the PRI, an approximate PRI (200 μs in the example of FIG. 4) is determined, and the PRI is set so that the PRI length changes by 2 μs around the PRI. The number of pulses in each frequency step is 10, and the bin interval is 1 / frame length. Here, interpolation is performed so that the spectrum is displayed smoothly. Interpolation uses a short time Fourier transform that interpolates with a trailing zero, but the number of zeros is common to all frequency steps. The frame length is defined by the length of only the pulse train not including zero.

  In the case of FIG. 4B, the peak position is clearly different for each frequency step, that is, for each PRI, as compared with FIG. Since the difference in peak position is not large, even if the addition is performed as it is (non-coherent integration), a peak indicating the approximate center Doppler frequency can be obtained. For this reason, if the SNR is relatively high and the target is a single target and there is little variation in the peak height for each frequency step, it is possible to estimate with a certain degree of accuracy even if the speed is estimated as it is. is there.

  However, the present embodiment aims to improve speed estimation accuracy at a low SNR, and it is very unlikely that the target is a single point target, and the peak height for each frequency step is large. Often different. For this reason, even if non-coherent integration is simply performed as described above, the peak of the integration result does not indicate the center of the whole, and the speed estimation accuracy deteriorates.

Therefore, in this embodiment, as shown in FIG. 5, the PRI length, the frame length, and the number of zeros added at the time of FFT are determined. That is, at each frequency step, the number of transmitted and received pulses is common in N _p number, PRI is different for each frequency step. Further, the relationship between the PRI / FFT frame length and the number of samples in the FFT frame is different from that in FIG. Thus, in order to enable non-coherent integration of spectra at a plurality of frequency steps having different PRIs, the bin intervals of the spectra may be aligned. For example, if DFT is used, the frequency of spectral bins can be determined completely freely. However, since the processing amount is very large in DFT, application of FFT is desirable here.

The bin interval by FFT is determined by the reciprocal of the frame length. Therefore, in order to align the frequency bin intervals of a plurality of frequency steps, the frame lengths may be aligned with each other. As shown in FIG. 5, although the PRI is different for each frequency step, the frame length is common and the number of samples in the integer FFT frame is the same in the FFT frame that is different for each PRI. It means that the number of samples. The FFT frame length is common to all frequency steps, and is assumed to be T _FFT . The PRIs are PRI _{1 to} PRI _Nf , respectively. The number of pulses N _p is common to all frequency steps, but the number of zeros added at the time of FFT differs from N _{Z1 to} N _{ZNf for} each frequency step. In order to process by FFT, the number of samples in a frame is an integer.

  As shown in FIG. 6, the frequency bin interval, which is the reciprocal of the FFT frame length, is common to all frequency steps, as shown in FIG. However, since the total number of bins in each spectrum is equal to the number of samples in the FFT frame, it differs for each frequency step. With a finite number of FFTs, the spectrum is folded at a period (spectrum period) corresponding to the reciprocal of the sampling interval. In addition, since the PRI corresponding to the sampling interval is different for each frequency step, the spectrum period is different for each frequency step. However, since the frequency bin intervals are common, Doppler peaks with the same speed appear at substantially the same position even if the non-coherent integration is performed on the result of the FFT. Of course, since there is a difference in carrier frequency for each frequency step, it is not exactly the same, but it can be a variation that is not significantly different from FIG.

  A method of performing a Fourier transform by adding zero after the sample sequence is called a short time Fourier transform. Normally, zeros are added to the end of the sample sequence. However, as long as the sample sequence is grouped, it is not always necessary to be behind, and it may be before or after. In the finite number of Fourier transforms, since the end and the beginning of the frame are considered to be continuous, zeros may be inserted collectively at one intermediate position of the sample sequence. In the application of using the obtained Fourier transform result for non-coherent integration, the position of zero can be variously changed as described above. However, in the synthetic band radar that proceeds from the Fourier transform result to more coherent processing, the phase after the Fourier transform changes at the position of the zero to be added, so it is necessary to correct the influence.

The simplest method for realizing the states shown in FIGS. 5 and 6 is a method of determining the FFT frame length in a least common multiple. To explain a specific example, the PRI is first determined as shown in FIG. For example, if the amount of change in each PRI step is 2 μs, the PRI of each frequency step is determined. Therefore, the number obtained by dividing each PRI by 2 μs, 95, 96,... 104 is calculated and these least common multiples are calculated. Since these numbers are relatively prime, the least common multiple is the product of all the numbers. In this case, it is about 9.47 * 10 ¹⁹ . The number obtained by multiplying this number by 2 μs is the frame length, and the number obtained by dividing the frame length by each PRI length is the number of samples in the FFT frame.

  However, as can be seen from the above numerical examples, in this method, the number of samples in the FFT frame is enormous, and even if it is FFT, the amount of calculation is enormous. One of the purposes of this embodiment is to lighten the processing by using FFT. Therefore, it is desired to reduce the number of samples in the FFT as much as possible. Therefore, PRI, the FFT frame length, and the number of samples in the FFT frame are determined by the following procedure.

  First, an approximate PRI standard is determined from the distance to the target, the CPI length, and the total number of pulses. For example, it is set to 200 μs. Next, the rate at which the PRI can be changed is determined. It is difficult to change the PRI extremely due to restrictions such as the distance to the target. For example, it can be changed by about 10%. This standard is the same as the example of FIG.

Next, an approximate number of samples in the FFT frame is determined. The number of frequency steps M _f , for example 10, is divided by 10% that is the PRI variable amount, that is, 0.1. As a result, it becomes 100. The number of samples in the FFT frame at each frequency step is determined so that the number of samples in the FFT frame is different by 1 around 100. For example, 96, 97, ..., 105 are determined. A value obtained by multiplying 200 μs, which is an indication of PRI, by an approximate number of samples in the FFT frame of 100, for example, 20 ms, is determined as the frame length, and divided by the number of samples in the FFT frame of each frequency step obtained, and this is divided into each frequency step. It is set as PRI.

  That is, in the FFT frame consisting of approximately 100 samples, the number of samples in the FFT frame of each frequency step is determined first so that the integer number of samples changes by about 10%, and the PRI is determined therefrom. To do. The PRI is not determined first, but is finally determined so as to be an integer as small as possible, and is determined so that the change in the number of samples in 10% of the FFT frames is approximately equal to the number of frequency steps of 10.

  In this way, as can be seen from the above example, if the rate of change in PRI is 10%, the number of samples can be about 10 times, and if it is 20%, the number of samples can be about 5 times. Can be fastened.

An example of the number of samples obtained in this way is shown in FIG. FIG. 7A shows the number of samples in the FFT frame, and FIG. 7B shows the PRI length and bin interval frequency. In this example, the number of samples in the FFT frame is different by 1 for each frequency step. The PRI length is about 190 μs to about 208 μs, which is similar to FIG. 4 but slightly different. Specifically, {190.476, 192.308, 194.175, 196.078, 198.020, 200, 202.020, 204.082, 206.186, 208.333} μs. The bin interval frequency is common to all frequency steps. The operation of adding a trailing zero is the same as the operation of interpolating the Doppler spectrum. In FIG. 4, the frequency interval determined by the number of samples Np before interpolation is _defined as the bin interval, but in FIG. 7, the frequency interval determined by the number of samples in the FFT frame after zero addition is displayed as the bin interval.

  In the PRI determination method as described above, the PRI is a very odd value. In signal processing, if the value exceeds the sampling rate and becomes odd, processing becomes more complicated than necessary. Therefore, if the time interval corresponding to the sampling rate of the signal processing unit 6, for example, the time interval corresponding to the sampling rate is 0.1 μs, the PRI obtained as described above is rounded to the unit of 0.1 μs. In the above example, {190.5, 192.3, 194.2, 196.1, 198.0, 200, 202.0, 204.1, 206.2, 208.3} μs. At that time, the number of samples in the FFT frame is not changed. As a result, the FFT frame length, that is, the spectral bin interval, is slightly different for each frequency step and is not exactly the same.

  FIG. 8 shows the PRI length and bin interval frequency when the example of FIG. 7 is rounded at a time interval corresponding to the sampling rate. Although the difference in the PRI length is not known in the figure, the bin interval frequency clearly has irregularities. However, this difference does not have a significant effect on the detection speed accuracy.

  FIG. 9 shows the result of calculating the spectrum of each frequency step by using the PRI and the number of samples in the FFT frame thus obtained. Similar to FIGS. 3 and 4, this is an example without Doppler ambiguity. FIG. 9A shows the overwritten waveform of the spectrum at all frequency steps, FIG. 9B shows the waveform obtained by non-coherent integration of FIG. 9A with the amplitude, and FIG. 9C shows the end of the Doppler spectrum. The waveform is shown enlarged. Comparing FIG. 9A with FIG. 4B, it can be seen that even when the rounding process is performed, the peak positions are aligned to the extent that they are not inferior to those in FIG. It can also be seen that the last bin number of the spectrum is different for each frequency step.

  FIG. 10 is an example of a detailed form of the speed estimation unit 63. When the Doppler spectrum of each frequency step obtained as shown in FIG. 9A is non-coherently integrated by the non-coherent integrator 631, it is as shown in FIG. 9B. The peak detector 632 detects the peak value of the Doppler frequency from the non-coherent integration result of the Doppler spectrum by maximum value detection, CFAR (Constant False Alarm), or the like. The speed calculator 633 calculates the speed from the peak value of the Doppler frequency obtained by the peak detector 632. In the case of the synthetic band radar, since the carrier frequency is different for each frequency step, the Doppler frequency is also strictly different. However, if the difference is as large as shown in FIG. There is no error.

  In the above example, the example in which the number of samples in the FFT frame changes by one for each frequency step is shown. However, this does not necessarily change in the order of frequencies, and the order may be random. For example, in the case of all 10 steps, there is no big problem even if one or two parts are skipped somewhere in the arrangement of the number of samples. Further, there is no problem even if there are some steps having the same PRI. The gist of this embodiment is that the difference between the maximum value and the minimum value of the number of samples in the FFT frame is approximately equal to the number of frequency steps. In a plurality of frequency steps, any two steps may be such that the difference between the maximum value and the minimum value of the number of samples in the FFT frame including zero is 1 or 0.

For example, even if the interval of the number of samples in the FFT frame is variously changed with respect to the basic number of samples N _lm (100 in the above example), it does not depart from the spirit of the present embodiment.
Example 1 N _lm + { _-5 , -4, -3, -2, -1, 0, 1, 2, 2, 4}
Example 2… N _lm + { _-5 , -4, -3, -2, -1, 0, 1, 2, 3, 5}
Example 3… N _lm + { _-6 , -4, -3, -2, -1, 0, 1, 2, 2, 5}
Example 4… N _lm + { _-5 , -4, -3, -2, -1, 0, 1, 2, 3, 3}
Example 5 ... N _lm + {-4, -4, -2, -2, 0, 0, 2, 2, 4, 4}
Example 5 can further reduce the number of samples in the FFT frame to a number obtained by dividing the above value by two. However, since there is a case where it is desired to set the number of samples _Nlm to be somewhat large due to a demand for peak detection accuracy obtained from the non-coherent integration result, the value as in Example 5 may be used. As described above, the order of the number of samples in the FFT frame is not related to the order of frequency steps or the order of transmission times, and may be in any order.

  The number of samples in the FFT frame obtained in this way is stored in correspondence with the PRI value, and the zero adding unit 61 in FIG. 1 converts the stored number of zeros into the input PRI value information. Select based on.

  Therefore, according to the radar apparatus of the present embodiment, even if the PRI is a variable PRI that handles a different pulse train for each frequency step, even if the SNR is low, the Doppler spectrum peaks at each frequency step are aligned and coherently integrated. The peak detection accuracy of the result is improved, and thereby high speed detection accuracy can be obtained.

(Second Embodiment)
Next, as a second embodiment, a case where the predicted speed is known with Low-PRF will be described. The basic configuration of the radar apparatus according to this embodiment is the same as that of the apparatus shown in FIG.
First, in Low-PRF, Doppler ambiguity exists and the Doppler spectrum appears in a folded state many times. For example, when the Doppler spectrum ends at 5 kHz but the Doppler frequency is 10 kHz.

  The process up to the calculation of the Doppler spectrum for each frequency step is the same as in the example of FIG. FIG. 11 shows a simulation example obtained by increasing the target speed so that the spectrum is folded for about five laps. Even though the frequency bin intervals of all frequency steps are aligned, the peak positions are completely disordered for each frequency step. This is because the frequency of the aliasing frequency is different for each frequency step, and the difference is integrated by the number of aliasing. Therefore, in Low-PRF, non-coherent integration cannot be performed simply by aligning the frequency bin intervals, so it is necessary to correct the frequency difference caused by folding in advance.

  There are several possible correction methods, but when Doppler ambiguity exists, if the predicted speed is known, a method that eliminates ambiguity by selecting the value closest to the predicted speed is common. It is.

  Therefore, in the present embodiment, a Doppler ambiguity elimination method when the predicted speed is known is provided.

FIG. 12 is a diagram showing a state in which the folding frequency varies depending on the frequency step. In FIG. 12, it is assumed that one of the sections indicated by double-pointed arrows on the two-dot chain line is one period of the Doppler spectrum, and the Doppler frequency corresponding to the target predicted speed is approximately at the frequency indicated by the vertical one-dot chain line. In the example of FIG. 12, there is a target predicted speed at the third turn back at frequency steps f ₁ , f ₂ , and f _{3 and} at the fourth turn at f _Nf . Also from this figure, it can be seen that the target peak does not appear at the same position by simply overlapping the folded spectra.

  One method in the case where the predicted speed is known is a method in which the same frequency in the spectrum of each frequency step is aligned in the same bin in the vicinity of the Doppler frequency corresponding to the predicted speed. That is, since the predicted speed is known, as can be seen from FIG. 12, it is also known how many bins in the circumference of the spectrum of each frequency step have the predicted speed. Therefore, for example, the spectrum of each frequency step is cyclically shifted so that the predicted speed comes to the bin No. 50, for example, approximately in the middle of about 100 bins.

  The processing by the above cyclic shift will be described with reference to FIG. FIG. 13 shows a state in which the number of the spectral component is cyclically shifted with respect to the bin number in the destination bin of the target predicted speed. Assume that the predicted speed of a frequency step with 100 bins is included in the 97th bin, for example. In order to move this to the 50th bin, the frequency is shifted to the right or left. Although the figure shows an example of shifting to the right, the result is the same in either direction as long as it is patroled. In the first state, the 50th spectral component exists in the 50th bin, but after the cyclic shift, the 97th component is located in the 50th bin. In the vicinity, the spectral components are arranged in order, and after 99, it returns to 0.

  Since the number of bins is different for each frequency step, the number of the central bin is also different, but when performing non-coherent integration, cyclic shift is performed so that the frequency of the prediction speed is the same bin number. As a method (1), as shown in FIG. 13, if the number is 50, the prediction value corresponding bins of all frequency steps are aligned to the number 50 so that a difference due to the variation in the number of bins appears at the end of the spectrum. is there. Further, as the method (2), there is a method of shifting the bin corresponding to the predicted value to the center of the spectrum for each frequency step, and shifting again so that the number of bins after the shift is aligned at all frequency steps.

  The method (2) will be described in more detail. For example, if the number of banks is 96, first, cyclic shift is performed so that a bin including the predicted speed comes in the 48th. Next, when the maximum number of bins is 105, nine zeros are added after 96, and the cyclic shift is performed so that the 48th component moves to the 53rd, which is about half of the 105. Similar processing is performed for the other frequency steps. In such a processing form, the difference in the number due to the difference in the number of bins at each frequency step appears as being divided into two at the beginning and the end of the spectrum.

  FIG. 14 shows the result of processing the Doppler spectrum for each frequency step in FIG. 11 by the latter. In FIG. 14, since the predicted speed is set to 10 m / s larger than the true value, there is no peak at the 53rd center, but it can be seen that the peaks of the target Doppler spectrum are still almost aligned. In this method, if the error between the predicted speed and the true value is within about half of the spectrum, a spectrum with substantially uniform peaks can be obtained.

  FIG. 15 shows the configuration of the speed estimation unit 63 in Low-PRF in the second embodiment. The speed estimation unit 63 illustrated in FIG. 15 has a configuration in which a folding correction unit 634 is added before the non-coherent integration unit 631 in the configuration illustrated in FIG. 10, and a predicted speed value is input to the folding correction unit 634. . In addition, the predicted speed value is also input to the speed calculator 633, the ambiguity is canceled, and the speed of the peak value obtained as a result of peak detection is calculated. That is, the speed corresponding to the peak is calculated from the difference between the Doppler frequency between the spectrum position corresponding to the predicted speed and the detected peak position.

  FIG. 16 shows a specific configuration example of the aliasing correction unit 634. In FIG. 16A, the shift number determination unit 6341 determines the shift number for each frequency step from the predicted speed, and the spectrum shift unit 6342 shifts the spectrum based on the value. The configuration of FIG. 16A is a form in which a difference in the number of bins appears at the end of the spectrum. The configuration of FIG. 16B is a form in which a difference in the number of bins appears at the beginning and end of the spectrum. The bin number including the predicted speed once in the vicinity of the center of each frequency step at the shift number determination unit 6341 and the spectrum shift unit 6342. Further, 0 is added by the zero addition unit 6343 so that the number of bins of all frequency steps becomes equal, and then the spectrum shift unit 6344 shifts to the center number of the spectrum having the maximum number of bins.

  As a modification of FIG. 16B, there is the following method. A spectrum is created by combining three spectra of the same frequency step side by side, and approximately the same number of bins on both sides is selected around the frequency bin corresponding to the predicted speed in the three central spectra. The number of bins to be selected is common to all frequency steps. Similar processing is applied by non-coherent integration of selected bins for all frequency steps.

  In the above configuration, it is expressed that the Doppler frequencies of the central bin are aligned. However, as described above, the bin interval frequencies are not exactly equal due to the rounding of the PRI, and thus are not completely aligned. Also, when multiplying the PRI by a coefficient so as to absorb the difference in Doppler frequency for each frequency step, which will be described later, it is not the correct Doppler frequency but the Doppler frequency scaled by the carrier frequency.

(Third embodiment)
Next, as a third embodiment, another mode when the predicted speed is known with Low-PRF will be described. The basic configuration of the radar apparatus according to this embodiment is the same as that of the apparatus shown in FIG.

  In the radar apparatus shown in FIG. 1, depending on the application destination, the process of changing the shift amount for each predicted speed may be complicated. In such a case, a plurality of shift destinations may be prepared in advance, and the shift destination may be selected so that the predicted speeds are aligned corresponding to the predicted speeds.

Specifically, first, a reference frequency step is determined. For example, in the example of FIG. 12, the _Nf- th frequency step with the shortest repetition period is used. The start of each turn and the center of each turn of this frequency step are determined in advance as the frequencies to be aligned. The amount of spectrum shift for all frequency steps is calculated in advance so that these frequencies become, for example, the 0th bin after the shift, and the value of the shift amount of each frequency step is stored in the table.

  Next, for each predicted speed, the number of turn-around turns at the reference frequency step (number of turns) is determined. In the example of FIG. 12, it is between 3 and 4 laps. Further, it is determined where in one lap starting from the center of the turn. In the example of FIG. 12, it is between 2.5 laps and 3.5 laps. Therefore, the shift amount for adjusting to the frequency of exactly three rounds is acquired from the table, and each frequency step is shifted. Next, the shift amount adjusted to 2.5 rounds is acquired from the table, and each frequency step is shifted. Two types of spectra are generated: 3 to 4 laps and 2.5 to 3.5 laps.

  In this method, it seems that only one system is required in accordance with the start point of the integer cycle. However, for the following reasons, it is possible to prepare a plurality of start points that are shifted by at least a half turn, and use two or more of them. desirable.

  That is, since the number of bins is different for each frequency step, it may not be correctly detected if a peak is applied to the end of the spectrum. An example is shown in FIG. In FIG. 17A, the frequency of the start point of the cycle is matched, but the predicted speed value is 5 m / s less than the true speed value (true value), and the true value and the predicted speed value are bin numbers. It is a spectrum in the case where the number of laps is over 0 and the number of laps is one less than the true value. In this example, each spectrum is given a 0 after the last bin of the spectrum so that the position of the last bin can be clearly read. In the case of this example, an incorrect peak appears around the beginning of the spectrum, but the peak position of each spectrum is disordered because of an incorrect lap.

  On the other hand, it can be seen that the peak positions are aligned at the end of the spectrum where there are no bins at some frequency steps. This occurs only with one system that matches the start point of the cycle. Therefore, if a plurality of start points that are shifted by at least a half circumference are prepared, the peak positions of the respective spectra can be aligned with either of them. An example is shown in FIG. FIG. 17B is a spectrum that is shifted based on the predicted speed, starting from the half circumference after the spectrum start point of FIG. Here you can see that the peaks are aligned. From this, the peak position can be correctly detected in any system by using two or more systems.

  Here, if the accuracy of the predicted speed is high, it is only necessary to select one start point whose predicted speed is closest to the center of the spectrum among the start points shifted by half a circle. However, since the start points are fixed, the predicted value may come to a position about 1/4 of both sides of the spectrum. In such a case, the peak may not appear depending on the accuracy of the predicted value. . Therefore, the certainty of detection is improved by using two systems.

  Each system has a clearly defined Doppler frequency corresponding to bin number 0. Therefore, after detecting a peak from the non-coherent integrated waveform, add the frequency of the difference from that peak to the frequency of bin number 0. Can eliminate the ambiguity.

  As can be seen from FIG. 17 (a), the peak obtained at both ends of the spectrum may not be the correct peak even if the frequency is not limited to the end portion without the bin in some frequency steps. Is expensive. Therefore, in this method, it is preferable that about half of the spectral line width at both ends of the spectrum be the detection prohibited area.

  Although it is possible to set the detection areas excluding the detection prohibition area so as not to overlap in a plurality of systems, they may overlap. However, it is necessary to ensure that all Doppler frequencies are detected in either system. When the detection areas overlap, only the same data including noise is shifted, so if the detection areas are correctly detected, the detection speed is exactly the same. When the values of the two systems detected in the overlapping area are different, there is a possibility that the error of the predicted value is large and the number of laps is wrong. Therefore, it is preferable to correct the result by a method such as selecting two of the selected laps that are shifted from both outer half cycles, performing the same processing on these, and selecting a value that has the same value in the overlapping region.

  Also, it is not always necessary to have two systems, and three systems shifted by 1/3 turn or a plurality of systems shifted at appropriate intervals may be used.

  FIG. 18 is a block diagram showing a configuration when two systems are used for the speed estimation unit 63 in the third embodiment. In FIG. 18, the FFT result of the FFT unit 62 is input to the speed estimation unit 63 to the speed estimation unit 63. On the other hand, a predicted speed value is also input. The FFT result and the predicted speed value are input to the aliasing correction unit 634. In the aliasing correction unit 634, the spectral region selection unit 6345 selects two spectral regions so as to include the predicted speed out of the spectral regions set in advance by deviating, for example, by a half turn as described above, from the predicted speed value. The selected result is notified to the shift number selection unit 6346. The shift number selection unit 6346 has a table regarding how many spectrums of each frequency step should be shifted in the selected spectrum region, and refers to this to determine the shift number of the spectrum of each frequency step. select. The result is notified to spectrum shifters 6347-1 and 6347-2.

  In spectrum shift sections 6347-1 and 6347-2, the FFT results are input, the start point is shifted by about a half turn, and the spectrum of each frequency step is based on the shift number selection result so that the Doppler frequencies of all frequency steps are aligned. Shift. The shift results are non-coherently integrated by the non-coherent integrators 631-1 and 631-2 and input to the peak detectors 632-1 and 632-2.

  The peak detectors 632-1 and 632-2 detect peak values from the added spectrum as a result of non-coherent integration. However, in this form, as described above, there is a possibility that correct detection cannot be performed at both ends of the spectrum. Accordingly, the peak detectors 632-1 and 632-2 use a predetermined number of bins at both ends as detection prohibited areas, and detect peak values within an area where correct detection is possible. The detected peak value is sent to the speed calculators 633-1 and 633-2.

  The speed calculators 633-1 and 633-2 calculate the speed from the detected peak value. The speed calculation units 633-1 and 633-2 are notified of the frequency of the spectrum region selected by the spectrum region selection unit 6345, that is, the number 0 bin of the non-coherently integrated spectrum. The speed calculators 633-1 and 633-2 calculate the speed at which the ambiguity is eliminated based on the value. The calculated speed is sent to the result fusion / selection unit 635.

  The result fusion / selection unit 635 fuses or selects the speeds detected in either of the two systems, or in some cases. As described above, if only one of them is detected as described above, select it, and if a target that seems to be the same target is detected on both sides, check whether the values are equal. Depending on the case, an instruction is fed back to the spectral region selection unit 6345 so as to reselect the spectral region.

  According to the present embodiment, there is an advantage that the number of spectrum shifts at each frequency step can be determined simply by referring to a prepared table.

(Fourth embodiment)
Next, as a fourth embodiment, a mode in the case where the predicted speed is unknown in Low-PRF will be described. The basic configuration of the radar apparatus according to this embodiment is the same as that of the apparatus shown in FIG.

  FIG. 12 illustrates the state of the spectrum including the aliasing, but this embodiment provides a method for generating a spectrum having the same state as that in FIG. That is, even if the predicted speed is unknown, in most cases the maximum relative speed that can be detected can be defined. Therefore, the same spectrum is repeatedly arranged from the absolute frequency 0 for each frequency step up to the number of turns including the Doppler frequency corresponding to the maximum relative speed.

  An example of the result is shown in FIG. In the example of FIG. 19A, the spectrum of 10 frequency steps is overwritten, but the peak positions of the plurality of frequency steps do not match except for the correct number of folds. FIG. 19B shows the result of non-coherent integration of the 10 spectra shown in FIG. Multiple peaks are detected, but the peak with the correct number of folds is the highest. Therefore, by detecting the highest peak value, it is possible to identify the Doppler frequency with the correct number of turns.

  FIG. 20 is a block diagram illustrating a configuration of the speed estimation unit 63 of the radar apparatus according to the fourth embodiment. The speed estimation unit 63 illustrated in FIG. 20 includes a spectrum duplication / combination unit 636, a non-coherent integration unit 637, a peak detection unit 638, an ambiguity elimination unit 639, and a speed calculation unit 63A.

  The spectrum duplicating / combining unit 636 duplicates and combines the spectrum of each frequency step from the input FFT result of each frequency step up to the number of folds including the maximum relative velocity that can be detected. The non-coherent integration unit 637 performs non-coherent integration on the copied and combined results. The peak detector 638 detects a plurality of peaks from the non-coherent integration result. The ambiguity elimination unit 639 selects the maximum peak as the peak of the correct number of times of folding from the plurality of peaks detected by the peak detection unit 638. The speed calculator 63A converts the frequency of the selected peak into a speed.

  As described above, according to the present embodiment, the spectrum of each frequency step is duplicated and combined side by side up to the number of folds including the maximum relative speed that can be detected, and then the non-coherent integration, peak detection is performed, and the correct fold is performed. Since the peak of the number of times is selected and converted into the speed, the relative speed can be detected even when the number of times of Doppler frequency folding is unknown in advance in Low-PRF.

  When a plurality of targets having different numbers of turns are mixed, the peak detection unit 638 detects all peak values exceeding the threshold for the time being and sends them to the ambiguity cancellation unit 639. Since the peak of the same target appears periodically, the ambiguity canceling unit 639 determines the periodicity of the detected plurality of peaks, collects the periodically appearing peaks into groups, and sets the maximum in each group. Select the peak value of. Thereby, it is possible to detect a plurality of targets having different numbers of times of folding.

(Fifth embodiment)
In the present embodiment, since it is essential to absorb the difference in PRI, it is not always necessary to use a radar type having a different carrier frequency for each step. The above embodiments can be applied even when the carrier frequencies of all steps are the same. However, in the synthetic band radar and the frequency agility radar in which the carrier frequency is different for each step, the speed estimation accuracy can be further improved by making further improvements.

  FIG. 21 is obtained by removing noise from the Doppler spectrum of FIG. As a result, the spectrum is very clear, and it can be seen that the position of the peak is slightly different depending on the frequency step. In FIG. 21, the target relative speed value is large and there is Doppler ambiguity. Therefore, the difference in Doppler frequency due to frequency steps appears to be relatively large. In this way, if the peak heights of all frequency steps are relatively uniform, even if non-coherent integration is performed as it is, the peak is detected and converted to speed using the center frequency of the entire bandwidth, to some extent. Speed can be detected with accuracy.

  However, when the SNR is low or internal interference occurs at the multipoint target, the peak height for each frequency step may vary greatly. In such a case, if non-coherent integration is simply performed, a value biased to a high peak frequency may be detected, which causes an error.

  Here, in FIG. 21, when the tails of the individual peak waveforms are viewed, the left side is widened and the right side is narrowed. This is because the spread width of the peak waveform varies depending on the frequency step. Since the number of pulses in all frequency steps is the same, the ratio of the spread of the peak waveform to the total spectrum width is the same in all frequency steps. On the other hand, in this embodiment, since the bin interval frequency of the spectrum of all frequency steps is made uniform, the number of bins is different for each frequency step. Since the ratio of the peak waveform spread to the entire spectrum width is the same, the peak waveform spread bin number varies depending on the frequency step. If this slight difference in the spread of the peak waveform occurs at the same time as the shift of the peak position, even if the peak heights of all frequency steps are substantially the same, it will contribute to an increase in error.

  Therefore, in the fifth embodiment, when there is room for modifying the PRI, all peak positions are aligned by performing the following. That is, the Doppler frequency of each frequency step for a certain target speed is proportional to the carrier frequency of each frequency step. In order for a peak to appear at the same position, the frequency scale itself may be proportional to the carrier frequency. That is, the bin interval of each frequency step may be proportional to each carrier frequency.

If the FFT frame length used in the configuration so far is T _FFT0 (sec), the bin interval B is 1 / T _FFT0 .

This is set to a value B _i proportional to the carrier frequency of each frequency step. Simply, it may do it over the carrier frequency f _i in the bin interval. However, to a value close to the value of the bank interval value up to now, may do it by multiplying the ratio of the carrier frequency f _i for example to the central frequency f _c of the band.

Of course, B _i only needs to be multiplied by a coefficient proportional to f _i , so the denominator may be a constant other than f _c . A frame length slightly changed for each frequency step i in accordance with the above equation is _defined as T _FFTi .

_Assuming that the number of samples in the FFT frame is the number N _Li for each frequency step i as determined so far, the value obtained by dividing T _FFTi by N _Li is the PRI for aligning the peaks at the same speed in all frequency steps. Become.

FIG. 22 shows the FFT frame configuration. In this example, it is assumed that the carrier frequency of each frequency step increases in ascending order of numbers. Compared with FIG. 5, the FFT frame length becomes shorter as the frequency step number increases. Thus, when trying to absorb the difference in Doppler frequency due to the difference in carrier frequency in the FFT frame, the FFT frame length is not completely the same in all frequency steps, and the degree of change in the carrier frequency is subtle. It becomes a different value. PRI _i (i = 1 to N _f ) is slightly shorter as compared with FIG. 5 as the frequency step number increases, but is not clearly shown in the figure.

  As described above, the PRI length is preferably rounded to a unit of time interval corresponding to the sampling rate of the signal processing unit 6, but rounding is preferably performed lastly after performing the above processing.

  As described above, the fact that the FFT frame length becomes short means that the bin interval after FFT becomes large. This is shown in FIG. Taking the horizontal axis as the absolute value of the frequency, the bin interval frequency is gradually increased as the frequency step, that is, the carrier frequency increases. Therefore, the Doppler frequencies of the same bin number in each frequency step are not equal to each other and increase / decrease in proportion to the carrier frequency.

  FIG. 24 shows the PRI length and bin interval frequency for each frequency step. The slope of the change in the PRI length is slightly less than in FIG. 8, and the bin interval frequency changes in proportion to the carrier frequency. The reason why the bin interval frequency is uneven is that the PRI is rounded at a time interval corresponding to the sampling rate.

  Thereafter, the result of calculating the spectrum of each frequency step in the same procedure as in the previous embodiments is shown in FIG. FIG. 25A is a peak waveform diagram when there is no noise as in FIG. Compared with FIG. 21, it can be seen that all the peaks clearly overlap. FIG. 25 (b) is a peak waveform diagram when there is noise as in FIG. Although some variations remain in height and peak position due to noise, it can be seen that the peak positions are well aligned as compared to FIG.

  This embodiment is almost the same as the embodiment and the description thereof except that the PRI length is slightly different, that is, the PRI length is corrected corresponding to the carrier frequency of the frequency step. However, in the description so far, the portion expressed as aligning the frequency is replaced with a form in which the frequency scaled by the carrier frequency is aligned.

  When calculating the speed from the detected Doppler frequency, calculate how many Doppler frequencies the bin number of the detected peak corresponds to at any one frequency step, and calculate the speed at the carrier frequency of that frequency step. Convert it. Furthermore, in order to absorb errors due to rounding, the frequency may be converted into speeds at all frequency steps and averaged.

(Application examples)
FIG. 26 is a block diagram showing a configuration when the radar apparatus according to the present embodiment is a synthetic band radar. In FIG. 26, the same parts as those in FIG. 1 are denoted by the same reference numerals, and different parts will be described here.

  In the above-described embodiment, only the part for detecting the speed has been described, but the radar often needs to detect the distance and / or the angle. Therefore, FIG. 26 shows the configuration of the radar apparatus including processing up to the generation of the composite band waveform based on the first embodiment shown in FIG.

  In FIG. 26, a synthetic band processing unit 7 is added after the signal processing unit 6. The combined band processing unit 7 receives the speed estimation result and the FFT result from the signal processing unit 6 and, in some cases, the peak bin number. In the synthesized band processing unit 7, the step representative value extracting unit 71 first extracts a frequency step representative value to which the synthesized band is applied from the FFT result based on the peak bin number or the estimated speed. Next, the speed estimation result and the frequency step representative value are input to the phase correction unit 72, and the phase change due to the passage of time of the frequency step is corrected. The phase correction result is input to the synthesized band waveform generation unit 73, and a synthesized band waveform is generated by inverse Fourier transform. Thereafter, it is subjected to processing such as distance detection and angle detection. According to this configuration, it is possible to detect a speed with a small error even with a variable PRI and a low SNR, and as a result, it is possible to improve the distance detection accuracy from the combined band waveform.

  FIG. 27 is a block diagram showing a configuration when the radar apparatus according to the present embodiment is a frequency agility radar. In FIG. 27, the same parts as those in FIG. 1 are denoted by the same reference numerals, and different parts will be described here.

  Usually, in a frequency agility radar, since the carrier frequency changes with each pulse, it is difficult to make it a pulse Doppler radar, but it is possible to detect Doppler as a form that transmits and receives multiple pulses at each frequency step. Is possible. In the configuration of FIG. 27, the signal processing unit 6 detects the Doppler frequency, and the distance detection unit 8 detects the distance from the output of the radar wave receiving unit 5. These results are sent to a subsequent pulse Doppler processing unit (not shown), and pulse Doppler detection processing is performed.

  Furthermore, as described above, in this embodiment, the frequency does not necessarily have to be different for each step. Even in a pulse Doppler radar in which the carrier frequencies of all the pulses are the same, when the pulse is divided into steps and the PRI is variable for each step, this embodiment is applied, so that high speed detection accuracy can be achieved with a low SNR. Realized.

  FIG. 28 is a block diagram showing a configuration when the radar apparatus of the present embodiment is used for a guidance apparatus. In FIG. 28, information detected by the radar apparatus 100 is output to the guidance signal generation unit 200. The guidance signal generation unit 200 generates a guidance signal for guiding the flying object from the input information, and controls a flying body driving unit (not shown). The radar apparatus 100 may be, for example, a synthetic band radar shown in FIG. 26 or a frequency agility radar shown in FIG. Alternatively, it may be a device that includes the radar device of the present embodiment and performs detection processing such as distance and angle not shown.

  FIG. 29 shows a flowchart of a radar signal processing method applied in the radar apparatus of this embodiment. First, a pulse representative value is extracted for each pulse from a signal obtained by digitizing a received radar wave (in some cases, a demodulated signal) (step S1). Next, for each step, an appropriate number of zeros corresponding to each PRI is added to the pulse representative value sequence of the corresponding step to form an FFT frame (step S2). Next, the FFT frame to which zeros in each step are added is subjected to FFT (step S3). Next, the FFT results of all steps are non-coherently integrated (step S4). Finally, a peak is detected from the non-coherent integration result, and the target speed is estimated (step S5).

  With the above processing, even if the PRI is a variable PRI that handles different pulse trains for each frequency step, even if the SNR is low, the Doppler spectrum peaks at each frequency step are aligned and coherent integration is performed, so the peak detection accuracy of the integration result is improved. Thus, high speed detection accuracy can be obtained.

  Note that the present embodiment is not limited to the above-described embodiment as it is, and can be embodied by modifying constituent elements without departing from the scope of the invention in the implementation stage. In addition, various inventions can be formed by appropriately combining a plurality of components disclosed in the embodiment. For example, some components may be deleted from all the components shown in the embodiment. Furthermore, constituent elements over different embodiments may be appropriately combined.

DESCRIPTION OF SYMBOLS 1 ... Synthesizer, 2 ... Radar wave transmission part, 3 ... Circulator, 4 ... Antenna, 5 ... Radar wave receiving part, 6 ... Signal processing part, 61 ... Zero addition part, 62 ... FFT part, 63 ... Speed estimation part, 631 , 631-1, 631-2 ... non-coherent integration unit, 632, 632-1, 632-2 ... peak detection unit, 633, 633-1, 633-2 ... speed calculation unit, 634 ... aliasing correction unit, 6341 ... Shift number determination unit, 6342 ... spectrum shift unit, 6343 ... zero addition unit, 6344 ... spectrum shift unit, 6345 ... spectrum region selection unit, 6346 ... shift number selection unit, 6347-1, 6347-2 ... spectrum shift unit, 635 ... result fusion / selection unit, 636 ... spectral replication / combination unit, 637 ... non-coherent integration unit, 638 ... peak detection unit, 639 ... a Bigyuiti canceller, 63A ... speed calculation unit, 7 ... synthetic band processing unit, 71 ... Step representative value extracting unit, 72 ... phase correction unit, 73 ... synthetic band waveform generation unit, 100 ... radar apparatus, 200 ... induction signal generator

Claims (15)

In each of a plurality of steps repeated within one transmission period (CPI) , a plurality of pulses are generated at different unique pulse repetition intervals (PRI) for each step, and the plurality of pulses at each step are sequentially used as radar waves. A transmission unit for transmission;
A receiver for receiving the reflected wave of the radar wave;
Signal processing for calculating a Doppler spectrum by performing Fast Fourier Transform (FFT) on a received signal corresponding to a plurality of pulses at each step obtained by the receiving unit, and estimating a target velocity from the calculation result of the Doppler spectrum And
Comprising
The signal processing unit
Applying the FFT after adding the number of zeros determined for each step around the value of the plurality of pulses of each of the plurality of steps,
A radar apparatus in which the PRI, the FFT frame length, and the number of zeros to be added are determined so that the time corresponding to the FFT frame length after adding the zeros is substantially the same in all steps. The radar apparatus according to claim 1, wherein the plurality of steps are frequency steps that generate the plurality of pulses at carrier frequencies different from each other .   3. The PRI of the frequency step is a value obtained by multiplying the value obtained by dividing the FFT frame length by the total number of samples including zero of the frequency step by a coefficient that is inversely proportional to the ratio of the carrier frequency of the frequency step. Radar equipment.   4. The radar device according to claim 1, wherein the PRI of the plurality of steps is a value further rounded at a time interval corresponding to the sampling rate of the FFT. 5.   The radar apparatus according to claim 1, wherein the plurality of steps further includes a difference between a maximum value and a minimum value of the number of samples in the FFT frame including zero approximately equal to the number of steps.   6. The radar apparatus according to claim 5, wherein the plurality of steps further differ from each other in the number of samples in the FFT frame including zero by one.   The radar apparatus according to claim 5, wherein the plurality of steps further includes a difference of 1 or 0 in the number of samples in the FFT frame including zero in any two steps.   5. The signal processing unit according to claim 1, wherein the signal processing unit performs non-coherent integration of the Doppler spectrum of the plurality of steps, detects a peak of the integration result, and estimates the target speed from the peak. The radar device according to any one of the above. When the PRI is of a low pulse repetition frequency (Low-PRF) type and the predicted relative velocity with respect to the target is known,
9. The non-coherent integration is performed according to claim 8, wherein the signal processing unit shifts the Doppler spectrum of each of the plurality of steps so that Doppler frequencies around the predicted relative velocity values are substantially uniform in any bin. Radar device.   The signal processing unit sets two or more Doppler frequencies aligned in the same bin around the predicted relative velocity value, shifts the Doppler spectrum of each of the plurality of steps for each Doppler frequency, and performs the non-coherent integration. The radar device according to claim 9. When the PRI is of a low pulse repetition frequency (Low-PRF) type,
The radar apparatus according to claim 8, wherein the signal processing unit generates a plurality of copies of the Doppler spectrum of each of the plurality of steps, arranges them in series, combines them, and integrates the Doppler spectra combined in each step with the non-coherent integration. When the bandwidth is divided into a plurality of frequency bands, and the divided bands are applied to the synthetic band radar corresponding to the step,
A composite band that generates a composite band waveform by inverse Fourier transform after correcting a phase change of a Doppler spectrum over time of the plurality of steps based on the target speed estimation result and FFT result obtained by the signal processing unit The radar apparatus according to claim 2, further comprising a processing unit. When applied to a frequency agility radar where the carrier frequency changes within the measurement period,
The radar apparatus according to claim 2, further comprising a distance detection unit that detects the target distance from the received signal. A radar device according to any one of claims 1 to 13,
A guidance device comprising: a guidance signal generation unit that generates a guidance signal based on target detection information including the speed from the radar device. In each of a plurality of steps repeated within one transmission period (CPI) , a plurality of pulses are generated at a unique pulse repetition interval (PRI) different from each other for each step, and the plurality of pulses are sequentially converted into radar waves for each step. A radar signal processing method for estimating a target speed by transmitting a radar signal transmitted and receiving a reflected wave of the radar wave,
Extracting a pulse representative value of each of the plurality of pulses from the radar signal;
Adding a number of zeros corresponding to the PRI of each of the plurality of steps to the pulse representative value sequence of the step to form a Fast Fourier Transform (FFT) frame;
FFT is performed on the FFT frame of each of the plurality of steps to calculate a Doppler spectrum;
Non-coherent integration of the multiple-step Doppler spectrum, and detecting a peak from the integration result,
Estimating the target speed from the peak value,
The radar characterized in that the PRI, the FFT frame length, and the number of zeros to be added are determined so that the time corresponding to the FFT frame length after adding the zeros is substantially the same in all steps. Signal processing method.

Images