[go: up one dir, main page]
More Web Proxy on the site http://driver.im/

CN110222748B - OFDM radar signal recognition method based on 1D-CNN multi-domain feature fusion - Google Patents

OFDM radar signal recognition method based on 1D-CNN multi-domain feature fusion Download PDF

Info

Publication number
CN110222748B
CN110222748B CN201910446014.9A CN201910446014A CN110222748B CN 110222748 B CN110222748 B CN 110222748B CN 201910446014 A CN201910446014 A CN 201910446014A CN 110222748 B CN110222748 B CN 110222748B
Authority
CN
China
Prior art keywords
radar signal
layer
ofdm radar
ofdm
network structure
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Fee Related
Application number
CN201910446014.9A
Other languages
Chinese (zh)
Other versions
CN110222748A (en
Inventor
葛鹏
金炜东
张文强
李冰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Southwest Jiaotong University
Original Assignee
Southwest Jiaotong University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Southwest Jiaotong University filed Critical Southwest Jiaotong University
Priority to CN201910446014.9A priority Critical patent/CN110222748B/en
Publication of CN110222748A publication Critical patent/CN110222748A/en
Application granted granted Critical
Publication of CN110222748B publication Critical patent/CN110222748B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
    • G01S13/88—Radar or analogous systems specially adapted for specific applications
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/25—Fusion techniques
    • G06F18/253—Fusion techniques of extracted features
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Remote Sensing (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Biomedical Technology (AREA)
  • Software Systems (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Mathematical Physics (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Electromagnetism (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

The invention discloses an OFDM radar signal identification method based on 1D-CNN multi-domain feature fusion, which comprises the following specific steps: 1. performing feature learning on the OFDM radar signal by adopting a two-path one-dimensional convolutional neural network structure, and respectively learning time domain and frequency domain features of the OFDM radar signal; 2. fusing the OFDM radar signal characteristics learned by the time domain characteristic network structure and the frequency domain characteristic network structure to form a multi-domain characteristic fusion network structure; 3. inputting an OFDM radar signal data set into a multi-domain feature fusion network structure for model training, and obtaining a radar signal classification model through training; 4. implanting the radar signal classification model obtained by training in the step 3 into an industrial control computer, and realizing the identification of the OFDM radar signal modulation mode of the real environment through the model; the method is more beneficial to quickly learning the deep-level characteristics of the characterization signals, and has better identification effect on the OFDM radar signal modulation mode.

Description

基于1D-CNN多域特征融合的OFDM雷达信号识别方法OFDM radar signal recognition method based on 1D-CNN multi-domain feature fusion

技术领域technical field

本发明属于雷达识别领域，具体涉及一种基于1D-CNN多域特征融合的OFDM雷达信号识别方法。The invention belongs to the field of radar identification, and in particular relates to an OFDM radar signal identification method based on 1D-CNN multi-domain feature fusion.

背景技术Background technique

由于无线电技术的发展，越来越多的新型调制雷达信号出现在电磁环境中，新体制雷达辐射源具有低截获性和抗干扰能力强等特点，被经常使用在电子作战中。而正交频分复用(Orthogonal Frequency Division Multiplexing，OFDM)雷达信号作为一种新体制雷达具有突出的优点：灵活的波形设计、距离分辨率高、良好的测量精度、良好的杂波抑制能力和低截获能力等。在信息化战场日益复杂的今天，OFDM雷达信号已经被广泛应用在雷达系统中，但对OFDM雷达信号进行电子侦察具有较大的困难，所以实现OFDM雷达信号的调制识别具有非常重要的意义。在电子侦查环境中，对OFDM雷达信号的识别，不仅需要识别出信号是否属于OFDM雷达系统，而且还需识别出是哪一种调制方式的OFDM雷达信号。Due to the development of radio technology, more and more new modulated radar signals appear in the electromagnetic environment. The new radar radiation source has the characteristics of low interception and strong anti-interference ability, and is often used in electronic warfare. As a new radar system, Orthogonal Frequency Division Multiplexing (OFDM) radar signal has outstanding advantages: flexible waveform design, high distance resolution, good measurement accuracy, good clutter suppression ability and Low interception capability, etc. In today's increasingly complex information battlefield, OFDM radar signals have been widely used in radar systems, but it is difficult to conduct electronic reconnaissance of OFDM radar signals, so it is of great significance to realize the modulation recognition of OFDM radar signals. In the electronic reconnaissance environment, the identification of OFDM radar signals not only needs to identify whether the signal belongs to the OFDM radar system, but also needs to identify which modulation mode OFDM radar signal it is.

而一般采用卷积神经网络对OFDM雷达信号识别的方法，多用二维卷积神经网络学习OFDM雷达信号的时频域特征以及一维卷积神经网络对OFDM雷达信号进行特征学习，二维卷积神经网络对OFDM雷达信号特征学习时需要对雷达信号进行时频变换，且网络参数相对较多，相应地对训练电脑的硬件配置要求很高，会耗费大量时间在预处理和训练过程中。而一维卷积神经网络不能充分的学习到OFDM雷达信号的特征，在噪声复杂的情况下难以准确识别出OFDM雷达信号。Generally, the convolutional neural network is used to identify the OFDM radar signal, and the two-dimensional convolutional neural network is used to learn the time-frequency domain characteristics of the OFDM radar signal and the one-dimensional convolutional neural network is used to learn the features of the OFDM radar signal. When the neural network learns the characteristics of OFDM radar signals, it needs to perform time-frequency transformation on the radar signals, and the network parameters are relatively large. Correspondingly, the hardware configuration of the training computer is very high, and it will consume a lot of time in the preprocessing and training process. However, the one-dimensional convolutional neural network cannot fully learn the characteristics of OFDM radar signals, and it is difficult to accurately identify OFDM radar signals in the case of complex noise.

发明内容Contents of the invention

为解决上述技术问题，本发明提供一种基于1D-CNN多域特征融合的OFDM雷达信号识别方法，具体包括以下步骤：In order to solve the above technical problems, the present invention provides a method for recognizing OFDM radar signals based on 1D-CNN multi-domain feature fusion, which specifically includes the following steps:

步骤1：采用两路一维卷积神经网络结构对OFDN雷达信号进行特征学习，分别学习OFDM雷达信号的时域和频域特征。Step 1: Two-way one-dimensional convolutional neural network structure is used to learn the features of OFDN radar signals, and the time domain and frequency domain features of OFDM radar signals are learned respectively.

步骤2：将时域特征网络结构和频域特征网络结构学习到的OFDM雷达信号特征进行融合，形成多域特征融合网络结构。Step 2: Fuse the OFDM radar signal features learned by the time-domain feature network structure and the frequency-domain feature network structure to form a multi-domain feature fusion network structure.

步骤3：将OFDM雷达信号数据集，输入到多域特征融合网络结构中进行模型训练，通过训练得到雷达信号分类模型。Step 3: Input the OFDM radar signal data set into the multi-domain feature fusion network structure for model training, and obtain the radar signal classification model through training.

步骤4：将步骤3中训练得到的雷达信号分类模型，植入到工控计算机中，通过该模型实现真实环境的OFDM雷达信号调制方式识别。Step 4: Implant the radar signal classification model trained in step 3 into the industrial control computer, and realize OFDM radar signal modulation recognition in the real environment through this model.

进一步的，在步骤1中学习时域特征的网络结构具体为：输入层、卷积层、池化层、卷积层、GlobalAveragePooling1D、全连接层、Dropout。学习频域特征的网络结构具体为：输入层、Lambda、卷积层、池化层、卷积层、GlobalAveragePooling1D、全连接层、Dropout。其中Lambda层为定义的傅里叶变换层；Dropout层为防止网络的过拟合。Further, the network structure for learning time-domain features in step 1 is specifically: input layer, convolutional layer, pooling layer, convolutional layer, GlobalAveragePooling1D, fully connected layer, and Dropout. The network structure for learning frequency domain features is specifically: input layer, Lambda, convolutional layer, pooling layer, convolutional layer, GlobalAveragePooling1D, fully connected layer, and Dropout. Among them, the Lambda layer is the defined Fourier transform layer; the Dropout layer is to prevent the network from overfitting.

进一步的，在步骤2中，将时域特征和频域特征进行融合，融合的方式是通过Add层做加法；然后连接全连接层、Dropout、输出层，从而形成多域特征融合网络结构。Further, in step 2, the time-domain features and frequency-domain features are fused, and the way of fusion is addition through the Add layer; then the fully connected layer, Dropout, and output layer are connected to form a multi-domain feature fusion network structure.

进一步的，OFDM雷达信号包括相位编码(Phase Shift Keying，PSK)OFDM雷达信号(PSK-OFDM)，线性调频(Linear Frequency Modulation，LFM)OFDM雷达信号(LFM-OFDM)和分段线性调频(Multi-segment Linear Frequency Modulation，MLFM)OFDM雷达信号(MLFM-OFDM)三类。Further, the OFDM radar signal includes phase code (Phase Shift Keying, PSK) OFDM radar signal (PSK-OFDM), linear frequency modulation (Linear Frequency Modulation, LFM) OFDM radar signal (LFM-OFDM) and segmental linear frequency modulation (Multi- segment Linear Frequency Modulation, MLFM) OFDM radar signal (MLFM-OFDM) three types.

与现有技术相比，本发明的有益技术效果为：Compared with the prior art, the beneficial technical effect of the present invention is:

1D-CNN网络结构的收敛速度要快于2D-CNN网络结构，二维网络把一维数据变换为二维数据需要更多的时间进行处理，而一维网络只需要对数据进行一次傅里叶变换，其所需时间远远小于二维网络所需时间；采用二维网络时，会产生较多的参数，对训练网络的平台要求远高于一维网络，同样配置的训练平台的批量处理大小比二维网络更大，更有利于快速学到表征信号的深层次特征。同时对OFDM雷达信号调制方式识别时，获得了很好的识别效果。The convergence speed of the 1D-CNN network structure is faster than that of the 2D-CNN network structure. It takes more time for the two-dimensional network to convert one-dimensional data into two-dimensional data, while the one-dimensional network only needs to perform Fourier transformation on the data once. Transformation, the time required is much less than the time required by the two-dimensional network; when the two-dimensional network is used, more parameters will be generated, and the platform requirements for the training network are much higher than those for the one-dimensional network. The batch processing of the training platform with the same configuration The size is larger than the two-dimensional network, which is more conducive to quickly learning the deep features of the representation signal. At the same time, when identifying the modulation mode of OFDM radar signals, a good identification effect is obtained.

附图说明Description of drawings

图1为本发明多域特征融合网络结构示意图。Fig. 1 is a schematic diagram of the multi-domain feature fusion network structure of the present invention.

图2为本发明对三种雷达信号识别成功率曲线图。Fig. 2 is a curve diagram of the success rate of recognition of three kinds of radar signals according to the present invention.

具体实施方式detailed description

下面结合附图和具体实施方式对本发明作进一步详细的说明。The present invention will be further described in detail below in conjunction with the accompanying drawings and specific embodiments.

本发明的基本思想是在复杂密集的电磁环境中安装雷达数据专业采集设备，实时采集OFDM雷达时域信号，被采集的OFDM雷达时域信号经过(Convolutional NeuralNetworks,CNN)卷积神经网络学习得到分类模型，本发明在OFDM雷达信号的识别体系中，通过设计的网络结构对信号进行时域和频域特征学习得到分类模型，用测试数据集在训练好的模型上进行预测，得到识别结果。因OFDM雷达信号的主要调制特征在频域中和时域中表现出非常明显的区别，故采用两路训练学习OFDM雷达信号特征的网络结构具有较好的识别效果。The basic idea of the present invention is to install professional radar data acquisition equipment in a complex and dense electromagnetic environment to collect OFDM radar time-domain signals in real time, and the collected OFDM radar time-domain signals are classified through (Convolutional Neural Networks, CNN) convolutional neural network learning Model, in the recognition system of OFDM radar signals, the present invention uses the designed network structure to learn the time domain and frequency domain features of the signal to obtain a classification model, and uses the test data set to predict on the trained model to obtain the recognition result. Because the main modulation characteristics of OFDM radar signals are very different in the frequency domain and time domain, the network structure using two-way training to learn the characteristics of OFDM radar signals has a better recognition effect.

详细包括以下步骤：Details include the following steps:

步骤1：根据目前采用卷积神经网络对雷达信号识别的特点，采用两路一维卷积神经网络结构对OFDN雷达信号进行特征学习，分别学习OFDM雷达信号的时域和频域特征。本发明实验中训练轮次设置为50次，批量大小设置为64。OFDM雷达信号的调制识别主要包括三种，分别为：相位编码(Phase Shift Keying，PSK)OFDM雷达信号(PSK-OFDM)，线性调频(Linear Frequency Modulation，LFM)OFDM雷达信号(LFM-OFDM)和分段线性调频(Multi-segment Linear Frequency Modulation，MLFM)OFDM雷达信号(MLFM-OFDM)。Step 1: According to the current characteristics of radar signal recognition using convolutional neural network, two-way one-dimensional convolutional neural network structure is used to learn the features of OFDN radar signals, and the time domain and frequency domain features of OFDM radar signals are learned respectively. In the experiment of the present invention, the number of training rounds is set to 50, and the batch size is set to 64. The modulation recognition of OFDM radar signals mainly includes three types, namely: phase encoding (Phase Shift Keying, PSK) OFDM radar signal (PSK-OFDM), linear frequency modulation (Linear Frequency Modulation, LFM) OFDM radar signal (LFM-OFDM) and Multi-segment Linear Frequency Modulation (MLFM) OFDM radar signal (MLFM-OFDM).

学习时域特征的网络结构具体为：输入层、卷积层、池化层、卷积层、GlobalAveragePooling1D、全连接层、Dropout。包含了两个卷积层，两个池化层和一个全连接层，一共5个学习层。全连接层输出维度为64*256。The network structure for learning time-domain features is specifically: input layer, convolutional layer, pooling layer, convolutional layer, GlobalAveragePooling1D, fully connected layer, and Dropout. Contains two convolutional layers, two pooling layers and a fully connected layer, a total of 5 learning layers. The output dimension of the fully connected layer is 64*256.

学习频域特征的网络结构具体为：输入层、Lambda、卷积层、池化层、卷积层、GlobalAveragePooling1D、全连接层、Dropout，其中Lambda层为定义的傅里叶变换层。包含了两个卷积层，两个池化层，一个全连接层和一个自定义层，一共6个学习层。全连接层输出维度为64*256。The network structure for learning frequency domain features is: input layer, Lambda, convolutional layer, pooling layer, convolutional layer, GlobalAveragePooling1D, fully connected layer, Dropout, where the Lambda layer is the defined Fourier transform layer. Contains two convolutional layers, two pooling layers, a fully connected layer and a custom layer, a total of 6 learning layers. The output dimension of the fully connected layer is 64*256.

特征学习步骤描述如下：The feature learning steps are described as follows:

对特征进行学习的过程实质是由前向传播和反向传播构成，在训练过程中的反向传播实质是计算损失函数的梯度，根据实际输出值与理想输出值的误差修正网络参数。误差定义为The process of learning features is essentially composed of forward propagation and back propagation. The essence of back propagation in the training process is to calculate the gradient of the loss function, and correct the network parameters according to the error between the actual output value and the ideal output value. Error is defined as

Figure BDA0002073655040000031
Figure BDA0002073655040000031

其中En表示第n个样本的误差值，

Figure BDA0002073655040000032
表示当前神经元输出值，
Figure BDA0002073655040000033
为当前样本的真实标签值。where E n represents the error value of the nth sample,
Figure BDA0002073655040000032
Indicates the current neuron output value,
Figure BDA0002073655040000033
is the real label value of the current sample.

设前向传播公式为Let the forward propagation formula be

Figure BDA0002073655040000034
Figure BDA0002073655040000034

其中

Figure BDA0002073655040000035
表示第l层的第k个特征矩阵，
Figure BDA0002073655040000036
表示第l-1层的第i个特征矩阵与l层第k特征矩阵间的权值，
Figure BDA0002073655040000037
表示第l层的第k个特征矩阵的偏置项，Nl+1为输出层的神经元数目，则对权值和偏置项的调整方向为in
Figure BDA0002073655040000035
Represents the k-th feature matrix of the l-th layer,
Figure BDA0002073655040000036
Represents the weight between the i-th feature matrix of the l-1 layer and the k-th feature matrix of the l layer,
Figure BDA0002073655040000037
Indicates the offset item of the kth feature matrix of the l-th layer, N l+1 is the number of neurons in the output layer, then the adjustment direction of the weight and offset item is

Figure BDA0002073655040000038
Figure BDA0002073655040000038

Figure BDA0002073655040000039
Figure BDA0002073655040000039

其中

Figure BDA0002073655040000041
表示第l层第k个特征矩阵的残差。in
Figure BDA0002073655040000041
Represents the residual of the kth feature matrix of layer l.

反向传播计算公式为The backpropagation calculation formula is

Figure BDA0002073655040000042
Figure BDA0002073655040000042

其中

Figure BDA0002073655040000043
Figure BDA0002073655040000044
表示表示第l层的第k个特征矩阵，
Figure BDA0002073655040000045
表示对卷积核翻转180度，conv1Dz(·)为全卷积运算。权值和偏置项计算公式为in
Figure BDA0002073655040000043
Figure BDA0002073655040000044
Indicates the k-th feature matrix representing the l-th layer,
Figure BDA0002073655040000045
Indicates that the convolution kernel is flipped 180 degrees, and conv1Dz(·) is a full convolution operation. The calculation formula of weight and bias term is

Figure BDA0002073655040000046
Figure BDA0002073655040000046

Figure BDA0002073655040000047
Figure BDA0002073655040000047

步骤2：将时域特征网络结构和频域特征网络结构学习到的OFDM雷达信号特征进行融合，形成多域特征融合网络结构。融合方式是通过Add层做加法，融合后的输出维度为64*256。融合后连接全连接层、Dropout、输出层。全连接层的输出维度是64*3(64是信号的样本数量，3是信号的类别数，有PSK-OFDM雷达信号、LFM-OFDM雷达信号和MLFM-OFDM雷达信号3类)。Step 2: Fuse the OFDM radar signal features learned by the time-domain feature network structure and the frequency-domain feature network structure to form a multi-domain feature fusion network structure. The fusion method is addition through the Add layer, and the output dimension after fusion is 64*256. After fusion, connect the fully connected layer, dropout, and output layer. The output dimension of the fully connected layer is 64*3 (64 is the number of samples of the signal, 3 is the number of categories of the signal, there are 3 categories of PSK-OFDM radar signal, LFM-OFDM radar signal and MLFM-OFDM radar signal).

设计的多域特征融合网络结构中时域特征学习结构的学习层有5个，频域特征学习结构的学习层有6个，加上融合后的全连接层一共有13个学习层，具体结构如图1所示。In the designed multi-domain feature fusion network structure, there are 5 learning layers in the time-domain feature learning structure, 6 learning layers in the frequency-domain feature learning structure, and a total of 13 learning layers with the fully connected layer after fusion. The specific structure As shown in Figure 1.

步骤3：将大量的OFDM雷达信号数据集作为样本数据，输入到多域融合网络中进行模型训练。学习率通过keras框架下提供的优化器Adadelta函数进行自动调整，本实施中训练轮次设置为50次，批量大小设置为64。当信噪比在-10-10dB每隔2dB变化时，在每个信噪比下产生2000个样本，其中训练集的样本数量为1600，测试集样本数量为400，分别包含PSK-OFDM雷达信号、LFM-OFDM雷达信号和MLFM-OFDM雷达信号3类。采用一维卷积神经网络对雷达信号进行特征学习，训练速度和收敛速度都非常快，通过训练得到雷达信号分类模型。选取信噪比为-6dB时的结果具体说明：3类雷达信号，共产生2000个样本，设置产生信号的长度为2048，将OFDM雷达时域信号经过1D-CNN网路结构进行数据特征的学习从而达到对OFDM雷达信号调制方式的识别，最终识别率可达93％。识别成功率曲线如图2所示。可以看出信噪比超过-2dB时，对OFDM雷达信号识别率基本达到100％。Step 3: Take a large number of OFDM radar signal datasets as sample data and input them into the multi-domain fusion network for model training. The learning rate is automatically adjusted by the optimizer Adadelta function provided under the keras framework. In this implementation, the number of training rounds is set to 50, and the batch size is set to 64. When the signal-to-noise ratio changes every 2dB at -10-10dB, 2000 samples are generated at each signal-to-noise ratio, of which the number of samples in the training set is 1600, and the number of samples in the test set is 400, respectively containing PSK-OFDM radar signals , LFM-OFDM radar signal and MLFM-OFDM radar signal 3 categories. Using one-dimensional convolutional neural network to learn the features of radar signals, the training speed and convergence speed are very fast, and the radar signal classification model is obtained through training. The results when the signal-to-noise ratio is selected as -6dB are specifically explained: 3 types of radar signals, a total of 2000 samples are generated, and the length of the generated signal is set to 2048. The OFDM radar time-domain signal is passed through the 1D-CNN network structure for data feature learning. In this way, the modulation mode of OFDM radar signal can be identified, and the final identification rate can reach 93%. The recognition success rate curve is shown in Figure 2. It can be seen that when the signal-to-noise ratio exceeds -2dB, the recognition rate of OFDM radar signals basically reaches 100%.

步骤4：将步骤3中训练得到的雷达信号分类模型，植入到工控计算机中，通过该模型实现真实环境的OFDM雷达信号调制方式识别。输出测试结果，并制作报表。Step 4: Implant the radar signal classification model trained in step 3 into the industrial control computer, and realize OFDM radar signal modulation recognition in the real environment through this model. Output test results and make reports.

Claims (4)

1.一种基于1D-CNN多域特征融合的OFDM雷达信号识别方法，其特征在于，包括以下步骤：1. a kind of OFDM radar signal recognition method based on 1D-CNN multi-domain feature fusion, is characterized in that, comprises the following steps: 步骤1：采用两路一维卷积神经网络结构对OFDN雷达信号进行特征学习，分别学习OFDM雷达信号的时域和频域特征；Step 1: Use two-way one-dimensional convolutional neural network structure to learn the features of OFDN radar signals, and learn the time domain and frequency domain features of OFDM radar signals respectively; 学习时域特征的网络结构具体为：输入层、卷积层、池化层、卷积层、GlobalAveragePooling1D、全连接层、Dropout；学习频域特征的网络结构具体为：输入层、Lambda、卷积层、池化层、卷积层、GlobalAveragePooling1D、全连接层、Dropout；所述Lambda层为定义的傅里叶变换层；The network structure for learning time domain features is: input layer, convolution layer, pooling layer, convolution layer, GlobalAveragePooling1D, fully connected layer, Dropout; the network structure for learning frequency domain features is: input layer, Lambda, convolution layer, pooling layer, convolutional layer, GlobalAveragePooling1D, fully connected layer, Dropout; the Lambda layer is a defined Fourier transform layer; 步骤2：将时域特征网络结构和频域特征网络结构学习到的OFDM雷达信号特征进行融合，形成多域特征融合网络结构；Step 2: Fuse the OFDM radar signal features learned by the time-domain feature network structure and the frequency-domain feature network structure to form a multi-domain feature fusion network structure; 步骤3：将OFDM雷达信号数据集，输入到多域特征融合网络结构中进行模型训练，通过训练得到雷达信号分类模型；Step 3: Input the OFDM radar signal data set into the multi-domain feature fusion network structure for model training, and obtain the radar signal classification model through training; 步骤4：将步骤3中训练得到的雷达信号分类模型，植入到工控计算机中，通过该模型实现真实环境的OFDM雷达信号调制方式识别。Step 4: Implant the radar signal classification model trained in step 3 into the industrial control computer, and realize OFDM radar signal modulation recognition in the real environment through this model. 2.根据权利要求1所述的一种基于1D-CNN多域特征融合的OFDM雷达信号识别方法，其特征在于，所述步骤2具体为：将时域特征和频域特征进行融合，融合的方式是通过Add层做加法；然后连接全连接层、Dropout、输出层。2. a kind of OFDM radar signal recognition method based on 1D-CNN multi-domain feature fusion according to claim 1, is characterized in that, described step 2 is specifically: the time-domain feature and the frequency-domain feature are fused, the fused The method is to add through the Add layer; then connect the fully connected layer, Dropout, and output layer. 3.根据权利要求1或2所述的一种基于1D-CNN多域特征融合的OFDM雷达信号识别方法，其特征在于，所述Dropout层为防止网络的过拟合。3. a kind of OFDM radar signal recognition method based on 1D-CNN multi-domain feature fusion according to claim 1 and 2, is characterized in that, described Dropout layer is to prevent the over-fitting of network. 4.根据权利要求1所述的一种基于1D-CNN多域特征融合的OFDM雷达信号识别方法，其特征在于，所述OFDM雷达信号包括PSK-OFDM雷达信号、LFM-OFDM雷达信号和MLFM-OFDM雷达信号三类。4. a kind of OFDM radar signal recognition method based on 1D-CNN multi-domain feature fusion according to claim 1, is characterized in that, described OFDM radar signal comprises PSK-OFDM radar signal, LFM-OFDM radar signal and MLFM- There are three types of OFDM radar signals.
CN201910446014.9A 2019-05-27 2019-05-27 OFDM radar signal recognition method based on 1D-CNN multi-domain feature fusion Expired - Fee Related CN110222748B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910446014.9A CN110222748B (en) 2019-05-27 2019-05-27 OFDM radar signal recognition method based on 1D-CNN multi-domain feature fusion

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910446014.9A CN110222748B (en) 2019-05-27 2019-05-27 OFDM radar signal recognition method based on 1D-CNN multi-domain feature fusion

Publications (2)

Publication Number Publication Date
CN110222748A CN110222748A (en) 2019-09-10
CN110222748B true CN110222748B (en) 2022-12-20

Family

ID=67818418

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910446014.9A Expired - Fee Related CN110222748B (en) 2019-05-27 2019-05-27 OFDM radar signal recognition method based on 1D-CNN multi-domain feature fusion

Country Status (1)

Country Link
CN (1) CN110222748B (en)

Families Citing this family (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110569928B (en) * 2019-09-23 2023-04-07 深圳大学 Micro Doppler radar human body action classification method of convolutional neural network
CN110825852B (en) * 2019-11-07 2022-06-14 四川长虹电器股份有限公司 Long text-oriented semantic matching method and system
CN110826630B (en) * 2019-11-08 2022-07-15 哈尔滨工业大学 Feature-level fusion recognition method of radar jamming signal based on deep convolutional neural network
CN111652040A (en) * 2020-04-14 2020-09-11 广西万云科技有限公司 Method and system for recognizing human body activity based on radio frequency and signal self-adaptive convolutional neural network
CN111624570B (en) * 2020-04-30 2023-09-26 南京理工大学 Radar target identification method based on two-dimensional convolutional neural network
CN114070684B (en) * 2020-08-05 2024-02-09 西南交通大学 Modulation signal identification method based on two-dimensional convolutional neural network
CN112287784B (en) * 2020-10-20 2022-05-31 哈尔滨工程大学 A radar signal classification method based on deep convolutional neural network and feature fusion
CN112308133A (en) * 2020-10-29 2021-02-02 成都明杰科技有限公司 Modulation identification method based on convolutional neural network
CN114021603B (en) * 2021-10-25 2024-05-31 哈尔滨工程大学 Radar signal modulation mode identification method based on model re-parameterization
CN114509736B (en) * 2022-01-19 2023-08-15 电子科技大学 A Radar Target Recognition Method Based on UWB Electromagnetic Scattering Features
CN114114227B (en) * 2022-01-27 2022-05-31 安徽京淮健锐电子科技有限公司 Radar signal modulation type identification method based on dual-channel heterogeneous fusion network

Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104537387A (en) * 2014-12-16 2015-04-22 广州中国科学院先进技术研究所 Method and system for classifying automobile types based on neural network
CN104677987A (en) * 2015-03-15 2015-06-03 何赟泽 Eddy radar defect detecting, quantifying and imaging method and system
CN107220606A (en) * 2017-05-22 2017-09-29 西安电子科技大学 The recognition methods of radar emitter signal based on one-dimensional convolutional neural networks
CN107832794A (en) * 2017-11-09 2018-03-23 车智互联（北京）科技有限公司 A kind of convolutional neural networks generation method, the recognition methods of car system and computing device
EP3352112A1 (en) * 2017-01-20 2018-07-25 Nokia Technologies Oy Architecture adapted for recognising a category of an element from at least one image of said element
CN109036467A (en) * 2018-10-26 2018-12-18 南京邮电大学 CFFD extracting method, speech-emotion recognition method and system based on TF-LSTM
CN109254274A (en) * 2018-11-23 2019-01-22 哈尔滨工程大学 A kind of Radar emitter discrimination method based on Fusion Features
CN109597043A (en) * 2018-11-16 2019-04-09 江苏科技大学 Radar Signal Recognition method based on quantum particle swarm convolutional neural networks

Patent Citations (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104537387A (en) * 2014-12-16 2015-04-22 广州中国科学院先进技术研究所 Method and system for classifying automobile types based on neural network
CN104677987A (en) * 2015-03-15 2015-06-03 何赟泽 Eddy radar defect detecting, quantifying and imaging method and system
EP3352112A1 (en) * 2017-01-20 2018-07-25 Nokia Technologies Oy Architecture adapted for recognising a category of an element from at least one image of said element
CN107220606A (en) * 2017-05-22 2017-09-29 西安电子科技大学 The recognition methods of radar emitter signal based on one-dimensional convolutional neural networks
CN107832794A (en) * 2017-11-09 2018-03-23 车智互联（北京）科技有限公司 A kind of convolutional neural networks generation method, the recognition methods of car system and computing device
CN109036467A (en) * 2018-10-26 2018-12-18 南京邮电大学 CFFD extracting method, speech-emotion recognition method and system based on TF-LSTM
CN109597043A (en) * 2018-11-16 2019-04-09 江苏科技大学 Radar Signal Recognition method based on quantum particle swarm convolutional neural networks
CN109254274A (en) * 2018-11-23 2019-01-22 哈尔滨工程大学 A kind of Radar emitter discrimination method based on Fusion Features

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Shallow Convolutional Neural Networks for Acoustic Scene Classification;LU Lu等;《Wuhan University Journal of Natural Sciences》;20180313(第02期);169-173 *
深度学习优化算法研究;仝卫国等;《计算机科学》;20181115;411-415 *

Also Published As

Publication number Publication date
CN110222748A (en) 2019-09-10

Similar Documents

Publication Publication Date Title
CN110222748B (en) OFDM radar signal recognition method based on 1D-CNN multi-domain feature fusion
Xu et al. A deep learning method based on convolutional neural network for automatic modulation classification of wireless signals
CN109597043B (en) Radar signal recognition method based on quantum particle swarm convolutional neural network
Li et al. IncepTCN: A new deep temporal convolutional network combined with dictionary learning for strong cultural noise elimination of controlled-source electromagnetic data
CN110516718B (en) Zero sample learning method based on deep embedding space
CN111105008A (en) Model training method, data recognition method and data recognition device
CN111160176B (en) Fusion feature-based ground radar target classification method for one-dimensional convolutional neural network
CN110728360A (en) Micro-energy device energy identification method based on BP neural network
CN110378205A (en) A kind of Complex Radar Radar recognition algorithm based on modified CNN network
CN112001270A (en) Automatic target classification and recognition method for ground radar based on one-dimensional convolutional neural network
CN114595732B (en) Radar radiation source sorting method based on depth clustering
CN113780242A (en) Cross-scene underwater sound target classification method based on model transfer learning
CN116388933B (en) Communication signal blind identification system based on deep learning
CN111340132A (en) Machine olfaction mode identification method based on DA-SVM
CN117746194A (en) Robustness testing method and system for aerial remote sensing image target detection model
CN112686297B (en) A method and system for classifying the motion state of a radar target
CN111612130B (en) Frequency shift keying communication signal modulation mode identification method
CN117978595A (en) Automatic modulation classification method and device, equipment and computer readable storage medium
CN108805206A (en) Improved L SSVM establishing method for analog circuit fault classification
CN113359091B (en) Deep learning-based multi-kernel function aliasing radar radiation source identification method and system
Ruan et al. Automatic recognition of radar signal types based on CNN-LSTM
Zhang et al. [Retracted] Optimization of College English Classroom Teaching Efficiency by Deep Learning SDD Algorithm
Si et al. Multi‐label hybrid radar signal recognition based on a feature pyramid network and class activation mapping
Si et al. Radar signal recognition and localization based on multiscale lightweight attention model
CN116340846A (en) Aliasing modulation signal identification method for multi-example multi-label learning under weak supervision

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20221220