Fusion and decision-making method based on multi-node spectrum data

Abstract

The invention discloses a fusion and decision-making method based on multi-node spectrum data. The method comprises the following steps: arranging a plurality of nodes in an application area, constructing a receiving signal spectrum sensing model of the nodes, constructing a receiving spectrum matrix, calculating the power of each frequency value in the receiving spectrum matrix, and constructing a receiving power matrix of the nodes by utilizing the compensated power; determining a mean value of each node in a frequency band based on the receiving power matrix of each node, and determining a threshold value of each node by using a variance based on the mean value of each node, thereby setting a weight for the receiving power matrix of each node; setting parameters required by spectrum data fusion of multiple nodes; the multi-node spectrum data is fused, the spectrum data in a certain frequency band and a certain time range is compared with a threshold value, then the idle proportion in each frequency band range is calculated, and the frequency band with the maximum idle proportion serves as the optimal frequency band to be output. The system has the characteristics of high data acquisition efficiency, high reliability and wide application range.

Description

Technical Field

[0001] The present invention relates to the field of multi-node spectrum sensing, and in particular to a decision-making method based on multi-node spectrum data fusion in a complex environment. Background Art

[0002] With the continuous advancement of wireless communication technology, 5G communication technology has brought great convenience to people's lives, but with the improvement of service quality and the advent of the big data era, people's demand for big data transmission has become more urgent. However, spectrum resources are limited, and exploring and realizing the efficient use of spectrum resources has become a key issue that needs to be solved urgently, and one of the key directions is to detect spectrum holes. Spectrum holes are detected in unoccupied frequency bands within a certain area, a certain time, and a certain frequency band. Due to the complexity and randomness of the signal detection environment and the openness of signal transmission, how to effectively detect spectrum holes and improve the accuracy and reliability of detection has become a problem that needs to be solved urgently today:

[0003] In the existing technology, the detection of spectrum holes mainly faces the following technical defects:

[0004] 1. Insufficient multi-dimensional spectrum sensing: Existing spectrum sensing theories are difficult to cover all dimensions of the electromagnetic environment and cannot obtain in-depth information such as signal characteristics. Traditional wireless communication electromagnetic environment perception theories and technologies mainly look for "spectrum holes" in the frequency domain and time domain, but lack the perception, understanding and planning of "holes" in wide-area complex electromagnetic environments such as space.

[0005] 2. Dynamic nature of spectrum sensing: The spectrum environment changes dynamically, and existing spectrum sensing technologies may find it difficult to adapt to such dynamic changes, resulting in inaccurate detection of spectrum holes.

[0006] 3. Complexity of electromagnetic situational analysis and intelligent decision-making: With the explosive growth in the number of electromagnetic frequency-using equipment and systems, the electromagnetic spectrum space has become intricate and complex, and spectrum order and safety management has become an important issue.

[0007] Therefore, how to provide a fusion and decision-making method based on multi-node spectrum data is an urgent problem that technicians in this field need to solve. Summary of the invention

[0008] The purpose of the present invention is to propose a fusion and decision-making method based on multi-node spectrum data to solve the problems existing in the prior art in the detection of spectrum holes.

[0009] In order to achieve the above tasks, the present invention adopts the following technical solutions:

[0010] A fusion and decision-making method based on multi-node spectrum data, comprising:

[0011] S1, arranging multiple nodes in the application area, constructing a spectrum perception model of the node's received signal, performing fast Fourier transform on the received signal represented by the model, and constructing a received spectrum matrix based on the transformed received signal; calculating the power of each frequency value in the received spectrum matrix, and performing gain compensation on the power, and using the compensated power to construct a received power matrix of the node;

[0012] S2, based on the received power matrix of the node, determine the mean of each node in the frequency band, and based on the mean of each node, use the variance to determine the threshold of the node, so as to set the weight for the received power matrix of each node;

[0013] S3, setting the parameters required for multi-node spectrum data fusion;

[0014] S4, fuses the spectrum data of multiple nodes, compares the spectrum data within a certain frequency band and a certain time range with the threshold, then calculates the idle ratio within each frequency band, and outputs the frequency band with the largest idle ratio as the optimal frequency band.

[0015] Furthermore, the received signal spectrum sensing model of the constructed node in S1 is expressed as:

[0016]

[0017] Where r(n) represents the signal received by the node at time n, h(m), m = 0, 1, ..., M-1 represents the discrete time channel impulse response between the transmitting signal and the receiving node, M represents the number of resolvable multipaths, s(n) represents the transmitted signal at time n, s(nm) represents the transmitted signal after the multipath delay at time m based on time n, v(m) represents the mean value is 0, and the variance is Gaussian white noise, denoted by

[0018] Furthermore, in S1, a fast Fourier transform is performed on the received signal represented by the model, and a received spectrum matrix is ​​constructed based on the transformed received signal; the power of each frequency value in the received spectrum matrix is ​​calculated, and gain compensation is performed on the power, and a received power matrix of the node is constructed using the compensated power, including:

[0019] The signal r(n) received by the node is converted to the frequency domain by fast Fourier transform, and the kth frequency value R of the received signal sampled j times after fast Fourier transform is obtained. jk ;k=1,2,…,K; j=1,2,…,J;

[0020] Construct the received spectrum matrix R based on the transformed signal:

[0021]

[0022] Where R represents the spectrum matrix of the Fast Fourier Transform FFT of K points collected each time after J samplings;

[0023] The frequency value R jk Converted to power:

[0024]

[0025] Among them, real (R jk ), imag(R jk ), respectively representing R jk The real and imaginary parts of Y jk represents the power of the kth frequency band of the jth sampling;

[0026] To compensate the power gain:

[0027] Y′ jk =Y jk +C

[0028] Among them, C is a constant, Y′ jk represents the power of the j-th sampling and k-th frequency band after gain compensation;

[0029] There are N nodes in total, so the received power matrix Y(i) of the i-th node is:

[0030]

[0031] where Y′ jk (i) represents the power of the i-th node, the j-th sampling, and the k-th frequency band.

[0032] Furthermore, the method described in S2 is based on the received power matrix of the node, determining the mean of each node in the frequency band, and determining the threshold of the node based on the mean of each node using the variance, so as to set the weight for the received power matrix of each node, including:

[0033] Calculate the mean of each frequency point in the frequency band:

[0034]

[0035] Among them, M k (i) represents the mean value of the i-th node and the k-th frequency band;

[0036] Calculate the mean of the i-th node:

[0037]

[0038] Where M(i) represents the mean of the i-th node;

[0039] Calculate the variance:

[0040]

[0041] Y′ jk (i)>M(i)+2σ is considered as a possible signal, and Y′ jk (i) The corresponding data points are removed; and the above judgment process is repeated with the remaining data. When the calculated variance σ is less than the preset reference value σ ref When , the loop ends. At this time, M(i) at the end of the loop is the threshold value γ of the i-th node i ;

[0042] Through the threshold γ of each node i The weight can be obtained, and the specific expression is as follows:

[0043]

[0044] where w i represents the weight of the i-th node;

[0045] Therefore, the node's received power matrix is ​​expressed as follows:

[0046] Y=[w1Y(1) w2Y(2) … w i Y(i)w N Y(N)].

[0047] Furthermore, the parameters required for setting the multi-node spectrum data fusion in S3 include:

[0048] In order to represent the actual frequency occupied by each frequency point, it is assumed that K frequency points occupy a total bandwidth of W:

[0049]

[0050] Among them, F interval Indicates the actual bandwidth occupied by a frequency point. Indicates rounding down X;

[0051] Assume the frequency window is B, then we can get the actual number of frequency points occupied by the frequency window:

[0052]

[0053] Set the step gap to search for the optimal frequency band:

[0054]

[0055] Among them, F gnum Indicates the actual number of frequency points occupied by the step path.

[0056] Furthermore, the spectral data of multiple nodes are integrated as described in S4, the spectral data within a certain frequency band and a certain time range are compared with the threshold, and then the idle ratio within each frequency band is calculated, including:

[0057] For the J×K receiving power matrix Y of N nodes, using the weight w i The final weighted matrix U can be obtained:

[0058]

[0059] where w i is the weight, Y(i) is the received power matrix, U jk It represents the weighted power sum of N nodes in the jth sampling and kth frequency band;

[0060] First divide it according to the frequency window:

[0061]

[0062] The matrix Z after partitioning col Contains rows 1 to J in the weight matrix U, (col-1)×F gnum +1 to (col-1)×F gnum +F wnum The elements of the column, represents the Jth row and the Jth row in the weight matrix U The elements of the column, col = 1, 2, 3, ... C, let

[0063] Then divide the time slots into Z col Divide into D sub-matrices:

[0064]

[0065] The sub-matrix Represents Z col The elements from ((row-1)*10+1) to (row*10) rows and all columns; Represents Z col In the middle row ((row-1)×10+1), the Column data;

[0066] Determine whether the sub-matrix block is busy or empty:

[0067] First, we get the submatrix The mean

[0068]

[0069] Then The mean With threshold γ i If the value is less than the threshold, it is marked as "1", indicating idle; otherwise, it is marked as "0", indicating busy:

[0070]

[0071] in, It is used to determine whether the spectrum resources in a certain frequency band and within a certain period of time are occupied;

[0072] Get the idle ratio within each frequency band:

[0073]

[0074] Among them, V rate (r) represents the idle ratio of the rth frequency band, and the frequency window F wnum Take the path F gnum Slide and repeat the above process of calculating the idle ratio until the idle ratios of C frequency bands are obtained.

[0075] Furthermore, the number of sub-matrices

[0076] Furthermore, the step of outputting the frequency band with the largest idle rate as the optimal frequency band includes:

[0077] Find the frequency band with the largest idle ratio and record the idle ratio value and index:

[0078] R max =Max{V rate (1) V rate (2)…V rate (r)…V rate (C-1), V rate (c)}

[0079] Index=Max_Index{V rate (1) V rate (2)…V rate (r)…V rate (C-1), V rate (C)}

[0080] Among them, R max Indicates the value corresponding to the maximum idle ratio, and Index indicates the index corresponding to the maximum idle ratio;

[0081] Output optimal frequency band:

[0082]

[0083] where F interval Indicates the actual bandwidth occupied by a frequency point, Start freq Indicates the starting frequency of the received signal. Start and End indicate the recommended starting frequency and ending frequency respectively.

[0084] Furthermore, if V rate (r) = V rate (r+s), where s=0,1,…,Cr, then after removing the busy sub-matrix blocks, calculate the idle mean Mean′(r) of the rth frequency band; if Mean′(r)>Mean′(r+s), then it is considered that V rate (r) <V rate (r+s).

[0085] A terminal device includes a processor, a memory, and a computer program stored in the memory; when the processor is executed by a computer, the fusion and decision-making method based on multi-node spectrum data is implemented.

[0086] A computer-readable storage medium stores a computer program; when the computer program is executed by a processor, the fusion and decision-making method based on multi-node spectrum data is implemented.

[0087] Compared with the prior art, the present invention has the following technical features:

[0088] 1. The present invention can restore the spectrum environment more realistically through multi-node spectrum data, especially for spectrum perception in complex electromagnetic environments. In natural disasters or other emergency situations, traditional communication networks may be damaged. The present invention can quickly map out the currently available spectrum resources to ensure that communications for rescue operations and critical tasks will not be interfered with. In the construction of smart cities, a large number of devices need to be interconnected. The present invention can provide efficient spectrum resource management for IoT devices and ensure the reliability and security of data transmission.

[0089] 2. The present invention adopts advanced data fusion algorithms that can integrate spectrum data from different geographical locations and different types of sensors. Through this fusion, the system can integrate the observation data of each node, reduce the errors and deviations of single-point measurements, and improve the overall perception accuracy. Through parallel processing technology, the present invention can simultaneously analyze data from multiple nodes, significantly improving the processing speed. Compared with traditional serial processing methods, the parallel processing strategy of the present invention reduces data processing time and achieves near real-time spectrum perception. For scenarios with high real-time requirements and high accuracy requirements for spectrum perception, such as: detection of illegal drones, illegal occupation and purchase of frequency bands, etc., reliable technical support is provided.

[0090] 3. The present invention can flexibly adjust the frequency window and frequency step of the detection to achieve spectrum perception under different bandwidths. The present invention can adjust the size of the frequency window according to the current network status (such as spectrum congestion) and user needs (such as different service quality requirements). This means that in areas or times where the spectrum is intensively used, the system can select a finer frequency window to improve the accuracy of perception, and vice versa, the window can be expanded to quickly cover a wider spectrum range. BRIEF DESCRIPTION OF THE DRAWINGS

[0091] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0092] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner, and therefore only show the components related to the present invention.

[0093] The spectrum data of multiple nodes can restore the electromagnetic environment more realistically, especially for spectrum perception in complex electromagnetic environments. In natural disasters or other emergency situations, traditional communication networks may be damaged. The present invention can quickly map out the currently available spectrum resources to ensure that rescue operations and communications for critical tasks will not be interfered with. In addition, the present invention adopts data fusion algorithms that can integrate spectrum data from different geographical locations and different types of sensors. Through this fusion, the system can integrate the observation data of each node, reduce the error and deviation of single-point measurement, and improve the overall perception accuracy. Through parallel processing technology, the present invention can simultaneously analyze data from multiple nodes, significantly improving the processing speed. Finally, the present invention can flexibly adjust the frequency window and frequency step of the detection to achieve spectrum perception under different bandwidths. The present invention can adjust the size of the frequency window according to the current network status (such as spectrum congestion) and user needs (such as different service quality requirements). This means that in areas or times where the spectrum is intensively used, the system can select a finer frequency window to improve the accuracy of perception, and vice versa, the window can be expanded to quickly cover a wider spectrum range. Therefore, the present invention has the advantages of wide application scenarios, low complexity, and high accuracy.

[0094] See also Figure 1The present invention provides a fusion and decision-making method based on multi-node spectrum data. In terms of hardware requirements, the present invention requires the use of a device that can receive signals, such as a software-defined radio (SDR) device available on the market, such as HackRF, etc. For the convenience of description, the device that receives the signal is referred to as a node here. First, a certain number of nodes are arranged in the targeted application area according to actual application requirements. The specific number and location of the nodes should be determined based on detailed on-site investigations and spectrum usage analysis; then, the spectrum data of each node is uploaded to the central processing unit (a machine with data processing and computing capabilities, such as a laptop computer, etc.). After the central processing unit receives the spectrum data uploaded by multiple nodes, it uses a fusion and decision-making method based on multi-node spectrum data to perform data fusion and analysis to identify the optimal frequency band (the most idle frequency band) and deploy it to each node to provide a high-quality communication environment for the subsequent period; the method specifically includes the following steps:

[0095] S1, arrange multiple nodes in the application area, build a spectrum perception model of the node's received signal, perform fast Fourier transform on the received signal represented by the model, and build a received spectrum matrix based on the transformed received signal; calculate the power of each frequency value in the received spectrum matrix, and perform gain compensation on the power, and use the compensated power to build a received power matrix of the node.

[0096] In this implementation, S1 specifically includes:

[0097] S11, in the present invention, the multipath fading environment that the signal may encounter during the transmission process is taken into account. This environment will have a significant impact on the propagation of the signal, resulting in a decrease in the received signal strength and quality. Assuming that the signal passes through a multipath fading environment from the transmitter to the node reception, a spectrum perception model of the node receiving signal is established:

[0098]

[0099] Among them, r(n) represents the signal received by the node at time n. This signal may include various unknown situations such as WiFi, Bluetooth, mobile phone signals, radio stations, etc. h(m), m = 0, 1, ..., M-1 represents the discrete time channel impulse response between the transmitting signal and the receiving node, M represents the number of distinguishable multipaths, which is related to the multipath fading environment, s(n) represents the user's transmitted signal at time n, s(nm) represents the user's transmitted signal after the multipath delay at time m based on time n, v(m) represents the mean value is 0, and the variance is Gaussian white noise, denoted by

[0100] S12, performing a fast Fourier transform (FFT) on the signal r(n) received by the node to convert it into the frequency domain for frequency domain analysis:

[0101]

[0102] Where K represents the number of frequency bands in one sampling, J represents the total number of sampling times, which is usually a multiple of 10 to facilitate the calculation of subsequent algorithms, and R jk It is the kth frequency value of the jth sampled received signal after fast Fourier transform (FFT), i is an imaginary unit, and e is a natural constant.

[0103] S13, construct the received spectrum matrix R based on the transformed signal:

[0104]

[0105] Where R represents the frequency spectrum matrix of the Fast Fourier Transform FFT after J samplings, with K points collected each time.

[0106] S14, since the result obtained by fast Fourier transform is the amplitude of the received signal, in order to intuitively see the loss and amplification of the signal during the transmission process, it needs to be converted into power:

[0107]

[0108] Among them, real (R jk ), imag(R jk ), respectively representing R jk The real and imaginary parts of Y jk Represents the power of the kth frequency band of the jth sampling.

[0109] S15, the actual received signal is attenuated, and gain compensation is used to enhance the signal detectability:

[0110] Y′ jk =Y jk +C

[0111] Among them, C is a constant, which is obtained based on actual measurements in different application scenarios, and Y′ jk Represents the power of the j-th sampling and k-th frequency band after gain compensation.

[0112] S16, assuming there are N nodes in total, then the receiving power matrix Y(i) of the i-th node is:

[0113]

[0114] where Y′ jk (i) represents the power of the i-th node, the j-th sampling, and the k-th frequency band.

[0115] S17, then we can get the received power matrix Y of N nodes:

[0116] Y=[w1Y(1) w2Y(2) … w i Y(i)w N Y(N)]

[0117] where w i represents the weight of the i-th node, which will be solved in S2.

[0118] S2, based on the received power matrix of the node, determine the mean of each node in the frequency band, and based on the mean of each node, use the variance to determine the threshold of the node, so as to set the weight for the received power matrix of each node.

[0119] In this step, the threshold and weight of each node are obtained based on the multi-node spectrum sensing model. The node threshold is determined mainly based on the characteristics of low noise power and high signal power. The core idea is to approach the background noise by calculating the mean and variance multiple times, and then determine the node threshold; the node weight is determined by the threshold; the node threshold and weight will be described in detail in S4.

[0120] In this implementation, S2 specifically includes:

[0121] S21, calculate the mean value of each frequency point in the frequency band:

[0122]

[0123] Among them, M k (i) represents the mean value of the i-th node and the k-th frequency band.

[0124] S22, calculate the mean of the i-th node:

[0125]

[0126] Where M(i) represents the mean of the i-th node.

[0127] S23, calculate the variance:

[0128]

[0129] Where σ is the variance, describing M k (i) and M(i); at this time, if Y′ jk (i)>M(i)+2σ can be regarded as a possible signal, and Y′ jk (i) The corresponding data points are removed; the calculation process of S21-S23 is repeated with the remaining data. When the calculated variance σ is less than a reference value σ specified by the user, ref When , the loop ends. Generally speaking, the reference variance σref =30; at this time, M(i) at the end of the loop is the threshold γ of the i-th node i , which is explained in detail in S3.

[0130] S24, through the threshold γ of each node i The weight can be obtained, and the specific expression is as follows:

[0131]

[0132] where w i represents the weight of the i-th node.

[0133] S3, setting the parameters required for multi-node spectrum data fusion, including the actual bandwidth occupied by a frequency point, the frequency window and the actual number of frequency points occupied by the frequency window, the step size and the actual number of frequency points occupied by the step size.

[0134] In this implementation, S3 specifically includes:

[0135] S31, in order to indicate the actual frequency occupied by each frequency point, assume that K frequency points occupy a total bandwidth of W (MHz):

[0136]

[0137] Among them, F interval Indicates the actual bandwidth occupied by a frequency point. Indicates that X is rounded down.

[0138] S32, in order to make the collected data meet the actual needs, it is necessary to splice the collected data into data of specified bandwidth. In order to distinguish it from the bandwidth in S31, the bandwidth specified by the user is called frequency window. Assuming that the frequency window is B (MHz), the actual number of frequency points occupied by the frequency window can be obtained:

[0139]

[0140] S33, this algorithm also needs to set the step size Gap (MHz) to search for the optimal frequency band:

[0141]

[0142] Among them, F gnum Indicates the actual number of frequency points occupied by the step path.

[0143] S4, fuses the spectrum data of multiple nodes, compares the spectrum data within a certain frequency band and a certain time range with the threshold, then calculates the idle ratio within each frequency band, and outputs the frequency band with the largest idle ratio as the optimal frequency band.

[0144] In this implementation, S4 specifically includes:

[0145] S41, for the J×K receiving power matrix Y of N nodes, using weight w i The final weighted matrix U can be obtained:

[0146]

[0147] where w i is the weight obtained by S24, Y(i) is the received power matrix in S16, It represents the weighted power sum of N nodes in the jth sampling and kth frequency band.

[0148] S42, first divide it according to the frequency window:

[0149]

[0150] Where Z col Contains rows 1 to J in the weight matrix U, (col-1)×F gnum +1 to (col-1)×F gnum +F wnum The elements of the column, represents the Jth row and the Jth row in the weight matrix U The elements of the column, col = 1, 2, 3, ... C, let Frequency window F wnum From S32, step F gnum Obtained from S33.

[0151] S43, then divide the time slots into Z col Divide into D 10×1 sub-matrices, generally J is a multiple of 10:

[0152]

[0153] in Represents Z col The elements of rows ((row-1)*10+1) to (row*10) and all columns in ; for example Represents Z col In the middle row ((row-1)×10+1), the Column data; that is represents the data of a certain frequency band and a certain time. For the convenience of subsequent description, it is called a submatrix. The number of submatrices D is given by get.

[0154] S44, determining whether the sub-matrix block is busy or empty.

[0155] Default Submatrix The data in is relatively stable, that is, the data fluctuation is small. First, we get the submatrix The mean

[0156]

[0157] S45, then The mean With threshold γ i If the value is less than the threshold, it is marked as "1", indicating idle; otherwise, it is marked as "0", indicating busy:

[0158]

[0159] Among them, the threshold γ i From S21-S23, It is used to judge whether the spectrum resources in a certain frequency band and within a certain time are occupied, that is, whether the data corresponding to the sub-matrix is ​​occupied.

[0160] S46, get the idle ratio within each frequency band:

[0161]

[0162] Among them, V rate (r) represents the idle ratio of the rth frequency band, and the frequency window F wnum Take the path F gnum Slide and repeat S42 to S46 until the idle ratios of C frequency bands are obtained.

[0163] S47, find the frequency band with the largest idle ratio, and record the idle ratio value and index:

[0164] R max =Max{V rate (1) V rate (2)…V rate (r)…V rate (C-1), V rate (C)}

[0165] Index=Max_Index{V rate (1) V rate (2)…V rate (r)…V rate (C-1), V rate (C)}

[0166] Among them, R maxIndicates the value corresponding to the maximum idle ratio, and Index indicates the index corresponding to the maximum idle ratio. rate (r) = V rate (r+s), that is, the idle ratios of the rth and r+sth frequency bands are the same, where s = 0, 1, ..., Cr, then the busy sub-matrix blocks are removed ( After the sub-matrix block marked as 0), the idle mean Mean′(r) of the r-th frequency band can be expressed by If Mean′(r)>Mean′(r+s), then V rate (r) <V rate (r+s).

[0167] S48, output optimal frequency band:

[0168]

[0169] Among them, F interval Indicates the actual bandwidth occupied by a frequency point, obtained from S31, Start freq Indicates the starting frequency of the received signal. Start and End respectively indicate the recommended starting frequency and ending frequency obtained based on this method.

[0170] Example

[0171] In order to verify the effectiveness and practical value of the present invention, we applied the present invention to a spectrum sensing experiment in a downtown commercial district. The main challenge of this experiment is the complex electromagnetic environment, which includes a large number of people and tall buildings, which bring great interference to spectrum detection. The purpose of the present invention is to improve the utilization rate of spectrum resources and the accuracy and real-time performance of spectrum detection through a data fusion and decision-making method with low complexity.

[0172] This experiment was deployed in the commercial district of the city center and lasted for one week. The multi-node data fusion and decision-making method of the present invention significantly improved the accuracy and real-time performance of spectrum detection. The following is a comparison of data before and after the application of the present invention to demonstrate the effect of the present invention:

[0173] Table 1 Spectrum sensing comparison test

[0174] Performance Indicators Traditional methods The novel method of the present invention Spectrum Utilization 70.3％ 81.5％ Interference rate 12.8％ 4.1％ Accuracy 82.7％ 91.6％ System response time 8.3s 4.8s

[0175] Spectrum utilization measures the ratio of effectively used spectrum resources to total available spectrum resources. From the data, the spectrum utilization of the new method of the present invention has increased by 11.2 percentage points, which shows that the new method can more effectively utilize the available spectrum and reduce spectrum waste, which is particularly important in an environment with tight spectrum resources. The interference rate reflects the interference caused to other users (such as primary users) due to inaccurate spectrum sensing. The new method of the present invention has significantly reduced the interference rate by 8.7 percentage points, showing that it has significantly improved the accuracy of spectrum sensing and interference avoidance. This is crucial to ensure the fairness and efficiency of communication systems. The accuracy rate measures the ability of the system to identify idle spectrum. Improving the accuracy rate means that the system has a lower misjudgment rate during spectrum sensing and can more accurately identify which frequency bands are available. The new method of the present invention has improved by nearly 9 percentage points compared with the traditional method, which can not only reduce resource waste, but also reduce interference to other communication systems. The system response time refers to the time interval from detecting idle spectrum to actually starting to use the spectrum. The response time of the new method of the present invention has been reduced by almost half, from 8.3 seconds to 4.8 seconds. This improvement significantly improves the flexibility of the system and the user experience, allowing the system to adapt to environmental changes more quickly and provide continuous services.

[0176] In summary, the new method of the present invention has significant improvements in key performance indicators such as spectrum utilization, interference rate, accuracy and system response time. These improvements make the system likely to have higher operating efficiency and better user experience in practical applications, and are particularly suitable for advanced communication networks that require highly dynamic spectrum management and high service quality. These results show that the new method has obvious technical advantages and is of great significance for promoting the development of cognitive radio and other intelligent communication technologies.

[0177] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A fusion and decision-making method based on multi-node spectrum data, characterized in that: include: S1, arranging multiple nodes in the application area, constructing a received signal spectrum perception model of the node, performing fast Fourier transform on the received signal represented by the model, and constructing a received spectrum matrix based on the transformed received signal; Calculate the power of each frequency value in the received spectrum matrix, perform gain compensation on the power, and use the compensated power to construct the node's receiving power matrix; S2, based on the received power matrix of the node, determine the mean of each node in the frequency band, and based on the mean of each node, use the variance to determine the threshold of the node, so as to set the weight for the received power matrix of each node; S3, setting the parameters required for multi-node spectrum data fusion; S4, fuses the spectrum data of multiple nodes, compares the spectrum data within a certain frequency band and a certain time range with the threshold, then calculates the idle ratio within each frequency band, and outputs the frequency band with the largest idle ratio as the optimal frequency band.

2. The fusion and decision-making method based on multi-node spectrum data according to claim 1 is characterized in that: The received signal spectrum sensing model of the constructed node in S1 is expressed as: Where r(n) represents the signal received by the node at time n, h(m), m = 0, 1, ..., M-1 represents the discrete time channel impulse response between the transmitting signal and the receiving node, M represents the number of resolvable multipaths, s(n) represents the transmitted signal at time n, s(nm) represents the transmitted signal after the multipath delay at time m based on time n, v(m) represents the mean value is 0, and the variance is Gaussian white noise, denoted by 3. The fusion and decision-making method based on multi-node spectrum data according to claim 1 is characterized in that: In S1, a fast Fourier transform is performed on the received signal represented by the model, and a received spectrum matrix is ​​constructed based on the transformed received signal; Calculate the power of each frequency value in the received spectrum matrix, perform gain compensation on the power, and use the compensated power to construct the node's receiving power matrix, including: The signal r(n) received by the node is converted to the frequency domain by fast Fourier transform, and the kth frequency value R of the received signal sampled j times after fast Fourier transform is obtained. jk ;k=1,2,…,K; j=1,2,…,I; Construct the received spectrum matrix R based on the transformed signal: Where R represents the spectrum matrix of the Fast Fourier Transform FFT of K points collected each time after J samplings; The frequency value R jk Converted to power: Among them, real (R jk ), imag(R jk ), respectively representing R jk The real and imaginary parts of Y jk represents the power of the kth frequency band of the jth sampling; To compensate the power gain: AND' jk =And jk +C Among them, C is a constant, Y′ jk represents the power of the j-th sampling and k-th frequency band after gain compensation; There are N nodes in total, so the received power matrix Y(i) of the i-th node is: where Y′ jk (i) represents the power of the i-th node, the j-th sampling, and the k-th frequency band.

4. The fusion and decision-making method based on multi-node spectrum data according to claim 1 is characterized in that: The method described in S2 is to determine the mean of each node in the frequency band based on the received power matrix of the node, and to determine the threshold of the node based on the mean of each node using the variance, so as to set the weight for the received power matrix of each node, including: Calculate the mean of each frequency point in the frequency band: Among them, M k (i) represents the mean value of the i-th node and the k-th frequency band; Calculate the mean of the i-th node: Where M(i) represents the mean of the i-th node; Calculate the variance: Y′ jk (i)>M(i)+2σ is considered as a possible signal, and Y′ jk (i) The corresponding data points are removed; and the above judgment process is repeated with the remaining data. When the calculated variance σ is less than the preset reference value σ ref When , the loop ends. At this time, M(i) at the end of the loop is the threshold value γ of the i-th node i ; Through the threshold γ of each node i The weight can be obtained, and the specific expression is as follows: where w i represents the weight of the i-th node; Therefore, the node's received power matrix is ​​expressed as follows: Y=[w1Y(1) w2Y(2)…w i Y(i) w N Y(N)]。 5. The fusion and decision-making method based on multi-node spectrum data according to claim 1 is characterized in that: The parameters required for setting the multi-node spectrum data fusion described in S3 include: In order to represent the actual frequency occupied by each frequency point, it is assumed that K frequency points occupy a total bandwidth of W: Among them, F interval Indicates the actual bandwidth occupied by a frequency point. Indicates rounding down X; Assume the frequency window is B, then we can get the actual number of frequency points occupied by the frequency window: Set the step gap to search for the optimal frequency band: Among them, F gnum Indicates the actual number of frequency points occupied by the step path.

6. The fusion and decision-making method based on multi-node spectrum data according to claim 1 is characterized in that: As described in S4, the spectrum data of multiple nodes are integrated, the spectrum data within a certain frequency band and a certain time range are compared with the threshold, and then the idle ratio within each frequency band is calculated, including: For the J×K receiving power matrix Y of N nodes, using the weight w i The final weight matrix U can be obtained: where w i is the weight, Y(i) is the received power matrix, U jk It represents the weighted power sum of N nodes in the jth sampling and kth frequency band; First divide it according to the frequency window: The matrix Z after partitioning col Contains rows 1 to J in the weight matrix U, (col-1)×F gnum +1 to (col-1)×F gnum +F wnum The elements of the column, represents the Jth row and the Jth row in the weight matrix U The elements of the column, col = 1, 2, 3, ... C, let Then divide the time slots into Z col Divide into D sub-matrices: The sub-matrix Represents Z col The elements from ((row-1)*10+1) to (row*10) rows and all columns; Represents Z col In the middle row ((row-1)×10+1), the Column data; Determine whether the sub-matrix block is busy or empty: First, we get the submatrix The mean Then The mean With threshold γ i Compare, if it is less than the threshold, it is marked as "1", indicating idle; otherwise it is marked as "0", indicating busy: in, It is used to determine whether the spectrum resources in a certain frequency band and within a certain period of time are occupied; Get the idle ratio within each frequency band: Among them, V rate (r) represents the idle ratio of the rth frequency band, and the frequency window F wnum Take the path F gnum Slide and repeat the above process of calculating the idle ratio until the idle ratios of C frequency bands are obtained.

7. The fusion and decision-making method based on multi-node spectrum data according to claim 6 is characterized in that: The number of sub-matrices 8. The fusion and decision-making method based on multi-node spectrum data according to claim 1 is characterized in that: The step of outputting the frequency band with the largest idle rate as the optimal frequency band includes: Find the frequency band with the largest idle ratio and record the idle ratio value and index: R max =Max{V rate (1)、V rate (2)…V rate (r)…V rate (C-1)、V rate (C)} Index=Max_Index{V rate (1)、V rate (2)…V rate (r)…V rate (C-1)、V rate (C)} Among them, R max Indicates the value corresponding to the maximum idle ratio, and Index indicates the index corresponding to the maximum idle ratio; Output optimal frequency band: Among them, F interval Indicates the actual bandwidth occupied by a frequency point, Start freq Indicates the starting frequency of the received signal. Start and End indicate the recommended starting frequency and ending frequency respectively.

9. The fusion and decision-making method based on multi-node spectrum data according to claim 8, characterized in that: If V rate (r) = V rate (r+s), where s=0,1,…,Cr, then after removing the busy sub-matrix blocks, calculate the idle mean Mean′(r) of the rth frequency band; if Mean′(r)>Mean′(r+s), then it is considered that V rate (r) <V rate (r+s).

10. A computer-readable storage medium, wherein a computer program is stored in the medium; when the computer program is executed by a processor, the fusion and decision-making method based on multi-node spectrum data according to any one of claims 1-9 is implemented.

Images