Self-adaptive compressed spectrum sensing method based on deterministic evaluation model

Abstract

The invention belongs to the technical field of cognitive radio based on compressed spectrum sensing, and particularly relates to a self-adaptive compressed spectrum sensing method based on a deterministic evaluation model. Firstly, a new signal reconstruction algorithm is provided in each sensing interval of adaptive compressed spectrum sensing, and the algorithm integrates priori knowledge into l2, 1 norm minimization of block sparse signals. Secondly, deducing a CDF of a reconstruction signal error, and taking the CDF as a stopping criterion of observation sample collection; and finally, processing the current reconstructed frequency spectrum signal by using an energy detection method to obtain a corresponding frequency spectrum use state. According to the method, the accuracy of spectrum sensing is improved, and a deterministic guarantee is provided for a sensing result.

Description

Technical Field

[0001] The present invention belongs to the technical field of cognitive radio based on compressed spectrum sensing, and in particular relates to an adaptive compressed spectrum sensing method based on a deterministic evaluation model. Background Art

[0002] Cognitive Radio (CR) is a flexible and intelligent wireless communication technology that aims to improve the utilization efficiency of wireless spectrum and solve the problem of scarce and unbalanced spectrum resource allocation. Traditional wireless communication systems rely on fixed spectrum allocation, which is inefficient in the context of growing spectrum demand. Cognitive radio can provide higher spectrum utilization without interfering with existing users by enabling wireless devices to dynamically sense, identify and utilize idle spectrum resources.

[0003] Compressed spectrum sensing, as an important part of cognitive radio technology, aims to quickly and efficiently detect and identify available spectrum through compressed sensing (CS) methods. Traditional spectrum sensing methods usually require long sampling times and high signal-to-noise ratios, which can lead to lower accuracy and real-time performance in dynamic environments and high interference conditions. Spectrum sensing based on compressed sensing can restore spectrum information through sparse representation at low sampling rates, greatly improving the speed and efficiency of the sensing process, especially in scenarios with limited resources and high real-time requirements.

[0004] With the continuous development of wireless communication technology, cognitive radio technology combining intelligent algorithms and compressed sensing is becoming one of the key technologies to improve spectrum sensing efficiency and cope with challenges in complex environments. It can not only realize spectrum sensing in traditional static environments, but also provide more accurate and timely spectrum information in dynamic and complex communication environments, such as high-speed mobile communication scenarios, providing theoretical basis and technical support for the design and optimization of next-generation wireless communication systems.

[0005] Adaptive Compressed Spectrum Sensing (ACSS) plays an important role in cognitive radio networks because it can reduce sampling rate and power consumption. Most existing ACSS schemes focus on improving spectrum sensing performance but do not provide deterministic guarantees of sensing results. Summary of the invention

[0006] Aiming at the problems existing in the prior art, the present invention designs an adaptive compressed spectrum sensing (ACSS) method based on a deterministic estimation model. By deriving a closed-form expression of the cumulative distribution function (CDF) of the reconstruction error, the confidence level of the reconstructed signal is provided. Specifically, firstly, in each perception interval of ACSS, the present invention proposes a novel signal reconstruction algorithm, which incorporates prior knowledge into the l 2,1 norm is minimized. Secondly, the CDF of the reconstruction error is derived and used as the stopping criterion for collecting observation samples. Finally, the energy detection method is used to process the reconstructed spectrum signal at the current moment to obtain the corresponding binary spectrum occupancy state.

[0007] Technical solution of the present invention:

[0008] An adaptive compressed spectrum sensing method based on a deterministic estimation model comprises the following steps:

[0009] Step 1. Based on l 2，1 The signal reconstruction model combining norm minimization and prior knowledge mining reconstructs the spectrum signal: After the secondary user SU compresses and samples the frequency domain signal, it uses l 2，1 The norm minimization solves the broadband signal, which is used as the prior knowledge and l 2，1 The norm minimization is combined with solving the spectrum signal of the current time slot to reconstruct the signal.

[0010] Step 2. Derive the CDF of the reconstructed signal error and use it as the stopping criterion for observation sample collection to achieve adaptive stopping of the perception process: if the CDF less than the preset error tolerance is greater than the preset probability value, stop sampling and obtain the reconstructed spectrum signal; otherwise, continue the next sampling process.

[0011] Step 3: Use the energy detection method to make spectrum judgment and obtain the spectrum occupation and idleness.

[0012] The beneficial effects of the present invention are as follows:

[0013] The present invention provides the certainty of compressed spectrum sensing results, realizes broadband spectrum sensing with adaptively adjusted sampling rate and reliable confidence of sensing results, ensures the certainty of sensing results while improving the accuracy of spectrum sensing, and improves the spectrum supply capacity of cognitive radio networks. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a process flow chart of the method of the present invention;

[0015] Figure 2 Schematic diagram of communication time frame division;

[0016] Figure 3 It is a schematic diagram of block sparse signal;

[0017] Figure 4 The spectrum sensing performance comparison between the ACSS-DEM method of the present invention and other algorithms is shown;

[0018] Figure 5 CDF curves of different numbers of sparse blocks in the embodiment of the present invention (number of sparse blocks: (a) 4, (b) 5, (3) 6, (4) 7);

[0019] Figure 6 NMSE comparison of reconstructed signals with different numbers of sparse blocks in the embodiments of the present invention (number of sparse blocks: (a) 4, (b) 5, (c) 6, (d) 7). DETAILED DESCRIPTION

[0020] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0021] Example

[0022] The present invention is an adaptive compressed spectrum sensing method based on a deterministic evaluation model, and the specific steps are as follows: (e.g. Figure 1 )

[0023] Step 1. Based on l 2，1 The method of combining norm minimization and prior knowledge mining to reconstruct the spectrum signal: After the secondary user SU (Secondary User) compresses and samples the frequency domain signal, it uses l 2，1 The norm minimization solves the broadband signal, which is used as the prior knowledge and l 2，1 The norm minimization is combined with solving the spectrum signal of the current time slot to reconstruct the signal.

[0024] The prior knowledge is only l 2，1 A broadband spectral signal for which norm minimization is solved.

[0025] Step (11) Figure 2 , using periodic compressed spectrum sensing, each communication time frame T is divided into a sensing time slot (duration: Ts), followed by a data transmission time slot Td, (i.e., T = Ts + Td). In the sensing stage, the total interval Ts is evenly divided into J time slots, and the length of each time slot τ = Ts / J. In each time slot τ, the secondary user samples the time domain signal x(t) at the same sampling rate and obtains ΔM compressed measurement values, where ΔM is the number of samples sampled in each time slot. After j times (j represents the number of sampling) of sensing processes, the signal is reconstructed using jΔM observation samples.

[0026] Specifically, the secondary user has an N×1 frequency domain signal Compressed sampling is performed, and the compressed observation can be expressed as

[0027]

[0028] Among them, Φ represents the observation matrix of jΔM×N, n~N(0,σ 2 I jΔM ) represents the observation noise.

[0029] Utilize l 2，1 The norm minimization model is used for the signal Reconstruct the reconstructed spectrum signal It can be expressed as:

[0030]

[0031] Here, δ represents the noise margin.

[0032] Reconstructed spectrum signal Provided as prior knowledge to step (12).

[0033] Step (12) Combine the signal l 2，1 The norm is minimized and the inner product of the prior knowledge obtained in step (11) is solved for the broadband signal as follows:

[0034] Utilize l 2，1 The inner product of the norm minimization model and prior knowledge on the signal Reconstruct the reconstructed spectrum signal It can be expressed as:

[0035]

[0036] Among them, β is a balancing factor, which is used to control the influence ratio of the two parts before and after the minus sign, and its value range is (0,1).

[0037] Step 2. Derive the CDF of the reconstructed signal error and use it as the stopping criterion for observation sample collection to achieve adaptive stopping of the perception process: if the CDF less than the preset error tolerance is greater than the preset probability value, stop sampling and obtain the reconstructed spectrum signal; otherwise, continue the next sampling process.

[0038] Step (21) Represents a real spectrum signal. If the number of sampled samples M = jΔM satisfies the following inequality

[0039]

[0040] Then at least 1-2exp(-γ2(Γ ξ ∩S N-1 )) has a probability that satisfy

[0041]

[0042] in, Representation function exist The tangent cone at S N-1 express The unit ball in space. γ(Γ ξ ∩S N-1 ) represents the set Γ ξ ∩S N-1 Gaussian complexity. ρ and C represent constants,

[0043] According to Gaussian complexity γ(Γ ξ ∩S N-1 ) and Gaussian width w(Γ ξ ∩S N-1 ), the sampling sample size M satisfies the following inequality

[0044]

[0045] Step (22) Gaussian width w(Γ ξ ∩S N-1 ) can be expressed as

[0046] w(Γ ξ ∩S N-1 )≤E[dist(g,N ξ )]

[0047] in, represents ξ(x f ) in x f Normal cone at g~N(0,I N ) is a random Gaussian vector. According to Jensen’s inequality,

[0048] w 2 (Γ ξ ∩S N-1 )

[0049]

[0050] For μ ≥ 0, fix any g and choose any available

[0051]

[0052] Define the parameters:

[0053]

[0054] It can be deduced that:

[0055]

[0056] therefore:

[0057]

[0058] Define the parameters:

[0059]

[0060] According to the convex geometry theory, it can be deduced that:

[0061]

[0062]

[0063] Among them, μ k Represents χ k The mean of the distribution, l represents the number of non-adjacent blocks into which the block-sparse signal is divided, and s represents the number of blocks with signals.

[0064] Then the cumulative distribution function (CDF) of the reconstruction error is derived:

[0065]

[0066] in, σ represents the standard deviation of the observation noise,

[0067] Stopping criteria for the perception process: e represents the preset error tolerance, p set Indicates the preset probability value.

[0068] When the stopping criterion is met, sampling stops and the reconstructed spectrum signal is obtained.

[0069] Otherwise, continue with the next sampling process.

[0070] Step 3. Spectrum status determination

[0071] Get the reconstructed spectrum signal After that, the corresponding binary spectrum state is obtained by the energy detection method. The energy detection method is specifically as follows: first, the receiving device samples in the target frequency band to obtain the received signal. The signal can be represented by two assumptions: when the frequency band is idle, only the noise signal is received, and when the frequency band is occupied, the superposition of noise and the main user signal is received. After sampling the signal, an energy value is obtained by calculating its energy value, that is, summing the square of the amplitude of the sampled signal. Subsequently, the energy value is compared with a preset threshold, which is usually designed based on the statistical characteristics of noise power or the detection probability and the false alarm probability. If the energy value is lower than the threshold, the frequency band is determined to be idle, otherwise it is determined to be occupied, thereby generating a binary state of the target frequency band (such as 0 for idle and 1 for occupied).

[0072] The method of the present invention is applicable to Figure 3 Compressed spectrum sensing of block sparse signals shown.

[0073] A simulation environment was built to compare the performance of the method of the present invention (ACSS-DEM) with typical methods. The compared methods include: ①ACSS-JL, an adaptive compressed spectrum sensing method based on Johnson-Lindenstrauss; ②ACSS-SOC, an adaptive compressed spectrum sensing method based on sparsity estimation; ③ACSS-CV, an adaptive compressed spectrum sensing method based on cross-validation.

[0074] The results are shown in the attached figure, where:

[0075] Figure 4 The spectrum sensing performance of the comparison method is compared with that of the comparison method, wherein the horizontal axis is the false alarm probability and the vertical axis is the correct detection probability. It can be seen that under the same false alarm probability, the method of the present invention has the highest correct detection probability;

[0076] Figure 5 The CDF curves under different numbers of sparse blocks (number of sparse blocks: (a) 4, (b) 5, (3) 6, (4) 7); it can be seen that the CDF theoretical value is basically consistent with the experimental value obtained by this method;

[0077] Figure 6 For comparison of NMSE (normalized mean square error) of reconstructed signals with different numbers of sparse blocks of comparative methods (number of sparse blocks: (a) 4, (b) 5, (c) 6, (d) 7), the method of the present invention has the smallest normalized mean square error.

[0078] In summary, compared with the comparison method, the method of the present invention has the best spectrum sensing performance.

[0079] The above is only a preferred embodiment of the present invention, and the scope of rights claimed by the present invention is not limited thereto. The present invention has many other embodiments, and those skilled in the art can make various corresponding changes and deformations according to the present invention without departing from the spirit and essence of the present invention, but these changes and deformations should all fall within the scope of protection of the claims attached to the present invention.

Claims

1. An adaptive compressed spectrum sensing method based on a deterministic evaluation model, characterized in that: The following steps are involved: Step 1. Based on The signal reconstruction model combining norm minimization and prior knowledge mining reconstructs the spectrum signal: After the secondary user SU compresses and samples the frequency domain signal, The norm minimization solves the broadband signal, which is used as a priori knowledge and The norm minimization is combined with solving the spectrum signal of the current time slot to reconstruct the signal; Step 2. Derive the CDF of the reconstructed signal error and use it as the stopping criterion for observation sample collection to achieve adaptive stopping of the perception process: if the CDF less than the preset error tolerance is greater than the preset probability value, stop sampling and obtain the reconstructed spectrum signal; otherwise, continue the next sampling process; Step 3: Use the energy detection method to make spectrum judgment and obtain the spectrum occupation and idleness.

2. The method according to claim 1, characterized in that: Step 1 includes the following steps: Step 11: Using periodic compressed spectrum sensing, each communication time frame T is divided into a sensing time slot of duration Ts, followed by a data transmission time slot Td; in the sensing stage, the total interval Ts is evenly divided into J time slots, and the length of each time slot τ = Ts / J; in each time slot τ, the secondary user samples the time domain signal x(t) at the same sampling rate and obtains ΔM compressed measurement values; after j sensing processes, the signal is reconstructed using jΔM observation samples; specifically, the secondary user samples the N×1 frequency domain signal Compressed sampling is performed, and the compressed observation can be expressed as Among them, Φ represents the observation matrix of jΔM×N, n~N(0,σ 2 I jΔM ) represents the observation noise; use The norm minimization model is used for the signal Reconstruct the reconstructed spectrum signal It can be expressed as: Where δ represents the noise tolerance; Reconstructed spectrum signal Provided as prior knowledge to step 12; Step 12: Combining Signals The norm minimization and the inner product of prior knowledge are used to solve the broadband signal as follows: use The inner product of the norm minimization model and prior knowledge on the signal Reconstruct the reconstructed spectrum signal It can be expressed as: Among them, β is a balancing factor, which is used to control the influence ratio of the two parts before and after the minus sign, and its value range is (0,1).

3. The method according to claim 1, characterized in that Step 2 includes the following steps: Step 21: Order Represents a real spectrum signal. If the number of samples M = jΔM satisfies the following inequality Then at least 1-2exp(-γ 2 (Γ ξ ∩S N-1 )) has a probability that satisfy in, Representation function exist The tangent cone at S N-1 express The unit ball in space; γ(Γ ξ ∩S N-1 ) represents the set Γ ξ ∩S N-1 Gaussian complexity; ρ and C represent constants, According to Gaussian complexity γ(Γ ξ ∩S N-1 ) and Gaussian width w(Γ ξ ∩S N-1 ), the sampling sample size M satisfies the following inequality Step 22: Gaussian width w(Γ ξ ∩S N-1 ) is expressed as w(Γ ξ ∩S N-1 )≤E[dist(g,N ξ )] in, represents ξ(x f ) in x f Normal cone at g~N(0,I N ) is a random Gaussian vector; according to Jensen’s inequality, For μ ≥ 0, fix any g and choose any get Define the parameters: Derived: therefore: Define the parameters: According to the convex geometry theory, it is deduced that: Among them, μ k Represents χ k The mean of the distribution, l represents the number of non-adjacent blocks into which the block-sparse signal is divided, and s represents the number of blocks with signals; Thus, the cumulative distribution function CDF of the reconstruction error is derived: in, σ represents the standard deviation of the observation noise, The stopping criterion of the perception process is: e represents the preset error tolerance, p set Indicates the preset probability value; If the stopping criterion is met, stop sampling and get the reconstructed spectrum signal Otherwise, continue with the next sampling process.

4. The method according to claim 1, characterized in that: In step 3, the energy detection method is specifically as follows: first, the receiving device samples within the target frequency band to obtain the received signal; the signal is represented by two assumptions: when the frequency band is idle, only the noise signal is received, and when the frequency band is occupied, the signal received is the superposition of noise and the main user signal; after sampling the signal, the energy value is calculated, that is, the square of the amplitude of the sampled signal is accumulated to obtain an energy value; then, the energy value is compared with a preset threshold, if the energy value is lower than the threshold, the frequency band is determined to be idle, otherwise it is determined to be occupied, thereby generating a binary state of the target frequency band.

Images