A dual-branch automatic modulation recognition method based on cross signal enhancement

Abstract

The present invention belongs to the technical field of wireless communication signal recognition, and specifically discloses a dual-branch automatic modulation recognition method based on cross-signal enhancement. First, the modulation signal is acquired, and the modulation signal is preprocessed to convert the modulation signal from I / Q data to A / P data; then, a modulation recognition model is built; the modulation recognition model includes four modules: cross-enhancement, channel feature fusion, temporal context feature capture, and classifier; the cross-enhancement module cross-enhances the I and Q channel data of the modulation signal and extracts I / Q features; the channel feature fusion module deeply fuses the I / Q features and the A / P features to obtain channel fusion enhancement features; the temporal context feature capture module extracts temporal context features from the channel fusion enhancement features, and the temporal context features are input into the classifier for classification to obtain the modulation type; finally, the modulation recognition model is trained and optimized, and the optimized modulation recognition model is used to recognize the modulation signal. This method can effectively reduce the imbalance between I / Q channels and improve recognition accuracy.

Description

Technical Field

[0001] The present invention belongs to the technical field of wireless communication signal recognition, and in particular is a dual-branch automatic modulation recognition method based on cross signal enhancement. Background Art

[0002] Automatic modulation recognition (AMR) is a key technology in cognitive radio. It aims to use prior knowledge of communication signals to identify modulation types, and plays an important role in spectrum perception, wireless interference identification, channel selection and management, etc.

[0003] Traditional AMR methods mainly include likelihood-based (LB) and feature-based (FB) methods. LB methods treat AMR as a hypothesis testing problem. While they can achieve high recognition accuracy, their computational complexity and weak robustness to model mismatch limit their real-time, low-cost application in practical communication scenarios. FB methods extract expert features from signals and input them into machine learning models for classification. While this method offers lower computational complexity, it often produces suboptimal solutions. However, the features extracted by FB methods are independent of the learning model, resulting in a lack of representativeness. Algorithm performance is heavily dependent on the training dataset, leading to poor model robustness.

[0004] Compared to traditional methods, deep learning-based AMR methods utilize neural networks to autonomously learn signal features, eliminating the feature design process and achieving higher recognition accuracy. Currently, signal features are typically represented in three types: I / Q data features, A / P data features, and image features. I / Q data features often contain amplitude and phase errors between the I and Q channels, leading to I / Q channel imbalance, necessitating I / Q channel imbalance calibration. Traditional I / Q channel imbalance calibration is primarily performed in the analog domain and the digital domain. Analog domain calibration is primarily achieved by improving device process capabilities and optimizing circuit layout, but is often limited by process limitations. Digital domain calibration uses back-end digital signal processing algorithms, such as the Gram-Schmidt algorithm or I / Q imbalance compensation methods based on adjacent channel detection. These methods are computationally complex. Existing deep learning-based AMR methods directly input I / Q signals into the model, failing to fully utilize the inherent interactions of the I / Q channels and ignoring the issue of I / Q channel imbalance.

[0005] In addition, existing deep learning-based AMR models usually use a single type of features or multimodal features (sequence features + image features) as input, which can lead to problems such as insufficient feature extraction or complex data preprocessing.

[0006] In view of the above problems, the present invention proposes a dual-branch automatic modulation recognition method based on cross signal enhancement. Summary of the Invention

[0007] In view of the deficiencies in the prior art, the technical problem to be solved by the present invention is to provide a dual-branch automatic modulation recognition method based on cross signal enhancement.

[0008] The present invention solves the technical problem by adopting the following technical solutions:

[0009] A dual-branch automatic modulation recognition method based on cross signal enhancement comprises the following steps:

[0010] Step 1: Obtain the modulated signal and pre-process it to convert the modulated signal from I / Q data to A / P data;

[0011] Step 2: Build a modulation recognition model; the modulation recognition model includes four modules: cross-enhancement, channel feature fusion, temporal context feature capture, and classifier; the cross-enhancement module cross-enhances the I and Q channel data of the modulated signal to extract I / Q features; the channel feature fusion module deeply fuses the I / Q features and A / P features to obtain channel fusion enhancement features; the temporal context feature capture module extracts temporal context features from the channel fusion enhancement features, and the temporal context features are input into the classifier for classification to obtain the modulation type;

[0012] The cross enhancement module includes an I channel feature extraction branch and a Q channel feature extraction branch with the same architecture but different weights; for the I channel feature extraction branch, the I channel data of the I / Q modulated signal is sequentially subjected to a one-dimensional convolution layer and reshaped to obtain an I channel feature; the I channel feature is subjected to a global average pooling layer to obtain an I channel global average pooling feature, and the I channel global average pooling feature is sequentially subjected to a fully connected layer, a ReLU activation function, a fully connected layer, and a Sigmoid activation function to obtain an I channel weight coefficient; similarly, the Q channel data of the I / Q modulated signal is subjected to a Q channel feature extraction branch to obtain a Q channel weight coefficient; after multiplying the I channel feature by the Q channel weight coefficient, the residual connection is performed with itself to obtain an I channel feature of information interaction; similarly, the Q channel feature of information interaction is obtained; the I channel feature of information interaction and the Q channel feature are spliced ​​to obtain an I / Q feature;

[0013] Step 3: Train and optimize the modulation recognition model, and use the optimized modulation recognition model to recognize the modulated signal.

[0014] Furthermore, in the channel feature fusion module, the A / P modulated signal passes through a two-dimensional convolution layer to obtain intermediate features. The intermediate features are residually connected with themselves after squeezing and stimulating attention mechanism to obtain A / P features; the I / Q features pass through a two-dimensional convolution layer and are spliced ​​with the A / P features in the channel dimension to obtain channel fusion features; the channel fusion features pass through a squeezing and stimulating attention mechanism, a two-dimensional convolution layer, a discard layer, and a squeezing and stimulating attention mechanism in sequence to obtain channel fusion enhanced features.

[0015] Furthermore, in the temporal context feature capture module, the channel fusion enhanced features are reshaped and then pass through the LSTM layer, the deformation LSTM layer and the LSTM layer in sequence to obtain the temporal context feature.

[0016] Furthermore, both LSTM layers contain 128 LSTM units, and the number of alternating rounds of the deformable LSTM layer is 5.

[0017] Compared with the prior art, the present invention has the following beneficial effects:

[0018] 1. The cross-enhancement module extracts the I-channel weight coefficient and the Q-channel weight coefficient respectively, cross-multiplies the I-channel weight coefficient with the Q-channel feature, and cross-multiplies the Q-channel weight coefficient with the I-channel feature to achieve interaction between different channel signals. It fully utilizes the interactive information between the I and Q channel data to enhance the signals of different channels to compensate for the difference between amplitude and phase, and performs imbalance calibration on the I / Q channels, thereby reducing the imbalance between the I / Q channels.

[0019] 2. The channel feature fusion module extracts A / P features and then concatenates them with I / Q features along the channel dimension, achieving deep feature fusion and leveraging the complementarity between I / Q and A / P data to extract more detailed and rich signal features. During the fusion process, squeezing and incentivized attention mechanisms are used to further extract and enhance features, thereby enhancing feature expression capabilities.

[0020] 3. For sequence signals, their features usually show long-term dependencies and rich temporal information, while the traditional LSTM model has limitations in extracting temporal features: the current input and hidden state are considered independent of each other and only interact within the gate unit. This interaction lacks effective use of contextual information before the gate unit is calculated, which may lead to the loss of some key information. Therefore, the temporal context feature capture module of the present invention utilizes the alternating stacking structure of LSTM and Mogrifier LSTM to gradually extract and enhance the temporal context information in the modulated signal. The Mogrifier LSTM adjusts the current input and hidden state to each other before the gate unit is calculated, so as to more comprehensively extract the temporal context information, more effectively capture the temporal features of the modulated signal, and enable the model to better learn the dynamic changes of the modulated signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is the overall flow chart;

[0022] Figure 2 It is the structural diagram of the modulation recognition model;

[0023] Figure 3 It is the recognition accuracy curve of different models under different signal-to-noise ratios;

[0024] Figure 4 It is the confusion matrix of the modulation recognition model for different modulation types when the signal-to-noise ratio is 12dB;

[0025] Figure 5 It is the confusion matrix of the CGDNN model for different modulation types when the signal-to-noise ratio is 12dB;

[0026] Figure 6 is the confusion matrix of the DAE model for different modulation types when the signal-to-noise ratio is 12dB;

[0027] Figure 7 This is the confusion matrix of the LSTM2 model for different modulation types when the signal-to-noise ratio is 12dB;

[0028] Figure 8 It is the confusion matrix of the MCLDNN model for different modulation types when the signal-to-noise ratio is 12dB;

[0029] Figure 9 It is the confusion matrix of the PET-CGDNN model for different modulation types when the signal-to-noise ratio is 12dB. DETAILED DESCRIPTION

[0030] Specific embodiments are given below in conjunction with the accompanying drawings. The specific embodiments are only used to further illustrate the technical solutions of the present invention in detail and are not intended to limit the scope of protection of the present application.

[0031] The present invention provides a dual-branch automatic modulation recognition method based on cross signal enhancement (hereinafter referred to as the method, see Figures 1 to 9 ), including the following steps:

[0032] Step 1: Obtain the modulated signal and preprocess it, converting the modulated signal from I / Q data to A / P data. Several modulated signal sequences form a data set.

[0033] The modulated signal is stored as I / Q data, and the I / Q modulated signal is expressed as:

[0034] (1)

[0035] Where, represents the I / Q modulated signal, Indicates the signal length, 2 indicates I and Q channels, and Respectively represent the I channel data and Q channel data of the I / Q modulated signal, represents the imaginary part;

[0036] Convert I / Q modulated signals to A / P modulated signals , the conversion formula is:

[0037] (2)

[0038] Where, and They represent the amplitude and phase of the A / P modulated signal respectively. The amplitude represents the size and amplitude of the signal, and the phase represents the periodic offset of the signal.

[0039] Step 2: Build a modulation recognition model; Figure 2 As shown in the figure, the modulation recognition model includes four modules: cross-enhancement (CE module), channel feature fusion (CFF module), temporal context feature capture (TCC module), and classifier module. The cross-enhancement module cross-enhances the I-channel and Q-channel data of the modulated signal to enhance the interaction information between the two channels and obtain I / Q features. The channel feature fusion module extracts A / P features and deeply fuses the I / Q features and A / P features in the channel dimension to obtain channel fusion enhanced features (CFF Fea). The temporal context feature capture module extracts key information in the time dimension from the channel fusion enhanced features to obtain temporal context features (TCC Fea). The classifier identifies the temporal context features and obtains the modulation type.

[0040] The cross enhancement module includes an I channel feature extraction branch and a Q channel feature extraction branch with the same architecture but different weights. For the I channel feature extraction branch, the I channel data of the I / Q modulated signal After feature extraction through the one-dimensional convolution layer, the I channel feature is obtained through shape reshaping. ;I channel characteristics After downsampling through the global average pooling layer, the I channel global average pooling feature is obtained ; I channel global average pooling feature After passing through the fully connected layer, ReLU activation function, fully connected layer and Sigmoid activation function in sequence, the I channel weight coefficient is obtained Similarly, the input of the Q channel feature extraction branch is the Q channel data of the I / Q modulated signal , the Q channel weight coefficient is obtained by the Q channel feature extraction branch ; Set the I channel feature and Q channel weight coefficient After multiplication, residual connection is performed to obtain the I channel feature of information interaction By cross-linking the I and Q channel information, the imbalance of amplitude and phase is corrected, so that they have similar amplitudes and corresponding phase relationships, which helps to restore the orthogonal relationship between the I / Q channels of the modulated signal; by adding the corrected features to the original features through residual connections, the model can better focus on the most relevant channels in the features, which helps to reduce the impact of irrelevant or noisy features and improve the expressiveness of the model. Similarly, the Q channel features of information interaction are obtained , respectively expressed as:

[0041] (3)

[0042] (4)

[0043] Where, Indicates Q channel characteristics;

[0044] The I channel features and Q channel features of the information interaction are spliced ​​to obtain the output of the cross enhancement module, namely the I / Q features .

[0045] A / P data is more suitable for representing time domain information because it can more comprehensively capture the amplitude and phase changes of the signal. I / Q data represents the signal in complex form, which supports the model in distinguishing nonlinear modulation. Therefore, by integrating these two types of data, the extracted features are more discriminative.

[0046] In the channel feature fusion module, the A / P modulated signal is extracted through a two-dimensional convolution layer to obtain the intermediate features. ; Intermediate features After squeezing and exciting the attention mechanism (SE), it is residually connected with itself to obtain the A / P feature ;I / Q characteristics After the two-dimensional convolution layer, and the A / P feature The channel-wise concatenation generates channel-fused features. These features are then subjected to a squeeze and incentivized attention mechanism, a two-dimensional convolutional layer, a dropout layer, and finally a squeeze and incentivized attention mechanism for further feature extraction and enhancement, resulting in enhanced channel-fused features, the output of the channel feature fusion module. The squeeze and incentivized attention mechanism enhances features, further enhancing their expressive power.

[0047] In the temporal context feature capture module, the channel fusion enhanced features undergo reshaping before passing through the LSTM layer, the Mogrifier LSTM layer, and the LSTM layer to obtain the temporal context features, which serve as the module's output. In this embodiment, both LSTM layers contain 128 LSTM units. The first LSTM layer is used to initially extract temporal features and capture dynamic changes in the modulated signal, while the second LSTM layer is used to further refine the temporal features. The Mogrifier LSTM layer has five alternating rounds to interactively adjust the input and hidden states before processing by the LSTM units, thereby more fully capturing the contextual information of the temporal features.

[0048] The classifier includes three fully connected layers, the first two fully connected layers are followed by a SELU activation function and a dropout layer, and the last fully connected layer is followed by a softmax activation function; the temporal context features are classified by the classifier and the modulation category of the modulated signal is output.

[0049] Step 3: Divide the data set into a training set and a validation set, use the training set to train the modulation recognition model, use the validation set to verify and optimize the trained modulation recognition model, and use the optimized modulation recognition model to recognize the modulated signal.

[0050] The Adam optimization algorithm and the categorical cross entropy loss function are used to train and validate the modulation recognition model. The initial learning rate is set to 0.001. During the training process, if the verification loss does not decrease within the time span of 5 consecutive epochs, the learning rate is multiplied by 0.5 to improve the training efficiency. If the verification loss still does not decrease after 30 cycles, the modulation recognition model with the smallest verification loss is saved as the optimized modulation recognition model.

[0051] To validate the performance of the modulation recognition model and simulate real-world conditions, experiments were conducted on the public RML2016.10a dataset. The RML2016.10a dataset contains 220,000 samples and includes eight digital modulation schemes: 8PSK, BPSK, CPFSK, GFSK, 4-PAM, 16-QAM, 64-QAM, and QPSK; and three analog modulation schemes: AM-DSB, AM-SSB, and WBFM. The signal-to-noise ratio ranges from -20dB to 18dB, with a 2dB increment.

[0052] In order to verify the effectiveness of the method of the present invention, the modulation recognition model (DB-CLDNN) of the present invention is compared with the current advanced AMR algorithms, including MCLDNN+AE-IEDNN [1] 、DAE [2] , PET-CGDNN [3] 、CGDNet [4] , MCLDNN [5] 、MCNET [6] and LSTM2 [7] , the test results are shown in Table 1. Figure 3 It is the recognition accuracy curve of different models under different signal-to-noise ratios.

[0053]

[0054] As can be seen from Table 1, the highest recognition accuracy and average recognition accuracy of the modulation recognition model of the present invention are higher than those of other models. Figure 3 It can be seen that after 0 dB, the modulation recognition model of the present invention performs well, and the recognition accuracy is significantly optimized compared with other models. At 12 dB, the highest recognition accuracy is 94.59%. Figures 4 to 9 The confusion matrices for different modulation types for the modulation recognition model of the present invention, CGDNN, DAE, LSTM2, MCLDNN, and PET-CGDNN at a signal-to-noise ratio of 12dB are shown. Comparison shows that the modulation recognition model of the present invention performs particularly well when processing easily confused modulation types (such as 16QAM and 64QAM, AM-DSB, and WBFM), achieving recognition accuracy rates of 74% and 79% for AM-DSB and WBFM, respectively, and 97% and 96% for 16-QAM and 64-QAM, respectively.

[0055] The above test results show that the modulation recognition model of the present invention has better recognition accuracy and optimal classification performance. This is because the present invention uses a cross-enhancement module to achieve interaction between I and Q channel signals, fully utilizing the interactive information between I and Q channel data to enhance the signals of different channels to reduce the imbalance between I / Q channels; the channel feature fusion module achieves a deep fusion of I / Q features and A / P features, fully utilizing the complementarity between I / Q and A / P data to extract more detailed and rich signal features; and the temporal context feature capture module uses the alternating stacking structure of LSTM and Mogrifier LSTM to gradually extract and enhance the temporal context information within the modulated signal, more effectively capturing the temporal features of the modulated signal.

[0056] The sources of each model are:

[0057] [1] F. Zhang, C. Luo, J. Xu, Y. Luo, "An Autoencoder-Based I / QChannel Interaction Enhancement Method for Automatic Modulation Recognition." IEEE Trans. Veh. Technol, vol. 72, no. 7, pp. 9620–9625, 1 July 2023, doi:10.1109 / tvt.2023.3248625.

[0058] [2] Z. Ke, H. Vikalo, "Real-Time Radio Technology and ModulationClassification via an LSTM Auto-Encoder." IEEE Trans. Wireless Commun, vol.21, no. 1, pp. 370–382, Jan. 2022, doi:10.1109 / twc.2021.3095855.

[0059] [3] F Zhang, C Luo, J Xu, Y Luo, “An Efficient Deep Learning Modelfor Automatic Modulation Recognition Based on Parameter Estimation andTransformation.” IEEE Commun Lett, vol. 25, no. pp. 3287–3290, 10, Oct. 2021,doi:10.1109 / lcomm.2021.3102656.

[0060] [4] JN. Njoku, ME. Morocho-Cayamcela and W. Lim, “CGDNet: EfficientHybrid Deep Learning Model for Robust Automatic Modulation Recognition,” IEEENetworking Letters, vol. 3, no. 2, pp. 47–51, June 2021, doi:10.1109 / lnet.2021.3057637.

[0061] [5] Xu J, Luo C, Parr G, Luo Y, “A spatiotemporal multi-channellearning framework for automatic modulation recognition”. EEE Wirel CommunLe, vol. 9, no. 10, pp 1629-1632, October. 2020, doi: 10.1109 / LWC.2020.2999453.

[0062] [6] T. Huynh-The, CH. Hua, QV. Pham and DS. Kim “MCNet: An EfficientCNN Architecture for Robust Automatic Modulation Classification,” IEEE CommunLett, vol. 24, no. 4, Apr. 2020, pp. 811815, doi:10.1109 / lcomm.2020.2968030.

[0063] [7] S. Rajendran, W. Meert, D. Giustiniano, V. Lenders and S. Pollin, "Deep Learning Models for Wireless Signal Classification with DistributedLow-Cost Spectrum Sensors," IEEE Trans. Cognit. Commun, vol. 4, no. 3, pp.433–445, Sept. 2018, doi:10.1109 / tccn.2018.2835460.

[0064] Any matters not described in the present invention are applicable to the prior art.

Claims

1. A dual-branch automatic modulation recognition method based on cross signal enhancement, characterized in that: The method comprises the following steps: Step 1: Obtain the modulated signal and pre-process it to convert the modulated signal from I / Q data to A / P data; Step 2: Build a modulation recognition model; the modulation recognition model includes four modules: cross-enhancement, channel feature fusion, temporal context feature capture, and classifier; the cross-enhancement module cross-enhances the I and Q channel data of the modulated signal to extract I / Q features; the channel feature fusion module deeply fuses the I / Q features and A / P features to obtain channel fusion enhancement features; the temporal context feature capture module extracts temporal context features from the channel fusion enhancement features, and the temporal context features are input into the classifier for classification to obtain the modulation type; The cross enhancement module includes an I channel feature extraction branch and a Q channel feature extraction branch with the same architecture but different weights; for the I channel feature extraction branch, the I channel data of the I / Q modulated signal is sequentially subjected to a one-dimensional convolution layer and reshaped to obtain an I channel feature; the I channel feature is subjected to a global average pooling layer to obtain an I channel global average pooling feature, and the I channel global average pooling feature is sequentially subjected to a fully connected layer, a ReLU activation function, a fully connected layer, and a Sigmoid activation function to obtain an I channel weight coefficient; similarly, the Q channel data of the I / Q modulated signal is subjected to a Q channel feature extraction branch to obtain a Q channel weight coefficient; after multiplying the I channel feature by the Q channel weight coefficient, the residual connection is performed with itself to obtain an I channel feature of information interaction; similarly, the Q channel feature of information interaction is obtained; the I channel feature of information interaction and the Q channel feature are spliced ​​to obtain an I / Q feature; Step 3: Train and optimize the modulation recognition model, and use the optimized modulation recognition model to recognize the modulated signal.

2. The dual-branch automatic modulation recognition method based on cross signal enhancement according to claim 1 is characterized in that: In the channel feature fusion module, the A / P modulated signal passes through a two-dimensional convolution layer to obtain intermediate features. The intermediate features are residually connected with themselves after the squeezing and excitation attention mechanism to obtain the A / P features; the I / Q features pass through a two-dimensional convolution layer and are spliced ​​with the A / P features in the channel dimension to obtain the channel fusion features; the channel fusion features pass through the squeezing and excitation attention mechanism, the two-dimensional convolution layer, the discard layer, the squeezing and excitation attention mechanism in turn to obtain the channel fusion enhanced features.

3. The dual-branch automatic modulation recognition method based on cross signal enhancement according to claim 1 or 2, characterized in that: In the temporal context feature capture module, the channel fusion enhanced features are reshaped and then pass through the LSTM layer, the deformation LSTM layer and the LSTM layer in sequence to obtain the temporal context feature.

4. The dual-branch automatic modulation recognition method based on cross signal enhancement according to claim 3 is characterized in that: Both LSTM layers contain 128 LSTM units, and the number of alternating rounds of the deformable LSTM layer is 5.

Images