Communication interference effect online evaluation method based on improved CRITIC-GAHP

Abstract

The invention discloses a communication interference effect online evaluation method based on improved CRITIC-GAHP, and the method specifically comprises the steps: selecting a proper communication party parameter as an evaluation index, carrying out the numeralization of the anti-interference capability of a modulation pattern, calculating the change amount of each evaluation index before and after interference, and carrying out the normalization of the change amount; using improved CRITIC to calculate the objective weight of each index, using GAHP to synthesize the opinions of multiple experts to calculate the subjective weight of each index, and combining the subjective weight and the objective weight according to the minimum information entropy principle to obtain a comprehensive weight; calculating to obtain an interference effect score according to the comprehensive weight and the normalization result, and grading the score to obtain an interference effect grade; and according to the test sample, dynamically updating the normalized parameter of each index to realize online evaluation of the interference effect. According to the method, the interference effect of a single interference scheme can be evaluated, self parameters can be adjusted according to a changing environment, and the interference scheme in an unknown environment can be evaluated.

Description

Technical Field

[0001] The invention relates to the technical field of communication countermeasures, and in particular to an online evaluation method for communication interference effects based on improved CRITIC-GAHP. Background Art

[0002] With the development of science and technology and information technology, the importance of electronic warfare in modern warfare is increasing day by day. Electronic warfare refers to various electronic measures and actions taken by hostile parties to weaken and destroy the effectiveness of the other party's electronic equipment and ensure the effectiveness of their own electronic equipment. It plays a vital role in achieving victory in combat. Interference effect evaluation is an important basis for adjusting interference strategies. Since the interference effect cannot be directly observed like the effect of weapon strikes, how to evaluate the interference effect is an urgent problem to be solved.

[0003] After the jammer implements the jamming, it is impossible to directly obtain the communication results of the communicating party, which makes the method of using bit error rate as the jamming effect evaluation standard unusable in wartime. With the improvement of reconnaissance technology, the reconnaissance system can obtain more communication party behavior parameters more accurately. Researchers began to comprehensively judge the jamming effect based on the changes in various communication behavior parameters of the sender before and after the jamming.

[0004] At present, interference effect evaluation methods are mainly divided into two categories, evaluation factor method and artificial intelligence evaluation method. The evaluation factor method mainly uses subjective and objective weight methods such as Analytic Hierarchy Process (AHP), entropy weight method, CRITIC, etc. to determine the weight of each evaluation indicator based on expert opinions and training samples, and then combines the Technique for Order Preference by Similarity to Ideal Solution (TOPSIS) or grey correlation analysis method to rank the interference effects of multiple interference schemes. The artificial intelligence evaluation method first divides the sample set into interference effect levels, and then uses classification methods such as Support Vector Machine (SVM) for training. However, these methods have the following defects:

[0005] (1) The evaluation factor method mainly ranks the interference effects of multiple interference schemes and cannot determine the interference effect of a single interference scheme;

[0006] (2) Artificial intelligence evaluation methods mainly study the classification of labeled samples and are unable to classify unlabeled samples;

[0007] (3) The current method cannot adaptively adjust itself when dealing with environmental changes, and has poor robustness, resulting in decreased accuracy.

[0008] Therefore, it is urgent to study an interference effect evaluation method to solve the problem that the current interference effect evaluation method can only classify label samples and cannot update its own method according to environmental changes. Summary of the invention

[0009] The purpose of the present invention is to provide a communication interference effect evaluation method that can evaluate the interference effect of a single interference scheme, classify unlabeled samples, and evaluate interference schemes in unknown environments.

[0010] The technical solution to achieve the purpose of the present invention is: an online evaluation method for communication interference effect based on improved CRITIC-GAHP, comprising the following steps:

[0011] Step 1: Select the communication party parameters as evaluation indicators;

[0012] Step 2: quantify the anti-interference capability of the modulation pattern;

[0013] Step 3: Calculate the change of each evaluation index before and after the disturbance, and normalize the change;

[0014] Step 4: Based on the normalized results of each indicator, the objective weight of each indicator is calculated using the improved weighting method, namely CRITIC;

[0015] Step 5, using the group hierarchy analysis method, or GAHP, to integrate the opinions of multiple experts and calculate the subjective weight of each indicator;

[0016] Step 6: According to the principle of minimum information entropy, combine the subjective weight and the objective weight to obtain the comprehensive weight;

[0017] Step 7: Calculate the interference effect score based on the comprehensive weight and normalization result, and classify the score to obtain the interference effect level;

[0018] Step 8: Dynamically update the normalized parameters of each indicator according to the test samples to achieve online evaluation of the interference effect.

[0019] Furthermore, the communication party parameters selected as evaluation indicators in step 1 are as follows:

[0020] The five parameters obtained by the reconnaissance system, including modulation style, low-density parity check code LDPC code rate, power, frequency, and duration of communication behavior, are selected as evaluation indicators.

[0021] Furthermore, the anti-interference capability of the modulation pattern described in step 2 is digitized as follows:

[0022] Adaptive modulation technology selects a modulation style with high throughput for communication while meeting the error rate requirement. The higher the modulation order, the greater the throughput. The modulation styles are sorted based on the bit error rate BER and modulation order, and the anti-interference capabilities of different modulation styles are mapped to a set of values.

[0023] Furthermore, the change amount of each evaluation index before and after the disturbance is calculated in step 3, and the change amount is normalized, as follows:

[0024] Calculate the changes in the evaluation indicators of modulation style, code rate, power, frequency and duration of communication behavior, and normalize the changes in the evaluation indicators to between -1 and 1. When the changes in the evaluation indicators reflect that the interference effect is invalid, normalize it to between -1 and 0; when it reflects that the interference is effective, normalize it to between 0 and 1; in this way, the interference effect score of the samples with invalid interference is less than or equal to 0, thereby distinguishing them from the samples with effective interference;

[0025] The changes and normalization methods of each indicator before and after disturbance are as follows:

[0026] 1) Modulation style

[0027] Modulation pattern before interference m b Corresponding to SNR b , after disturbance m a Corresponding to SNR a , the change in modulation pattern before and after disturbance c m The calculation formula is:

[0028] c m =snr b -snr a ,m a ≠None (1)

[0029] Because the stronger the anti-interference ability of the modulation style, the smaller the corresponding SNR value, so SNR is used b -snr a To ensure that the interference is effective, the time difference is a positive number; when m a =None, the communicating parties stop communicating and directly normalize it to 1 without calculating the change amount;

[0030] Normalized result of modulation pattern change m The calculation formula is:

[0031]

[0032] Among them, max m =max(|max cm |,|min cm |), ensure e mThe result is between -1 and 1;

[0033] 2) Bit rate

[0034] The code rate before disturbance is r b , the code rate after interference is r a , the same as the modulation pattern setting, when the communication party stops communicating, r a =0, directly normalize the bit rate change to 1;

[0035] The bit rate change before and after the disturbance c r The calculation formula is:

[0036] c r =r a -r b ,r a ≠0 (3)

[0037] The normalized result of the bit rate change is e r The calculation formula is:

[0038]

[0039] Among them, max r =max(|max cr |,|min cr |), ensure e r The result is between -1 and 1;

[0040] 3) Power

[0041] The power before the disturbance is p b , after being disturbed, it is p a , the change in power before and after the disturbance c p The calculation formula is:

[0042] c p =p a -p b ,p a ≠0 (5)

[0043] Measurement errors can cause fluctuations in power measurement results, so a power change decision threshold δ is set. p , if the power difference before and after the disturbance exceeds δ p , the interference is considered effective; otherwise it is considered to be a measurement error; therefore c p Need and δ p For comparison, if the threshold is not exceeded, it needs to be normalized to 0;

[0044] Normalized result of power change e p The calculation formula is:

[0045]

[0046] Among them, max p =max(|max cp |,|min cp |), ensure e p The result is between -1 and 1;

[0047] 4) Frequency

[0048] The frequency before disturbance is f b , after being disturbed, it is f a , the change in frequency before and after the disturbance c f The calculation formula is:

[0049] c f =|f a -f b | (7)

[0050] Taking into account the measurement error, the decision threshold δ of the frequency change is set f , the frequency has nothing to do with the anti-interference ability. When the frequency change exceeds the threshold, e f =1; otherwise e f =0;

[0051] The normalized result of the frequency change is e f The calculation formula is:

[0052]

[0053] 5) Duration of communication behavior

[0054] After the communication behavior changes, the duration t of the previous state can be obtained. Therefore, there is no need to calculate the change amount. It is only necessary to calculate the maximum and minimum duration T in the sample. max and T min Normalize the duration, and the normalized result of the communication behavior duration is e t The calculation formula is:

[0055]

[0056] After normalizing the values ​​of each evaluation index in the sample data, the normalized result E=(e m ,e r ,e p ,e f ,e t ).

[0057] Furthermore, according to the normalized results of each indicator in step 4, the objective weight of each indicator is calculated using the improved weighting method, namely CRITIC, as follows:

[0058] Step 4.1: Calculate the standard deviation and information entropy of each indicator

[0059] Suppose there are p evaluation indicators in total, where the standard deviation of the i-th indicator is S i and information entropy E i The calculation formula is:

[0060]

[0061]

[0062] Among them, x ki Represents the value corresponding to the i-th indicator in the k-th sample; is the mean of all samples corresponding to the ith index table,

[0063] Step 4.2: Calculate the conflict of each indicator

[0064] The calculation formula of each index conflict R is:

[0065]

[0066] In the formula, r ij It represents the correlation coefficient between indicators i and j, which is calculated using the Pearson correlation coefficient. The calculation formula is:

[0067]

[0068] Step 4.3: Calculate the weight of each indicator

[0069] By standard deviation S i 、Information entropy E i and conflict R i Calculate the information content C of each indicator i and objective weight The calculation formula is:

[0070] C i =(1-E i +S i )R i (14)

[0071]

[0072] Step 4.4, finally get the total objective weight:

[0073]

[0074] Furthermore, the group analytic hierarchy process (GAHP) described in step 5 is used to combine the opinions of multiple experts to calculate the subjective weight of each indicator, as follows:

[0075] Step 5.1: Calculate the similarity of each expert

[0076] First, AHP is used to determine the subjective weight vectors according to the opinions of each expert. The subjective weight W determined by the i-th expert is (i) and the subjective weight W determined by the jth expert (j) The degree of similarity d ij It is calculated using the Euclidean distance, and the calculation formula is:

[0077]

[0078] where d ij Satisfy ij =0,i=j;d ij =d ji ≥0;

[0079] The similarity between the lth expert and other experts among L experts is d l The calculation formula is:

[0080]

[0081] Step 5.2: Calculate the decision weight of each expert

[0082] The decision weight coefficient λ of the lth expert l The calculation formula is:

[0083]

[0084] Among them, d l The larger the value is, the greater the disagreement between the first expert and other experts is, and the smaller the decision weight coefficient is; l The smaller it is, the closer the opinions of the first expert are to those of other experts, and the larger the decision weight coefficient is; l =0, the opinions of the L experts are the same, so the decision weights are also the same, which is 1 / L;

[0085] Step 5.3: Calculate subjective weight

[0086] The weight coefficient λ is determined by experts l And each weight vector weighted to get the subjective weight The calculation formula is:

[0087]

[0088] Furthermore, according to the minimum information entropy principle described in step 6, the subjective weight and the objective weight are combined to obtain a comprehensive weight, which is as follows:

[0089] The information entropy between the combined weight and the subjective weight, objective weight, and the constraint formula of the combined weight are:

[0090]

[0091] According to formula (21), the Lagrange equation shown in formula (22) is established:

[0092]

[0093] L vs w i Find the partial derivative and set the result to 0, and we get:

[0094]

[0095] Finally, the combined weight w is obtained i , the calculation formula is:

[0096]

[0097] Furthermore, in step 7, the interference effect score is calculated based on the comprehensive weight and the normalized result, and the score is graded to obtain the interference effect level, which is as follows:

[0098] The interference effect is divided into 5 levels from 0 to 4, where 0 represents ineffective interference and 1 to 4 represents effective interference. The higher the level, the better the interference effect. The calculation formula for the interference effect score S and the interference effect level G is:

[0099] S=EW T (25)

[0100]

[0101] S'=e m w m +e r w r +e p w p +e f w f (27)

[0102] Among them, E is the normalized result, W is the subjective and objective combined weight, E = (e m ,e r ,e p ,e f ,e t ), e i ∈[-1,1],w i ≥0, i=m,r,p,f,t, so S∈[-1,1]; S' is the interference effect score calculated after removing the duration of the communication behavior; when S' is less than or equal to 0, the interference is judged to be invalid and the interference effect level G=0; when S' is greater than 0, the interference effect level is divided according to the overall interference effect score S.

[0103] Furthermore, in step 8, the normalized parameters of each indicator are dynamically updated according to the test sample to realize the online evaluation of the interference effect, as follows:

[0104] In actual applications, the max value is dynamically updated based on the test samples. m 、max r 、max p , δ f 、T min The normalization parameter of f Used to determine whether the frequency has changed, and set the parameter value according to the reconnaissance equipment and environment used; other indicators change dynamically according to the change of the test sample. When the change of the test sample exceeds the indicator range, the indicator is updated. The update method is:

[0105]

[0106] Among them, c i is the change of indicator i in the test sample, and t is the duration of the communication behavior in the test sample.

[0107] A communication interference effect online evaluation system based on improved CRITIC-GAHP, the system is used to implement the communication interference effect online evaluation method, the system includes a parameter selection module, a numerical module, a variable calculation module, an objective weight calculation module, a subjective weight calculation module, a comprehensive weight calculation module, an interference effect level classification module and an online evaluation module, wherein:

[0108] The parameter selection module selects the communication party parameters as evaluation indicators;

[0109] A digitization module digitizes the anti-interference capability of the modulation pattern;

[0110] The variable calculation module calculates the changes of each evaluation index before and after the disturbance and normalizes the changes;

[0111] The objective weight calculation module uses the improved CRITIC to calculate the objective weight of each indicator based on the normalized results of each indicator;

[0112] The subjective weight calculation module uses GAHP to combine the opinions of multiple experts to calculate the subjective weight of each indicator;

[0113] The comprehensive weight calculation module combines subjective weight and objective weight according to the principle of minimum information entropy to obtain the comprehensive weight;

[0114] The interference effect level classification module calculates the interference effect score based on the comprehensive weight and normalization result, and classifies the score to obtain the interference effect level;

[0115] The online evaluation module dynamically updates the normalized parameters of each indicator according to the test samples to realize the online evaluation of the interference effect.

[0116] Compared with the prior art, the present invention has the following significant advantages: (1) the interference effect is judged based on the change in the communication parameters before and after the interference, and the interference effect of a single interference scheme can be evaluated; (2) the interference effect score is calculated based on the normalized results of each evaluation indicator and the indicator weight, and the unlabeled samples can be classified; (3) the dynamic update normalization parameter method is adopted, and the parameters can be adjusted according to the changing environment, so that the interference scheme in an unknown environment can be evaluated. BRIEF DESCRIPTION OF THE DRAWINGS

[0117] Figure 1 It is a flow chart of an online evaluation method for communication interference effect based on improved CRITIC-GAHP of the present invention.

[0118] Figure 2 It is a BER curve diagram of different modulation modes in an embodiment of the present invention.

[0119] Figure 3 It is a graph of subjective and objective weights and combined weight curves calculated using different methods in the embodiments of the present invention.

[0120] Figure 4 It is a classification accuracy curve diagram of the four classification methods in different parameter ranges in the embodiment of the present invention.

[0121] Figure 5 is a classification accuracy curve diagram under different numbers of training samples in an embodiment of the present invention. DETAILED DESCRIPTION

[0122] The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

[0123] Combination Figure 1 The present invention provides an online evaluation method for communication interference effect based on improved CRITIC-GAHP, comprising the following steps:

[0124] Step 1: Select appropriate communication party parameters as evaluation indicators, as follows:

[0125] The five parameters available to the reconnaissance system, namely modulation style, LDPC code rate, power, frequency, and duration of communication behavior, are selected as evaluation indicators for the following reasons:

[0126] In order to increase the data transmission rate, adaptive modulation technology is widely used. Adaptive modulation technology can intelligently change the modulation style according to the channel environment, and use high-order modulation styles to improve throughput while ensuring communication quality. When the communication party is interfered and the communication quality decreases, the communication party will change the modulation style, reduce the modulation order, and improve transmission reliability.

[0127] Error correction coding is a commonly used anti-interference technology. It increases supervision symbols in the transmission sequence and reduces the transmitted information symbols at the cost of reducing the information transmission rate, thereby improving the reliability of transmission. With its error correction performance close to the Shannon limit, LDPC has become the channel coding scheme for wireless communication standards such as DVB-S2, IEEE802.16e, CCSDS, and 5G.

[0128] For the jammer to successfully interfere, the interference signal to communication signal ratio must meet certain conditions. When the communicating party is interfered with, increasing the communication power can effectively reduce the interference signal ratio, reduce the impact of the interference signal on the communication signal, and ensure the communication quality.

[0129] When the communicating party is interfered with, it will choose to change the frequency of the signal to prevent interference. Therefore, aligning the frequency of the interference signal with the frequency of the communication signal is a prerequisite for successful interference.

[0130] The duration of communication behavior refers to the time from when the communication party is interfered to when it changes its communication behavior. When the enemy is interfered, it will change the communication behavior parameters, and the duration will become shorter. The stronger the interference to the enemy, the shorter the communication time. The duration of communication status can also be used as one of the indicators to evaluate the effect of interference.

[0131] The parameter ranges of the evaluation indicators selected by the present invention are shown in Table 1. The present invention selects 10 modulation styles as references and 5 LDPC code rates. For the convenience of calculation, the present invention will not use LDPC coding and use code rate 1 to represent it. The power range is set to 20W-200W, and the frequency range is set to 30MHz-400MHz. The duration of the communication behavior refers to the time from the start of the interference by the interferer to the change of the communication behavior by the sender. The present invention sets an upper limit for the duration of the communication behavior. When the duration exceeds the upper limit and the communication behavior of the sender has not changed, the interference is considered invalid and the interferer needs to reselect the interference strategy.

[0132] Table 1 Evaluation index parameter range

[0133]

[0134] Step 2: digitize the anti-interference capability of the modulation pattern, as follows:

[0135] Other evaluation indicators can reflect the interference effect through numerical changes, so it is necessary to digitize the anti-interference ability of the modulation style to make it easier to calculate. The anti-interference ability of different modulation styles is mapped to a set of numerical values, and the modulation styles are sorted in combination with the bit error rate BER and the modulation order. The adaptive modulation technology selects the modulation style with a large throughput for communication under the premise of meeting the error rate. The higher the modulation order, the greater the throughput. Except for the modulation style, other evaluation indicators can reflect the interference effect through numerical changes. In order to calculate the difference in the change of the modulation style before and after the interference, it is necessary to map the anti-interference ability of different modulation styles to a set of numerical values. The standard for adaptive modulation technology to select the modulation style is to select the modulation style with a large throughput for communication under the premise of meeting the error rate. The higher the modulation order, the greater the throughput. Based on this criterion, the present invention sorts the modulation styles in combination with the bit error rate BER and the modulation order. Figure 2 The curves of BER changing with signal-to-noise ratio (SNR) for different modulation styles in the additive white Gaussian noise (AWGN) channel are shown.

[0136] The present invention selects the SNR values ​​corresponding to different modulation modes when BER=0.1 as the anti-interference capability values ​​of different modulation modes. The sorting results of the modulation modes and the corresponding SNR values ​​are shown in Table 2.

[0137] Table 2 The anti-interference ability ranking results of the modulation patterns and the corresponding SNR values

[0138]

[0139] Step 3: Calculate the change of each evaluation index before and after the disturbance, and normalize the change for subsequent calculation of objective weights and interference effect scores, as follows:

[0140] Calculate the changes in the evaluation indicators of modulation style, code rate, power, frequency and duration of communication behavior, and normalize the changes in the evaluation indicators to between -1 and 1. When the changes in the evaluation indicators reflect that the interference effect is invalid, normalize it to between -1 and 0; when it reflects that the interference is effective, normalize it to between 0 and 1; in this way, the interference effect score obtained by the samples with invalid interference is less than or equal to 0, which is convenient for distinguishing them from the samples with effective interference;

[0141] The changes and normalization methods of each indicator before and after disturbance are as follows:

[0142] 1) Modulation style

[0143] Modulation pattern before interference m b Corresponding to SNR b , after disturbance m a Corresponding to SNR a , the change in modulation pattern before and after disturbance cm The calculation formula is:

[0144] c m =snr b -snr a ,m a ≠None (1)

[0145] Because the stronger the anti-interference ability of the modulation style, the smaller the corresponding SNR value, so SNR is used b -snr a To ensure that the interference is effective, the time difference is a positive number; when m a =None, the communicating parties stop communicating and directly normalize it to 1 without calculating the change amount;

[0146] Normalized result of modulation pattern change m The calculation formula is:

[0147]

[0148] Among them, max m =max(|max cm |,|min cm |), ensure e m The result is between -1 and 1;

[0149] 2) Bit rate

[0150] The code rate before disturbance is r b , the code rate after interference is r a , the same as the modulation pattern setting, when the communication party stops communicating, r a =0, directly normalize the bit rate change to 1;

[0151] The bit rate change before and after the disturbance c r The calculation formula is:

[0152] c r =r a -r b ,r a ≠0 (3)

[0153] The normalized result of the bit rate change is e r The calculation formula is:

[0154]

[0155] Among them, max r =max(|max cr |,|min cr |), ensure e r The result is between -1 and 1;

[0156] The power before the disturbance is p b , after being disturbed, it is p a , the change in power before and after the disturbance c p The calculation formula is:

[0157] c p =p a -p b ,p a ≠0 (5)

[0158] Measurement errors can cause fluctuations in power measurement results, so a power change decision threshold δ is set. p , if the power difference before and after the disturbance exceeds δ p , the interference is considered effective; otherwise it is considered to be a measurement error; therefore c p Need and δ p For comparison, if the threshold is not exceeded, it needs to be normalized to 0;

[0159] Normalized result of power change e p The calculation formula is:

[0160]

[0161] Among them, max p =max(|max cp |,|min cp |), ensure e p The result is between -1 and 1;

[0162] 4) Frequency

[0163] The frequency before the disturbance is f b , after being disturbed, it is f a , the change in frequency before and after the disturbance c f The calculation formula is:

[0164] c f =|f a -f b | (7)

[0165] Taking into account the measurement error, the decision threshold δ of the frequency change is set f , the frequency has nothing to do with the anti-interference ability. When the frequency change exceeds the threshold, e f =1; otherwise e f =0;

[0166] The normalized result of the frequency change is e f The calculation formula is:

[0167]

[0168] 5) Duration of communication behavior

[0169] After the communication behavior changes, the duration t of the previous state can be obtained. Therefore, there is no need to calculate the change amount. It is only necessary to calculate the maximum and minimum duration T in the sample. max and T min Normalize the duration, and the normalized result of the duration of the communication behavior is e t The calculation formula is:

[0170]

[0171] After normalizing the values ​​of each evaluation index in the sample data, the normalized result E=(e m ,e r ,e p ,e f ,e t ).

[0172] Step 4: Based on the normalized results of each indicator, use the improved CRITIC to calculate the objective weight of each indicator, as follows:

[0173] CRITIC is an objective weighting method that determines weights based on the volatility of each indicator sample value and the conflict between indicators. Volatility is expressed in the form of standard deviation. The larger the standard deviation, the greater the difference in sample values ​​of the evaluation indicator, and the higher the weight. Conflict is calculated using the correlation coefficient. If there is a strong correlation between two indicators, the smaller the conflict of the indicators, the lower the weight. In order to more comprehensively evaluate the volatility and dispersion of indicators, information entropy is introduced into CRITIC. The improved CRITIC method for calculating weights includes the following steps:

[0174] Step 4.1: Calculate the standard deviation and information entropy of each indicator

[0175] Suppose there are p evaluation indicators in total, where the standard deviation of the i-th indicator is S i and information entropy E i The calculation formula is:

[0176]

[0177] Among them, x ki Represents the value corresponding to the i-th indicator in the k-th sample; is the mean of all samples corresponding to the ith index table,

[0178] Step 4.2: Calculate the conflict of each indicator

[0179] The calculation formula of the conflict R of each indicator is:

[0180]

[0181] In the formula, r ij It represents the correlation coefficient between indicators i and j, which is calculated using the Pearson correlation coefficient. The calculation formula is:

[0182]

[0183] Considering the possible negative correlation between indicators, it is necessary to add the absolute value to the correlation coefficient;

[0184] Step 4.3: Calculate the weight of each indicator

[0185] By standard deviation S i 、Information entropy E i and conflict R i Calculate the information content C of each indicator i and objective weight The calculation formula is:

[0186] C i =(1-E i +S i )R i (14)

[0187]

[0188] Step 4.4, finally get the total objective weight:

[0189]

[0190] Step 5: Use the group hierarchy analysis method (GAHP) to combine the opinions of multiple experts and calculate the subjective weight of each indicator, as follows:

[0191] AHP determines the subjective weight of each indicator based on expert opinions, but the traditional AHP method does not take into account the degree of difference in expert opinions when synthesizing multiple expert opinions. The present invention uses GAHP as a subjective weight calculation method. First, AHP is used to determine multiple subjective weight vectors based on the opinions of each expert, and then the decision weights of different experts are determined based on the similarity between the subjective weight vectors. Finally, the weighted synthesis is performed to obtain the comprehensive subjective weight. The calculation process of GAHP includes the following steps:

[0192] Step 5.1: Calculate the similarity of each expert

[0193] First, AHP is used to determine the subjective weight vectors according to the opinions of each expert. The subjective weight W determined by the i-th expert is (i) and the subjective weight W determined by the jth expert (j) The degree of similarity d ij It is calculated using the Euclidean distance, and the calculation formula is:

[0194]

[0195] where d ij Satisfy ij =0,i=j;d ij =d ji ≥0;

[0196] The similarity between the lth expert and other experts among L experts is d l The calculation formula is:

[0197]

[0198] Step 5.2: Calculate the decision weight of each expert

[0199] The decision weight coefficient λ of the lth expert l The calculation formula is:

[0200]

[0201] Among them, d l The larger the value is, the greater the disagreement between the first expert and other experts is, and the smaller the decision weight coefficient is; l The smaller it is, the closer the opinions of the first expert are to those of other experts, and the larger the decision weight coefficient is; l =0, the opinions of the L experts are the same, so the decision weights are also the same, which is 1 / L;

[0202] Step 5.3: Calculate subjective weight

[0203] The weight coefficient λ is determined by experts l And each weight vector weighted to get the subjective weight The calculation formula is:

[0204]

[0205] Step 6: Based on the minimum information entropy principle, combine the subjective and objective weights to obtain the comprehensive weight, as follows:

[0206] The objective weight is calculated from the sample data and is easily affected by the sample. When the sample data is incomplete, the objective weight result will also be inaccurate. Therefore, it is necessary to combine the subjective weight W s and objective weight W o Get the combined weight W, W = (w1, w2, ..., w p ). In order to make the combined weight as close as possible to the subjective and objective weight, the present invention calculates the combined weight based on the minimum information entropy principle. The information entropy between the combined weight and the subjective and objective weight, and the constraint formula of the combined weight are:

[0207]

[0208] According to formula (21), the Lagrange equation shown in formula (22) can be established:

[0209]

[0210] L vs w i Find the partial derivative and set it to 0, and we get:

[0211]

[0212] Finally, the combined weight w is obtained i , the calculation formula is:

[0213]

[0214] Step 7: Calculate the interference effect score based on the comprehensive weight and normalization result, and classify the score to obtain the interference effect level, as follows:

[0215] The interference effect is obtained by combining the change amount of each indicator before and after the sender is disturbed and the indicator weight. The more the number of indicator changes and the greater the change amount, the higher the interference effect score and the higher the interference effect level. The present invention divides the interference effect into 0-4, a total of 5 levels, 0 represents ineffective interference, 1-4 represents effective interference, the higher the level, the better the interference effect; the calculation formula for the interference effect score S and the interference effect level G is:

[0216] S=EW T (25)

[0217]

[0218] S'=e m w m +e r w r +e p w p +e f w f (27)

[0219] Among them, E is the normalized result, W is the subjective and objective combined weight, E = (e m ,e r ,e p ,e f ,e t ), e i ∈[-1,1],w i≥0, i=m,r,p,f,t, so S∈[-1,1]; S' is the interference effect score calculated after removing the duration of the communication behavior; because when the interference of the interferer is invalid, when the sender changes the communication behavior due to the improvement of channel quality, the duration of the communication behavior will also change. Therefore, when judging whether it is invalid interference, it is necessary to remove the influence of the duration of the communication behavior; when S' is less than or equal to 0, the interference is judged to be invalid and the interference effect level G=0; when S' is greater than 0, the interference effect level is divided according to the overall interference effect score S.

[0220] Step 8: Dynamically update the normalization parameters of each indicator according to the test sample so that it can cope with the changing environment, as follows:

[0221] When evaluating the interference effect level of new samples, it is also necessary to normalize the change before and after the interference to between -1 and 1. However, when the training sample data is incomplete and the change of the new sample exceeds the maximum change in the training sample, the normalized result will be greater than 1 or less than -1, resulting in incorrect classification of the interference effect level. Therefore, in actual applications, the maximum value is dynamically updated according to the test sample. m 、max r 、max p , δ f 、T min The normalization parameter of f Used to determine whether the frequency has changed. Parameters need to be set according to the reconnaissance equipment and environment used. Other indicators need to change dynamically according to the change amount of the test sample. When the change amount of the test sample exceeds the indicator range, it needs to be updated to the corresponding value. The update method is:

[0222]

[0223] Among them, c i is the change of indicator i in the test sample, and t is the duration of the communication behavior in the test sample.

[0224] The present invention also provides a communication interference effect online evaluation system based on improved CRITIC-GAHP, which is used to implement the communication interference effect online evaluation method. The system includes a parameter selection module, a numerical module, a variable calculation module, an objective weight calculation module, a subjective weight calculation module, a comprehensive weight calculation module, an interference effect level classification module and an online evaluation module, wherein:

[0225] The parameter selection module selects the communication party parameters as evaluation indicators;

[0226] A digitization module digitizes the anti-interference capability of the modulation pattern;

[0227] The variable calculation module calculates the changes of each evaluation index before and after the disturbance and normalizes the changes;

[0228] The objective weight calculation module uses the improved CRITIC to calculate the objective weight of each indicator based on the normalized results of each indicator;

[0229] The subjective weight calculation module uses GAHP to combine the opinions of multiple experts and calculate the subjective weight of each indicator;

[0230] The comprehensive weight calculation module combines subjective weight and objective weight according to the principle of minimum information entropy to obtain the comprehensive weight;

[0231] The interference effect level classification module calculates the interference effect score based on the comprehensive weight and normalization result, and classifies the score to obtain the interference effect level;

[0232] The online evaluation module dynamically updates the normalized parameters of each indicator according to the test samples to realize the online evaluation of the interference effect.

[0233] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0234] Example 1

[0235] In order to demonstrate the effectiveness and advantages of the minimum information entropy combined subjective and objective weights proposed in the present invention, the proposed method is compared with other methods. The comparison methods include: the present invention adopts the minimum information entropy combined subjective and objective weights, linear weighted division, multiplication weighting method, etc. The calculation methods of the linear weighted method and the multiplication weighted method are:

[0236] W=αW s +(1-α)W o (29)

[0237]

[0238] The weighting coefficient α=0.5 is obtained by minimizing the distance between the combined weight and the subjective and objective weights.

[0239] The subjective and objective weights calculated by GAHP and improved CRITIC, as well as the combined weights calculated by the three methods are shown in Table 3. Figure 3 shown.

[0240] Table 3 Results of subjective and objective weights and combined weights

[0241]

[0242] The weights of each indicator obtained by the minimum information entropy weighting method and the linear weighting method are between the subjective and objective weights, and the weights of the power, frequency and duration of communication behavior of the two are very close, indicating that both methods can take into account the subjective and objective weights well and alleviate the conflict between the two. The weights obtained by the multiplication weighting method are outside the subjective and objective weights. This is because the multiplication weighted combination weights need to multiply the corresponding subjective and objective weights, so that the weights of the indicators that account for a large proportion of the subjective and objective weights are multiplied. The combined weight is larger, and the weights that account for a small proportion are multiplied. The combined weight is smaller. For example, the frequency has the largest weight in the subjective and objective weights, which are 0.4931 and 0.277 1 respectively. After the weights are combined, the weight accounts for a larger proportion and becomes 0.594 5; in contrast, the weight of the duration of the communication behavior, the subjective and objective weight distribution is 0.059 3 and 0.129 2, and after weighting, it becomes 0.033 4. In comparison, the results of the minimum information entropy combined weight method and the linear weighting method are more reasonable. Based on the above analysis, the minimum information entropy combined weight method used in the present invention can reasonably balance the differences in subjective and objective weights and give play to the advantages of both.

[0243] Example 2

[0244] In order to demonstrate the effectiveness and advantages of the interference effect evaluation method and the method for dynamically updating normalized parameters proposed in the present invention, the proposed method is compared with other methods. The comparison methods include: the method of the present invention (OUR_dynamic), the classification method without dynamically updating normalized parameters (OUR_static), GA-ELM and GA-SVM. Among them, the GA-ELM method and GA-SVM are commonly used interference effect level classification methods. The training parameters of the two methods are shown in Table 4.

[0245] Table 4 GA-ELM and GA-SVM training parameters

[0246]

[0247] The normalized parameters of each index of the training sample of this embodiment and the normalized parameters of the index after the change are shown in Table 5. Origin represents the normalized parameters of the index of the training sample of the present invention. The middle 5 groups only change the normalized parameters of the corresponding indexes, for example, M represents only changing the normalized parameters of the modulation style. Since the normalized result of the frequency is only related to the threshold δ f Therefore, δ f From 5 to 0.5, and the frequency range is changed to 1.5-30MHz. All1 and All2 represent that the normalized parameters of all indicators have changed, and the parameter change of All2 is larger than that of All1.

[0248] Table 5 Normalization parameters of each index

[0249]

[0250] In this embodiment, 2,000 samples are generated for each group according to the parameters in Table 5, and training classification is performed on them, and the classification results are used as the real classification results. On this basis, the samples are tested using the above four methods to obtain the predicted classification results, and the real classification results are compared with the predicted classification results to obtain the classification accuracy. In order to reduce the influence of errors, this embodiment performs 1,000 Monte Carlo simulations.

[0251] Figure 4 The mean and standard deviation of the classification accuracy of 1,000 tests of four classification methods under different parameter ranges are shown. As can be seen from the figure, the classification accuracy of OUR_dynamic remains above 97.5%, which can effectively cope with the situation of changes in the indicator range. The classification accuracy of the other three methods decreased to varying degrees due to the different changes in the number and range of indicator parameters. In the All2 test group with the largest number and range of indicator parameter changes, the classification accuracy of OUR_static decreased by about 10%, and the classification accuracy of GA-ELM and GA-SVM decreased by about 20%, and the standard deviation of the test results was also larger. This is because these three methods can only classify test samples based on training experience, and cannot continuously improve their own classification methods according to changes in test samples. ELM is prone to overfitting, and SVM has difficulties in solving multi-classification tasks. Compared with the method of the present invention, it is more susceptible to changes in the range of indicator parameters.

[0252] Based on the above analysis, compared with the traditional classification method, the dynamic update normalization parameter method adopted in the present invention can effectively solve the problem of parameter range change and better improve the classification accuracy.

[0253] Example 3

[0254] The more samples there are, the higher the reliability of the calculated objective weights and the more accurate the interference effect classification results. However, the samples that the interfering party can obtain during the game with the communicating party are limited. In order to verify whether the method proposed in the present invention can be used when the number of samples is small, this experiment changes the number of training samples for experiments.

[0255] This experiment uses 1,000 samples as test data. In order to reduce the impact of errors, 1,000 Monte Carlo simulations are performed first, generating 3,000 sample data each time, and the objective weights are obtained respectively. The mean and subjective weights are combined to obtain the combined weights. After that, the interference effect level of the test samples is evaluated, and the obtained results are used as the true classification results of the test data. This experiment uses 200, 800, 1,400, 2,000, and 2,500 sample data to train the method of the present invention, GA-ELM, and GA-SVM, respectively, and then classifies the test data. The classification accuracy is obtained by comparing with the true classification results, and 1,000 Monte Carlo simulations are performed. The classification accuracy results under different numbers of training samples are shown in Figure 2. Figure 5 shown.

[0256] from Figure 5 It can be seen that as the number of training samples increases, the classification accuracy of different methods generally shows an upward trend. The classification accuracy of the method of the present invention is the highest and the standard deviation is the smallest. When the number of training samples is 200, the accuracy is above 98%. SVM is suitable for small sample classification, and an accuracy of about 92% can be achieved using 200 sample training. ELM uses neural network training, which is more susceptible to the impact of the number of training samples, has a larger standard deviation, and when the number of samples is small, the accuracy is less than 50%.

[0257] Based on the above analysis, the classification ability of the method of the present invention for unlabeled samples continues to improve with the increase in the number of collected samples, and it still has stable performance when the number of samples is small.

[0258] The above are only preferred embodiments of the present invention. It should be pointed out that, for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A communication interference effect online evaluation method based on improved CRITIC-GAHP, characterized in that: The following steps are involved: Step 1: Select the communication party parameters as evaluation indicators; Step 2: quantify the anti-interference capability of the modulation pattern; Step 3: Calculate the change of each evaluation index before and after the disturbance, and normalize the change; Step 4: Based on the normalized results of each indicator, the objective weight of each indicator is calculated using the improved weighting method, namely CRITIC; Step 5, using the group hierarchy analysis method, or GAHP, to integrate the opinions of multiple experts and calculate the subjective weight of each indicator; Step 6: According to the principle of minimum information entropy, combine the subjective weight and the objective weight to obtain the comprehensive weight; Step 7: Calculate the interference effect score based on the comprehensive weight and normalization result, and classify the score to obtain the interference effect level; Step 8: Dynamically update the normalized parameters of each indicator according to the test samples to achieve online evaluation of the interference effect.

2. According to claim 1, the communication interference effect online evaluation method based on improved CRITIC-GAHP is characterized in that: The communication party parameters selected as evaluation indicators in step 1 are as follows: The five parameters obtained by the reconnaissance system, including modulation style, low-density parity check code LDPC code rate, power, frequency, and duration of communication behavior, are selected as evaluation indicators.

3. The communication interference effect online evaluation method based on improved CRITIC-GAHP according to claim 1 is characterized in that: The anti-interference capability of the modulation pattern described in step 2 is quantified as follows: Adaptive modulation technology selects a modulation style with high throughput for communication while meeting the error rate requirement. The higher the modulation order, the greater the throughput. The modulation styles are sorted based on the bit error rate BER and modulation order, and the anti-interference capabilities of different modulation styles are mapped to a set of values.

4. The communication interference effect online evaluation method based on improved CRITIC-GAHP according to claim 1 is characterized in that: Step 3 calculates the change of each evaluation index before and after the disturbance, and normalizes the change, as follows: Calculate the changes in the evaluation indicators of modulation style, code rate, power, frequency and duration of communication behavior, and normalize the changes in the evaluation indicators to between -1 and 1. When the changes in the evaluation indicators reflect that the interference effect is invalid, normalize it to between -1 and 0; when it reflects that the interference is effective, normalize it to between 0 and 1; in this way, the interference effect score of the samples with invalid interference is less than or equal to 0, thereby distinguishing them from the samples with effective interference; The changes and normalization methods of each indicator before and after disturbance are as follows: 1) Modulation style Modulation pattern before interference m b Corresponding to SNR b , after disturbance m a Corresponding to SNR a , the change in modulation pattern before and after disturbance c m The calculation formula is: c m =snr b -snr a ,m a ≠None (1) Because the stronger the anti-interference ability of the modulation style, the smaller the corresponding SNR value, so SNR is used b -snr a To ensure that the interference is effective, the time difference is a positive number; when m a =None, the communicating parties stop communicating and directly normalize it to 1 without calculating the change amount; Normalized result of modulation pattern change m The calculation formula is: Among them, max m =max(|max cm |,|min cm |), ensure e m The result is between -1 and 1; 2) Bit rate The code rate before disturbance is r b , the code rate after interference is r a , the same as the modulation pattern setting, when the communication party stops communicating, r a =0, directly normalize the bit rate change to 1; The bit rate change before and after the disturbance c r The calculation formula is: c r ＝r a -r b ,r a ≠0 (3) The normalized result of the bit rate change is e r The calculation formula is: Among them, max r =max(|max cr |,|min cr |), ensure e r The result is between -1 and 1; 3) Power The power before the disturbance is p b , after being disturbed, it is p a , the change in power before and after the disturbance c p The calculation formula is: c p =p a -p b ,p a ≠0 (5) Measurement errors can cause fluctuations in power measurement results, so a power change decision threshold δ is set. p , if the power difference before and after the disturbance exceeds δ p , the interference is considered effective; otherwise it is considered to be a measurement error; therefore c p Need and δ p For comparison, if the threshold is not exceeded, it needs to be normalized to 0; Normalized result of power change e p The calculation formula is: Among them, max p =max(|max cp |,|min cp |), ensure e p The result is between -1 and 1; 4) Frequency The frequency before the disturbance is f b , after being disturbed, it is f a , the change in frequency before and after the disturbance c f The calculation formula is: c f =|f a -f b | (7) Taking into account the measurement error, the decision threshold δ of the frequency change is set f , the frequency has nothing to do with the anti-interference ability. When the frequency change exceeds the threshold, e f =1; otherwise e f =0; The normalized result of the frequency change is e f The calculation formula is: 5) Duration of communication behavior After the communication behavior changes, the duration t of the previous state can be obtained. Therefore, there is no need to calculate the change amount. It is only necessary to calculate the maximum and minimum duration T in the sample. max and T min Normalize the duration, and the normalized result of the communication behavior duration is e t The calculation formula is: After normalizing the values ​​of each evaluation index in the sample data, the normalized result E=(e m ,e r ,e p ,e f ,e t ).

5. The communication interference effect online evaluation method based on improved CRITIC-GAHP according to claim 4 is characterized in that: According to the normalized results of each indicator as described in step 4, the objective weight of each indicator is calculated using the improved weighting method, namely CRITIC, as follows: Step 4.1: Calculate the standard deviation and information entropy of each indicator There are p evaluation indicators in total, among which i The standard deviation of the index S i and information entropy E i The calculation formula is: Among them, x ki Represents the value corresponding to the i-th indicator in the k-th sample; is the mean of all samples corresponding to the ith index table, Step 4.2: Calculate the conflict of each indicator The calculation formula of the conflict R of each indicator is: In the formula, r ij It represents the correlation coefficient between indicators i and j, which is calculated using the Pearson correlation coefficient. The calculation formula is: Step 4.3: Calculate the weight of each indicator By standard deviation S i 、Information entropy E i and conflict R i Calculate the information content C of each indicator i and objective weight The calculation formula is: C i =(1-E i +S i )R i (14) Step 4.4, finally get the total objective weight:

6. The communication interference effect online evaluation method based on improved CRITIC-GAHP according to claim 5 is characterized in that: Step 5 described the use of group analysis hierarchy process (GAHP), combining the opinions of multiple experts, to calculate the subjective weight of each indicator, as follows: Step 5.1: Calculate the similarity of each expert First, AHP is used to determine the subjective weight vectors according to the opinions of each expert. The subjective weight W determined by the i-th expert is (i) and the subjective weight W determined by the jth expert (j) The degree of similarity d ij It is calculated using the Euclidean distance, and the calculation formula is: where d ij Satisfy ij =0,i=j;d ij =d ji ≥0; The similarity between the lth expert and other experts among L experts is d l The calculation formula is: Step 5.2: Calculate the decision weight of each expert The decision weight coefficient λ of the lth expert l The calculation formula is: Among them, d l The larger the value is, the greater the disagreement between the first expert and other experts is, and the smaller the decision weight coefficient is; l The smaller it is, the closer the opinions of the first expert are to those of other experts, and the larger the decision weight coefficient is; l =0, the opinions of the L experts are the same, so the decision weights are also the same, which is 1 / L; Step 5.3: Calculate subjective weight The weight coefficient λ is determined by experts l And each weight vector weighted to get the subjective weight The calculation formula is:

7. The communication interference effect online evaluation method based on improved CRITIC-GAHP according to claim 6 is characterized in that: According to the minimum information entropy principle described in step 6, the subjective weight and the objective weight are combined to obtain the comprehensive weight, which is as follows: The information entropy between the combined weight and the subjective weight, objective weight, and the constraint formula of the combined weight are: According to formula (21), the Lagrange equation shown in formula (22) is established: L vs w i Find the partial derivative and set the result to 0, and we get: Finally, the combined weight w is obtained i , the calculation formula is:

8. The communication interference effect online evaluation method based on improved CRITIC-GAHP according to claim 7 is characterized in that: The interference effect score is calculated based on the comprehensive weight and normalized result in step 7, and the score is graded to obtain the interference effect level, which is as follows: The interference effect is divided into 5 levels from 0 to 4, where 0 represents ineffective interference and 1 to 4 represents effective interference. The higher the level, the better the interference effect. The calculation formula for the interference effect score S and the interference effect level G is: S=EW T (25) S'=e m w m +e r w r +e p w p +e f w f (27) Among them, E is the normalized result, W is the subjective and objective combined weight, E = (e m ,e r ,e p ,e f ,e t ), e i ∈[-1,1],w i ≥0, i=m,r,p,f,t, so S∈[-1,1]; S' is the interference effect score calculated after removing the duration of the communication behavior; when S' is less than or equal to 0, the interference is judged to be invalid and the interference effect level G=0; when S' is greater than 0, the interference effect level is divided according to the overall interference effect score S.

9. The communication interference effect online evaluation method based on improved CRITIC-GAHP according to claim 8 is characterized in that: The normalized parameters of each indicator are dynamically updated according to the test sample in step 8 to realize the online evaluation of the interference effect, as follows: In actual applications, the max value is dynamically updated based on the test samples. m 、max r 、max p ,δ f , T min The normalization parameter of f Used to determine whether the frequency has changed, and set the parameter value according to the reconnaissance equipment and environment used; other indicators change dynamically according to the change of the test sample. When the change of the test sample exceeds the indicator range, the indicator is updated. The update method is: Among them, c i is the change of indicator i in the test sample, and t is the duration of the communication behavior in the test sample.

10. An online evaluation system for communication interference effects based on improved CRITIC-GAHP, characterized in that: The system is used to implement the communication interference effect online evaluation method according to any one of claims 1 to 9, and the system includes a parameter selection module, a numerical module, a variable calculation module, an objective weight calculation module, a subjective weight calculation module, a comprehensive weight calculation module, an interference effect level classification module and an online evaluation module, wherein: The parameter selection module selects the communication party parameters as evaluation indicators; A digitization module digitizes the anti-interference capability of the modulation pattern; The variable calculation module calculates the changes of each evaluation index before and after the disturbance and normalizes the changes; The objective weight calculation module uses the improved CRITIC to calculate the objective weight of each indicator based on the normalized results of each indicator; The subjective weight calculation module uses GAHP to combine the opinions of multiple experts and calculate the subjective weight of each indicator; The comprehensive weight calculation module combines subjective weight and objective weight according to the principle of minimum information entropy to obtain the comprehensive weight; The interference effect level classification module calculates the interference effect score based on the comprehensive weight and normalization result, and classifies the score to obtain the interference effect level; The online evaluation module dynamically updates the normalized parameters of each indicator according to the test samples to realize the online evaluation of the interference effect.

Images