A collaborative spectrum sensing method based on sticky hidden Markov model

Abstract

The present invention discloses a cooperative spectrum sensing method based on a viscous hidden Markov model. On the basis of the spectrum sensing model, a power matrix composed of the measured powers of signals received by multiple secondary users in multiple sensing time slots is obtained; a state transition probability matrix, an initial probability distribution, and a state parameter set of the viscous hidden Markov model are initialized, and the power matrix is ​​regarded as an observation sequence; hidden states of each sensing time slot are obtained when three model parameters and the observation sequence are known; the state transition probability matrix and the state parameter set are updated when the observation sequence and a hidden state vector composed of all hidden states are known; a final hidden state vector is obtained after multiple iterations; a test statistic of each secondary user is calculated according to the final hidden state vector, and the test statistic of each secondary user is compared with a threshold value to determine whether the secondary user can access the authorized frequency band of the primary user; the method has the advantage of having higher detection performance.

Description

Technical Field

[0001] The present invention relates to a cognitive radio technology, and in particular to a cooperative spectrum sensing method based on a sticky hidden Markov model (SHMM). Background Art

[0002] The limited radio spectrum resources and the greatly increased demand for radio spectrum resources constitute a major contradiction in the development of wireless communications today. A large number of studies have shown that the existing fixed spectrum resource allocation strategy makes it impossible to fully utilize radio spectrum resources, and fully improving the utilization rate of radio spectrum resources will effectively alleviate the above contradiction. Dynamic spectrum access technology provides a feasible solution to improve the utilization rate of radio spectrum resources and alleviate the shortage of radio spectrum resources. Dynamic spectrum access means that wireless devices can interact with the communication environment in which they are located and change their own transmission parameters according to the results of the interaction, so as to flexibly utilize potential idle spectrum in a dynamic and adaptive manner. Cognitive radio technology is a key technology for dynamic spectrum access. Its performance is affected by spectrum sensing performance and channel selection performance. Incorrect identification of spectrum opportunities, coupled with subsequent suboptimal channel selection, will lead to unnecessary delays, reduced throughput and overall performance degradation of cognitive radio networks. As the foundation and important step of cognitive radio technology, the research on spectrum sensing technology has received more and more attention.

[0003] Most existing spectrum sensing algorithms (non-cooperative) only analyze from the perspective of time domain to consider whether the primary user (PU) signal exists, such as traditional energy detection (ED), matched filter detection and feature detection algorithms. In fact, whether a secondary user (SU) can use the current frequency band depends not only on whether the primary user (PU) signal exists, but also on the distance between the secondary user (SU) and the primary user (PU). A secondary user (SU) located outside the communication range of the primary user (PU) can use the authorized frequency band of the primary user (PU) when the primary user (PU) signal exists. Traditional spectrum sensing algorithms (including cooperative spectrum sensing algorithms and non-cooperative spectrum sensing algorithms) cannot maximize spectrum resource utilization in heterogeneous spectrum environments where primary users (PU) at different locations have different access opportunities. In order to solve the spectrum sensing problem in heterogeneous spectrum environment, the literature (Han W, Li J, Li Z, Si J, et al. Spatial False Alarmin Cognitive Radio Network [j]. IEEE Transactions on Signal Processing, 2013, 61 (6): 1375-1388. (Spatial False Alarm in Cognitive Radio Networks)) proposed the concept of spatial false alarm, that is, the secondary user (SU) outside the communication range of the primary user (PU) loses the opportunity to use the frequency band because it senses that the primary user (PU) is occupied. A collaborative spectrum sensing algorithm is proposed to solve the spectrum sensing problem in heterogeneous spectrum environment by using the energy detection method to detect the received signal power of the current secondary user (SU) to determine whether the current frequency band is available. However, this collaborative spectrum sensing algorithm is more likely to generate a large number of spatial false alarms in heterogeneous spectrum environment, resulting in low detection performance. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a cooperative spectrum sensing method based on a viscous hidden Markov model, which has higher detection performance.

[0005] The technical solution adopted by the present invention to solve the above technical problems is: a collaborative spectrum sensing method based on a viscous hidden Markov model, characterized by comprising the following steps:

[0006] Step 1: In the cognitive radio system, there is a primary user and Num secondary users. From the perspective of time domain, the secondary user can dynamically access the authorized frequency band of the primary user only when the primary user is idle. From the perspective of air domain, the secondary user outside the communication range of the primary user uses the authorized frequency band of the primary user without affecting the normal communication of the primary user. Then, combining the time domain and air domain, a spectrum sensing model is constructed, which is described as: Where Num represents the number of secondary users in the cognitive radio system, Num ≥ 2. The spectrum sensing model is actually a binary hypothesis testing problem. S0 represents the situation in which the secondary user can access the authorized frequency band of the primary user in a sensing time slot. i (n) represents the nth sample value of the signal received by the i-th secondary user, 1≤i≤Num, 1≤n≤N, N represents the number of samples obtained by sampling the signal received by each secondary user, N≥25, w i (n) represents r i The noise contained in (n), w i (n) has a mean of 0 and a variance of The complex Gaussian distribution, r i (n) = w i (n) represents the situation where the i-th secondary user dynamically accesses the authorized frequency band of the primary user when the primary user is idle, and g i represents the channel gain of the ith secondary user, g i Obeying the free space loss model, h i (n) represents the channel coefficient of the i-th secondary user, h i (n) obeys Rayleigh distribution, s i (n) represents r i (n) contains the signal sent by the primary user, s i (n) has a mean of zero and a variance of The complex Gaussian distribution, d i represents the Euclidean distance from the ith secondary user to the primary user, r pu represents the communication range of the primary user, d i >r pu Indicates that the i-th secondary user is outside the communication range of the primary user, d i >r pu represents the situation where the i-th secondary user outside the communication range of the primary user uses the authorized frequency band of the primary user without affecting the normal communication of the primary user, S1 represents the situation where the secondary user cannot access the authorized frequency band of the primary user when the primary user sends a signal in a sensing time slot, and d i ≤r pu Indicates that the i-th secondary user is within the communication range of the primary user, d i ≤r pu represents the situation where the i-th secondary user located within the communication range of the primary user cannot access the authorized frequency band of the primary user;

[0007] Step 2: For consecutive T slotsensing time slots, calculate the measured power of the signal received by each secondary user in each sensing time slot, and record the measured power of the signal received by the i-th secondary user in the t-th sensing time slot as y t,i , Then the power vector composed of the measured power of the signals received by Num secondary users in the tth sensing time slot is recorded as y t ,y t =[y t,1 ,...,y t,i ,...,y t,Num ] T ; Then in T slot The power matrix composed of the measured power of the signals received by Num secondary users in a sensing time slot is denoted as Y. Among them, T slot ≥4, 1≤t≤T slot , the symbol “||” is the modulus operator, r t,i (n) represents the nth sample value of the signal received by the i-th secondary user in the t-th sensing time slot, y t,1 represents the measured power of the signal received by the first secondary user in the tth sensing time slot, y t,Num represents the measured power of the signal received by the Num-th secondary user in the t-th sensing time slot, the superscript "T" represents the transpose of the vector or matrix, y1 represents the power vector composed of the measured powers of the signals received by the Num-th secondary users in the 1st sensing time slot, Indicates that in the T slot The power vector composed of the measured power of the signals received by Num secondary users in a sensing time slot, y1, y t , The dimension of is Num×1, and the dimension of Y is Num×T slot ;

[0008] Step 3: The three model parameters of the viscous hidden Markov model, namely the state transition probability matrix, the initial probability distribution, and the state parameter set, are recorded as A, π, and θ, where θ = {θ1, θ2}; then A, π, θ1, and θ2 are initialized, and the initialization value of A is The initialization value of π is [0.5 0.5], and the initialization values ​​of θ1 and θ2 are random numbers. Then Y is regarded as the observation sequence of the viscous hidden Markov model. Among them, θ1 represents the parameter of the first type of state, and θ2 represents the parameter of the second type of state. μ1 represents the mean of the first type of state, represents the variance of the first category of states, μ2 represents the mean of the second category of states, represents the variance of the second type of state;

[0009] Step 4: When Y, A, π, and θ are known, the hidden state of each perception time slot is calculated in turn, and the hidden state of the tth perception time slot is recorded as z t , Then T slot The hidden state of each sensing time slot constitutes a hidden state vector, denoted as Z. in, Indicates that The value of z is maximum when t The value of p(z t ,Y 1:t |A,π,θ) means that when A,π,θ are known, z t , Y 1:t The probability of Indicates that z is given when A, π, and θ are known. t , The probability of Y 1:t represents the power matrix composed of the measured powers of the signals received by Num secondary users in the first sensing time slot to the tth sensing time slot, Y 1:t It also represents the 1st to tth observations in the observation sequence Y of the viscous hidden Markov model, Indicates that from the t+1th sensing time slot to the Tth slot The power matrix is ​​composed of the measured powers of the signals received by Num secondary users in a sensing time slot, It also represents the observation sequence Y of the viscous hidden Markov model from the t+1th to the Tth slot observations, z t The value of is 1 or 2, z1 represents the hidden state of the first sensing time slot, Indicates the T slot The hidden state of the sensing time slot, the dimension of Z is 1×T slot ;

[0010] Step 5: When Y and Z are known, update A and θ, and record the updated values ​​of A and θ as A * and θ * , A * The value of the element in the jth row and j'th column is denoted as a * j,j' , θ * The mean of the j-th state and the variance of the j-th state are denoted by μ j * and Where, 1≤j≤2, 1≤j'≤2, κ is the viscosity factor of the viscous hidden Markov model, δ() is the Kronecker function, when j=j', the value of δ(j,j') is 1, when j≠j', the value of δ(j,j') is 0, when j=1, the value of δ(j,1) is 1, when j≠1, the value of δ(j,1) is 0, when j=2, the value of δ(j,2) is 1, when j≠2, the value of δ(j,2) is 0, N j,j' Indicates T slot The number of occurrences of two adjacent sensing time slots in the sensing time slots where the hidden state of the previous sensing time slot is j and the hidden state of the next sensing time slot is j', N j,1 Indicates T slot The number of occurrences of two adjacent sensing time slots in the sensing time slots where the hidden state of the previous sensing time slot is j and the hidden state of the next sensing time slot is 1, N j,2 Indicates T slot The number of occurrences of two adjacent sensing slots in a sensing slot where the hidden state of the previous sensing slot is j and the hidden state of the next sensing slot is 2, express The number of elements in express The average value of all elements in represents the set of all observations corresponding to the position in Y and the position of the hidden state with value j in Z;

[0011] Step 6: Let A = A * , θ=θ * ; Then return to step 4 to continue executing. After at least 10 iterations, the obtained hidden state vector is used as the final hidden state vector, denoted as Z final , Among them, A=A * and θ = θ * The "=" in is the assignment symbol, z1 final represents the final hidden state of the first sensing time slot, z t final represents the final hidden state of the t-th sensing slot, Indicates the T slot The final hidden state of the sensing slot;

[0012] Step 7: Calculate the test statistic for each secondary user and record the test statistic of the i-th secondary user as Then the test statistic of each secondary user is compared with the threshold value to determine whether the secondary user can access the authorized frequency band of the primary user. For the i-th secondary user, if If the value is greater than the threshold, it means that the i-th secondary user cannot access the authorized frequency band of the primary user; is less than or equal to the threshold value, it means that the i-th secondary user can access the authorized frequency band of the primary user; where 1≤b≤Num, ω b represents the weight of the b-th secondary user, e is the natural base number, β0 is the fusion parameter, β0 is adjustable, Representation vector The bth element in Representation vector The i-th element in Indicates that Z is obtained final The first The mean of the class state, the threshold value is calculated based on the given false alarm probability.

[0013] The value of κ in step 5 is 0.1.

[0014] In step 7, the value of β0 is 2.0.

[0015] Compared with the prior art, the advantages of the present invention are:

[0016] The perception data of secondary users on multiple consecutive perception time slots are clustered by a clustering algorithm based on the viscous hidden Markov model, and the correlation of the perception data on consecutive perception time slots is used to reduce the impact of noise uncertainty and improve the perception performance. The current state is judged by weighted fusion of the mean of each secondary user cluster, and the perception performance is improved by the correlation of the perception data of adjacent secondary users. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 It is a flowchart of the overall implementation process of the method of the present invention;

[0018] Figure 2 The present invention and the receiver operating characteristic (ROC) curves of the existing energy detection method (ED), the existing collaborative fusion energy detection method (CED) and the existing time domain weighted fusion detection method (TF) are respectively shown. DETAILED DESCRIPTION

[0019] The present invention is further described in detail below with reference to the accompanying drawings.

[0020] The present invention proposes a collaborative spectrum sensing method based on a viscous hidden Markov model, and its overall implementation flow chart is as follows: Figure 1 As shown, it includes the following steps:

[0021] Step 1: In the cognitive radio system, there is a primary user and Num secondary users. From the perspective of time domain, the secondary user can dynamically access the authorized frequency band of the primary user only when the primary user is idle. From the perspective of air domain, the secondary user outside the communication range of the primary user uses the authorized frequency band of the primary user without affecting the normal communication of the primary user. Then, combining the time domain and air domain, a spectrum sensing model is constructed, which is described as: Wherein, Num represents the number of secondary users in the cognitive radio system, Num≥2, and in this embodiment, Num=20. The spectrum sensing model is actually a binary hypothesis testing problem. S0 represents the situation in which the secondary user can access the authorized frequency band of the primary user in a sensing time slot, r i (n) represents the nth sample value of the signal received by the i-th secondary user, 1≤i≤Num, 1≤n≤N, N represents the number of samples obtained by sampling the signal received by each secondary user, N≥25, in this embodiment, N=1000, the value of N can be considered comprehensively based on accuracy and computational complexity, in general, the larger the value of N, the higher the accuracy, but when the value of N reaches a certain value, there is little room for accuracy improvement or even no improvement, at this time, the computational complexity should be considered, and the upper limit of the value of N is also limited by the sampling device, w i (n) represents r i The noise contained in (n), w i (n) has a mean of 0 and a variance of The complex Gaussian distribution, r i (n) = w i (n) represents the situation where the i-th secondary user dynamically accesses the authorized frequency band of the primary user when the primary user is idle, and g i represents the channel gain of the ith secondary user, g i Obeying the free space loss model, h i (n) represents the channel coefficient of the i-th secondary user, h i (n) obeys Rayleigh distribution, s i (n) represents r i (n) contains the signal sent by the primary user, s i (n) has a mean of zero and a variance of The complex Gaussian distribution, d i represents the Euclidean distance from the ith secondary user to the primary user, r pu represents the communication range of the primary user, d i >r pu Indicates that the i-th secondary user is outside the communication range of the primary user, d i >r purepresents the situation where the i-th secondary user outside the communication range of the primary user uses the authorized frequency band of the primary user without affecting the normal communication of the primary user, S1 represents the situation where the secondary user cannot access the authorized frequency band of the primary user when the primary user sends a signal in a sensing time slot, and d i ≤r pu Indicates that the i-th secondary user is within the communication range of the primary user, d i ≤r pu It represents the situation that the i-th secondary user located within the communication range of the primary user cannot access the authorized frequency band of the primary user.

[0022] Step 2: For consecutive T slot sensing time slots, calculate the measured power of the signal received by each secondary user in each sensing time slot, and record the measured power of the signal received by the i-th secondary user in the t-th sensing time slot as y t,i , Then the power vector composed of the measured power of the signals received by Num secondary users in the tth sensing time slot is recorded as y t ,y t =[y t,1 ,...,y t,i ,...,y t,Num ] T ; Then in T slot The power matrix composed of the measured power of the signals received by Num secondary users in a sensing time slot is denoted as Y. Among them, T slot ≥4, in this embodiment, T slot =20, T slot If the value of is too large, the computational complexity will increase. slot If the value of is too small, the performance will be degraded. 1≤t≤T slot , the symbol “||” is the modulus operator, r t,i (n) represents the nth sample value of the signal received by the i-th secondary user in the t-th sensing time slot, y t,1 represents the measured power of the signal received by the first secondary user in the tth sensing time slot, y t,Num represents the measured power of the signal received by the Num-th secondary user in the t-th sensing time slot, the superscript "T" represents the transpose of the vector or matrix, y1 represents the power vector composed of the measured powers of the signals received by the Num-th secondary users in the 1st sensing time slot, Indicates that in the T slot The power vector composed of the measured power of the signals received by Num secondary users in a sensing time slot, y1, y t ,y TslotThe dimension of is Num×1, and the dimension of Y is Num×T slot , according to the central limit theorem y t,i It obeys the following Gaussian distribution: When the primary user does not send a signal, When sending a signal on behalf of the primary user, γ t,i represents the signal-to-noise ratio of the signal received by the i-th secondary user in the t-th sensing time slot, g t,i represents the channel gain of the ith secondary user in the tth sensing time slot, h t,i represents the average value of the channel coefficient of the ith secondary user in the tth sensing time slot, h t,i (n) represents the channel coefficient of the i-th secondary user in the t-th sensing time slot.

[0023] Step 3: The three model parameters of the viscous hidden Markov model, namely the state transition probability matrix, the initial probability distribution, and the state parameter set, are recorded as A, π, and θ, where θ = {θ1, θ2}; then A, π, θ1, and θ2 are initialized, and the initialization value of A is The initialization value of π is [0.5 0.5], and the initialization values ​​of θ1 and θ2 are random numbers. Then Y is regarded as the observation sequence of the viscous hidden Markov model. Among them, θ1 represents the parameter of the first type of state, and θ2 represents the parameter of the second type of state. μ1 represents the mean of the first type of state, represents the variance of the first category of states, μ2 represents the mean of the second category of states, represents the variance of the second class state.

[0024] Step 4: When Y, A, π, and θ are known, the hidden state of each perception time slot is calculated in turn, and the hidden state of the tth perception time slot is recorded as z t , Then T slot The hidden state of each sensing time slot constitutes a hidden state vector, denoted as Z. in, Indicates that The value of z is maximum when t The value of p(z t ,Y 1:t |A,π,θ) means that when A,π,θ are known, z t , Y 1:t The probability of Indicates that z is given when A, π, and θ are known. t , The probability of Y 1:trepresents the power matrix composed of the measured powers of the signals received by Num secondary users in the first sensing time slot to the tth sensing time slot, Y 1:t It also represents the 1st to tth observations in the observation sequence Y of the viscous hidden Markov model, Indicates that from the t+1th sensing time slot to the Tth slot The power matrix is ​​composed of the measured powers of the signals received by Num secondary users in a sensing time slot, It also represents the observation sequence Y of the viscous hidden Markov model from the t+1th to the Tth slot Observations, in p(z t ,Y 1:t |A,π,θ) is the forward information, is the backward information, z t The value of is 1 or 2, z1 represents the hidden state of the first sensing time slot, Indicates the T slot The hidden state of the sensing time slot, the dimension of Z is 1×T slot .

[0025] Step 5: When Y and Z are known, update A and θ, and record the updated values ​​of A and θ as A * and θ * , A * The value of the element in the jth row and j'th column is denoted as a * j,j' , θ * The mean of the j-th state and the variance of the j-th state are denoted by μ j * and Wherein, 1≤j≤2, 1≤j'≤2, κ is the viscosity factor of the viscosity hidden Markov model, in this embodiment, κ=0.1, δ() is the Kronecker function, when j=j', the value of δ(j,j') is 1, when j≠j', the value of δ(j,j') is 0, when j=1, the value of δ(j,1) is 1, when j≠1, the value of δ(j,1) is 0, when j=2, the value of δ(j,2) is 1, when j≠2, the value of δ(j,2) is 0, N j,j' Indicates T slot The number of occurrences of two adjacent sensing time slots in the sensing time slots where the hidden state of the previous sensing time slot is j and the hidden state of the next sensing time slot is j', N j,1 Indicates T slot The number of occurrences of two adjacent sensing time slots in the sensing time slots where the hidden state of the previous sensing time slot is j and the hidden state of the next sensing time slot is 1, N j,2 Indicates Tslot The number of occurrences of two adjacent sensing slots in a sensing slot where the hidden state of the previous sensing slot is j and the hidden state of the next sensing slot is 2, express The number of elements in express The average value of all elements in represents the set of all observations corresponding to the position in Y and the position of the hidden state with value j in Z.

[0026] Step 6: Let A = A * , θ=θ * ; Then return to step 4 to continue executing. After at least 10 iterations, the obtained hidden state vector is used as the final hidden state vector, denoted as Z final , Among them, A=A * and θ = θ * The "=" in is the assignment symbol, z1 final represents the final hidden state of the first sensing time slot, z t final represents the final hidden state of the t-th sensing slot, Indicates the T slot The final hidden state of each sensing slot.

[0027] Step 7: Calculate the test statistic for each secondary user and record the test statistic of the i-th secondary user as Then the test statistic of each secondary user is compared with the threshold value to determine whether the secondary user can access the authorized frequency band of the primary user. For the i-th secondary user, if If the value is greater than the threshold, it means that the i-th secondary user cannot access the authorized frequency band of the primary user; is less than or equal to the threshold value, it means that the i-th secondary user can access the authorized frequency band of the primary user; where 1≤b≤Num, ω b represents the weight of the b-th secondary user, e is the natural cardinality, e=2.71..., β0 is the fusion parameter, β0 is adjustable, and in this embodiment, β0=2.0, Representation vector The bth element in Representation vector The i-th element in Indicates that Z is obtained final The first The mean of the class state (i.e., the vector composed of the means of each secondary user cluster), if 1 Equal to get Z final The mean of the first type of state in θ corresponding to θ is μ1, if 2 Equal to get Z final The mean value of the second type of state in θ corresponding to the time is μ2, and the threshold value is calculated based on the given false alarm probability.

[0028] The method of the present invention clusters the perception data (i.e., the measured power of the received signal) on multiple consecutive perception time slots of secondary users through a clustering algorithm based on a viscous hidden Markov model (i.e., steps 3 to 6), and reduces the influence of noise uncertainty by utilizing the correlation of the perception data on consecutive perception time slots, thereby improving the perception performance; and judges the current state by weighted fusion of the mean of each secondary user cluster, thereby improving the perception performance by utilizing the correlation of the perception data of adjacent secondary users.

[0029] In order to verify the feasibility and effectiveness of the method of the present invention, experiments were conducted on the method of the present invention.

[0030] The existing energy detection method (ED), the existing cooperative fusion energy detection method (CED) and the existing time domain weighted fusion detection method (TF) are used as comparison methods. The ED method is cited from Han W, Li J, Li Z, Si J, et al. Spatial False Alarmin Cognitive Radio Network[j]. IEEE Transactions on Signal Processing, 2013, 61(6): 1375-1388. (Spatial false alarm in cognitive radio networks), the CED method is cited from Ma J, Zhao G, Li, Y. (.Soft Combination and Detection for Cooperative Spectrum Sensing in Cognitive Radio Networks[J]. IEEE Transactions on Wireless Communications, 2008, 7(11): 4502-4507. (Soft combination detection for cooperative spectrum sensing in cognitive radio networks), and the TF method is cited from Shen F, Ding G, Wang Z, et al. UAV-Based 3D Spectrum Sensing in Spectrum-Heterogeneous Networks[J]. IEEE Transactions on Vehicular Technology, 2019, 68(6): 5711-5722. (UAV-based 3D spectrum sensing in spectrum heterogeneous networks).

[0031] Figure 2 The receiver operating characteristic (ROC) curves of the method of the present invention, the existing energy detection method (ED), the existing collaborative fusion energy detection method (CED) and the existing time domain weighted fusion detection method (TF) are given. Figure 2 It can be seen that when the false alarm probability is 0.1, the detection probability of the method of the present invention reaches 0.96, which is 0.08 higher than that of the energy detection method and 0.53 higher than that of the collaborative fusion energy detection method, which is enough to show that the method of the present invention is significantly superior to other algorithms.

Claims

1. A collaborative spectrum sensing method based on a viscous hidden Markov model, characterized in that The following steps are involved: Step 1: In the cognitive radio system, there is a primary user and Num secondary users. From the perspective of time domain, the secondary user can dynamically access the authorized frequency band of the primary user only when the primary user is idle. From the perspective of air domain, the secondary user outside the communication range of the primary user uses the authorized frequency band of the primary user without affecting the normal communication of the primary user. Then, combining the time domain and air domain, a spectrum sensing model is constructed, which is described as: Where Num represents the number of secondary users in the cognitive radio system, Num ≥ 2. The spectrum sensing model is actually a binary hypothesis testing problem. S0 represents the situation in which the secondary user can access the authorized frequency band of the primary user in a sensing time slot. i (n) represents the nth sample value of the signal received by the i-th secondary user, 1≤i≤Num, 1≤n≤N, N represents the number of samples obtained by sampling the signal received by each secondary user, N≥25, w i (n) represents r i The noise contained in (n), w i (n) has a mean of 0 and a variance of The complex Gaussian distribution, r i (n) = w i (n) represents the situation where the i-th secondary user dynamically accesses the authorized frequency band of the primary user when the primary user is idle, and g i represents the channel gain of the ith secondary user, g i Obeying the free space loss model, h i (n) represents the channel coefficient of the i-th secondary user, h i (n) obeys Rayleigh distribution, s i (n) represents r i (n) contains the signal sent by the primary user, s i (n) has a mean of zero and a variance of The complex Gaussian distribution, d i represents the Euclidean distance from the ith secondary user to the primary user, r pu represents the communication range of the primary user, d i >r pu Indicates that the i-th secondary user is outside the communication range of the primary user, d i >r pu represents the situation where the i-th secondary user outside the communication range of the primary user uses the authorized frequency band of the primary user without affecting the normal communication of the primary user, S1 represents the situation where the secondary user cannot access the authorized frequency band of the primary user when the primary user sends a signal in a sensing time slot, and d i ≤r pu Indicates that the i-th secondary user is within the communication range of the primary user, d i ≤r pu represents the situation where the i-th secondary user located within the communication range of the primary user cannot access the authorized frequency band of the primary user; Step 2: For consecutive T slot sensing time slots, calculate the measured power of the signal received by each secondary user in each sensing time slot, and record the measured power of the signal received by the i-th secondary user in the t-th sensing time slot as y t,i , Then the power vector composed of the measured power of the signals received by Num secondary users in the tth sensing time slot is recorded as y t ,y t =[y t,1 ,…,y t,i ,…,y t,Num ] T ; Then in T slot The power matrix composed of the measured power of the signals received by Num secondary users in a sensing time slot is denoted as Y. Among them, T slot ≥4, 1≤t≤T slot , the symbol "||" is the modulus operator, r t,i (n) represents the nth sample value of the signal received by the i-th secondary user in the t-th sensing time slot, y t,1 represents the measured power of the signal received by the first secondary user in the tth sensing time slot, y t,Num represents the measured power of the signal received by the Num-th secondary user in the t-th sensing time slot, the superscript "T" represents the transpose of the vector or matrix, y1 represents the power vector composed of the measured power of the signal received by Num-th secondary users in the 1st sensing time slot, y Tslot Indicates that in the T slot The power vector composed of the measured power of the signals received by Num secondary users in a sensing time slot, y1, y t ,y Tslot The dimension of is Num×1, and the dimension of Y is Num×T slot ; Step 3: The three model parameters of the viscous hidden Markov model, namely the state transition probability matrix, the initial probability distribution, and the state parameter set, are recorded as A, π, and θ, where θ = {θ1, θ2}; then A, π, θ1, and θ2 are initialized, and the initialization value of A is The initialization value of π is [0.5 0.5], and the initialization values ​​of θ1 and θ2 are random numbers. Then Y is regarded as the observation sequence of the viscous hidden Markov model. Among them, θ1 represents the parameter of the first type of state, and θ2 represents the parameter of the second type of state. μ1 represents the mean of the first type of state, represents the variance of the first category of states, μ2 represents the mean of the second category of states, represents the variance of the second type of state; Step 4: When Y, A, π, and θ are known, the hidden state of each perception time slot is calculated in turn, and the hidden state of the tth perception time slot is recorded as z t , Then T slot The hidden state of each sensing time slot constitutes a hidden state vector, denoted as Z. in, Indicates that The value of z is maximum when t The value of p(z t ,Y 1:t |A,π,θ) means that when A,π,θ are known, z t , Y 1:t The probability of Indicates that z is given when A, π, and θ are known. t , The probability of Y 1:t represents the power matrix composed of the measured powers of the signals received by Num secondary users in the 1st sensing time slot to the tth sensing time slot, Y 1:t It also represents the 1st to tth observations in the observation sequence Y of the viscous hidden Markov model, Indicates that from the t+1th sensing time slot to the Tth slot The power matrix is ​​composed of the measured powers of the signals received by Num secondary users in a sensing time slot, It also represents the observation sequence Y of the viscous hidden Markov model from the t+1th to the Tth slot observations, z t The value of is 1 or 2, z1 represents the hidden state of the first sensing time slot, Indicates the T slot The hidden state of the sensing time slot, the dimension of Z is 1×T slot ; Step 5: When Y and Z are known, update A and θ, and record the updated values ​​of A and θ as A * and θ * , A * The value of the element in the jth row and j'th column is denoted as a * j,j' , θ * The mean of the j-th state and the variance of the j-th state are denoted by μ j * and Where, 1≤j≤2, 1≤j'≤2, κ is the viscosity factor of the viscous hidden Markov model, δ( ) is the Kronecker function, when j=j', the value of δ(j,j') is 1, when j≠j', the value of δ(j,j') is 0, when j=1, the value of δ(j,1) is 1, when j≠1, the value of δ(j,1) is 0, when j=2, the value of δ(j,2) is 1, when j≠2, the value of δ(j,2) is 0, N j,j' Indicates T slot The number of occurrences of two adjacent sensing time slots in the sensing time slots where the hidden state of the previous sensing time slot is j and the hidden state of the next sensing time slot is j', N j,1 Indicates T slot The number of occurrences of two adjacent sensing time slots in the sensing time slots where the hidden state of the previous sensing time slot is j and the hidden state of the next sensing time slot is 1, N j,2 Indicates T slot The number of occurrences of two adjacent sensing slots in a sensing slot where the hidden state of the previous sensing slot is j and the hidden state of the next sensing slot is 2, express The number of elements in express The average value of all elements in represents the set of all observations corresponding to the position in Y and the position of the hidden state with value j in Z; Step 6: Let A = A * , θ=θ * ; Then return to step 4 to continue execution. After at least 10 iterations, the obtained hidden state vector is used as the final hidden state vector, denoted as Z final , Among them, A=A * and θ = θ * The "=" in is the assignment symbol, z1 final represents the final hidden state of the first sensing time slot, z t final represents the final hidden state of the t-th sensing slot, Indicates the T slot The final hidden state of the sensing time slot; Step 7: Calculate the test statistic for each secondary user and record the test statistic of the i-th secondary user as Then the test statistic of each secondary user is compared with the threshold value to determine whether the secondary user can access the authorized frequency band of the primary user. For the i-th secondary user, if If the value is greater than the threshold, it means that the i-th secondary user cannot access the authorized frequency band of the primary user; is less than or equal to the threshold value, it means that the i-th secondary user can access the authorized frequency band of the primary user; where 1≤b≤Num, ω b represents the weight of the b-th secondary user, e is the natural base number, β0 is the fusion parameter, β0 is adjustable, Representation vector The bth element in Representation vector The i-th element in Indicates that Z is obtained final The first The mean of the class state, the threshold value is calculated based on the given false alarm probability.

2. The collaborative spectrum sensing method based on a viscous hidden Markov model according to claim 1 is characterized in that The value of κ in step 5 is 0.

1.

3. A collaborative spectrum sensing method based on a viscous hidden Markov model according to claim 1 or 2, characterized in that In step 7, the value of β0 is 2.0.

Images