Frequency spectrum sensing detection method based on channel

Abstract

The invention relates to the technical field of digital signals, in particular to a spectrum sensing detection method based on a channel. The method comprises a digital channelization design method of the fast filter bank and a spectrum sensing algorithm. The structure of the fast filter bank is formed by cascading i stages of half-sideband filters, the input signal frequency spectrum of the fast filter is uniformly divided into k sub-channels, and the hardware implementation complexity of the fast filter bank is estimated by calculating the parameters of each stage of half-sideband filter, so that the design of the fast filter bank is realized; the step of designing the spectrum sensing algorithm comprises the steps of S1, estimating a channel power spectrum mean value, S2, estimating a noise mean value, and S3, calculating a detection threshold value T. Digital channelization of input signals is realized by designing an optimized fast filter bank, and channel occupation conditions are accurately judged in combination with the spectrum sensing algorithm. And the signal processing efficiency and the accuracy of spectrum sensing are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of wireless communication technology, and in particular to a channel-based spectrum sensing detection method. Background Technology

[0002] In digital signal processing, digital channelization and spectrum sensing are important and complex tasks. Traditional digital channelization methods, due to the complexity of filter bank design and the difficulty of hardware implementation, are difficult to meet the requirements of modern signal processing systems. Meanwhile, the accuracy of spectrum sensing is affected by various factors, leading to frequent false positives and false negatives. Therefore, this invention proposes a digital channelization and spectrum sensing method and system based on fast filter banks to address the shortcomings of existing technologies.

[0003] For example, in 2023, Tu Zhiyi from the University of Electronic Science and Technology of China published "Design and Implementation of Broadband Spectrum Sensing Based on Dynamic Digital Channelization Technology," which designed a broadband spectrum sensing scheme based on an analysis-synthesis filter bank. This scheme utilizes time-division multiplexing and pulsating array technologies to achieve real-time and efficient digital channelization. Furthermore, it can dynamically sort and reconstruct cross-channel signals and estimate parameters such as bandwidth, center frequency, and signal-to-noise ratio (SNR) of all sensed signals without prior information. The scheme is deployed on a QT7012 board and uses a high-resolution, high-sampling-rate ADC FMC daughterboard, QT7126. The entire processing flow includes an analysis filter bank module, an autocorrelation detection module, a morphological processing module, a time-frequency graph information extraction module, and a synthesis filter bank. First, the analysis filter bank module uses time-division multiplexing and data multiplexing methods to channelize 32 channels of sampled data with a sampling rate of 4.8 GHz, generating 128 sub-channel output signals. 56 of these signals are then sent to the autocorrelation detection module for judgment, while unprocessed signals are buffered for reconstruction. The autocorrelation detection module uses an adaptive algorithm to set a threshold to determine the presence of a signal in the sub-channel and outputs a valid flag signal and time-frequency graph information as input to the time-frequency graph information extraction module. The morphological processing module optimizes the time-frequency graph, filtering out interference and improving the accuracy of information extraction. The time-frequency graph information extraction module determines the sub-channel of the valid signal based on the information in the time-frequency graph and sends it to the integrated filter bank for cross-channel signal reconstruction. The integrated filter bank selects a two-in-one or four-in-one reconstructor for signal reconstruction based on the cross-channel status of the valid signal; non-cross-channel signals are directly buffered and output.

[0004] This technical solution mainly uses polyphase filtering for channelized reception. This reception method is difficult to dynamically configure the number of channels, and an increase in the number of channels requires a large number of FIR filters. Using a fast filter bank can alleviate this problem well. In addition, the autocorrelation detection algorithm used in this solution is operated in the time domain. When the signal-to-noise ratio is below -10dB, the detection performance drops significantly. A long signal is prone to long breaks in the time domain. Even after morphological processing, it is impossible to synthesize a complete signal or to detect the signal directly. Therefore, frequency domain signals are used for signal detection to improve signal detection performance. Summary of the Invention

[0005] The purpose of this invention is to overcome the problems of long detection time and complex reconstruction of channelized signals in the prior art, and to provide a channel-based spectrum sensing detection method.

[0006] In a first aspect, the present invention provides a channel-based spectrum sensing detection method. The present invention uses a fast filter bank structure based on half-sideband filters to realize spectrum sensing detection. The channel-based spectrum sensing detection method includes a digital channelization design method for fast filter banks and a spectrum sensing algorithm. The fast filter bank structure consists of i-stage cascaded half-sideband filters. The spectrum of the input signal of the fast filter is uniformly divided into k sub-channels. The hardware implementation complexity of the fast filter bank is estimated by calculating the parameters of each stage of the half-sideband filter, thus realizing the design of the fast filter bank. The steps for designing the spectrum sensing algorithm include: S1: Channel power spectrum mean estimation: In each channel, QW sampled data are extracted in real time and divided into Q groups, each group containing W sample points. The mean power spectrum estimate of the received channel is then calculated. (ω); S2: Noise Mean Estimation: Sort the estimated mean values ​​of the channel power spectrum and take the minimum. The average value of each point is used as the mean value for the noise signal power spectrum estimation. The estimated value , Calculate the largest positive integer less than L / 5. and will participate in the calculation The channel corresponding to (ω) is marked as an idle channel; First, from a hardware implementation perspective, the half-sideband filter structure is non-recursive in the time domain, eliminating issues such as stability and ringing. Furthermore, its design is relatively simple, optimizing resource utilization and reducing system complexity. The half-sideband filter can simultaneously filter out half of the passband of the previous stage filter, ensuring the sum of the passbands encompasses all passbands of the previous stage. This effectively reduces unnecessary computational and processing steps, thereby improving the overall system efficiency.

[0007] Furthermore, the introduction of the half-sideband filter structure improves filtering performance. Because the transition bandwidth of each sub-filter is precisely controlled, the transition between the passband and stopband is smoother, reducing signal distortion and interference. This performance improvement is crucial for digital channelization systems, as it directly affects the accuracy and reliability of signal processing. By comparing the performance metrics of traditional filters and half-sideband filters, such as inter-channel interference and signal distortion rate, it is found that half-sideband filters have significant advantages in reducing system complexity and improving signal processing performance.

[0008] Preferably, a half-sideband filter is a digital filter that divides the spectrum of the input signal into two parts for filtering. Its frequency response is flat in one band and exhibits higher attenuation in the other. In MATLAB, the `firhalfband` function can be used to design and implement half-sideband filters.

[0009] By optimizing the design of the prototype filter, it is possible to ensure that the fast filter bank can maintain good filtering performance, such as the specified passband ripple and stopband ripple requirements, while reducing complexity.

[0010] A two-dimensional search method is used to find the optimal number of channels and stages, ensuring that the configuration with the lowest hardware implementation complexity is found while meeting system performance requirements.

[0011] In a second aspect, the present invention provides a spectrum sensing algorithm for estimating the power spectrum of a received signal and simultaneously estimating the power spectrum and mean of a noise signal; a smoothing method is used to improve the accuracy of power spectrum estimation, and the signal sample data is processed in groups to calculate the power spectrum estimation mean of each group of signal sample data.

[0012] By employing a smoothing method to group a certain number of signal sample data and perform power spectrum estimation, the accuracy of power spectrum estimation can be significantly improved. This method effectively reduces the impact of noise on power spectrum estimation, making the estimation results more reliable. Spectrum sensing algorithms can dynamically track channel occupancy in real time. By estimating the channel power spectrum in real time, changes in channel occupancy can be detected promptly. This is crucial for dynamic spectrum management and resource allocation. Through real-time estimation and sorting of channel power spectra, the algorithm can efficiently detect occupied and idle channels. Especially for sparse signals in broadband spectrum sensing, the system can accurately identify channels with low occupancy rates, providing strong support for the effective utilization of spectrum resources. The algorithm designs a flexible detection threshold, introducing a correction factor to adapt to fluctuations in the mean of idle channel power spectrum estimation in practical applications. This flexible threshold setting method improves the accuracy and robustness of detection.

[0013] The algorithm derives an expression for the false alarm probability through theoretical analysis and calculation, and provides a method for calculating a suitable correction factor based on a given false alarm probability requirement. This helps control the false alarm probability in practical applications and improves the reliability of spectrum sensing. Simultaneously, it fully utilizes the statistical properties of power spectrum estimation results, such as the normal distribution, to further improve detection accuracy and efficiency. This statistically based detection method has significant advantages when handling large-scale data and complex spectrum environments. The system design is geared towards the needs of broadband spectrum sensing and can handle signals and spectrum resources within a wide bandwidth. This makes the system more adaptable and flexible in dealing with the complex spectrum environment and diverse spectrum resources in modern communication systems.

[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention provides a channel-based spectrum sensing detection method. By designing and optimizing a fast filter bank to achieve digital channelization of the input signal, and combining it with a spectrum sensing algorithm to accurately determine the channel occupancy, the method can significantly improve the spectrum detection rate, avoid the picket fence effect, and reduce the signal loss rate. Attached Figure Description

[0015] Figure 1 This is a flowchart of the two-dimensional search algorithm described in Embodiment 1 of the present invention; Figure 2 The spectrum of the multi-tone signal as described in Embodiment 1 of the present invention; Figure 3 The channelized spectrum of the conventional polyphase filtering method described in Embodiment 1 of this invention; Figure 4 The channelized spectrum of the FFB fast filter bank method described in Embodiment 1 of the present invention; Figure 5The performance of the algorithm described in Embodiment 1 of the present invention under single-tone interference (STJ), multi-tone interference (MTJ), and narrowband interference (NBJ) conditions is shown. Detailed Implementation

[0016] The present invention will be further described in detail below with reference to experimental examples and specific embodiments. However, this should not be construed as limiting the scope of the above-mentioned subject matter of the present invention to the following embodiments; all technologies implemented based on the content of the present invention fall within the scope of the present invention.

[0017] Example 1 Combination Figure 1 This embodiment describes a channel-based spectrum sensing detection method, which includes the following steps: S1: This invention includes a digital channelization design section and a spectrum sensing algorithm design section; the digital channelization design section includes a fast filter bank design section and a fast filter bank optimization section. The design of the fast filter bank includes calculating the parameters of each half-sideband filter and cascading multiple half-sideband filters to form a fast filter bank; the optimization of the fast filter bank includes calculating the hardware complexity and using a two-dimensional search method to find the minimum number of channels and stages of the fast filter bank.

[0018] A half-sideband filter is a special type of FIR filter whose order can only be even and whose length is odd. Except for the middle value of 0.5, all other even-numbered coefficients are 0, thus saving the need for multiplication and addition operations during filtering. Furthermore, the frequency response of a half-sideband filter is symmetrical: the passband and stopband are symmetrical. If the passband cutoff frequency is ω, then the stopband cutoff frequency is (π-ω).

[0019] Assume the first The passband ripple of the stage filter is The stopband ripple is The passband cutoff frequency is The stopband cutoff frequency is After normalizing the frequency, the following applies to the half-sideband filter structure: Assume the passband ripple of each channel frequency response of FFB is The stopband ripple is Therefore: Combining the characteristics of a half-sideband filter, we get: As can be seen from the characteristics of FFB, each stage of sub-filters filters out half of the passband of the previous stage filter, but the sum of their passbands can encompass the entire passband of the previous stage, that is: To reduce the complexity of hardware implementation, Then we can get: in, It is the passband cutoff frequency of the 0th stage of the half-sideband filter.

[0020] Then the transition bandwidth of each filter stage It can be represented as: in, It is the passband cutoff frequency of the 0th-order filter.

[0021] In addition, in half-sideband filters Since the transfer function is half that of an FFT filter bank, there is no need to normalize the FFB filter bank designed with a half-sideband filter. However, when designing an FFB filter bank with a half-sideband filter structure, the transfer function should be transformed.

[0022] In the complementary transfer function of each sub-filter, only the coefficients at time 0 need to be kept unchanged, and the rest can be inverted: in, The transfer function representing the sub-filter. This represents the complementary transfer function of the corresponding sub-filter. .

[0023] Using traditional polyphase filtering channelization methods for Figure 2 The multi-tone signal spectrum shown is channelized to obtain the following result: Figure 3 The channelized spectrum shown; Using the FFB fast filter bank method to Figure 2 The multi-tone signal spectrum shown is channelized to obtain the following result: Figure 4 The channelized spectrum shown; Comparing the results of the two channelization methods, it can be seen that the FFB fast filter bank method has better dynamic configuration of the number of channels in hardware, while achieving the same channelization effect as the traditional polyphase filter bank.

[0024] S2: The steps involved in designing the spectrum sensing algorithm include the following: The design of half-sideband (FFB) filters can be implemented using MATLAB's `firhalfband` function. Their order depends on the passband cutoff frequency and passband ripple. However, in hardware implementation, the complexity of an FFB filter bank depends not only on the order of the prototype filter but also on the number of channels in the FFB filter bank. This complexity can be addressed using... To estimate, Indicates the first The number of non-zero tap coefficients in the prototype filter is used, since the number of tap coefficients is related to the multiplication resources required during implementation. The total number of non-zero tap coefficients in the filter bank is used instead. To estimate the complexity of its hardware implementation: Therefore, it is necessary to specify a reasonable passband ripple value based on the actual maximum channel bandwidth required, and find the value that minimizes the hardware implementation complexity. Minimum number of channels , ,series The problem then becomes finding the optimal solution. and .

[0025] Specify maximum channel bandwidth as The normalized value indicates that the maximum value of the passband ripple after cascading the specified filter banks is [value missing]. .

[0026] According to the formula It can be seen that the maximum passband ripple value after cascading the filter bank is _____. Combined with formula Filter channel bandwidth The sum of transition bandwidth and effective bandwidth: The optimization objective formula and its constraints can be obtained: As can be seen from the equation, this is a two-dimensional optimal problem

[68] . Generally, the passband bandwidth of the filter is smaller than the transition bandwidth, so in this problem, it is actually possible to estimate the bandwidth. : In the formula Round up to the nearest whole number.

[0027] Use a two-dimensional search method to find the largest With the smallest : 1) Finding the optimal series: From Start loop calculation The size after each loop When the result calculated in the current loop Greater than the result calculated in the previous loop Exit the loop and make .

[0028] 2) Finding the best : Find the largest The value problem is equivalent to finding the minimum. .exist In each loop that takes a fixed value Reduce the step value Find within the search range that makes smallest , The search scope is as follows: The algorithm graph traverses from the minimum required level. For example, if the required bandwidth is 100MHz and the sampling rate is 4.8GHz, then 4800 / 100=48. Since 64>40>32, at least 64 channels are needed, where 2 is taken as a power of L, that is, L is at least 6.

[0029] That is, start traversing from the smallest L, initially using L as a fixed value, and continuously changing it. Find the value that minimizes G. Then compare the current value of G with the previously calculated G. If the current value of G is less than the previous value, assign the value of G to G', increment L by one, and repeat the loop until the current value of G is found to be greater than G'. This indicates that the complexity obtained in the previous loop was lower, i.e., G was lower, so the loop is exited, and the L used in the previous loop (i.e., the current L minus 1) is compared with... Output.

[0030] Furthermore, the hardware implementation requires... According to a suitable length To do quantitative analysis, and The larger the size, the greater the resource consumption. A smaller value will lead to excessive error and degrade filter performance; therefore, setting the passband ripple value is important. Since the coefficients of the half-sideband filter are all less than 1, let It contains 1 sign bit. Decimal places, initialization , constantly increasing The value ensures that the filter passband ripple meets the requirements. The algorithm flow is as follows: Figure 1 As shown.

[0031] Analyze the signal power spectrum: Let the received signal The power spectrum is estimated as : The equality holds when the channel is not occupied, i.e., there is no active signal. Let the noise signal... The power spectrum is estimated as Its mean is Then we have: In the formula .

[0032] Based on practical experience, the accuracy of power spectrum estimation is affected by noise. Therefore, this paper combines a smoothing method to improve the accuracy of power spectrum estimation. A certain number of signal sample data are used... Divided into Q groups, each group contains W signal sample data, respectively. : ( ) represents the amplitude spectrum value of the q-th group of signals. This represents the power spectral density value of the q-th group of signals.

[0033] However, in practical applications It will fluctuate over time, therefore it needs to be adjusted. To perform real-time dynamic tracking.

[0034] Since the system designed in this paper is for broadband spectrum sensing, and based on practical experience, the signals in broadband spectrum sensing are sparse signals with a channel occupancy rate of less than 80%, it can be deduced that there exists... And the quantity is greater than L / 5.

[0035] right Sort in ascending order, and denote the sorted sequence as follows: Take the first few digits after sorting. Each point as The estimated value , Find the largest positive integer less than L / 5. S3: Calculate the detection threshold T. According to theoretical derivation, the mean of the channel power spectrum estimate under channel occupancy conditions must be greater than... The mean channel power spectrum estimate under idle channel conditions is equal to Therefore, it is only necessary to Setting a threshold is sufficient. However, in practical applications, the mean channel power spectrum estimate of the idle channel may be in the range of... There are fluctuations, so a correction factor is added. Design detection threshold : Then we have: That is, power spectrum estimation for the k-th sub-channel. If the value is greater than the detection threshold, a valid signal is considered to exist, and the sub-channel is not empty. Assumptions; conversely, it is assumed that no valid signal exists and the sub-channel is not occupied, i.e. Assumption.

[0036] Due to the formula With formula The mean power spectrum when the channel is occupied is always greater than It can be deduced that The probability of a false early warning. It can be represented as: and They all follow a normal distribution: Assumption for exist Mean under channel, for exist The variance under the channel, and using and express and exist Mean under channel, and express and exist The variance under the channel is then: in: have to: Therefore, the probability of false early warning expression: make Then there is Known If given The appropriate correction factor can then be calculated using the following formula. : like Figure 5 As shown, the detection performance of this method is good. The curve reflects the accuracy of the algorithm under different signal-to-noise ratios. It can be seen that when the signal-to-noise ratio is greater than 3dB, the correct detection probability under single-tone interference, multi-tone interference and narrowband interference all reach 1.

[0037] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A channel-based spectrum sensing detection method, characterized in that, This method uses a fast filter bank structure based on half-sideband filters to achieve spectrum sensing detection. The channel-based spectrum sensing detection method includes a digital channelization design method for fast filter banks and a spectrum sensing algorithm. The fast filter bank structure consists of i-stage cascaded half-sideband filters. The spectrum of the input signal of the fast filter is uniformly divided into k sub-channels. The hardware implementation complexity of the fast filter bank is estimated by calculating the parameters of each stage of the half-sideband filter, thus realizing the design of the fast filter bank. The steps for designing the spectrum sensing algorithm include: S1: Channel power spectrum mean estimation: In each channel, QW sampled data are extracted in real time and divided into Q groups, each group containing W sample points. The mean power spectrum estimate of the received channel is then calculated. (ω); S2: Noise Mean Estimation: Sort the estimated mean values ​​of the channel power spectrum and take the minimum. The average value of each point is used as the mean value for the noise signal power spectrum estimation. The estimated value , Calculate the largest positive integer less than L / 5. and will participate in the calculation The channel corresponding to (ω) is marked as an idle channel; S3: Calculate the detection threshold T, and then... (ω) is compared with T to determine the occupancy status of non-idle channels.

2. The channel-based spectrum sensing detection method according to claim 1, characterized in that, The digital channelization design method for the fast filter bank includes the design and optimization of the fast filter bank; the optimization of the fast filter bank adopts a two-dimensional search method to find the minimum number of channels and stages of the fast filter bank.

3. The channel-based spectrum sensing detection method according to claim 2, characterized in that, The design of the fast filter bank specifically includes the following steps: S31: Set the parameters according to actual application requirements, including passband ripple, stopband ripple, passband cutoff frequency and stopband cutoff frequency; S32: Calculate the transition bandwidth of each half-sideband filter to ensure that the filter bank can cover the required frequency range; S33: Based on the design principle of half-sideband filters, design the transfer function of each stage of the filter, and transform the transfer function of each stage of the half-sideband filter to ensure that the coefficients at time 0 in the complementary transfer function remain unchanged, and the remaining coefficients are inverted. S34: Verify the performance of the fast filter bank, including whether the passband ripple, stopband ripple, and transition bandwidth meet the design requirements.

4. The channel-based spectrum sensing detection method according to claim 2, characterized in that, The optimization of the fast filter bank specifically includes the following steps: S41: Set the total number of non-zero tap coefficients in the fast filter bank to estimate its hardware implementation complexity as G. Combined with the actual maximum channel bandwidth required, specify the maximum value of the passband ripple value after the filter bank is cascaded. S42: Based on the maximum passband ripple value of the cascaded filter bank, find the minimum number of channels N, the passband cutoff frequency of the 0th stage filter, and the stage number L that minimizes the complexity of the hardware implementation of G. For an N-channel fast filter bank, its corresponding stage number... ; S43: Calculate the maximum passband ripple value of the fast filter bank based on the passband ripple, and obtain the channel bandwidth of the filter by combining the transition bandwidth of each half-sideband filter. Obtain the optimization objective formula and its constraints based on the channel bandwidth of the filter.

5. The channel-based spectrum sensing detection method according to claim 2, characterized in that, The two-dimensional search method specifically includes the following steps: S51: Find the optimal level. Start by traversing from the minimum required level. First, use the level as a fixed value and continuously change the value of the transition bandwidth of the 0th level filter. Find the value of the transition bandwidth of the 0th level filter that minimizes the hardware complexity. Then compare it with the hardware complexity calculated in the previous iteration. If the hardware complexity calculated in the current iteration is greater than the hardware complexity calculated in the previous iteration, exit the loop and set the level to the level minus one. S52: Finding the optimal passband cutoff frequency of the 0th-order filter: The problem of finding the maximum passband cutoff frequency of the 0th-order filter is equivalent to finding the minimum transition bandwidth of the 0th-order filter. In each iteration where the order is fixed, the transition bandwidth of the 0th-order filter is reduced by a step value of Δ. Within the search range, the transition bandwidth of the 0th-order filter that minimizes the hardware complexity is found.

6. The channel-based spectrum sensing detection method according to claim 1, characterized in that, The formula for calculating the detection threshold is: Where T is the detection threshold, As a correction factor, The mean of the power spectrum estimate of the noise signal. This is an estimate of the mean of the power spectrum of the noise signal.

7. The channel-based spectrum sensing detection method according to claim 1, characterized in that, The specific method for determining the occupancy status of non-idle channels is as follows: If the power spectrum estimate of the k-th sub-channel is greater than or equal to the detection threshold, the non-idle channel is occupied. If the power spectrum estimate of the k-th sub-channel is less than the detection threshold, then the non-idle channel is not occupied.

8. The channel-based spectrum sensing detection method according to claim 1, characterized in that, The estimated mean of the noise signal power spectrum and the estimated mean of the channel power spectrum both follow a normal distribution.

9. The channel-based spectrum sensing detection method according to claim 1, characterized in that, The formula for calculating the mean power spectrum estimate of the channel is as follows: in, This is the power spectrum estimate mean, where Q is the number of signal sample data groups, and W is the number of signal sample data in each group of Q, where q = 0, 1, ..., Q-1. The power spectrum represents the signal sample group q of the k-th sub-channel. It is the signal amplitude of the q-th group of signal samples in the k-th sub-channel.

10. The channel-based spectrum sensing detection method according to claim 4, characterized in that, The formulas for the optimization objective and the constraints are as follows: in, It represents the number of non-zero tap coefficients in the i-th stage prototype filter. It is the normalized passband bandwidth. It is the passband cutoff frequency of the 0th-order filter. It is a reasonable passband ripple value. The maximum value of the passband ripple after the filter bank is cascaded.

Images