A multipath identification method based on Bayesian reasoning

Abstract

The present invention belongs to the technical field of multipath identification, and relates to a multipath identification method based on Bayesian reasoning. During the flight of a reconnaissance aircraft, a passive receiving reconnaissance mission is performed on a stationary radiation source. There is strong multipath reflection wave interference in the VHF channel, which is difficult to distinguish in the time and frequency domain. The arrival angle information of the direct wave signal and the multipath reflection signal is obtained through direction finding technology, and the multipath signal is regarded as a false radiation source target to track multiple targets together with the real target to extract its course information. The difference in course characteristics between the real target and the multipath target extraction is studied through the VHF channel characteristics, a likelihood model is constructed, and the Bayesian reasoning algorithm is used to calculate the posterior probability at each moment, thereby achieving real-time identification and inference effects, and as the amount of data increases, the inference result will be more accurate, achieving the purpose of eliminating the multipath interference trajectory. The present invention can effectively identify the course corresponding to the real target and the multipath false target, and the algorithm effect is good and the robustness is strong.

Description

Technical Field

[0001] The present invention belongs to the technical field of multipath identification, and specifically relates to a multipath identification method for detecting a radiation source in a VHF channel with a strong multipath effect, using the characteristics of the arrival angle variation of a direct wave and a multipath reflected wave within a period of time. Background Art

[0002] During the flight of the reconnaissance aircraft, passive receiving reconnaissance missions are carried out on stationary radiation sources. There is strong multipath reflection wave interference in the VHF channel, which is difficult to distinguish in the time and frequency domain. However, the arrival angle information of the direct wave signal and the multipath reflection signal can be obtained through direction finding technology. The multipath signal is regarded as a false radiation source target and multi-target tracking is performed together with the real target to extract its process information. The difference in process characteristics between the real target and the multipath target is studied through the characteristics of the VHF channel. The identification criteria and algorithms are studied to achieve the purpose of eliminating the multipath interference trajectory.

[0003] In traditional multipath identification algorithms, the arrival angle information of the course at multiple times is taken, and then the speed characteristics or trajectory correlation characteristics are calculated, and then the courses corresponding to the multipath false targets and the courses corresponding to the real targets are identified based on the characteristics. Traditional identification algorithms do not make full use of data and have strong randomness, and are prone to misjudgment when the measurement noise is large. Summary of the invention

[0004] For the above scenario, in order to better smooth out the randomness at a certain moment or a certain period of time and make full use of the extracted feature data at each moment, the present invention uses a probability model to describe the feature data at the current moment, constructs a likelihood model through the arrival angle characteristics in the VHF channel, and uses the Bayesian reasoning algorithm to calculate the posterior probability at each moment, thereby achieving real-time identification and inference effects, and as the amount of data increases, the inference results will be more accurate.

[0005] The technical solution adopted by the present invention is:

[0006] (1) Analysis of the arrival angle characteristics of direct waves and multipath reflected waves in VHF channels.

[0007] like Figure 1 As shown in the figure, the radiation source is stationary and the reconnaissance aircraft is in a horizontal flight motion scene, where A represents the meter-wave radar radiation source, T represents the position of the reconnaissance receiver at the kth moment, T' represents the position of the reconnaissance receiver at the k+1th moment, v represents the horizontal flight speed of the reconnaissance aircraft, and there is a direct wave signal at each moment, θ d (k) represents the arrival angle of the direct wave signal at the kth moment, It represents the arrival angle of the jth multipath signal at the kth moment. The first Fresnel reflection area will change with the movement of the reconnaissance aircraft, and the multipath signal reflection point at each moment is not necessarily the same. Its reflection point obeys a uniform distribution in the first Fresnel reflection area at that moment. The number of multipath reflection signals is also a random variable that obeys a certain probability distribution model, indicating that the number of multipath signals may change during the movement, that is, the multipath reflection signals may be born and die.

[0008] It can be known that when the multipath reflected signal is in the process of the reconnaissance aircraft moving, since the meter-wave radar radiation source is in a static state, the angle change rate of the direct wave's arrival angle measured in a short period of time is not much different and tends to be stable. However, the multipath reflected signal is in the first Fresnel reflection zone. Due to the complexity of the terrain, its reflection point will change accordingly and the position is relatively random, so the arrival angle change rate of the reflected signal at this time fluctuates greatly. This characteristic is called the volatility characteristic of the signal arrival angle in the VHF channel, which will be used as an important characteristic for identifying direct waves and multipath reflected waves.

[0009] The reconnaissance aircraft can always detect the direct wave signal during the movement, but due to the generation and disappearance of multipath reflection signals, the continuous detection time of each multipath is short. It can be considered that as long as the continuous detection time is long, the continuous detection time corresponding to the direct wave signal is longer than that of the multipath reflection signal. This characteristic is called the continuity characteristic of the signal arrival angle in the VHF channel, which will also be an important characteristic for identifying direct waves and multipath reflection waves.

[0010] (2) Construction of multipath identification likelihood model

[0011] It is known that the reconnaissance aircraft measured the radiation source and the arrival angle data stream of the multipath reflection wave at 1 to k moments, and the data stream was subjected to track extraction and multi-target tracking filtering to obtain the course data. It represents the arrival angle filtering result of the ith track in the time period from 1 to k, and defines the arrival angle filtering result of the ith track at the kth time as

[0012] The intermediate parameter expression defined at the kth moment of the ith track is as follows:

[0013]

[0014] At this time, the continuity index of the i-th track at the k-th moment is:

[0015]

[0016] Assume that the upper threshold of continuity is set to η cmax , the lower threshold is set to η cmin The continuity probability can be set as

[0017]

[0018] For each track, the volatility can be characterized by the variance of the arrival angle change rate of each track. For the i-th track, its volatility index b is defined as i for:

[0019]

[0020] in, represents the average rate of change of the arrival angle of the ith track at time k, that is,

[0021]

[0022] Assume that the upper threshold of volatility is set to η bmax , the lower threshold is set to η bmin , usually η bmin =b min σ,η bmax =b max σ, where σ is the standard deviation of the angle measurement, b max ,b min is a constant set according to the system. The volatility probability can be set to

[0023]

[0024] Define l1(z1|H) as the likelihood function of volatility, and l2(z2|H) as the likelihood function of continuity, which can be calculated based on the volatility probability and continuity probability respectively.

[0025]

[0026]

[0027] The assumption that the path is a direct wave is H0, and the assumption that the path is a multipath is H1. Here, f1 is the volatility likelihood function under the two assumptions of volatility probability, and here f2 is the continuity likelihood function under the two assumptions of volatility probability, which represents the continuity probability. and volatility probability The upper and lower thresholds of the calculation need to be set accordingly based on assumptions H0 and H1. as well as

[0028] (3) Bayesian reasoning

[0029] According to the characteristic differences between direct waves and multipath signals under typical channels, combined with the multi-source identification likelihood model, multi-source identification decision-making is completed, and judgment can be made through the framework of Bayesian reasoning.

[0030]

[0031]

[0032] Here P(H) is the prior probability. The initial prior probability needs to be set in advance. Generally, the initial prior probability is set to p(H0) = 0.5, p(H1) = 0.5. The prior probability at the subsequent moment is expressed by the posterior probability p(H|Z) of the previous moment. Z = {z1, z2} represents statistical information, and p(Z|H) is the likelihood probability.

[0033] Since for each path, its continuity and volatility properties have different degrees of determination for whether the path is a direct signal or a multipath signal, in order to better describe the likelihood function, the weight parameter w will be introduced to achieve it. The likelihood probability calculation can be rewritten as

[0034] p(Z|H)=w1l1(z1|H)+w2l2(z2|H) (11)

[0035] Where w1 is the weighted coefficient of volatility, w2 is the weighted coefficient of continuity, and satisfies

[0036] w1+w2=1 (12)

[0037] By introducing the characteristic data of each moment, the likelihood probability is calculated by the multipath identification likelihood model and formula (11), and the likelihood probability is substituted into formula (9) to calculate the posterior probability. The prior probability P(H) of the initial moment is set by itself, and the prior probability of the subsequent moment is expressed by the posterior probability of the previous moment. In this way, the Bayesian reasoning iteration calculation is completed, and the posterior probability of the last moment corresponding to the process data is calculated. Then, according to the numerical value of the posterior probability, it is judged whether the process is a multipath signal. If the probability is greater than a certain threshold, the setting of the threshold is determined according to the identification accuracy requirements, and is generally set to 0.9, that is, it can be identified whether the process corresponds to a multipath signal. If the threshold is not reached, it can be selected to continue to detect the arrival angle data for Bayesian reasoning iteration until the threshold is reached to obtain the identification result. The identification result of this method is accurate and the effect is good. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A schematic diagram of the scene;

[0039] Figure 2 Single radiation source tracking filtering results;

[0040] Figure 3 Continuity and volatility index curves of the single radiation source tracking process; (a) continuity index curve; (b) waveform index curve

[0041] Figure 4 Posterior probability curves of Bayesian inference for single radiation source tracking; (a) posterior probability curve of process a; (b) posterior probability curve of process b; (c) posterior probability curve of process c

[0042] Figure 5 Filtering results for false tracking of single radiation source tracking;

[0043] Figure 6 Continuity and volatility index curves of the process of mistracking in single radiation source tracking; (a) continuity index curve; (b) waveform index curve

[0044] Figure 7 Bayesian reasoning posterior probability curves for single-radiator tracking with mistracking; (a) posterior probability curve for process a; (b) posterior probability curve for process b; (c) posterior probability curve for process c DETAILED DESCRIPTION

[0045] The practicability of the present invention is demonstrated below in conjunction with a simulation example:

[0046] Assuming that it is used in a single radiation source tracking scenario, assuming that the horizontal distance between the reconnaissance aircraft and the radiation source is relatively close, the center frequency of the radiation source is 100MHz, the height of the radiation source is 300 meters, and it is stationary, the reconnaissance aircraft flies at a horizontal speed of 50m / s, and flies horizontally toward the meter-wave radar. The reconnaissance aircraft flies from 4000 meters to 1500 meters from the radiation source. Assuming that the multipath signal is evenly distributed in the first Fresnel zone, and there is a multipath between 1 and 20 seconds, and then disappears, and a multipath signal is re-established at 30 seconds. The direct signal always exists. The multi-target tracking filtering results are as follows: Figure 2 As shown, process a corresponds to the direct signal, and processes b and c correspond to multipath signals.

[0047] Extract the process data from the filtering results, and calculate the continuity and volatility index curves of each process at each time point through formulas (1) and (3): Figure 3 shown.

[0048] According to the continuity and volatility indicators of the above different processes, and set η cmin =0, η cmax is 1, and η is set bmin is 0.1, η bmax=0.6, and the likelihood probability of each indicator is calculated by formula (2) and formula (5), and the volatility weight w1 is set to 0.5, the continuity weight w2 is set to 0.5, and the likelihood probability of Bayesian reasoning is calculated by formula (10). The simulation uses the identification method of Bayesian reasoning, and sets the prior probability p(H0) of the direct signal corresponding to each process to 0.5, and the prior probability p(H1) of the multipath signal to 0.5. The Bayesian reasoning posterior probability curve obtained at each time for each process is as follows: Figure 4 shown.

[0049] The posterior probability shows at the last moment that the probability that process a is a direct signal is close to 1, the probability that process b is a multipath signal is close to 1, and the probability that process c is a multipath signal is close to 1. The identification result is that process a corresponds to a direct signal, and processes b and c correspond to multipath signals. The identification results of each process are correct.

[0050] Mistracking occurs in single-radiator tracking scenarios: Mistracking may occur for a target, i.e., the filtering result is as follows: Figure 5 As shown in the figure, process a initially corresponds to the real target, process b initially corresponds to the multipath signal, at this time, the mistracking phenomenon occurs, process b corresponds to the real target at the subsequent time, process a corresponds to the multipath at the subsequent time, and terminates at time 20, and process c always corresponds to the multipath signal. At this time, the continuity and volatility index curves of each process at each time point are calculated as follows Figure 6 shown.

[0051] At this time, the posterior probability curve of Bayesian inference is as follows Figure 7 As shown in the figure, the posterior probability shows that at the last moment, the probability that process a is a multipath signal is close to 1, the probability that process b is a direct signal is close to 1, and the probability that process c is a multipath signal is close to 1. The identification result is that process b corresponds to a direct signal, and processes a and c correspond to multipath signals. Even if there is a false tracking phenomenon, the identification result of each process is correct.

[0052] Identification effect: The simulation experiment verifies that the multipath identification method based on Bayesian reasoning has correct identification results for the process in the single radiation source reconnaissance scenario and the scenario of mis-following, which verifies the effectiveness and robustness of the method in the present invention.

Claims

1. A multipath identification method based on Bayesian reasoning is used for the scene of a moving reconnaissance aircraft detecting a stationary radiation source. It defines a data stream of the radiation source and the arrival angle of the multipath reflection wave measured by the reconnaissance aircraft at 1 to k moments, and uses the data stream to extract the track and filter the multi-target tracking to obtain the course data. It represents the arrival angle filtering result of the ith track in the time period from 1 to k, and defines the arrival angle filtering result of the ith track at the kth time as It is characterized in that The multipath identification method comprises: The intermediate parameter expression defined at the kth moment of the ith track is as follows: The continuity index of the ith track at the kth moment is: Set the upper threshold of the continuity index to η cmax , the lower threshold is set to η cmin , setting the continuity probability to: For each track, the volatility is characterized by the variance of the arrival angle change rate of each track. For the i-th track, its volatility index b is defined as i for: in, represents the average rate of change of the arrival angle of the ith track at time k, that is, Set the upper threshold of volatility to η bmax , the lower threshold is set to η bmin , η bmin =b min σ,η bmax =b max σ, where σ is the standard deviation of the angle measurement, b max ,b min Constants set for the system; set the volatility probability to: Define l1(z1|H j ) is the likelihood function of volatility, l2(z2|H j ) is the likelihood function of continuity, which is calculated based on the volatility probability and continuity probability respectively: The path is defined as H0 for direct waves and H1 for multipath, where f1 is the volatility likelihood function under the two assumptions of volatility probability, and f2 is the continuity likelihood function under the two assumptions of continuity probability, which represents the continuity probability. and volatility probability The upper and lower thresholds of the calculation need to be set accordingly based on the assumptions H0 and H1; define as well as Decision making through the framework of Bayesian reasoning: Where P(H) is the prior probability. The initial prior probability needs to be set in advance. The prior probability at the subsequent moment is represented by the posterior probability p(H|Z) at the previous moment. Z = {z1, z2} represents statistical information, and p(Z|H) is the likelihood probability. Introducing the weight parameter w rewrites the likelihood probability calculation as: p(Z|H)=w1l1(z1|H j )+w2l2(z2|H j ) (11) Where w1 is the weighted coefficient of volatility, w2 is the weighted coefficient of continuity, and satisfies w1+w2=1 (12) Using the history data at each moment, the likelihood probability is calculated through formula (7), formula (8) and formula (11), and the likelihood probability is substituted into formula (9) to calculate the posterior probability. The prior probability P(H) at the initial moment is set by itself, and the prior probability at the subsequent moment is expressed by the posterior probability at the previous moment. In this way, the Bayesian reasoning iteration calculation is completed, and the posterior probability at the last moment corresponding to the history data is calculated. Then, according to the numerical value of the posterior probability, it is judged whether the history data is a multipath signal. If the probability is greater than the set threshold, the history data corresponds to a multipath signal. If the threshold is not reached, the arrival angle data is continuously detected for Bayesian reasoning iteration until the threshold is reached to obtain the identification result.

Images