MIMO-OFDM space-frequency coding mode identification method, system, device and medium under alpha stable distribution noise

Abstract

The application discloses a method, system, device and medium for identifying MIMO-OFDM space-frequency coding mode under alpha stable distribution noise, which comprises the following steps: firstly, performing space-time-frequency coding inter-class identification based on fractional low-order correlation on a received signal; then, extracting a cyclic correlation entropy spectrum feature map of the received signal; and finally, inputting the extracted feature map into a trained deep forest network to identify the space-frequency coding mode; the system and device are based on space-time-frequency coding inter-class identification based on fractional low-order time lag correlation, inter-class identification of unknown signals is realized, and whether space-frequency coding is used is confirmed; the application extracts a cyclic correlation entropy spectrum feature map of a received signal, combines a deep forest network, and converts a coding identification problem into an image identification problem; the application can effectively identify MIMO-OFDM space-frequency coding mode under alpha stable distribution noise, solves the problem that a traditional method is limited in function in non-cooperative communication, and is also applicable to a Gaussian noise environment.

Description

Technical Field

[0001] This invention belongs to the field of communication signal demodulation technology, and particularly relates to a method, system, device and medium for identifying MIMO-OFDM space-frequency coding schemes under Alpha stable distributed noise. Background Technology

[0002] MIMO-OFDM technology, combining Orthogonal Frequency Division Multiplexing (OFDM) with MIMO, is an efficient solution to address performance degradation in new scenarios and has become a research hotspot in the field of wireless communication in recent years. In addition to space-time coding, MIMO-OFDM systems employ space-frequency coding schemes, utilizing space-frequency diversity. The combination of Space-Frequency Block Codes (SFBCs) and MIMO-OFDM can improve the effectiveness and reliability of communication. Identification of SFBC schemes is a crucial aspect of deep sensing technology in non-cooperative MIMO systems. Traditional SFBC identification algorithms are designed to extract relevant information from received signals in Gaussian noise environments. However, Gaussian noise is not the only case; for example, non-Gaussian impulse noise modeled using an Alpha stable distribution typically has a heavy-tailed probability distribution, making it difficult to model using a Gaussian distribution effectively. In the case of non-Gaussian noise, performance degrades significantly due to noise model mismatch. Therefore, research on MIMO-OFDM SFBC scheme identification under Alpha stable noise is of great significance.

[0003] There is relatively little research on SFBC coding identification methods: Marey et al. extended the method of detecting the peak value of the cross-correlation function of two receiving antenna signals with specific time delays, and used the spatial redundancy of SFBC-OFDM to realize the identification of AL signals and SM signals (Marey M, Dobre O A. Automatic Identification of Space-Frequency Block Coding for OFDM Systems[J].IEEE Transactions on Wireless Communications,2016,16(1):117-128.). However, this algorithm does not make full use of the frequency domain redundancy of SFBC signals. Therefore, when the number of OFDM carriers increases, its performance does not improve, but the computational complexity increases exponentially. Gao et al. utilized the spatial-frequency redundancy of SFBC signals, extracted the subspace rank features of adjacent subcarriers, solved the features using random matrix theory, and used the minimum distance criterion to distinguish SFBC-OFDM signals (Gao M, Li Y, Dobre OA, et al. Blind Identification of SFBC-OFDM Signals Using Subspace Decompositions and Random Matrix Theory[J].IEEE Transactions on Vehicular Technology,2018,67(10):9619-9630.). They also proposed using multiple receiving antennas to construct detection statistics, thereby efficiently utilizing frequency domain redundancy to complete SFBC-OFDM signal identification (Gao M, Li Y, Dobre OA, et al. Blind Identification of SFBC-OFDM Signals Based on the Central Limit Theorem[J].IEEE Transactions on Wireless Communications,2019,18(7):3500-3514.). Kun et al. proposed a MIMO-SFBC blind identification algorithm based on symbol eigenvalues.Based on the symbol correlation characteristics of different space-frequency block codes in the frequency domain, the feature vector sequences of different space-frequency block codes are derived. The symbol eigenvalues ​​are estimated using binary hypothesis testing, and different coding types are distinguished by decision tree classification and recognition algorithm (KunJin, JinKun, Yu Keyuan, Yan Wenjun. Blind recognition of MIMO-SFBC based on Symbolic eigenvalue*[J]. Journal of Physics: Conference Series, 2020, 1650(3).). Zhang Yuyuan et al. used the correlation function feature map of the signal in the frequency domain at the receiving end, performed spatial scale preprocessing to transform it into two dimensions, and finally used an extended dense convolutional network to realize SFBC coding classification and recognition (Zhang Yuyuan, Zhang Limin, Yan Wenjun. SFBC-OFDM recognition method based on cross-correlation feature map and extended dense convolutional network[J]. Systems Engineering and Electronics Technology, 2021, 43(09):2657-2664.). Although the above algorithms can effectively achieve SFBC coding classification, current research is based on spatial frequency coding intra-class recognition under the premise of knowing the major categories of diversity coding. In addition, most of them are based on Gaussian noise models. In the context of increasingly complex noise environments, the performance degradation problem of traditional recognition methods under non-Gaussian noise urgently needs to be solved.

[0004] Based on the above analysis, the problems and shortcomings of the existing technology are as follows:

[0005] (1) Existing space frequency block coding identification methods are intra-class identification based on the premise of knowing the major categories of diversity coding. In non-cooperative communication, they face functional limitations and cannot perform subsequent intra-class identification.

[0006] (2) Current research on spatial frequency block coding identification methods focuses on the Gaussian noise environment assumption. For non-Gaussian Alpha stable distribution noise that better reflects the actual environment, its identification performance will significantly decrease;

[0007] The difficulties in solving the above problems and defects are as follows: In non-cooperative communication, there are functional limitations, and it is necessary to first confirm whether the signal uses space-frequency coding before subsequent intra-class identification can be performed; secondly, under non-Gaussian noise interference, the recognition performance of traditional identification methods will be weakened. Therefore, the importance of inter-class identification, the distinguishing features applicable to Alpha stable distributed noise environment, and the construction of classifier models are the technical difficulties of MIMO-OFDM space-frequency coding identification methods under Alpha stable distributed noise. Summary of the Invention

[0008] To address the problems existing in the prior art, the present invention proposes a method, system, device, and medium for identifying MIMO-OFDM space-frequency coding schemes under Alpha stable distributed noise. Based on fractional low-order time-delay correlation (FLD) inter-class identification of space-time-frequency coding, it achieves inter-class identification of unknown signals to confirm whether space-frequency coding has been used. The invention extracts the cyclic correlation entropy spectrum feature map of the received signal and, combined with a deep forest network, transforms the coding identification problem into image recognition. This invention can effectively achieve MIMO-OFDM space-frequency coding scheme identification under Alpha stable distributed noise, solving the problem of limited functionality of traditional methods in non-cooperative communication.

[0009] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0010] A method for identifying the space-frequency coding scheme in a MIMO-OFDM system under alpha-stable distributed noise includes the following steps:

[0011] Step 1: Perform inter-class identification of the received signal from the receiving antenna based on fractional low-order time-delay correlation space-time-frequency coding;

[0012] Step 2: For the received signals of different receiving antennas identified as having space frequency coding in Step 1, calculate the cyclic correlation entropy of the received signals and extract the cyclic correlation entropy spectral feature map of the received signals.

[0013] Step 3: Input the cyclic correlation entropy spectrum feature map extracted in Step 2 into the trained deep forest network to identify the space-frequency coding method.

[0014] The specific process of step one is as follows:

[0015] 1.1) The signal sample received by the i-th receiving antenna is represented as:

[0016]

[0017] Among them, h fi (j) represents the channel parameters between the transmitting and receiving antennas, L h w represents the number of propagation paths. i (n) represents non-Gaussian Alpha stable distribution noise;

[0018] 1.2) To address the impact of non-Gaussian Alpha stable distribution noise on the signal, fractional low-order processing is performed on the received signal to suppress the influence of non-Gaussian noise on the useful signal. Fractional low-order time-delay correlations are calculated for different receiving antennas:

[0019]

[0020] Among them, z ＝|z|p-1z*,z∈C,k是编码方式的块长度，τ代表不同时滞,Nk表示OFDM长度，y(·)表示不同接收天线的接收信号；

[0021] 在时滞相关函数中，寻找峰值，并提取峰值对应的时滞值作为峰值特征，绘制时滞相关峰值特征图；

[0022] 1.3)之后，提取多时滞(τ＝1,2,3,4)下的时滞相关峰值特征图，并将特征图划分为训练集和测试集,利用训练集中的特征图样本训练双通道网络(Dual Path Network，DPN)，用以对时滞相关峰值特征图进行识别；在DPN网络识别时滞相关峰值特征图的过程中，将输出切分为两路，一路和原始输入特征累加构成残差结构，减少原始输入特征的冗余度；另一路和原始输入特征并联，使得当前网络层能够直接获得上一级网络层的输出，并进一步从该输出中提取更深层的特征，提升模型的分类准确率，实现空时频编码类间识别。

[0023] 所述步骤二的具体过程为：

[0024] 对于不同接收天线的接收信号，计算其相关熵定义为：

[0025]

[0026] 将其写成傅里叶级数形式，则有循环相关熵函数为：

[0027]

[0028] 对循环相关熵求取傅里叶变换，进一步得到循环相关熵谱(CCES)函数：

[0029]

[0030] 在此基础上取固定循环频率f的循环相关熵谱切片作为该接收信号的特征图。

[0031] 所述步骤三中将提取到的特征图输入训练好的深度森林网络对空频编码方式进行识别的具体过程为：

[0032] 深度森林网络的每一层都由级联的多个随机森林组成；将步骤二提取到的循环相关熵谱特征图作为输入，使用多粒度扫描对输入的循环相关熵谱特征图进行预处理得到特征向量，将得到的特征向量输入到级联的多个随机森林中进行训练；通过随机森林学习输入特征向量的特征信息，将该特征信息输入到深度森林网络的下一层；为了增强模型的泛化能力，每一层选取不同类型的随机森林，以适应于不同大小的数据集，将上一步骤提取到的循环相关熵谱输入训练好的深度森林网络，得到最终的分类结果。

[0033] 一种实施上述Alpha稳定分布噪声下MIMO-OFDM空频编码方式识别的系统，包括：

[0034] 类间识别模块，用于步骤一中，对接收信号进行基于分数低阶相关的空时频编码类间识别，识别出接收信号是否为空频编码方式；

[0035] 特征提取模块，用于步骤二中计算接收信号的循环相关熵，提取循环相关熵谱特征图；

[0036] 空频编码方式识别模块，在步骤三中，用于将步骤二提取到的循环相关熵谱特征图投入到训练好的深度森林网络实现对空频编码方式的分类识别。

[0037] 一种实施如上述Alpha稳定分布噪声下MIMO-OFDM空频编码方式识别的设备，包括存储器和处理器，所述存储器存储有计算机程序，所述计算机程序被所述处理器执行时，能够实现步骤一至步骤三所述Alpha稳定分布噪声下MIMO-OFDM空频编码方式识别方法。

[0038] 一种接收用户输入程序的存储介质，所述存储介质存储的计算机程序被处理器执行时能够基于步骤一至步骤三所述Alpha稳定分布噪声下MIMO-OFDM空频编码方式识别方法，进行Alpha稳定分布噪声下MIMO-OFDM空频编码方式识别。

[0039] 结合上述的所有技术方案，本发明所具备的优点及积极效果为：

[0040] 本发明提出了基于分数低阶时滞相关的空时频编码类间识别，根据提取的时滞相关峰值特征，实现对未知信号的类间识别，用于确认是否使用了空频编码；

[0041] 本发明结合了深度森林网络，通过计算不同接收信号的循环相关熵谱，并将循环相关熵谱切片作为该接收信号的特征图，将编码识别问题转换为图像识别；具有特征快速提取和识别的效果。

[0042] 本发明可以有效实现Alpha稳定分布噪声下MIMO-OFDM空频编码方式识别，通过提取信号的时滞相关峰值特征和循环相关熵谱特征，能够表征接收信号的内部信息，解决了传统方法在非合作通信中功能受限的问题。

[0043] 综上，本发明采用基于分数低阶时滞相关的空时频编码类间识别方法确认空频编码的使用状态，通过提取信号的循环相关熵谱特征并输入到深度森林网络，能够准确表征接收信号的空频编码信息，实现了对Alpha稳定分布噪声下MIMO-OFDM空频编码方式的识别，有效提高了识别准确率和识别速度。附图说明

[0044] 图1是本发明实施例提供的Alpha稳定分布噪声下MIMO-OFDM空频编码方式识别方法流程图。

[0045] 图2是本发明实施例提供的Alpha稳定分布噪声下MIMO-OFDM空频编码方式识别系统结构示意图；

[0046] 图2中：1、类间识别模块；2、特征提取模块；3、空频编码方式识别模块。

[0047] 图3是本发明实施例提供的Alpha稳定分布噪声下MIMO-OFDM空时频编码方式类间识别的仿真实验结果示意图。具体实施方式

[0048] 为了使本发明的目的、技术方案及优点更加清楚明白，以下结合实施例，对本发明进行进一步详细说明。应当理解，此处所描述的具体实施例仅仅用以解释本发明，并不用于限定本发明。

[0049] 下面结合附图对本发明作详细的描述。

[0050] 本发明提供的Alpha稳定分布噪声下MIMO-OFDM空频编码方式识别方法业内的普通技术人员还可以采用其他的步骤实施，图1的本发明提供的Alpha稳定分布噪声下MIMO-OFDM空频编码方式识别方法仅仅是一个具体实施例而已。

[0051] 如图1所示，本发明实施例提供的一种Alpha稳定分布噪声下MIMO-OFDM系统中空频编码方式识别方法，具体包括以下步骤：

[0052] S101，对接收天线的接收信号进行基于分数低阶时滞相关的空时频编码类间识别；具体过程为：

[0053] 1.1)将第i个接收天线接收的信号样本表示为：

[0054]

[0055] 其中，hfi(j)表示发射天线与接收天线间的信道参数，Lh表示传播路径数目，wi(n)为非高斯Alpha稳定分布噪声；

[0056] 1.2)针对非高斯Alpha稳定分布噪声对信号的影响，对接收信号做分数低阶化处理，抑制非高斯噪声对有用信号的影响，计算不同接收天线下的分数低阶时滞相关：

[0057]

[0058] 其中，z＝|z|p-1z*,z∈C,k是编码方式的块长度，τ代表不同时滞,Nk表示OFDM长度，y(·)表示不同接收天线的接收信号；

[0059] 在时滞相关函数中，寻找峰值，并提取峰值对应的时滞值作为峰值特征，绘制时滞相关峰值特征图；

[0060] 1.3)之后，提取多时滞(τ＝1,2,3,4)下的时滞相关峰值特征图，并将特征图划分为训练集和测试集,利用训练集中的特征图样本训练双通道网络(Dual Path Network，DPN)，用以对时滞相关峰值特征图进行识别；在DPN网络识别时滞相关峰值特征图的过程中，将输出切分为两路，一路和原始输入特征累加构成残差结构，减少原始输入特征的冗余度；另一路和原始输入特征并联，使得当前网络层能够直接获得上一级网络层的输出，并进一步从该输出中提取更深层的特征，使得模型对学到的特征利用更加充分，提升模型的分类准确率，实现空时频编码类间识别。

[0061] S102，对于步骤一识别为空频编码的不同接收天线的接收信号，计算接收信号的循环相关熵，提取接收信号的循环相关熵谱特征图；具体过程为：

[0062] 对于不同接收天线的接收信号，计算其相关熵定义为：

[0063]

[0064] 将其写成傅里叶级数形式，则有循环相关熵函数为：

[0065]

[0066] 对循环相关熵求取傅里叶变换，进一步得到循环相关熵谱(CCES)函数：

[0067]

[0068] 在此基础上取固定循环频率f的循环相关熵谱切片作为该接收信号的特征图。

[0069] S103，将提取到的循环相关熵谱特征图输入训练好的深度森林网络对空频编码方式进行识别；具体过程为：

[0070] 深度森林网络的每一层都由级联的多个随机森林组成；将S102提取到的循环相关熵谱特征图作为输入，使用多粒度扫描对输入的循环相关熵谱特征图进行预处理得到特征向量，将得到的特征向量输入到级联的多个随机森林中进行训练；通过随机森林学习输入特征向量的特征信息，将该特征信息输入到深度森林网络的下一层；为了增强模型的泛化能力，每一层选取不同类型的随机森林，以适应于不同大小的数据集，将上一步骤提取到的循环相关熵谱输入训练好的深度森林网络，得到最终的分类结果。

[0071] 一种实施上述Alpha稳定分布噪声下MIMO-OFDM空频编码方式识别的系统，包括：

[0072] 类间识别模块，用于步骤S101中，对接收信号进行基于分数低阶相关的空时频编码类间识别，识别出接收信号是否为空频编码方式；

[0073] 特征提取模块，用于步骤S102中计算接收信号的循环相关熵，提取循环相关熵谱特征图；

[0074] 空频编码方式识别模块，在步骤S103中，用于将步骤S102提取到的循环相关熵谱特征图投入到训练好的深度森林网络实现对空频编码方式的分类识别。

[0075] 一种实施如上述Alpha稳定分布噪声下MIMO-OFDM空频编码方式识别的设备，包括存储器和处理器，所述存储器存储有计算机程序，所述计算机程序被所述处理器执行时，能够实现步骤S101至步骤S103所述Alpha稳定分布噪声下MIMO-OFDM空频编码方式识别方法。

[0076] 一种接收用户输入程序的存储介质，所述存储介质存储的计算机程序被处理器执行时能够基于步骤S101至步骤S103所述Alpha稳定分布噪声下MIMO-OFDM空频编码方式识别方法，进行Alpha稳定分布噪声下MIMO-OFDM空频编码方式识别。

[0077] 下面结合仿真实验对本发明的技术效果作详细的描述。

[0078] 为了评估本发明的性能，进行仿真验证。在仿真实验中，考虑非高斯Alpha稳定噪声下采用STBC / SFBC编码的MIMO-OFDM系统，待识别的SFBC编码方式包括4种Θ∈{SM,AL,SFBC3,SFBC4}。接收信号可能采用的空时编码集和空频编码集分别为ΩSTBC＝{AL-STBC,STBC3,STBC4}和ΩSFBC＝{AL-SFBC,SFBC3,SFBC4}，仿真中每个GSNR下进行800次，DPN网络训练和测试集比例为8：2，epoch＝30。其他参数如下设置：OFDM块数block＝8000,子载波数N＝64,循环前缀v＝N / 4，特征指数α＝1.5，接收天线数Nr＝2，调制模式为QPSK。信道选择为未知的频率选择性无线信道，非高斯噪声选择标准α稳定分布噪声，图3给出了不同特征指数α下所提类间识别方法精度变化，子载波数N＝64,循环前缀v＝N / 4，特征指数α∈{1.1,1.3,1.5,2.0}，从图3中可以看出，平均识别精度随着Alpha稳定分布特征指数α的增大而提高，在GSNR＝2dB，α＝1.1时，平均识别概率是60％，而在α＝1.5时，平均识别概率提高到90％左右，因为随着特征指数α的不断增大，Alpha稳定分布的拖尾会逐渐变浅，脉冲强度随之降低，对有用信号的干扰也会变弱，所以识别性能会随着特征指数α的增大而变高。

[0079] 应当注意，本发明的实施方式可以通过硬件、软件或者软件和硬件的结合来实现。硬件部分可以利用专用逻辑来实现；软件部分可以存储在存储器中，由适当的指令执行系统，例如微处理器或者专用设计硬件来执行。本领域的普通技术人员可以理解上述的设备和方法可以使用计算机可执行指令和 / 或包含在处理器控制代码中来实现，例如在诸如磁盘、CD或DVD-ROM的载体介质、诸如只读存储器(固件)的可编程的存储器或者诸如光学或电子信号载体的数据载体上提供了这样的代码。本发明的设备及其模块可以由诸如超大规模集成电路或门阵列、诸如逻辑芯片、晶体管等的半导体、或者诸如现场可编程门阵列、可编程逻辑设备等的可编程硬件设备的硬件电路实现，也可以用由各种类型的处理器执行的软件实现，也可以由上述硬件电路和软件的结合例如固件来实现。

[0080] 以上所述，仅为本发明的具体实施方式，但本发明的保护范围并不局限于此，任何熟悉本技术领域的技术人员在本发明揭露的技术范围内，凡在本发明的精神和原则之内所作的任何修改、等同替换和改进等，都应涵盖在本发明的保护范围之内。

Claims

1. A method for identifying the space-frequency coding scheme in a MIMO-OFDM system under Alpha stable distributed noise, characterized in that, Specifically, the following steps are included: Step 1: Perform inter-class identification of the received signal from the receiving antenna based on fractional low-order time-delay correlation space-time-frequency coding; 1.1) The first The signal samples received by each receiving antenna are represented as follows: in, This represents the channel parameters between the transmitting and receiving antennas. Indicates the number of propagation paths. Alpha-stable distributed noise; 1.2) To address the impact of alpha-stabilized noise on the signal, fractional low-order processing is applied to the received signal to suppress the influence of non-Gaussian noise on the useful signal. Fractional low-order time-delay correlations are calculated for different receiving antennas: in, , It is the block length of the encoding method. Representing different time delays, Indicates the length of the OFDM. This indicates the received signals from different receiving antennas; In the time-delay correlation function, find the peak value and extract the time delay value corresponding to the peak value as the peak feature, and draw the time-delay correlation peak feature map; 1.3) Following this, extract the multi-delay time. The time-delay related peak feature map is divided into training and test sets. The feature map samples in the training set are used to train a dual-channel network DPN to identify the time-delay related peak feature map. During the process of the DPN network identifying the time-delay related peak feature map, the output is split into two paths. One path is accumulated with the original input features to form a residual structure, reducing the redundancy of the original input features. The other path is connected in parallel with the original input features, so that the current network layer can directly obtain the output of the previous network layer and further extract deeper features from the output, thereby improving the classification accuracy of the model and realizing inter-class recognition of spatiotemporal coding. Step two: For the received signals from different receiving antennas identified as having space-frequency coding in step one, calculate the cyclic correlation entropy of the received signals and extract the cyclic correlation entropy spectral feature map of the received signals; the specific process is as follows: For signals received by different receiving antennas, the correlation entropy is defined as follows: Expressed in Fourier series form, the cyclic entropy function is: Taking the Fourier transform of the cyclic correlation entropy, we can further obtain the cyclic correlation entropy spectral function: Based on this, a fixed cycle frequency is taken. The cyclic correlation entropy spectrum slice is used as a feature map of the received signal; Step three, the specific process of inputting the cyclic correlation entropy spectrum feature map extracted in step two into the trained deep forest network to identify the space-frequency coding method is as follows: Each layer of the deep forest network consists of multiple cascaded random forests. The cyclic correlation entropy spectrum feature map extracted in step two is used as input. Multi-granularity scanning is used to preprocess the input cyclic correlation entropy spectrum feature map to obtain a feature vector. This feature vector is then input into the cascaded random forests for training. The random forests learn the feature information of the input feature vector, and this feature information is then input into the next layer of the deep forest network. To enhance the model's generalization ability, different types of random forests are selected for each layer to adapt to datasets of different sizes. The cyclic correlation entropy spectrum extracted in step two is then input into the trained deep forest network to obtain the final classification result.

2. A system for implementing the MIMO-OFDM space-frequency coding scheme identification method under Alpha stable distributed noise as described in claim 1, characterized in that, The MIMO-OFDM space-frequency coding scheme identification system under Alpha stable distributed noise includes: The inter-class identification module is used in step one to perform inter-class identification of the received signal based on fractional low-order correlation space-time-frequency coding, and to identify whether the received signal is space-frequency coded. The feature extraction module is used in step two to calculate the cyclic correlation entropy of the received signal and extract the cyclic correlation entropy spectral feature map. In step three, the space-frequency coding mode identification module is used to input the cyclic correlation entropy spectrum feature map extracted in step two into the trained deep forest network to achieve classification and identification of space-frequency coding modes.

3. An apparatus for implementing the MIMO-OFDM space-frequency coding scheme identification method under Alpha stable distributed noise as described in claim 1, characterized in that, It includes a memory and a processor. The memory stores a computer program, which, when executed by the processor, enables the identification method of MIMO-OFDM space-frequency coding under Alpha stable distributed noise as described in claim 1.

4. A storage medium for receiving user input, characterized in that, When the computer program stored in the storage medium is executed by the processor, it can identify the MIMO-OFDM space-frequency coding mode under Alpha stable distributed noise based on the MIMO-OFDM space-frequency coding mode identification method under Alpha stable distributed noise as described in claim 1.

Images