JP4549929B2 - Sensor signal processing system - Google Patents

Description

  The present invention relates to a sensor signal processing system that uses a signal processing algorithm that performs measurement processing of observation information after estimating the number of signal sources.

  As a signal processing algorithm for measuring observation information after estimating the number of signal sources, for example, Non-Patent Document 1 proposes a signal processing device using a MUSIC (MUltiple SIgnal Classification) algorithm. In this MUSIC algorithm, the correlation matrix of the received signal is calculated, and the correlation matrix is further subjected to eigenvalue decomposition to obtain eigenvalues and eigenvectors. The eigenvalues and eigenvectors are composed of signal eigenvalues and signal eigenvectors representing signal components from the signal source, and noise eigenvalues and noise eigenvectors representing noise components. In general, a signal eigenvalue and a noise eigenvalue are discriminated from the magnitude of the eigenvalue, and the number of signal sources is estimated from the number of signal eigenvalues. Based on the estimation result, an evaluation function, for example, in the case of MUSIC, a MUSIC spectrum is calculated, and observation information is obtained from the peak value.

  The signal eigenvalues and noise eigenvalues can be distinguished by arranging the eigenvalues in descending order, and setting the eigenvalues larger than the eigenvalues whose magnitude has decreased most rapidly as signal eigenvalues. And a method in which the noise of the receiver is measured in advance, an appropriate threshold value is estimated from the power, and a higher eigenvalue is used as the signal eigenvalue.

  In the conventional sensor signal processing system, correct observation information can be obtained if the number of signal sources is correctly estimated, but the number of signal sources is not suitable for estimation of the number of signal sources, for example, in a very low S / N environment. As a result, there is a problem that a false target or incorrect observation information may be generated.

  In order to solve this problem, for example, Patent Document 1 proposes a method of estimating the correct number of signal sources by using past tracking information for estimating the number of signal sources. The sensor signal processing system described in Patent Literature 1 acquires a received signal from a signal source using a receiver having a sensor, and estimates the number of signal sources when acquiring observation information of the signal source. Is used to measure the observation information from the signal source number hypothesis memory that stores the number of signal source hypotheses in advance and the received signal via the A / D converter. A signal preprocessing unit to calculate, an observation information extraction unit to calculate all desired observation information in the number of signal sources based on a plurality of hypotheses in the signal source number hypothesis memory, using the output information of the signal preprocessing unit, and an observation Based on the observation information extracted by the information extraction unit, a trajectory estimation unit that creates a temporal trajectory in the signal space of the signal source in each hypothesis using a tracking filter, and the trajectory created by the trajectory estimation unit is a tracking filter Exercise A trajectory evaluation unit that obtains the likelihood of a trajectory that indicates how much the data matches a hypothesis, and a temporary evaluation unit that determines the number of signal sources postponed by comparing the fitness of a hypothesis calculated from the likelihood of the trajectory A hypothesis deletion unit that deletes a hypothesis whose hypothesis suitability has decreased from the signal source number hypothesis memory.

  In the method disclosed in Patent Document 1, the trajectory obtained by processing the observation information extracted by the observation information estimation algorithm based on the hypothesis of the number of signal sources with the tracking filter matches the motion model set in advance in the tracking filter. This is a method of deferring the correct number of signal sources, assuming that the hypothesis is more reliable. According to this method, since past information can also be used for estimating the number of signal sources, if the signal sources are in a state that can be decomposed by an algorithm, the signal sources can be correctly estimated even in an environment where it is difficult to estimate the number of signal sources. The number can be estimated.

  Here, the number of signal sources estimated by these signal source number estimation methods is the number of wave sources that can be resolved by an algorithm. For example, when two signal sources are approaching a state where they cannot be disassembled, a correct observation value cannot be obtained even if detection processing is performed assuming that the number of signal sources is two. In this case, generally only one observation value is obtained, and the other one detects a false image that is unrelated to the signal source. For this reason, when the state is changed from a resolvable state to an unresolvable state due to the approach of a signal source or the like, it is necessary to quickly estimate the number of resolvable signal sources and obtain observation information. Conversely, the same applies to the case where the number of signal sources that can be decomposed changes due to separation of a plurality of signal sources that could not be resolved.

Ralph.O.Schmidt, "Multiple Emitter Location and Signal Parameter Estimation", IEEE Trans.Antennas Propagat., Vol.AP-34, no.3, Mar. 1986. JP 2004-309209 A (paragraph 0019)

  The conventional sensor signal processing system is configured as described above. In the method of Patent Document 1, the number of signal sources that can be decomposed is estimated using past tracking information. When the number of signal sources is changed, there is a problem that it takes time to detect the signal source. Specifically, taking as an example a case where two signal sources cannot approach and resolve, in the hypothesis where the number of signal sources is 2, one observation value and one false image can be obtained. For this reason, the trajectory is maintained by memory track processing (tracking filter processing that continues the trajectory with the predicted value) immediately after disassembly, but if the trajectory of either one of the signal sources is interrupted, the reliability decreases and the signal It is detected that the number of sources has changed. Since the past tracking information is used in this way, when the number of signal sources changes, detection takes time instead. In the meantime, the problem that the target detection performance and the tracking maintenance performance deteriorate also occurs.

  In addition, since the change in the number of signal sources due to such approach and separation is temporary, it is desirable that the signal source can be continuously tracked during the approach and separation. However, the conventional method has a problem that the trajectory is interrupted to recognize that the number of signal sources has changed.

  Further, in the method of Patent Document 1, the tracking results of each signal source number hypothesis are all managed independently, and the tracking results are not shared between hypotheses. For this reason, the number of signal source hypotheses that have been correctly answered may not improve tracking performance due to the influence of past erroneous observation values, and the tracking information in the previous correct answer hypothesis is not effectively utilized. There was also a problem.

  The present invention has been made to solve the above-described problems, and an object thereof is to obtain a sensor signal processing system capable of correctly estimating the number of signal sources and improving target detection performance and tracking maintenance performance. .

  It is another object of the present invention to provide a sensor signal processing system that can effectively utilize tracking information by sharing tracking results among signal source number hypotheses and can quickly cope with changes in the number of signal sources.

  The sensor signal processing system according to the present invention uses a signal processing algorithm that performs measurement processing of observation information after estimating the number of signal sources, and can be decomposed at the previous time stored in the decomposable locus memory. The tracking information of the tracking signal source predicted as follows and stored in the unresolvable track memory for determining whether the tracking signal source can be resolved at the current time for each signal source hypothesis. Whether the signal source being tracked is separated for each signal source hypothesis from the tracking information of the signal source being tracked that is predicted to be indecomposable at the previous time and the observed value extracted at the current time Based on the determination results of the separation signal source determination unit, the resolvable trajectory prediction unit, and the separation signal source determination unit, the signal source being tracked and the observation value extracted at the current time It is obtained by a correlation determination unit for determining the relationship.

  According to the present invention, it is possible to correctly estimate the number of signal sources and to improve the target detection performance and tracking maintenance performance.

Embodiment 1 FIG.
First, the basic concept of the present invention will be described using the MUSIC algorithm as an example. The present invention takes into account a plurality of hypotheses assuming the number of signal sources (hereinafter referred to as signal source number hypotheses), and based on the result of tracking the observation values obtained by each signal source number hypothesis, the likelihood of each signal source number hypothesis. The present invention relates to a method of calculating the hypothesis reliability, which is a measure of the likelihood, and deferring the correct signal source number hypothesis. In particular, when a change in the number of signal sources occurs, a method is shown in which correct observation information is acquired by detecting the change early and performing corresponding signal processing.

Here, the following cases are assumed as the number of signal sources changes.
(A) When the number of signal sources that can be decomposed decreases because the signal source approaches and becomes incapable of decomposition, (B) When the number of signal sources that can be decomposed increases because the signal source that has approached separates ( C) When the signal source is separated (eg, disconnecting the booster)
(D) When the signal source disappears (E) When a new signal source enters the coverage area Among these, for (E), the method of estimating the number of signal sources by the conventional method is effective. However, with regard to the above (A) to (D), the application of the present invention makes it possible to estimate the number of signal sources and maintain tracking faster and more accurately than the conventional method. As for (E), the number of signal sources is estimated by the conventional method. However, by applying a part of the present invention, improvement in tracking performance can be expected. This will be described later.

Hereinafter, an outline of a specific concept will be described using an example of the MUSIC algorithm.
FIG. 1 is a diagram showing an example of a time-series change of the MUSIC spectrum when the signal sources (A) and (B) are approaching and separating. In (A) above, the two signal sources gradually approach in FIGS. 1 (a) and 1 (b), and the state in FIG. The MUSIC spectrum in the signal source number hypothesis (signal source number 2) is shown. When the signal source cannot be resolved, only one peak is obtained and a false image is generated. However, when looking at the time series, it can be predicted that if the movement of the signal source is tracked, it will be approached in advance, so that it cannot be resolved. In such a case, a single peak is assigned to two signal sources and a false image is discarded, or a peak obtained by a hypothesis of the number of signal sources with one signal source is assigned to two signal sources. By doing so, tracking of the signal source can be maintained.

  On the other hand, in (B) above, the time-series change of the MUSIC spectrum when the approaching signal source is separated is as shown in FIGS. 1 (c) to 1 (a), contrary to (A) above. . In the state where disassembly is impossible in FIG. 1C, the position where the false image appears is not constant. On the other hand, when the state becomes decomposable, as shown in FIGS. 1 (b) and 1 (a), a new peak can be obtained near the unresolvable signal source being tracked and can be resolved. It can be detected that the number of signal sources has changed.

  FIG. 2 is a diagram showing an example of a time-series change of the MUSIC spectrum in the case of separation of the signal source of (C) above, from the state where the number of true signal sources in FIG. Represents the MUSIC spectrum of the hypothesis that the number of signal sources is 2 when the number of true signal sources is 2 due to separation of targets (other than separation). In the state before separation, a false image is generated at an indefinite position under the hypothesis that the number of signal sources is two. On the other hand, after separation, a false image generated at an indefinite position disappears, and a new peak is generated near the signal source before separation, and it is possible to detect that the signal source is separated. In FIG. 2, the state of the spectrum change is the same as in the case of separation of the signal source described in (B) above, but in (B), the true target number itself has not changed, and in the past, Only when there is a decrease in the number of signal sources as in (A). On the other hand, in (C), the true target number, that is, the number of correct signal source hypotheses is different. Therefore, it is possible to take over the tracking information from the correct signal source number hypothesis up to that time to the new correct signal source number hypothesis, and the past information can be used more efficiently.

  FIG. 3 is a diagram showing a time-series change example of the MUSIC spectrum in the case of the disappearance of the signal source (D), and assumes a case where one of the two signal sources disappears. FIG. 3A is before the disappearance, and FIG. 3B is after the disappearance. If one signal source disappears, a false image is generated in the signal source number hypothesis where the number of signal sources is 2, so that the memory track continues in the signal source that has disappeared. By monitoring the number of track memory tracks corresponding to the lost signal source, it can be detected that the number of signal sources has changed. Even in the conventional method, it is possible to detect the disappearance of the signal source by waiting for the likelihood of the signal source number hypothesis to be lowered, but particularly when application to the MUSIC algorithm is assumed. In the case of the hypothesis of the number of signal sources in which one signal source is mistaken a lot, since one observation value and one false image of the correct signal source are obtained, tracking of the remaining signal sources can often be continued. Since the influence of the memory track of one wake on the likelihood of the signal source number hypothesis is small, it takes time to reduce the likelihood of the hypothesis. Therefore, the present invention which focuses on a certain trajectory rather than the entire signal source number hypothesis has an application effect. When one signal source disappears because it has gone out of the radar coverage, it is also possible to predict in advance that the signal source will leave the radar coverage by looking at the time series of the tracking results. Thereby, a change in the number of signal sources can be detected quickly.

  By the way, with regard to (E), it is impossible to detect a change in the number of signal sources in advance from the past tracking information. However, after detecting a new signal source by the conventional method, the number of signal sources newly hypothesized becomes a correct answer. As tracking information, the tracking information of the signal source hypothesis that was correct until then can be transferred and used, so that the tracking performance can be improved more quickly than the conventional method. it can.

A specific system configuration will be described below with regard to the best mode for carrying out the present invention to which the above concept is applied.
First, definitions of terms used in the description of the present invention will be given.
“The number of true signal sources” represents the number of actual signal sources regardless of whether or not decomposition is possible. The “number of signal sources that can be decomposed” indicates the number of signal sources that can be decomposed by a signal processing algorithm to be used. For example, if there are actually two signal sources but only one signal source can be seen because it is too close, the number of true signal sources = 2 and the number of signal sources that can be resolved = 1.

  The “signal source number hypothesis” expresses how many true signal sources exist as a hypothesis. Here, the “trajectory evaluation function” refers to the degree to which the motion model assumes in advance how the signal source will move with respect to the trajectory of each signal source created in each signal source number hypothesis. It refers to a function that evaluates whether it is doing. The “hypothesis reliability” is expressed by using the above evaluation function, normalized so that each signal source number hypothesis is close to the correct answer so that the signal source number hypotheses can be compared. “Tracking information” refers to all information attached to the trajectory such as predicted values, estimated values (smooth values), gain matrix, error covariance matrix, residual covariance matrix, and correlation observation values that are output from the tracking filter. Yes.

Next, the best mode for carrying out the present invention will be described.
FIG. 4 is a block diagram showing the configuration of the sensor signal processing system according to Embodiment 1 of the present invention. In addition, although the case where the MUSIC algorithm is used as a signal processing algorithm is assumed, the present invention can be applied as long as the observation value is extracted based on the information after estimating the number of signal sources. is there.

  The sensor signal processing system shown in FIG. 4 includes a reception unit 101, an A / D conversion unit 102, a signal preprocessing unit 103, an unresolvable locus memory 104, a decomposable locus memory 105, a signal source number hypothesis memory 106, and an observation information extraction unit. 107, resolvable trajectory prediction unit 108, separation signal source determination unit 109, correlation determination unit 110, trajectory estimation unit 111, trajectory prediction unit 112, trajectory evaluation unit 113, hypothesis evaluation unit 114, hypothesis deletion unit 115, hypothesis setting unit 116 , A display trajectory selection unit 117, a display unit 118, and an inter-hypothesis memory transfer unit 119.

  In FIG. 4, the receiving unit 101 receives a reflected wave obtained by reflecting a transmission wave from a transmission source (not shown) on a target. The A / D converter 102 performs A / D conversion on the received signal acquired by the receiver 101. The signal preprocessing unit 103 performs preprocessing necessary for estimating the number of signal sources from the A / D converted received signal. The unresolvable trajectory memory 104 stores tracking information of the signal source being tracked that is predicted to be unresolvable. The resolvable trajectory memory 105 stores the tracking information of the signal source being tracked that is predicted to be resolvable. The signal source number hypothesis memory 106 stores a signal source number hypothesis as to how many true signal sources there are. The observation information extraction unit 107 extracts observation values as many as the number of signal source number hypotheses stored in the signal source number hypothesis memory 106 from the preprocessing result of the signal preprocessing unit 103.

  The decomposable trajectory predicting unit 108 performs tracking for each signal source number hypothesis stored in the signal source number hypothesis memory 106 and is predicted to be decomposable at the previous time stored in the decomposable trajectory memory 105. From the predicted value in the tracking information of the signal source, it is determined whether the signal source being tracked can be decomposed at the current time, and the predicted value of the signal source being tracked determined to be disassembled is output. The separation signal source determination unit 109 stores the tracking information of the signal source being tracked that is predicted to be unresolvable and is stored in the unresolvable trajectory memory 104 from the observation values extracted by the observation information extraction unit 107. The observation value existing in the correlation gate indicating the predetermined range from the predicted value in the signal is extracted for each signal source number hypothesis stored in the signal source number hypothesis memory 106, and the signal source being tracked is separated. It determines whether or not, and outputs the predicted value of the signal source being tracked and its determination result.

  The correlation determination unit 110 receives the observed value output from the observation information extraction unit 107, the predicted value of the signal source being tracked determined to be decomposable output from the decomposable locus prediction unit 108, and the separated signal source determination unit 109. The estimated value of the output signal source being tracked and the determination result thereof, and the past estimated value in the tracking information of the tracked signal source predicted to be unresolvable stored in the unresolvable trajectory memory 104 For each signal source number hypothesis stored in the signal source number hypothesis memory 106, a predetermined range from the predicted value of the signal source being tracked that is determined to be decomposable output from the decomposable locus prediction unit 108 Correlation indicating a predetermined range from the predicted value of the signal source being tracked that is determined to be non-separated and is determined to be correlated with the observed value present in the correlation gate indicating Within the gate In the correlation gate indicating a predetermined range from a predicted value obtained from a past estimated value of the signal source being tracked that is determined to be separated and is output from the separated signal source determination unit 109. By determining that there is a correlation between the observed value and the observed value, the correlation between the signal source being tracked and the observed value is determined, and the signal being tracked output from the decomposable trajectory prediction unit 108 and determined to be decomposable The predicted value of the source, the predicted value of the signal source being tracked output from the separation signal source determination unit 109, the determination result, and the observed value for which the correlation is determined are output.

  The trajectory estimation unit 111 outputs the predicted value of the signal source being tracked that has been determined to be decomposable from the decomposable trajectory prediction unit 108 and the signal being tracked from the separation signal source determination unit 109 that is output from the correlation determination unit 110 For each signal source number hypothesis stored in the signal source number hypothesis memory 106 based on the predicted value of the source, its determination result, the observed value for which the correlation has been determined, and the gain matrix held therein, Estimate the temporal trajectory and output the estimated value. The trajectory prediction unit 112 predicts the trajectory at the next observation based on the estimated value output from the trajectory estimation unit 111 and the transition matrix held therein, calculates the predicted value of the signal source, and is incapable of being decomposed. Tracking information of the signal source being predicted (including the predicted value of the signal source at the next observation) is stored in the unresolvable trajectory memory 104, and tracking information of the signal source being tracked that is predicted to be resolvable is stored. (Including the predicted value of the signal source at the next observation) is stored in the resolvable trajectory memory 105.

  The trajectory evaluation unit 113 outputs the predicted value of the signal source being tracked that has been determined to be decomposable from the decomposable trajectory prediction unit 108 and the signal being tracked from the separation signal source determination unit 109 that is output from the correlation determination unit 110 Based on the predicted value of the source, its determination result, and the observed value for which the correlation is determined, the detection probability of the actual observed value is obtained for each locus of the signal source number hypothesis stored in the signal source number hypothesis memory 106. The likelihood of each trajectory is calculated using a predetermined evaluation function based on the actual detection value detection probability obtained and the signal source detection probability defined in advance. The hypothesis evaluation unit 114 calculates the hypothesis reliability for each signal source number hypothesis stored in the signal source number hypothesis memory 106 based on the likelihood of each trajectory calculated by the trajectory evaluation unit 113, thereby generating each signal source. The number hypothesis is evaluated, and the hypothesis reliability of each signal source number hypothesis is output as an evaluation result.

  The hypothesis deletion unit 115 deletes the signal source number hypothesis determined to be deleted based on the hypothesis reliability of each signal source number hypothesis evaluated by the hypothesis evaluation unit 114 from the signal source number hypothesis memory 106. Based on the hypothesis reliability of each signal source number hypothesis evaluated by the hypothesis evaluation unit 114, the hypothesis setting unit 116 stores no suitable signal source number hypothesis and determines that a new installation is required. Based on an instruction from the operator or automatically, a new signal source number hypothesis is set in the signal source number hypothesis memory 106. The display trajectory selection unit 117 is based on the estimated value of the temporal trajectory of the signal source estimated by the trajectory estimation unit 111 and the hypothesis reliability of each signal source number hypothesis evaluated by the hypothesis evaluation unit 114. Is selected as a display locus to be displayed on a PPI (Plan-Position Indicator), for example. The display unit 118 displays the display trajectory selected by the display trajectory selection unit 117.

  The inter-hypothesis memory transfer unit 119 is for display at the display trajectory selected by the display trajectory selection unit 117, that is, the number of signal source hypotheses selected to be displayed at the current time and the previous time. And the number of the selected signal source number hypothesis, and when the display trajectory different from the previous time is selected at the current time, the past stored in the unresolvable trajectory memory 104 and the resolvable trajectory memory 105 The past tracking information in the hypothetical signal source number given a high hypothesis reliability is transferred to the new signal source hypothesis given a high hypothesis reliability this time.

Next, the operation will be described.
FIG. 5 is a flowchart showing a process flow of the sensor signal processing system according to the first embodiment of the present invention.
In step ST1, before starting the observation, first, the hypothesis setting unit 116 sets a predetermined signal source number hypothesis in the signal source number hypothesis memory 106. The hypothesis of the number of signal sources set here may be the maximum number of signal sources determined by the number of sensors, etc., and all possible signal source numbers may be used as the signal source number hypothesis, or any value may be used as the signal source number hypothesis. good.

The processing from step ST2 to step ST20 is a time loop, and shows a processing procedure for each time.
In step ST2, a time acquisition unit (not shown) of the sensor signal processing system acquires the current time. In step ST3, the reception unit 101 receives a reflected wave obtained by reflecting a transmission wave from a transmission source on a target, and the A / D conversion unit 102 digitally converts an analog reception signal received by the reception unit 101. A / D conversion to the received signal.

  In step ST4, the signal pre-processing unit 103 performs pre-processing necessary for estimating the number of signal sources from the A / D converted digital received signal. This pre-processing is necessary signal processing regardless of the number of signal sources. In the case of the MUSIC algorithm, it corresponds to correlation matrix calculation and eigenvalue analysis of a digital received signal. As an example applied to other algorithms, for example, when applying to peak detection of beat signals of FMCW (Frequency Modulated Continuous Wave) radar, the beat frequency spectrum is obtained by FFT (Fast Fourier Transform) of the beat signals. Equivalent to.

The processing from step ST5 to step ST15 is a hypothesis loop, and the observation value extracted for each signal source number hypothesis is different, so that the processing is performed for each signal source number hypothesis.
In step ST <b> 5, the observation information extraction unit 107 extracts as many observation values as the number of signal source hypotheses stored in the signal source number hypothesis memory 106 from the preprocessing result of the signal preprocessing unit 103. Here, when the MUSIC algorithm is used, a MUSIC spectrum is calculated from the results of correlation matrix calculation of the digital received signal and eigenvalue analysis, and peaks corresponding to the number of signal source hypotheses are extracted as observation values. When application to FMCW radar is assumed, peaks corresponding to the number of signal source hypotheses are extracted as observed values from the beat frequency spectrum after FFT processing.

The processing from step ST6 to step ST14 is a trajectory loop, which is performed for each trajectory.
In step ST6, when the signal source currently being tracked is determined to be decomposable at the previous time and stored in the decomposable trajectory memory 104, the process proceeds to step ST7, where it is determined that the signal source cannot be resolved at the previous time. If it is stored in the unresolvable trajectory memory 105, the process proceeds to step ST9.

  If it is determined in step ST6 that the signal source currently being tracked is decomposable at the previous time, the decomposable trajectory prediction unit 108 is stored in the signal source number hypothesis memory 106 in step ST7. For each signal source number hypothesis, the signal source currently being tracked is determined from the predicted value in the tracking information of the signal source being tracked that is predicted to be decomposable at the previous time stored in the resolvable trajectory memory 105. It is determined whether or not it can be resolved at the current time, and a predicted value of the signal source being tracked that is determined to be disassembled is output. The decomposable trajectory memory 105 stores data such as time, signal source number hypothesis number, trajectory number, and result of whether or not each miracle can be decomposed. The decomposable trajectory prediction unit 108 based on these data, It is determined whether the signal source being tracked can be resolved at the current time. In step ST8, the resolvable trajectory prediction unit 108 estimates the number of resolvable signal sources at the current time.

  FIG. 6 is a block diagram showing an internal configuration of the resolvable trajectory prediction unit 108. The decomposable trajectory prediction unit 108 includes a prediction specification difference calculation unit 201, a decomposability determination unit 202, a switch 203, an unresolvable trajectory predicted value calculation unit 204, a decomposable signal source number estimation unit 205, and an output unit 206. Yes.

  In FIG. 6, a prediction specification difference calculation unit 201 calculates a difference in prediction specifications (prediction value) between trajectories based on a prediction value at the previous time included in the tracking information stored in the decomposable trajectory memory 105. . The disassembly possibility determination unit 202 determines whether or not the trajectory can be disassembled at the current time based on the difference between the prediction items (predicted values) between the trajectories calculated by the prediction item difference calculation unit 201, and Calculate the predicted value. The switch 203 inputs the disassembly possibility determination result and the decomposable trajectory predicted value from the disassembly possibility determination unit 202, and outputs the decomposable trajectory prediction value for the decomposable trajectory group to the synthesizing unit 206. Are output to the unresolvable trajectory predicted value calculation unit 204.

  The non-decomposable trajectory predicted value calculation unit 204 uses the predicted value at the previous time included in the tracking information stored in the decomposable trajectory memory 105 for each trajectory of the non-decomposable trajectory group, and the group of the non-decomposable trajectory group at the current time. Calculate the predicted value of. The decomposable signal source number estimation unit 205 inputs the predicted value of the decomposable trajectory from the decomposability determination unit 202 and estimates the number of decomposable signal sources. The output unit 206 is capable of decomposing from the predicted value of the decomposable trajectory from the switch 203, the predicted value of the group of non-decomposable trajectory from the predictable trajectory predicted value calculation unit 204, and the decomposable signal source number estimating unit 205. The number of signal sources is collected and output as a predicted value of a signal source determined to be disassembled. Note that the predicted value of the signal source determined to be decomposable output by the output unit 206 includes the predicted value of the group of the non-decomposable trajectory, but the trajectory that is the predicted value of the group is the group. Since it is separable from other trajectories, it is included in the predicted value of the signal source determined to be decomposable.

Prediction specifications difference calculating unit 201, the following equation (1), the prediction specifications of difference dXp _ij between locus i, _j, calculates the k.
dXp _ij, k = | Xp _{i ,} k−Xp _{j ,} k | (1)
_{Here, Xp i, k, Xp j} , k is the predicted value of the locus stored in the degradable trajectory memory 105.

The decomposition possibility determination unit 202 determines whether or not decomposition is possible based on the difference dXp _ij, k of the prediction items between the trajectories i and j calculated by the prediction item difference calculation unit 201. For example, as shown in the following equation (2), when the difference dXp _ij, k between the prediction parameters between the trajectories i and j is smaller than a preset threshold value Th _div for determining whether or not decomposition is possible. There is a method to make it impossible to disassemble.
dXp _ij, k <Th _div (2)
Here, when making Th _div variable, for example, there is a method of setting it small when the S / N is relatively high, and setting it large when the S / N is extremely low.

ここで、Ｘpq,kは分解不能軌跡の個々の予測値（ｑ＝Ｇ１、Ｇ２、…、ＧＮndiv）で、Ｎndivは分解不能と判定された軌跡数で、Ｗ_q は分解不能軌跡予測値算出手段２０４が保持する重み係数である。 If the decomposability determination unit 202 determines that the decomposability is impossible, the undecomposable trajectory predicted value calculation unit 204 calculates the predicted value Xgpm, k of the nondecomposable trajectory by the following equation (3) via the switch 203. This predicted value plays the role of the predicted value as a whole group when the group of trajectories that have become unresolvable is considered as one group. It becomes the center of the correlation gate.

Here, Xpq, k is an individual predicted value (q = G1, G2,..., GNndiv) of the non-decomposable trajectory, Nndiv is the number of trajectories determined not to be decomposable, and W _q is a non-decomposable trajectory predicted value calculation means. 204 is a weighting coefficient held.

The weighting factor _Wq may be set to 1, or may be calculated so that the weight becomes larger as the variance is smaller according to the variance of the prediction error and the variance of the residuals of each trajectory. Specifically, it can be given by the following equation (4).
W _q = 1 / Ppq (4)
Here, Ppq is the variance of the prediction error or the residual variance of the trajectory q, and is stored in the resolvable trajectory memory 105.

  On the other hand, if it is determined in step ST6 of FIG. 5 that the signal source currently being tracked cannot be disassembled at the previous time, the signal source may be separated at the current time, and the process proceeds to step ST9. In step ST9, the separation signal source determination unit 109 estimates the number of signal sources that separate at the current time. In step ST10, the separation signal source determination unit 109 estimates which is the locus of the signal source that separates at the current time.

  FIG. 7 is a block diagram showing an internal configuration of the separation signal source determination unit 109. The separation signal source determination unit 109 includes an unresolvable locus predicted value calculation unit 301, a separation determination unit 302, a switch 303, a separation locus determination unit 304, a locus prediction value recalculation unit 305, and an output unit 306.

  In FIG. 7, the unresolvable locus predicted value calculation unit 301 tracks the unresolvable locus at the previous time stored in the unresolvable locus memory 104 for each signal source number hypothesis stored in the signal source number hypothesis memory 106. The predicted value included in the information is read, and the predicted value of the unresolvable trajectory at the current time is calculated. The separation determination unit 302 determines whether or not there is two or more observation values extracted by the observation information extraction unit 107 in the gate centered on the prediction value obtained by the indivisible trajectory prediction value calculation unit 301. The number of separation signal sources is estimated based on the number of observation values in the gate, and the determination result of the presence or absence of separation, the estimated number of separation signal sources, and the indecomposability at the previous time from the undecomposable locus memory 104 The predicted value of the trajectory, the predicted value of the unresolvable trajectory determined to be separated, and the predicted value of the unresolvable trajectory determined to be free of separation are output.

  The switch 303 switches between the predicted value of the unresolvable trajectory determined to be separated and the predicted value of the unresolvable trajectory determined to be not separated according to the determination result of the presence or absence of separation. The separation trajectory determination unit 304 is an observation value extracted by the observation information extraction unit 107 in the gate centering on a predicted value of the non-decomposable trajectory determined to be separated from the indecomposable trajectories determined to be separation. And the number of separation signal sources estimated by the separation determination unit 302 are determined. The predicted trajectory value recalculation unit 305 removes the separated trajectory determined by the separated trajectory determination unit 304 from the predicted value of the indissolvable trajectory determined to be separated from the switch 303, and obtains the predicted value of the incomparable trajectory. Recalculate. The output unit 306 compiles the predicted value of the unresolvable trajectory determined as having no separation from the switch 303 and the predicted value of the unresolvable trajectory calculated by the trajectory predicted value recalculation unit 305 to determine the signal source being tracked. While outputting as a predicted value, the determination result of the presence or absence of separation is output.

Next, detailed operations of step ST9 to step ST10 in FIG. 5 will be described.
First, the unresolvable trajectory predicted value calculation unit 301 calculates a predicted value Xgpm, k of the unresolvable trajectory. The predicted value Xgpm, k of the non-decomposable locus can be obtained by the above equation (3), and the smooth value Xgsm, k-1 of the unresolvable locus at the previous time included in the tracking information from the undecomposable locus memory 104 is obtained. The value extrapolated using the transition matrix Φ can be used to obtain the following equation (5).
Xgpm, k = Φ · Xgsm, k-1 (5)
Here, the transition matrix Φ is a matrix determined by a motion model that prescribes how the signal source moves (for example, at a constant speed linear motion, etc.), and is stored in the indecomposable trajectory predicted value calculation unit 301 internally. is doing.

The separation determination unit 302 checks whether or not the signal source is separated. For example, when two signal sources that could not be resolved are separated, the signal source number hypothesis of the number of signal sources 2 is as shown in FIG. 1 above, in the vicinity of the predicted value Xgpm, k of the undecomposable trajectory. Two pieces of observation information correlating with an impossible signal source and observation information correlating with a separated signal source can be obtained. That is, if there are a plurality of observed values Zi, k satisfying the following expression (6) within the specified gate size Gate _div centered on the predicted value Xgpm, k of the non-separable locus, there is a signal source that is separated. judge.
| Xgpm, k-Zi, k | <Gate _div (6)
Here, Zi, k is an observed value (i = 1, 2,..., M), and M is the number of observed values.
The separation determination unit 302 outputs the predicted value of the trajectory with separation and the predicted value of the trajectory without separation determined by the above equation (6).

  As a result of the determination by the separation determination unit 302, when it is determined that there is a signal source that is separated, the predicted trajectory value with separation and the information with separation and the number of signal sources that are separated and the previous time are determined via the switch 303. Tracking information is input to the separation locus determination unit 304. The separation trajectory determination unit 304 determines which trajectory is separated from the indissolvable trajectories using the observation value from the observation information extraction unit and each information input via the switch 303. As a specific example of the determination method, the difference between the extrapolated predicted value obtained by extrapolating the smooth value of the trajectory immediately before the time when disassembly is impossible and the current observation value is compared. There is a method to select the one closest to the observed value. The extrapolated predicted value is calculated from the smooth value of the trajectory and the transition matrix held by the separation trajectory determination unit 304.

FIG. 8 is a diagram for explaining a method for determining a separation locus by the separation locus determination unit 304. In FIG. 8, Xgpm, k is a predicted value of a group of unresolvable trajectories composed of a trajectory h, a trajectory i, and a trajectory j. Two observed values Zon1, k and Zon2, k are obtained in the gate centered at Xgpm, k. That is, the trajectory h, the trajectory i, and the trajectory j cannot be decomposed at the time kstart, and two observation values Zon1, k, Zon2, k are obtained at the current time k, so that any one of the three can be decomposed. It represents the state that became. The extrapolated predicted values Xph_ex, k, Xpi_ex, k, and Xpj_ex, k of each locus are obtained by the following equation (7).
Xptemp_ex, k = Φ (k-kstart) / Xstemp, kstart (7)
Here, Xptemp_ex, k is the extrapolated predicted value (temp = h, i, j) of the trajectory temp at the current time (k), and Xstemp, kstart is the smoothing of the trajectory temp at the time (kstart) immediately before the decomposition becomes impossible. Φ (k-kstart) is a transition matrix from time kstart to time k.

In this case, the trajectory temp selected as being separated is selected to satisfy the following equation (8). Further, the correlated observed values are Zj, k, and in FIG. 8, the locus j and the observed values Zon2, k correspond, respectively.
min (| Xptemp_ex, k−Zj, k |) (8)
That is, in FIG. 8, since the observation value closest to the predicted value Xgpm, k of the group is Zon1, k, Zon1, k is assigned as the correlated observation value to the group that cannot be decomposed, and is closest to the remaining observation value Zon2, k. Since the extrapolated predicted value Xpj_ex, k is the extrapolated predicted value of the trajectory j, the trajectory selected as being separated is determined as the trajectory j.

  When the separated trajectory j is determined, the trajectory predicted value recalculation unit 305 recalculates the predicted value of the indissolvable trajectory using the indissolvable trajectory excluding the trajectory j. A specific calculation method is the same as the calculation method of the unresolvable trajectory predicted value calculation unit 301. In addition, when the state which cannot leave | separate continues for a long time, the precision of the predicted value of each locus | trajectory may deteriorate, and it is also possible to abbreviate | omit the recalculation in the locus | trajectory estimated value recalculation part 305 in that case.

  If it is determined that there is no signal source that is separated, the predicted value of the unresolvable locus is not processed, and the predicted value of the unresolvable locus is output as it is as the predicted value of the signal source via the switch 303.

In step ST11 of FIG. 5, the correlation determining unit 110 determines the correlation between the observed value and the trajectory. For the trajectory that is separated this time, since the correlation has already been determined in steps ST9 and ST10, no work is actually required. As a specific determination method, for example, there is a method in which the observation value Zj, k from the observation information extraction unit 107 closest to the prediction value Xpi, k from the resolvable trajectory prediction unit 108 is assigned on the assumption that it correlates with the trajectory. . In this case, it is considered that the observed value Zj, k satisfying the following expression (9) correlates with the locus i.
min (| Xpi, k−Zj, k |) (9)
Here, if Xpi, k is a predicted value of a trajectory that cannot be resolved, the same observed value Zj, k is assigned to all trajectories that cannot be resolved. Note that the observation values to be assigned may be those obtained by the apparent signal source number hypothesis or those obtained by the true signal source number hypothesis.

Note that when the MUSIC algorithm is used as an algorithm, an erroneous observation value may be generated in an incorrect hypothesis with a small number of signal sources, but a false image corresponding to the number of erroneous signal sources in an erroneous hypothesis of many signal sources. However, since correct observations can be obtained, there is no problem even if the observations obtained by the true signal source number hypothesis are used as long as the correlation can be correctly determined and false images can be eliminated. However, when using the observation values obtained by the true signal source number hypothesis, if the same observation value is assigned to a plurality of trajectories, the observation value remains. Since this corresponds to a false image, after assigning correlation observation values to all the trajectories, a new trajectory is not created from the surplus observation values exceeding the number of hypothesis signal sources, and is discarded.
Note that if there are extra observation values because the number of hypothetical target trajectories has not been created, the observation values are not discarded because they are new signal sources.

Steps ST12 to ST14 are processes performed by the trajectory estimation unit 111.
In step 12, the trajectory estimation unit 111 performs signal source disappearance determination. In step ST11, for a trajectory in which a state in which no correlation is obtained continues for a certain period, it is considered that the signal source has disappeared, and the past information of the trajectory disappeared at the next time is not used. As an example of the disappearance determination method, for example, the following equation (10) is used to count the number of times TQ of the observed value detected within a certain time, and when TQi, k falls below the threshold value Th _TQ , it is considered that the disappearance has occurred. is there.
At detection: TQi, k = TQi, k-1 + tq_dtct
When not detected: TQi, k = TQi, k-1 + tq_nondtct (10)
Here, the number of detections TQi, k-1 of the observed value at the previous time is included in the tracking information determined to be decomposable stored in the decomposable locus memory 105, and tq_dtct is the TQi at the time of detecting the observed value The amount of increase, tq_nondtct is the amount of increase in TQi when the observed value is not detected (with a difference such as a negative value when not detected).

In step ST13, the trajectory estimation unit 111 estimates the temporal trajectory of the signal source and outputs an estimated value Xsi, k. When a Kalman filter or α-β filter is used as the estimation method, the estimated value Xsi, k is given by the following equation (11).
Xsi, k = Xpi, k + Gi, k (Zi, k-H.Xpi, k) (11)
Here, the predicted value Xpi, k and the observed value Zj, k of the trajectory are given from the correlation determination unit 110, H is an observation matrix, Gi, k is a gain matrix, and the trajectory estimation unit 111 holds both inside. ing.
The smooth value of the non-decomposable trajectory can be obtained by the equation (11) using the individual predicted values Xpi, k and the observed values (all the same) for the individual trajectories constituting the non-decomposable trajectory. Alternatively, the trajectories constituting the indecomposable trajectory are all considered to be the same smooth value, and may be obtained by the equation (11) using the predicted value Xgpm, k of the indecomposable trajectory.
It should be noted that in step ST11, since the number of hypotheses for the hypothetical target number has not been created, the observation value (new signal source) is left in the case where the observation value remains, and in this step 13, the locus estimation is performed. The unit 111 creates a new locus.

In step ST14, the trajectory prediction unit 112 predicts the trajectory at the next observation k + 1 based on the estimated value Xsi, k output from the trajectory estimation unit 111 and the transition matrix Φ held therein, and the predicted value of the signal source Xpi, k + 1 is calculated. The prediction formula is given by the following formula (12).
Xpi, k + 1 = Φ · Xsik (12)
As with the smooth value, the predicted value of the non-decomposable trajectory may be obtained by the equation (12) using the individual smoothed values (estimated values) Xsi, k. These may all be considered to be the same smooth value, and may be obtained from the equation (12) using the smooth value Xgsm, k of the non-decomposable locus.
In addition, the prediction can be performed by sampling set in advance prior to the next observation as described above, or after obtaining the next observation value, the sampling interval that is the difference between the current observation time and the next observation time is set. It can also be done after calculating.

  In step ST15, the trajectory evaluation unit 113 calculates and outputs the likelihood f of the trajectory using a predetermined evaluation function using the tracking result. For example, in Patent Document 1, assuming that the likelihood distribution of an observed value is a normal distribution centered on a predicted value, the value of the likelihood f of the locus is larger as the locus is obtained closer to the predicted value. It is trying to become. In addition, as a method of defining the likelihood f of the trajectory, it is possible that the trajectory matches the motion model as the number of detections increases, and TQ itself given by the above equation (10) can be used as the likelihood f of the trajectory.

ここで、ｋｓｔａｒｔは検出確率Ｐdo の計算開始カウント（任意に設定可）で、ＮはＰdo の計算対象カウント数である。信号源の検出確率Ｐd は固定でも良いし、Ｓ／Ｎが高いほど高く、低い場合は低く設定するようにする等、可変としても良い。また、Ｄtct_k は観測値検出時は１、観測値非検出時は０となる変数であり、例えば、ある目標を１秒に１回、５秒間観測し、５回の観測で全て検出した場合は５／５＝１で、３回検出した場合は３／５＝０．６となる。 Further, the detection probability Pd of the signal source (the probability that the observation value is in the correlation gate) Pd is defined in advance, and the likelihood f of the trajectory increases as the detection probability Pdo of the actual observation value is closer to Pd. It can also be. In this case, the likelihood f of the trajectory is given by the following equation (13), for example.
f = Pd / | Pd-Pdo | (13)
Here, the actual observation value detection probability Pdo is obtained by the following equation (14) using the presence or absence of detection, that is, the result of whether or not a correlated observation value is detected among the tracking results.

Here, kstart is a calculation start count (can be arbitrarily set) of the detection probability Pdo, and N is a calculation target count number of Pdo. The detection probability Pd of the signal source may be fixed, or may be variable such that it is set higher as the S / N is higher and is set lower when the S / N is lower. Dtct _k is a variable that is 1 when the observed value is detected and 0 when the observed value is not detected. For example, when a certain target is observed once per second for 5 seconds and all are detected in 5 observations. Is 5/5 = 1, and 3/5 = 0.6 when detected three times.

ここで、ｍは仮説番号（１〜Ｍ）で、Ｍは全仮説数で、Ｎmは信号源数仮説ｍの仮説信号源数で、ｎは軌跡番号で、Ｎtrk_m,k は信号源数仮説ｍの軌跡数である。 The processing after step ST16 is processing after the tracking processing of the total number of signal sources hypothesis is completed.
In step ST16, the hypothesis evaluation unit 114 calculates the hypothesis reliability Rel _{m, k} based on the calculation result of the likelihood f of the trajectory. Basically, all of the loci in the hypothesis that have a high likelihood f are considered to be signal source hypotheses with high reliability, and high hypothesis reliability is given. As a specific example, there is a method of defining by the average value of the likelihoods f of all the trajectories in the signal source number hypothesis, as shown in the following equation (15).

Here, m is a hypothesis number (1 to M), M is the number of all hypotheses, Nm is the number of hypothesis signal sources of the signal source number hypothesis m, n is a trajectory number, and Ntrk _{m, k} is a signal source number hypothesis. The number of trajectories of m.

  Alternatively, a threshold may be provided for the likelihood f of the trajectory, and only the likelihood f of the trajectory greater than or equal to the threshold may be an evaluation target in the equation (15). For example, when the detection count TQ is used as the likelihood f of the trajectory, a trajectory immediately after the start of tracking, such as a trajectory with the detection count TQ of a certain value or less, is not highly reliable, and should be excluded from the evaluation target. Because there is also.

  In step ST17, the display trajectory selection unit 117 selects a signal source number hypothesis that seems to be correct for the purpose of displaying the trajectory on the radar PPI. The hypothesis reliability hypothesis reliability may be selected simply by selecting the hypothesis reliability. However, a threshold is set for the hypothesis reliability, and the number of signal sources is the highest among the signal hypotheses with hypothesis reliability equal to or greater than the threshold. It is also possible to select a signal source hypothesis with a large number of signals. In general, the larger the number of signal sources, the more difficult it is to track all the signal sources. Therefore, the purpose of preventing the hypothesis that the number of signal sources with a large number of signal sources is unfairly evaluated is low. is there. Here, the number of signal sources in the hypothesis of the number of signal sources selected for display corresponds to the number of true signal sources.

ここで、mv_aveＲel_m,k は再計算した仮説信頼度で、ｋstartは開始時刻で、Ｎはｋstartから現在時刻までのカウント数である。 It is also possible to select a hypothesis at the current time in consideration of past hypothesis reliability selection results. Specifically, the hypothesis reliability is recalculated by the following equation (16) by moving average processing, and the one with the highest hypothesis reliability or the one with the number of signal sources greater than the threshold is selected.

Here, mv_aveRel _{m, k} is the recalculated hypothesis reliability, kstart is the start time, and N is the count number from kstart to the current time.

  In step ST18, the hypothesis deleting unit 115 deletes the signal source number hypothesis from which the hypothesis reliability is remarkably lowered from the signal source number hypothesis memory 106. The deletion may be determined based on a preset hypothesis reliability threshold.

  In step ST19, when it is determined that the signal source number hypothesis having a high hypothesis reliability is not in the signal source number hypothesis memory 106, the hypothesis setting unit 116 automatically selects a new signal source based on an instruction from the operator. The number hypothesis is set in the signal source number hypothesis memory 106. The case where it is determined that a hypothesis reliability with a high hypothesis reliability is not found in the signal source number hypothesis memory 106 is that the hypothesis reliability of all the signal source hypotheses is almost the same, or the number of signal sources with high hypothesis reliability. Even if a hypothesis exists, it is assumed that the signal source cannot be correctly tracked and maintained, for example, when the number of times of detection of the trajectory is low.

  In step ST20, the inter-hypothesis memory transfer unit 119 displays the display trajectory selected by the display trajectory selection unit 117, that is, the number of signal source hypotheses selected to be displayed at the current time and the previous time. The number of signal source hypotheses selected for display is compared, and if a display trajectory different from the previous time is selected at the current time, stored in the unresolvable trajectory memory 104 and the resolvable trajectory memory 105 The past tracking information in the signal source number hypothesis given a high hypothesis reliability in the past is transferred to the signal source number hypothesis newly given a high reliability this time.

  For example, consider a case where the number of signal sources is changed from 3 to 2 because the number of signal sources is 2 and the number of signal sources is 2 from the state of 3 signal sources. . At this time, the past tracking result of the trajectory that has not disappeared among the tracking trajectories in the number of signal sources 3 is used in place of the past tracking result in the signal source number hypothesis of the number 2 of signal sources. Then, since the past accurate tracking locus in the signal source number hypothesis of the signal source number 3 can be used also in the signal source number hypothesis of the signal source number 2, the hypothesis reliability of the signal source number hypothesis of the signal source number 2 is increased. As a result, the correct signal source hypothesis can be estimated quickly. In this way, the number of signal sources with high hypothesis reliability is improved because the data is effectively used and tracking performance is improved by using the past information of the signal source hypotheses with high hypothesis reliability between the signal source hypotheses. It becomes possible to quickly determine the hypothesis.

  According to the first embodiment, the signal source is determined based on the tracking information of the signal source that is being tracked that the resolvable trajectory prediction unit 108 is predicted to be resolvable at the previous time stored in the resolvable trajectory memory 105. For each of several hypotheses, it is determined whether the signal source being tracked can be resolved at the current time, and the separation signal source determination unit 109 is predicted to be unresolvable at the previous time stored in the unresolvable trajectory memory 104. From the tracking information of the signal source being tracked and the extracted observation value, it is determined whether or not the signal source being tracked is separated for each signal source hypothesis, and the correlation determining unit 110 determines the resolvable trajectory predicting unit 108. And by determining the correlation between the signal source being tracked and the observation value extracted by the observation information extraction unit based on the determination result of the separation signal source determination unit 109, the number of signal sources can be correctly estimated, and the target Detection performance There is an advantage that it is possible to improve the fine tracking maintenance performance.

  In addition, according to the first embodiment, the inter-hypothesis memory transfer unit 119 stores the number of hypotheses in the signal source hypothesis that has been given a high hypothesis reliability in the past stored in the unresolvable locus memory 104 and the decomposable locus memory 105. By transferring the past tracking information to the new signal source hypothesis that has been given a high level of hypothesis reliability this time, tracking information can be used effectively and changes in the number of signal sources can be dealt with quickly. The effect that it can be obtained.

It is a figure which shows the example of a time-sequential change of a MUSIC spectrum in the case of the approach / separation of a signal source. It is a figure which shows the example of a time-sequential change of a MUSIC spectrum in the case of isolation | separation of a signal source. It is a figure which shows the example of a time-sequential change of a MUSIC spectrum in the case of the loss | disappearance of a signal source. It is a block diagram which shows the structure of the sensor signal processing system by Embodiment 1 of this invention. It is a flowchart which shows the flow of a process of the sensor signal processing system by Embodiment 1 of this invention. It is a block diagram which shows the internal structure of the decomposable locus | trajectory estimation part of the sensor signal processing system by Embodiment 1 of this invention. It is a block diagram which shows the internal structure of the separation signal source determination part of the sensor signal processing system by Embodiment 1 of this invention. It is a figure explaining the separation trace determination method by the separation trace determination part of the separation signal source determination part of the sensor signal processing system by Embodiment 1 of this invention.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 101 Reception part, 102 A / D conversion part, 103 Signal pre-processing part, 104 Decomposable locus memory, 105 Decomposable locus memory, 106 Signal source number hypothesis memory, 107 Observation information extraction part, 108 Decomposable signal source estimation part, 109 separation signal source determination unit, 110 correlation determination unit, 111 locus estimation unit, 112 locus prediction unit, 113 locus evaluation unit, 114 hypothesis evaluation unit, 115 hypothesis deletion unit, 116 hypothesis setting unit, 117 inter-hypothesis memory transfer unit, 118 Display trajectory selection unit, 119 display unit, 201 prediction specification difference calculation unit, 202 decomposability determination unit, 203 switch, 204 unresolvable trajectory predicted value calculation unit, 205 decomposable signal source number estimation unit, 206 output unit, 301 Unresolvable locus predicted value calculation unit, 302 separation determination unit, 303 switch, 304 separation locus determination unit, 305 Trajectory predicted value recalculation unit, 306 output unit.

Claims (7)

In a sensor signal processing system that uses a signal processing algorithm that performs measurement processing of observation information after estimating the number of signal sources,
Whether the signal source being tracked can be resolved at the current time for each signal source number hypothesis from the tracking information of the signal source being tracked that is predicted to be decomposable at the previous time stored in the resolvable trajectory memory A resolvable trajectory prediction unit for determining
The signal source being tracked for each hypothesis of the number of signal sources from the tracking information of the signal source being tracked that was predicted to be unresolvable at the previous time stored in the unresolvable trajectory memory and the observed value extracted at the current time A separation signal source determination unit that determines whether or not the
A correlation determining unit that determines a correlation between the signal source being tracked and the observed value extracted at the current time based on the determination results of the resolvable trajectory predicting unit and the separation signal source determining unit; Sensor signal processing system. The decomposable trajectory predicting unit determines that the trajectory is not resolvable when a difference in predicted values between trajectories in the tracking information of the signal source being tracked stored in the decomposable trajectory memory is smaller than a predetermined threshold. Group multiple trajectories that have been determined to be unresolvable,
2. The correlation determination unit determines a correlation by assigning the same observation value extracted at a current time to a plurality of trajectories grouped by the resolvable trajectory prediction unit. The described sensor signal processing system.   The separation signal source determination unit is incapable of resolving at the current time from the predicted value included in the tracking information of the signal source being tracked that is predicted to be unresolvable at the previous time stored in the unresolvable trajectory memory. Calculates the predicted value of the signal source being tracked that is predicted to exist, and if there are multiple observation values extracted at the current time within the predetermined range from the calculated predicted value of the current time, the signal source being tracked The sensor signal processing system according to claim 1, wherein the sensor signal processing system is determined to be separated.   The separation signal source determination unit is performing tracking that is predicted to be unresolvable at the current time among the predicted values calculated by extrapolating the state quantity of each locus immediately before determined to be unresolvable to the current time. 4. The sensor according to claim 3, wherein the trajectory of the predicted value closest to the observed value other than the observed value extracted at the current time correlated with the predicted value of the signal source is determined to be a separated trajectory. Signal processing system. Based on the correlation between the signal source being tracked determined by the correlation determination unit and the observed value extracted at the current time, the detection probability of the actual observed value is obtained, and the likelihood of each trajectory is calculated using a predetermined evaluation function A trajectory evaluation unit to perform,
The sensor signal processing system according to claim 1, further comprising: a hypothesis evaluation unit that calculates a hypothesis reliability of the signal source number hypothesis based on the likelihood of each locus calculated by the locus evaluation unit.   6. The sensor according to claim 5, wherein the trajectory evaluation unit uses an evaluation function that increases the likelihood of each trajectory as the detection probability of the actual observation value is closer to the detection probability of the signal source defined in advance. Signal processing system.   The hypothesis evaluation unit obtains tracking information of the signal source hypothesis that has been given a high hypothesis reliability in the past by the hypothesis evaluation unit and the tracking information of the past signal sources stored in the decomposable locus memory. 6. The sensor signal processing system according to claim 5, further comprising an inter-hypothesis memory transfer unit for transferring to a signal source number hypothesis newly given a high hypothesis reliability.

Images