An automatic modulation recognition method based on a lightweight spatio-temporal fusion combination network

Abstract

The application relates to an automatic modulation recognition method based on a lightweight space-time fusion combination network, and belongs to the field of automatic modulation recognition of communication signals. The method shapes I / Q components of a signal to be recognized into time streams and space streams respectively, inputs the time streams and the space streams into a time feature extraction channel and a space feature extraction channel respectively, extracts time features and space features, connects the extracted features and inputs the connected features into a classification module, and finally outputs a recognition result, i.e. a modulation type of the signal to be recognized. The lightweight space-time fusion combination network model provided by the application reduces the model parameter quantity and the floating point calculation quantity by reducing the convolution channel number and the hidden layer dimension of an LSTM network, while ensuring sufficient feature extraction, so that a high recognition accuracy is realized. Compared with the prior art, the application has the advantages of small parameter quantity, low calculation quantity, high recognition accuracy and the like, and is suitable for resource-limited devices.

Description

Technical Field

[0001] The present invention belongs to the field of automatic modulation recognition of communication signals and relates to an automatic modulation recognition method based on a lightweight space-time fusion combined network. Background Art

[0002] Automatic Modulation Recognition (AMR) is an important research area in wireless communications. Its goal is to automatically identify the modulation type of a received signal. AMR technology is widely used in communications, military, and signal monitoring. Traditional AMR algorithms can be divided into two categories: likelihood-based and feature-based. Likelihood-based methods are based on statistics and typically classify the modulation type by maximizing the likelihood function of the received signal under certain assumptions. Likelihood-based methods have high theoretical optimality and can achieve low error probabilities, especially under ideal channel conditions. However, because they require complex statistical modeling and calculations of the signal, they are computationally expensive for high-dimensional signals or complex scenarios, which can limit their real-time and robustness. Feature-based methods extract signal features such as amplitude, phase, instantaneous frequency, and waveform, and use these features to construct a recognition model to classify the modulation type. Compared with likelihood-based methods, feature-based methods have lower computational complexity and are more suitable for real-time processing in practical scenarios. However, their performance depends heavily on the quality of feature design and the choice of classifier, and may degrade in conditions with low signal-to-noise ratios or complex channel conditions.

[0003] In recent years, deep learning has developed rapidly, achieving remarkable success in many traditional fields. For example, convolutional neural networks (CNNs) have demonstrated outstanding performance in image processing, bringing breakthroughs to tasks such as image classification, object detection, and semantic segmentation. In 2016, O'Shea et al. first introduced the CNN deep learning algorithm into automatic modulation recognition to classify the modulation modes of radio signals. The model demonstrated improved performance under low signal-to-noise ratio (SNR) conditions. O'Shea et al. also provided a benchmark dataset that comprehensively considers various harsh radio channel interference conditions, facilitating the comparison of automatic modulation recognition algorithms. Unlike traditional feature-based methods that separate feature extraction and classifiers, deep learning-based automatic modulation recognition algorithms can perform both feature extraction and classification tasks. CNN and RNN models are widely used in deep learning algorithms for automatic modulation recognition. Over time, CNN and RNN models have been combined in deep learning algorithms for automatic modulation recognition, further improving recognition accuracy. To explore temporal and spatial correlations, Wu et al. combined a CNN and LSTM network to form a CLDNN network model, with I / Q signal components as input. Zhang et al. comprehensively considered the I / Q and A / P representations of the signal, extracting features using a tandem CNN and LSTM respectively. The features learned from the two streams interacted pairwise, increasing feature diversity. Considering that amplitude and phase imbalance degrade the orthogonality between the I and Q channels, Xu et al. proposed an efficient multi-stream structure to process the I / Q multi-channel data of the original modulated signal. This structure, supplemented by independent I / Q channel inputs, fully extracted features, significantly improving modulation recognition accuracy. While this network achieved excellent performance, it had a large number of parameters and was computationally intensive, making it unsuitable for deployment on compute-constrained platforms. Ke et al. proposed a learning framework based on an LSTM-based autoencoder to extract stable and robust features and use the learned features to infer modulation type. The algorithm designed a compact neural network architecture that is easy to implement on low-cost computing platforms. Although this model solves the problem of a large number of model parameters, the network needs to calculate the phase and amplitude of the IQ components and normalize the data, which will bring additional delay and computational complexity.

[0004] In summary, current modulation recognition suffers from numerous challenges: large model parameters, heavy floating-point computation, low accuracy, and some models require preprocessing of the received I / Q components, which results in additional computational overhead and latency. Therefore, a lightweight, highly accurate modulation recognition network that does not require additional I / Q component preprocessing is urgently needed. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide an automatic modulation recognition method based on a lightweight spatiotemporal fusion combined network. This method can significantly reduce the number of model parameters while improving the recognition accuracy under different signal-to-noise ratio conditions. Unlike the prior art, the present invention does not require additional preprocessing operations on the I / Q components, but directly inputs them into the network model for automatic modulation recognition. The method provided by the present invention includes constructing and training a lightweight spatiotemporal fusion combined network model, obtaining an input model of the signal I / Q components to be identified, and outputting the modulation type after network calculation.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] An automatic modulation recognition method based on a lightweight spatiotemporal fusion combined network includes the following steps:

[0008] Obtaining I / Q components of the signal to be identified;

[0009] Reshaping the I / Q components into a time stream and a spatial stream respectively;

[0010] Inputting the time stream into a time feature extraction channel to extract time features;

[0011] Inputting the spatial stream into a spatial feature extraction channel to extract spatial features;

[0012] Connecting the temporal features and the spatial features;

[0013] The concatenated features are input into a classification module to output a recognition result, wherein the recognition result is the modulation type of the signal to be recognized.

[0014] Furthermore, the time feature extraction channel includes:

[0015] One layer of one-dimensional convolution;

[0016] A one-dimensional batch normalization layer;

[0017] Two-layer LSTM network.

[0018] Furthermore, the spatial feature extraction channel includes:

[0019] One layer of 2D convolution;

[0020] One 2D batch normalization layer;

[0021] A convolution block comprising three one-dimensional convolution layers and three one-dimensional batch normalization layers;

[0022] One-layer LSTM network;

[0023] One adaptive global average pooling layer.

[0024] Furthermore, the classification module includes:

[0025] One fully connected layer.

[0026] Furthermore, the I / Q components are respectively shaped into a time stream and a spatial stream, specifically:

[0027] The I / Q components are connected in series and in parallel respectively to form a time stream and a space stream.

[0028] Furthermore, the time stream is input into the time feature extraction channel, specifically:

[0029] The time stream is subjected to the one-dimensional convolution to obtain feature data with reduced step dimension;

[0030] The feature data is passed through the two-layer LSTM network to extract time information, and the feature data of the last time step is taken as the input of the concatenate layer.

[0031] Furthermore, the spatial stream is input into the spatial feature extraction channel, specifically:

[0032] The spatial stream is subjected to the two-dimensional convolution to extract spatial features, and then subjected to the convolution block to deeply extract spatial features to obtain feature data;

[0033] The feature data is extracted through the LSTM network, and then reshaped and passed through the adaptive global average pooling layer to obtain feature data, which is then input into the concatenate layer.

[0034] Furthermore, the temporal feature and the spatial feature are connected, specifically:

[0035] The temporal features and spatial features are connected through a concatenate layer.

[0036] Furthermore, the connected features are input into a classification module to output a recognition result, specifically:

[0037] The connected features are classified through the batch normalization layer and the fully connected layer to obtain the recognition result.

[0038] Furthermore, the parameters of the temporal feature extraction channel, the spatial feature extraction channel and the classification module are obtained through training.

[0039] The beneficial effects of the present invention are as follows: the present invention reshapes the I / Q and sends them to the time feature extraction channel and the spatial feature extraction channel respectively, then connects the features extracted from the time channel and the spatial channel using a concatenate layer and then obtains the classification result through a full connection. While fully extracting the time and spatial features of the I / Q components, the number of parameters and floating-point calculations are greatly reduced, and a high recognition accuracy is guaranteed. At the same time, the model can directly identify the I / Q signals received by the receiver without the need for additional preprocessing, thereby reducing additional calculations and delays.

[0040] Other advantages, objects, and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art upon examination of the following description or may be learned from practice of the present invention. The objects and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below with reference to the accompanying drawings, in which:

[0042] Figure 1 This is the structural diagram of the LW-STCLDNN model of the present invention;

[0043] Figure 2 This is the internal structure diagram of the convolution block ConvBlock of the present invention;

[0044] Figure 3 is the recognition accuracy of the signal under different signal-to-noise ratios of the LW-STCLDNN and MCDLNN models;

[0045] Figure 4 Confusion matrices for LW-STCLDNN and MCDLNN models at snr=4dB. DETAILED DESCRIPTION

[0046] The following describes the embodiments of the present invention by means of specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0047] Among them, the accompanying drawings are only for illustrative purposes and represent only schematic diagrams rather than actual pictures, and should not be understood as limiting the present invention. In order to better illustrate the embodiments of the present invention, some parts of the accompanying drawings may be omitted, enlarged or reduced, and do not represent the dimensions of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions may be omitted in the accompanying drawings.

[0048] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "left", "right", "front", "back", etc. indicating directions or positional relationships, they are based on the directions or positional relationships shown in the drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operate in a specific direction. Therefore, the terms describing the positional relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0049] This invention provides an automatic modulation recognition method based on a lightweight spatiotemporal fusion network. The method includes constructing and training a lightweight spatiotemporal fusion network (LW-STCLN) model, obtaining the I / Q components of the modulation signal to be identified, inputting the trained LW-STCLN model, and outputting the identified modulation type. The training steps of the LW-STCLN are as follows:

[0050] 1. Obtain a data set of I / Q component signals of different modulation types in an environment with characteristic transmission parameters similar to those of the real channel.

[0051] Specifically, for the received radio signal r(t):

[0052]

[0053] Automatic modulation recognition is an intermediate process between receiver detection and demodulation. Its goal is to predict the modulation scheme of the received signal, that is, the function g (·), then provides modulation information, from the received signal r ( t ) to recover the modulated signal s ( tThe I / Q components are the real and imaginary parts of the received signal r(t). The datasets include RadioML2016.10a and RadioML2016.10b, which were generated using simulated channels in GNU Radio. RadioML2016.10a includes 11 modulation types: 8PSK, AM-DSB, AM-SSB, BPSK, CPFSK, GFSK, 4-PAM, 16-QAM, 64-QAM, QPSK, and WBFM. Signal-to-noise ratios range from -20dB to 18dB, and there are 220,000 I / Q component signal samples. Compared to RadioML2016.10a, the RadioML2016.10b dataset includes one less modulation type, AM-SSB, and contains 1,200,000 more samples. Each sample in the two datasets is a 2x128 matrix, indicating I / Q dual-channel features and 128 sampling points.

[0054] 2. Encode the different modulation types of the acquired data set and divide the entire data set into training set, validation set, and test set according to the ratio of 8:1:1.

[0055] Specifically, the modulation types of the acquired dataset are encoded. The modulation type encoding results for the RadioML2016.10a dataset are as follows: "8PSK": 0,"BPSK": 1,"CPFSK": 2,"GFSK": 3,"PAM4": 4,"QAM16": 5,"QAM64": 6,"QPSK": 7,"AM-DSB": 8,"AM-SSB": 9,"WBFM": 10. The modulation type encoding results for the RadioML2016.10b dataset are as follows: "8PSK": 0,"BPSK": 1,"CPFSK": 2,"GFSK": 3,"PAM4": 4,"QAM16": 5,"QAM64": 6,"QPSK": 7,"AM-DSB": 8,"WBFM": 9.

[0056] 3. Build a lightweight spatiotemporal fusion combined network model.

[0057] Specifically, the model structure is as follows Figure 1 As shown in Figure 2, the lightweight spatiotemporal fusion combined network model includes a temporal feature extraction channel, a spatial feature extraction channel, and a classification module.

[0058] The temporal feature extraction pipeline consists of one 1D convolution layer with a kernel size of 2 and 16 kernels, one batch normalization layer, and two LSTM layers with both input and hidden dimensions of 16. The spatial feature extraction pipeline consists of one 2D convolution layer with a kernel size of 2 and a width of 1, and 16 kernels. It also consists of one convolution block containing three 1D convolutions and three batch normalization layers with a kernel size of 3 and 16 kernels, and one LSTM layer with both input and output feature dimensions of 16. The classification module consists of a single fully connected layer with 32 input features and an output feature size equal to the number of modulation types. All batch normalization layers have EPS set to 1e-5 and momentum set to 0.1.

[0059] Batch normalization is a regularization technique used in deep neural networks to reduce internal covariate shift, thereby accelerating training and improving model stability. It normalizes the input of each layer to have zero mean and unit variance, making model training more efficient. For a small batch input , the process of batch normalization is as follows:

[0060] a: Calculate the mean and variance:

[0061]

[0062] Where m is the batch size.

[0063] b: Normalization:

[0064]

[0065] in is a small constant that prevents division by zero.

[0066] c: Rescale and offset: in, and These are learnable parameters used for scaling and offsetting, and are fixed after training. The network uses batch normalization layers for the following purposes:

[0067] a. Accelerate training convergence: By normalizing the input of each layer, the problem of activation values ​​being too large or too small is avoided, making gradient propagation more stable.

[0068] b. Improve model robustness: Reduce the sensitivity of weight initialization to training, so that the model performs well under a wider range of hyperparameter settings.

[0069] c. Reduce overfitting: Although batch normalization itself is not a regularization method, its normalization effect and small batch noise can indirectly play a regularization role.

[0070] d. Reduce sensitivity to learning rate: Batch normalization allows the model to use larger learning rates without causing gradient explosion or gradient vanishing problems.

[0071] 4. Set the initial learning rate, learning rate adjustment strategy, training rounds, batch size, and loss function type.

[0072] Specifically, we use the Adam optimizer with an initial learning rate of 0.001 and a batch size of 512. If the validation loss does not decrease after 5 epochs during training, we adjust the learning rate with a decay factor of 0.8. The main purpose is to allow the optimizer to escape the local optimum and find a better solution. The training rounds are set to a maximum of 200. If the validation loss does not improve after 20 epochs, we stop training. The loss function uses the cross entropy loss function. For a sample x , whose true category label is y , the probability distribution predicted by the model is , the cross entropy loss function is defined as:

[0073]

[0074] C is the total number of categories, yi is the encoding of the true category, The model is i The predicted probability of the class.

[0075] 5. A lightweight spatiotemporal fusion combined network model is constructed by shaping the I / Q components in the training set into time stream and spatial stream inputs.

[0076] Specifically, I / Q can be expressed by the following formula:

[0077]

[0078]

[0079] The I / Q components are shaped into time streams and spatial streams as follows:

[0080]

[0081]

[0082] Where K represents the number of K samples, N Indicates the number of sampling points of the signal. According to the parameters of the open source dataset, Nis 128, so the shape of the spatial stream tensor is (1, 2, 128), and the shape of the temporal stream tensor is (1, 1, 256). The spatial stream needs to pass through the spatial channel to extract spatial features. The spatial stream (1, 2, 128) first passes through a two-dimensional convolution with a convolution kernel height of 2, a width of 1, and an output channel of 16. Among them, 128 means there are 128 time steps, and 2 means there are two features, I and Q, at each time step. Therefore, a convolution kernel with a height of 2 is used to help the model capture the spatial (feature) dependency in the data, and the output channel 16 is to capture different feature patterns in the input signal. The data shape obtained after a layer of two-dimensional convolution is feature data of (1, 16, 1, 128). In order to reduce the amount of calculation of the subsequent convolution block and minimize feature loss as much as possible, a two-dimensional maximum pooling of size (1, 2) is used. After the pooling layer, the time step dimension of the feature data is reduced by half. Then enter the convolution block with three layers of one-dimensional convolution (see Figure 2), the convolution kernel size of each one-dimensional convolution is 3, and the output channel is 16; the shape of the feature data does not change after processing by the convolution block. Its main function is to perform deeper feature extraction on the features extracted after the first layer of convolution, and then extract the temporal dependency in its sequence through a layer of LSTM network. Its input feature size and hidden layer dimension are both 16. Since convolution only extracts the spatial features of I / Q, the extracted spatial features are output as a time series. Therefore, LSTM is used for further spatial feature extraction, and the obtained feature data shape is (1, 64, 16). Then the Transpose function is used to exchange the dimensions to make the shape become (1, 16, 64). The purpose is to adapt to the adaptive global average pooling, so that the entire time series of each input channel is averaged and pooled, and finally the spatial feature data with a shape of (1, 16) is output. The time stream needs to pass through a temporal feature extraction channel to extract temporal features. Therefore, LSTM is primarily used to extract its temporal dependencies. First, to maintain the input step size of 128 and extract temporal features from multiple feature representations of the input signal, the time stream undergoes a one-dimensional convolution with a kernel size of 2 and an output channel size of 16. A two-layer LSTM network is then used, with both the input feature size and hidden layer dimension of 16. By stacking multiple LSTM layers, the model's expressive power is increased, enabling it to capture more complex and multi-layered time series dependencies and better extract the temporal features of the I / Q signal. Finally, the temporal features extracted from the time channel and the spatial features extracted from the spatial channel are concatenated using a concatenation algorithm with an input shape of (1, 32). The concatenation is then passed through a fully connected layer with an input feature size of 32 and an output feature size of 11 to obtain the classification result. Most traditional combined networks for automatic modulation recognition directly input the I / Q signals into convolution to extract their features, and then use the LSTM network to extract features after the convolution is complete. Although the two networks are combined to take advantage of their advantages, the accuracy is not significantly improved. At the same time, the hidden dimensions of the convolution channel and the LSTM network are large, resulting in a huge amount of model parameters and floating-point calculations, which is not conducive to deployment on resource-constrained devices. The combined network model proposed in the present invention reduces the number of model parameters by reducing the number of convolution channels and the hidden layer dimension of the LSTM network. At the same time, it fully considers the spatial feature extraction advantages of convolution and the temporal feature advantages of LSTM, shapes the input I / Q components into time streams and spatial streams, and then sends them to the time channel and spatial channel for feature extraction respectively. The dual-channel features are then concatenated and sent to the fully connected classification module for classification. While reducing the number of model parameters, it ensures that the features are fully extracted to achieve a higher accuracy.

[0083] 6. Calculate the training loss function and backpropagate to update the model parameters until the model converges.

[0084] 7. Input the test set into the trained lightweight spatiotemporal fusion combined network model to obtain the modulation recognition result.

[0085] In order to illustrate the effect of the present invention, a simulation experiment is conducted to further illustrate:

[0086] Simulation conditions: NVIDIA GeForce A6000 GPU is used for training, the network model is implemented using PyTorch, and the dataset uses the open source datasets RadioML 2016.10a and RadioML 2016.10b.

[0087] Simulation results: The model of the present invention is used to train on two training sets respectively, calculate the recognition accuracy of the test set, and draw a curve chart of the recognition accuracy changing with the signal-to-noise ratio. In order to highlight the advantages of the present invention, the network of the present invention is compared with the currently excellent combined network benchmark model MCLDNN. The MCLDNN network input adopts three-channel input, which are I / Q parallel, and a separate I path and a separate Q path. The three paths are respectively connected together after convolution using concatenate, and then pass through two layers of LSTM for time feature extraction, and then pass through two layers of fully connected layers, and finally classify and output the results. Figure 3 (a) shows the curve of the recognition accuracy of the LW-STCLN and MCLDNN model test set changing with the signal-to-noise ratio under the RadioML2016.10a dataset. Figure 3 (b) shows the curve of the recognition accuracy of the LW-STCLN and MCLDNN model test set changing with the signal-to-noise ratio under the RadioML2016.10b dataset. Figure 4 (a) Figure 4 (b) Figure 4 (c) Figure 4 (d) shows the confusion matrices for the LW-STCLN and MCLDNN models for a signal sample with SNR = 4dB in the RadioML2016.10a and RadioML2016.10b datasets, respectively. To demonstrate the advantages of our present invention over other benchmark models, we compared the parameters, floating-point calculations, and recognition accuracy of our present invention with those of the CNN1, CNN2, CNN_LSTM, and LSTM models. The results are shown in Table 1.

[0088] Table 1 Comparison of the parameters, floating-point calculation number, and recognition accuracy of the proposed method and the current high-recognition rate model on the RadioML2016.10a dataset.

[0089] Table 1

[0090]

[0091] Depend on Figure 3 (a) and Figure 3 (b) It can be seen that the accuracy of the model of the present invention is basically close to that of the MCLDNN model in both data sets. Figure 4 (a) and Figure 4 (b) It can be seen that the LW-STCLDNN model has confusion with 16-QAM and 64-QAM signals and WBFM signals on the RML2016.10a dataset, but other modulation types can be classified relatively correctly, which is the same problem as the MCLDNN model. Figure 4 (c) and Figure 4 (d) As can be seen, the LW-STCLDNN model significantly improves 16-QAM and 64-QAM signal recognition confusion on the RML2016.10b dataset, but WBFM recognition confusion still exists. Table 1 shows that while maintaining accuracy close to that of the excellent combined network model MCLDNN, the proposed model significantly reduces both the number of parameters and the number of floating-point computations compared to MCLDNN. Furthermore, compared to other benchmark models, it not only has certain advantages in terms of parameter count and floating-point computations, but also exceeds these benchmark models in recognition accuracy.

[0092] In summary, this invention leverages the strengths of CNN and LSTM networks to extract temporal and spatial features from I / Q signal components for modulation recognition. Simulation results demonstrate that while maintaining recognition accuracy comparable to current leading combined network models, this invention significantly reduces the number of model parameters and floating-point computations, demonstrating exceptional efficiency and practicality. Compared to other benchmark models, this invention not only offers significant advantages in model complexity but also achieves further improvements in recognition accuracy.

[0093] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions, which should all be included in the scope of the claims of the present invention.

Claims

1. An automatic modulation recognition method based on a lightweight spatiotemporal fusion combined network, characterized by: The following steps are involved: Obtaining I / Q components of the signal to be identified; Reshaping the I / Q components into a time stream and a spatial stream respectively; Inputting the time stream into a time feature extraction channel to extract time features; Inputting the spatial stream into a spatial feature extraction channel to extract spatial features; Connecting the temporal features and the spatial features; Inputting the concatenated features into a classification module to output a recognition result, wherein the recognition result is the modulation type of the signal to be identified; The time feature extraction channel includes: One layer of one-dimensional convolution; A one-dimensional batch normalization layer; Two-layer LSTM network; The spatial feature extraction channel includes: One layer of 2D convolution; One 2D batch normalization layer; A convolution block comprising three one-dimensional convolution layers and three one-dimensional batch normalization layers; One-layer LSTM network; One layer of adaptive global average pooling; The classification module includes: One fully connected layer; The I / Q components are respectively shaped into a time stream and a spatial stream, specifically: Connecting the I / Q components in series and in parallel respectively to form a time stream and a space stream; The time stream input time feature extraction channel is specifically: The time stream is subjected to the one-dimensional convolution to obtain feature data with reduced step dimension; Pass the feature data through the two-layer LSTM network to extract time information, and take the last time step feature data as the concatenate layer input; The spatial stream is input into the spatial feature extraction channel, specifically: The spatial stream is subjected to the two-dimensional convolution to extract spatial features, and then subjected to the convolution block to deeply extract spatial features to obtain feature data; The feature data is extracted through the LSTM network, and then reshaped and passed through the adaptive global average pooling layer to obtain feature data, which is then input into the concatenate layer; The connecting of the temporal feature and the spatial feature is specifically as follows: Connecting the temporal features and spatial features through a concatenate layer; The connected features are input into a classification module to output a recognition result, specifically: The connected features are classified through the batch normalization layer and the fully connected layer to obtain the recognition result.

2. The automatic modulation recognition method based on a lightweight spatiotemporal fusion combined network according to claim 1 is characterized in that: The parameters of the temporal feature extraction channel, the spatial feature extraction channel and the classification module are obtained through training.

Images