阅读说明：本技术 一种基于自适应滤波器的直达波对消方法 (Direct wave cancellation method based on adaptive filter ) 是由 刘金龙 吴雨欣 金海艳 宋晓凯 吴芝路 于 2020-07-01 设计创作，主要内容包括：本发明提出一种基于自适应滤波器的直达波对消方法,所述方法对参考信号进行抽头,并假设一组抽头系数,此时就得到了自适应滤波器的起始状态,然后令信号通过该滤波器,利用滤波后的信号与监视信号进行比对得到一个误差,并根据该误差调整这一组抽头系数进行迭代,误差会随着迭代次数的增加而减小。利用本发明所述方法实现了直达波对消的良好效果。(The invention provides a direct wave cancellation method based on an adaptive filter, which taps a reference signal, assumes a group of tap coefficients, obtains the initial state of the adaptive filter at the moment, then leads the signal to pass through the filter, compares the filtered signal with a monitoring signal to obtain an error, adjusts the group of tap coefficients according to the error to carry out iteration, and reduces the error along with the increase of the iteration times. The method of the invention realizes the good effect of direct wave cancellation.)

1. A direct wave cancellation method based on an adaptive filter is characterized in that: utilizing the self-adaptive filter to perform direct wave cancellation on an echo channel to obtain an echo signal, and performing parameter modification by calculating a difference value between an actual output result and an expected output result of the self-adaptive filter to finally obtain expected output; the adaptive filter needs two signals, namely an input signal x (n) and an expected signal d (n), generates an output signal y (n) after the input signal passes through the adaptive filter, compares the y (n) with the expected signal d (n) to obtain an error signal e (n), adjusts a tap coefficient according to an adaptive filtering algorithm to enable the e (n) to be minimum under a minimum mean square error criterion, and enables the output signal y (n) to approach the expected signal d (n) as much as possible under the minimum mean square error criterion, and the method specifically comprises the following steps:

the input signal is first represented by phasors, i.e., x (n) ═ x (n), x (n-1)]^TThe tap coefficients are then expressed in phasors, i.e. w (k) ═ w₀(k),w₁(k),...,w_L-1(k)]^TWhere L is the order of the adaptive filter, w (k) represents the set of tap coefficients for the kth iteration of the adaptive filter, thereby yielding an output signal:

y_k(n)＝W^T(k)X(n)

wherein y is_k(n) representing the output signal of the kth iteration, the estimated error signal being obtained by comparing the desired signal with the output signal:

e(n)＝d(n)-W^T(k)X(n)

the mean square error is thus obtained as:

J＝E[W^T(n)]＝E[d²(n)]-2E[d(n)W^T(n)X(n)]+E[W^T(n)X(n)X^T(n)W(n)]

when the tap coefficient is adjusted, in order to adjust the coefficient most effectively in each iteration, a steepest descent algorithm is adopted, that is, the adaptive filter weighting vector is adjusted along the negative gradient direction, so that the formula of the iterative tap coefficient is as follows:

the essence of iteration is that each adjustment is carried out to make the adjustment approach to an optimal state, the degree of each adjustment is determined by the adjustment step length, and in the formula of the iterative tap coefficient, mu is the adjustment length of each adjustment tap coefficient, namely the adjustment step length; the steepest descent algorithm based on the minimum mean square error criterion chooses to differentiate each tap vector by using the mean of the squares of the errors in the instantaneous time, and the mean square error gradient can be obtained as follows:

therefore, the iterative equation of the tap coefficient of the steepest descent algorithm based on the minimum mean square error criterion is simplified into the following equation:

W(k+1)＝W(k)+2μe(n)X(n)

then, in the subsequent operation, multiple iterations are performed according to the formula of the iterative tap coefficient, after k iterations, when the distance between the result after the k iteration and the tap coefficient vector of the (k-1) th iteration is smaller than a certain threshold, the iteration is ended, a final tap coefficient vector w (k) can be obtained, and then the final approximation signal of the expected signal can be obtained:

y_k(n)＝W^T(k)X(n)。

2. the method of claim 1, wherein: in the method, direct wave cancellation is required to be carried out on an echo channel by using the direct wave recovered from the reference channel.

3. The method of claim 2, wherein: the adaptive filter is a discrete time signal adaptive filter.

4. The method of claim 3, wherein: the number of iterations is greater than or equal to the maximum delay of the multipath effect and less than the minimum delay of the echo signal.

5. The method of claim 4, wherein: the value of the adjustment step size mu is specifically as follows:

Technical Field

The invention belongs to the technical field of low-altitude detection, and particularly relates to a direct wave cancellation method based on a self-adaptive filter.

Background

Effective detection and tracking of low, small and slow targets are one of four major problems in reconnaissance technology, in recent years, research in relevant aspects is carried out at home and abroad, and the general idea for solving the low-altitude detection problem at present comprises three methods of radar detection, acoustic detection and optical detection, but in the acoustic detection method, because the flying power of the low, small and slow targets is mainly electrodynamic force and the flying speed is slow, the noise generated in flying is very small, and if urban environment is considered, the noise can hardly be identified; for optical detection, because the engine with small size and slow target is small in size and an infrared wave-absorbing material is generally adopted, the optical detection is extremely difficult, the existing technology is still mainly researched based on a radio detection method, and the main ideas of the radio detection are as follows: 1. the existing defect technology inhibition method is researched, 2, a radar detection technology (backward reflection) of a target optimization algorithm is designed, and 3, an electromagnetic wave diffraction (forward scattering) technology is applied. However, the prior art cannot effectively solve the accurate detection and identification of the low-slow small target.

Disclosure of Invention

The invention aims to solve the problems in the prior art and provides a direct wave cancellation method based on an adaptive filter.

The invention is realized by the following technical scheme, the invention provides a direct wave cancellation method based on a self-adaptive filter, the self-adaptive filter is utilized to perform direct wave cancellation on an echo channel to obtain an echo signal, and the difference value between the actual output result and the expected output result of the self-adaptive filter is calculated to perform parameter modification to finally obtain the expected output; the adaptive filter needs two signals, namely an input signal x (n) and an expected signal d (n), generates an output signal y (n) after the input signal passes through the adaptive filter, compares the y (n) with the expected signal d (n) to obtain an error signal e (n), adjusts a tap coefficient according to an adaptive filtering algorithm to enable the e (n) to be minimum under a minimum mean square error criterion, and enables the output signal y (n) to approach the expected signal d (n) as much as possible under the minimum mean square error criterion, and the method specifically comprises the following steps:

the input signal is first represented by phasors, i.e., x (n) ═ x (n), x (n-1)]^TThe tap coefficients are then expressed in phasors, i.e. w (k) ═ w₀(k),w₁(k),...,w_L-1(k)]^TWherein L is the order of the adaptive filter, and W (k) denotes adaptationThe tap coefficient set of the k iteration of the filter, thereby yielding an output signal:

y_k(n)＝W^T(k)X(n)

wherein y is_k(n) representing the output signal of the kth iteration, the estimated error signal being obtained by comparing the desired signal with the output signal:

e(n)＝d(n)-W^T(k)X(n)

the mean square error is thus obtained as:

J＝E[W^T(n)]＝E[d²(n)]-2E[d(n)W^T(n)X(n)]+E[W^T(n)X(n)X^T(n)W(n)]

when the tap coefficient is adjusted, in order to adjust the coefficient most effectively in each iteration, a steepest descent algorithm is adopted, that is, the adaptive filter weighting vector is adjusted along the negative gradient direction, so that the formula of the iterative tap coefficient is as follows:

the essence of iteration is that each adjustment is carried out to make the adjustment approach to an optimal state, the degree of each adjustment is determined by the adjustment step length, and in the formula of the iterative tap coefficient, mu is the adjustment length of each adjustment tap coefficient, namely the adjustment step length; the steepest descent algorithm based on the minimum mean square error criterion chooses to differentiate each tap vector by using the mean of the squares of the errors in the instantaneous time, and the mean square error gradient can be obtained as follows:

therefore, the iterative equation of the tap coefficient of the steepest descent algorithm based on the minimum mean square error criterion is simplified into the following equation:

W(k+1)＝W(k)+2μe(n)X(n)

then, in the subsequent operation, multiple iterations are performed according to the formula of the iterative tap coefficient, after k iterations, when the distance between the result after the k iteration and the tap coefficient vector of the (k-1) th iteration is smaller than a certain threshold, the iteration is ended, a final tap coefficient vector w (k) can be obtained, and then the final approximation signal of the expected signal can be obtained:

y_k(n)＝W^T(k)X(n)。

further, in the method, the direct wave cancellation needs to be performed on the echo channel by using the direct wave recovered from the reference channel.

Further, the adaptive filter is a discrete-time signal adaptive filter.

Further, the number of iterations is greater than or equal to the maximum time delay of the multipath effect and less than the minimum time delay of the echo signal.

Further, the value of the adjustment step size μ is specifically:

drawings

FIG. 1 is a schematic diagram of a K-means iterative clustering algorithm;

FIG. 2 is a diagram showing the effect of direct wave purification;

FIG. 3 is a flow chart of signal processing for a discrete-time signal adaptive filter system;

FIG. 4 is a schematic block diagram of direct wave cancellation;

fig. 5 is a waveform diagram after cancellation.

Detailed Description

The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the accompanying drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

5G technology has developed rapidly, and there are many areas that have 5G base stations, and 5G base stations are core devices of 5G networks and provide wireless communication signals for 5G users, so that signal conversion and transmission between a wired communication network and a wireless communication terminal are realized. For larger bandwidth and faster data transmission, the frequency band applied by 5G base station signals is far higher than that of the previous generation mobile communication network in frequency band selection, generally, 5G signals have two common frequency bands, one is 450MHz-6GHz, the other is 24.25GHz-52.6GHz, wherein 24.25GHz-52.6GHz is a microwave band, which is not considered in the invention, and the frequency selection of the signals is carried out in the range of 450MHz-6 GHz. Since the higher the signal frequency, the more obvious the skin effect, and the worse the penetration capability of the signal, in order to realize wireless communication in the urban area of building, the base stations of the 5G network will be very densely arranged, which provides an environment for the application of FSR.

In an actual 5G signal, a symbol rate of 15kB is often used; for each path of subcarriers, different modulation modes can be provided for different application scenes, wherein QAM is a typical common system; in practical application, 5G signals occupy a wide frequency band and have dozens or even hundreds of paths of subcarriers, but in the invention, for the convenience of subsequent analysis, only one rb (resource block) signal is selected for analysis, that is, only 12 paths of subcarriers are included, in the invention, the symbol rate is selected to be 15kHz, the frequency difference of each path of subcarriers is 15kHz, the modulation mode is 16QAM, and the lowest subcarrier frequency of the selected frequency band is 600 MHz.

When the reflected signal of the 5G base station signal is analyzed, two channels are supposed, one is a reference channel and the other is an echo channel, both of which are influenced by the multipath effect, but the thinking for solving the problem of the two channels is different, firstly, a direct wave purification technology is used for extracting relatively pure direct waves in the reference channel, and then, a direct wave cancellation technology is used for extracting relatively pure echo signals in the echo channel.

Multipath effect can cause signal envelope fluctuation, and when a signal affected by the multipath effect is simulated, the range of multipath delay needs to be considered first, so that one-step approximate calculation is performed first. For simplicity, it is assumed that there are only two multipath signals, pair blockFading phenomena were analyzed. Assuming that there are two signals and they have the same amplitude, i.e. attenuation, but different time delays, when the original signal is f (t), it is propagated through two paths, and the signals received by the receiving end are Af (t- τ) respectively₀) And Af (t- τ)₀τ) where A is the attenuation caused in propagation, τ₀For the time delay of the first path, τ is the time delay difference of the two paths, and then, discussing the transfer function of the multipath transmission channel, assuming that the frequency spectrum of the original signal F (t) is F (ω), it can be known that:

the transfer function H (ω) of the multipath channel is obtained by dividing the right end of equation (3) by F (ω):

considering the modulus of the transfer function, one can obtain:

analysis of the modulus of the transfer function shows that the frequency component at ω ═ (2n +1) pi/τ is 0, i.e. frequency selective attenuation occurs at the frequency component, so for a broadband signal, the effect of multipath effect is very serious, for a bistatic radar, it is always considered in the face of echo whether the signal originates from a target desired to be detected or from an ineffective reflector, in the present invention, a reflector within 10m from a 5G base station transmitter is considered to be an ineffective reflector, and the time delay generated by a target at 10m is t_τ＝(10×2)/(3×10⁸)＝6.67×10^-8s is that the sampling rate of the signal is 4800MHz, when the discrete signal analysis is performed, the time delay is considered to be less than n_τ＝t_τ×f_sThe echo with the time delay larger than 320 is considered as the echo generated by the effective target.

Clustering refers to a process of grouping together data sets having similar characteristics in some respect for a group of data sets and considering them as a class, and the difference between clustering and classification is that clustering is an unsupervised learning process, and it is not known at the beginning what the specific intrinsic relation of each class is, which is quite consistent with the case that a fluctuation signal is received, i.e. both random envelope and random phase. The K-means clustering algorithm is particularly suitable for the situation that the data set is known to be divided into several classes, namely the K value is known, and the K value is known to be 16 because the known modulation mode is 16QAM, so that the K-means clustering algorithm is quite suitable and reasonable.

The K-means clustering algorithm (K-means clustering algorithm) is an algorithm for approximating the best clustering point of a data set by continuous iteration, and the specific steps are that after the data is divided into K groups, K objects in the data set are randomly selected as initial clustering points, then the distance between each object and each initial clustering point is calculated, each object is assigned to the closest clustering point, namely, the objects are considered to be one class, then the centroid of each class of data set is calculated as the clustering point after the first processing, but if the current process is finished, the clustering result is not ideal because of only one iteration, as shown in fig. 1(d), the position of the cross of two colors is not a reasonable clustering point intuitively. In order to obtain the best clustering point, the distance between each data point and the new clustering point is calculated for a plurality of times, each data point is divided to the closest clustering point to the new clustering point to serve as a class, then the centroid of each class of data points is calculated to serve as the clustering point of the second processing, iteration is carried out by analogy, and when the distance between the clustering point obtained after the N +1 th processing and the clustering point obtained after the N th processing is found to be smaller than a certain threshold value, the processing can be stopped, so that the purified direct wave signal is obtained. FIG. 1 is a schematic diagram of a K-means algorithm for realizing clustering through three times of iteration processing.

In the positioning system of the exogenous radiation radar, obtaining an accurate source signal is a key problem, and the problem is directly related to whether the subsequent operation of the system can be smoothly carried out and finally related to the positioning accuracy of the system. In an exogenous radiation radar positioning system, a receiving end generally receives a signal which directly reaches the receiving end without reflection, and the signal is generally called direct wave. Therefore, the direct wave extraction is essentially a problem of researching how to recover the original transmission signal under the interference of multipath clutter and additive noise. In wireless communication, time domain equalization techniques are generally used to recover the original signal against the waveform fluctuations caused by multipath interference. For constant envelope signals, a commonly used time domain equalization technique is called constant modulus algorithm, but QAM is not a constant envelope signal, however it is noted that the envelope of the in-phase and quadrature components of QAM is constant over one symbol duration, so that the quadrature and in-phase components of QAM signals can be extracted for each symbol, the consequence of multipath signal interference being an offset in the signal constellation. The direct wave signal with the deviation is subjected to direct wave purification by using a K-means clustering algorithm, and a purification effect graph is shown in fig. 2, wherein fig. 2(a) is an original signal, fig. 2(b) is a waveform under multipath clutter and additive noise interference, and fig. 2(c) is a purified direct wave.

As can be seen from fig. 2, the recovered direct wave is very close to the original signal, which can already meet the requirement of the subsequent processing, and the original RB 5G base station signal can be obtained by purifying and superposing each path of subcarriers according to the method, and since the direct wave is regenerated, the interference of the additive noise can be completely suppressed.

With reference to fig. 4, after a proper direct wave is recovered, the adaptive filter can be designed to perform direct wave cancellation on the echo channel, so as to eliminate the direct wave and multipath propagation effects.

The adaptive filter may be a continuous-time signal filter or a discrete-time signal filter. In the invention, in order to facilitate the signal processing of a computer, a discrete time signal adaptive filter model is adopted, which consists of a group of tapped delay lines, variable weighting coefficients and an automatic adjustment coefficient algorithm module. Fig. 3 shows a signal processing flow diagram of a discrete-time signal adaptive filter system. The system needs two signals, namely an input signal x (n) and an expected signal d (n), generates an output signal y (n) after the input signal passes through the system, then compares the y (n) with the expected signal d (n) to obtain an error signal e (n), and then adjusts tap coefficients according to an adaptive filtering algorithm to enable the e (n) to be minimum under a minimum mean square error criterion, so that the output signal y (n) is close to the expected signal d (n) as much as possible under the minimum mean square error criterion.

The input signal is first represented by phasors, i.e., x (n) ═ x (n), x (n-1)]^TThe tap coefficients are then expressed in phasors, i.e. w (k) ═ w₀(k),w₁(k),...,w_L-1(k)]^TWhere L is the order of the adaptive filter and w (k) represents the set of tap coefficients for the kth iteration of the adaptive filter, generally, the more iterations, the closer the final output signal is to the desired signal, thereby yielding an output signal:

y_k(n)＝W^T(k)X(n)

wherein y is_k(n) representing the output signal of the kth iteration, the estimated error signal being obtained by comparing the desired signal with the output signal:

e(n)＝d(n)-W^T(k)X(n)

the mean square error is thus obtained as:

J＝E[W^T(n)]＝E[d²(n)]-2E[d(n)W^T(n)X(n)]+E[W^T(n)X(n)X^T(n)W(n)]

when the tap coefficient is adjusted, in order to adjust the coefficient most effectively in each iteration, a steepest descent algorithm is adopted, that is, the adaptive filter weighting vector is adjusted along the negative gradient direction, so that the formula of the iterative tap coefficient is as follows:

the essence of iteration is that each adjustment is carried out to make the adjustment approach to an optimal state, the degree of each adjustment is determined by the adjustment step length, and in the formula of the iterative tap coefficient, mu is the adjustment length of each adjustment tap coefficient, namely the adjustment step length; the steepest descent algorithm based on the minimum mean square error criterion chooses to differentiate each tap vector by using the mean of the squares of the errors in the instantaneous time, and the mean square error gradient can be obtained as follows:

therefore, the iterative equation of the tap coefficient of the steepest descent algorithm based on the minimum mean square error criterion is simplified into the following equation:

W(k+1)＝W(k)+2μe(n)X(n)

then, in the subsequent operation, multiple iterations are performed according to the formula of the iterative tap coefficient, after k iterations, when the distance between the result after the k iteration and the tap coefficient vector of the (k-1) th iteration is smaller than a certain threshold, the iteration is ended, a final tap coefficient vector w (k) can be obtained, and then the final approximation signal of the expected signal can be obtained:

y_k(n)＝W^T(k)X(n)。

because in the echo channel, except the target reflection signal, there are multipath interference and stronger direct wave, in order to analyze the echo signal better, it needs to use the direct wave restored from the reference channel to cancel the direct wave, the basic idea of the direct wave cancellation is to use the direct wave signal to approach the multipath interference signal, the invention realizes the idea that the direct wave cancellation mainly uses the adaptive filter, i.e. taps the reference signal, and assumes a group of tap coefficients, at this moment, the initial state of the adaptive filter is obtained, then the signal passes through the filter, compares the filtered signal with the monitoring signal to obtain an error, and adjusts the group of tap coefficients according to the error to iterate, the error will decrease with the increase of the number of iterations, but will bring large calculation amount at the same time, it is worth noting that the number of iterations should be greater than or equal to the maximum delay of the multipath effect and less than the minimum delay of the echo signal, otherwise, the multipath effect is not completely inhibited or the echo signal is obviously inhibited, the invention utilizes the step-length-variable algorithm to adjust the coefficient greatly to quickly reduce the error when the error is larger, and can carry out fine adjustment when the error is smaller so as to achieve the purpose of realizing the best approximation under the condition of enough iteration times.

The value of the adjustment step size mu is specifically as follows:

fig. 5 shows the waveform after cancellation, and observing fig. 5, the amplitude of the signal is reduced after the direct wave is cancelled, and the echo signal under the noise coverage can be seen in a hidden way at the tip, and the next analysis can be carried out according to the signal. However, it has to be said that the signal-to-noise ratio of the echo channel is relatively high, and when the signal-to-noise ratio is greater than 15dB, the cancellation effect is relatively good.

After the direct wave and the multipath clutter of the echo receiving channel are eliminated, the echo signal still cannot be directly extracted from a noise environment because the power of the echo signal is very small, and the direct wave component of the channel is eliminated at the moment, so that the pure direct wave signal and the echo signal can be subjected to cross-correlation processing. In order to obtain the time delay information simply and quickly, the cross-correlation processing can be carried out on the pure direct wave signal and the echo signal. Target position information can be obtained according to the azimuth angle and the time delay information, cross-correlation processing can be understood as a cross-fuzzy function of two signals, and a mathematical model of the cross-fuzzy function is as follows:

where τ is the time delay, f₀(t) a fuzzy function representing the clean direct wave signal, f₁ ^*(t + τ) represents the blurring function of the echo signal, and t represents time.

The present invention provides a direct wave cancellation method based on an adaptive filter, which is described in detail above, and the principle and the implementation of the present invention are explained in this document by applying specific examples, and the description of the above examples is only used to help understanding the method of the present invention and the core idea thereof; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.