electronics

Article
Signals Recognition by CNN Based on Attention Mechanism
Feng Tian, Li Wang * and Meng Xia

School of Communication and Information Engineering, Xi’an University of Science and Technology,
Xi’an 710054, China; [email protected] (F.T.); [email protected] (M.X.)
* Correspondence: [email protected]; Tel.: +86-029-1959-156-5212

Abstract: Automatic modulation recognition is a key technology in non-collaborative communication.

However, it is affected by complex electromagnetic environments, leading to low recognition accuracy.
To address this problem, this paper develops a ResNext signal recognition model based on an attention
mechanism. Firstly, a channel, including additive Gaussian white noise (AWGN), Rician multipath
fading, and clock offset, is created to simulate the complex electromagnetic environment, and
transmission-impaired modulated signals with various signal-to-noise ratios (SNRs) are synthesized
as a dataset. Secondly, using parallel stacked residual blocks of the same topology, instead of the
residual blocks of ResNet, and introducing the attention layer (CBAM), the types of feature extraction
are enriched without significantly increasing the parameter order of magnitude and avoiding the
over-fitting phenomenon caused by depth deepening. The results show that the signal recognition
method, based on the improved neural network framework, outperformed other deep learning
methods, and the recognition rate obtained of 10 different modulation types of signals was above 90%
at SNRs greater than 0 dB. The proposed signal recognition method achieved accurate recognition in
complex electromagnetic environments.

Keywords: complex electromagnetic environments; modulation recognition; residual blocks; ResNext;

CBAM






1. Introduction
Citation: Tian, F.; Wang, L.; Xia, M.
Signals Recognition by CNN Based
Automatic modulation recognition (AMR) is a pivotal technique of non-collaborative
on Attention Mechanism. Electronics
communication, which refers to automatically recognizing the modulation type of the sig-
2022, 11, 2100. https://doi.org/ nal by the receiver with limited or no prior information, providing the basis for subsequent
10.3390/electronics11132100 signal extraction and processing [1,2]. With the development of software-defined radio tech-
nology and the increasingly complex electromagnetic environment, the wireless channels
Received: 23 May 2022
of various kinds of noise and interference are gradually increasing. Traditional modulation
Accepted: 2 July 2022
recognition technology for communication signals has been unable to effectively carry out
Published: 5 July 2022
recognition, which represents a severe test for modulation recognition technology. There-
Publisher’s Note: MDPI stays neutral fore, how to efficiently and accurately realize the modulation recognition of communication
with regard to jurisdictional claims in signals has attracted more and more attention. Typical modulation recognition techniques
published maps and institutional affil- can broadly be divided into three categories: likelihood ratio recognition methods based on
iations. decision theory (LB), pattern recognition methods based on feature extraction (FB), and
deep learning recognition methods [3,4].
LB methods [5–9] make decisions by calculating the likelihood function of the re-
ceived signals and then comparing it to a certain threshold. Although LB methods can
Copyright: © 2022 by the authors.
minimize the error rate, their high computational complexity make them unsuitable for
Licensee MDPI, Basel, Switzerland.
applications such as unknown channels and clock offsets due to inaccurate internal clock
This article is an open access article
distributed under the terms and
sources between the transmitter and the receiver [2]. FB methods [10–13] require the man-
conditions of the Creative Commons
ual calculation of certain features of the received signal, such as the normalized central
Attribution (CC BY) license (https:// amplitude mean, standard deviation and kurtosis, normalized absolute instantaneous
creativecommons.org/licenses/by/ frequency, higher-order moments, higher-order volume accumulation, cyclic moments, and
4.0/). other features. Although the computational complexity of these features is relatively low,

Electronics 2022, 11, 2100. https://doi.org/10.3390/electronics11132100 https://www.mdpi.com/journal/electronics

Electronics 2022, 11, 2100 2 of 13

the characteristics are overly dependent on manual analysis for their selection. It is diffi-
cult to characterize multiple modulation types in complex electromagnetic environments.
Therefore, AMR is a very challenging task, especially when there is no prior information
about the received signal in non-collaborative communication [1].
In recent years, due to the development of neural network units, such as hidden
layers and non-linear activation, deep neural networks have been particularly prominent
in image classification, machine translation, and natural language processing [14–20], as
deep learning models can extract deeper information hidden in the data. At present, deep
learning is also progressively being applied to wireless communication and radio signal
processing. In modulation recognition, deep learning methods obtain better performance
than FB methods. For instance, ref. [21] used a convolutional neural network (CNN) for
modulation recognition, which was shown experimentally to be close to the FB method
and had greater flexibility in detecting various modulation types. To further improve
performance, ref. [22] introduced a densely connected network (DenseNet) to deepen the
feature propagation in deep neural networks by creating shortcut paths between different
layers of the network. A convolutional long short-term deep neural network (CLDNN) was
introduced in [23], which exploits the complementary nature of CNN and LSTM to combine
the architectures of CNN and long short-term memory (LSTM) into deep neural networks.
The main difference between the deep learning-based modulation recognition system and
the traditional modulation recognition system is that the feature extraction is automatically
learned by the neural network, which avoids the feature design process and is more
appropriate for application in non-collaborative communication scenario requirements.
Existing methods are often inaccurate in estimating the reality of complex electro-
magnetic environments and signal-to-noise ratios (SNRs). Realistic channel SNRs may
be unstable or rapidly changing under certain circumstances. Although the use of sim-
ulated and synthetic data sets for learning is not favored in deep learning, the field of
radio communication is a special case. As the real complex electronics environment is
quantified as much as possible, and as simulation methods are refined, the gap between
synthetic and real data sets will narrow. This will facilitate modulation identification in
complex electromagnetic environments. This paper simulates a complex electromagnetic
environment by constructing channels containing additive white Gaussian noise (AWGN),
Rician multipath fading, and clock offset. Modulated signals with 10 kinds of transmission
impairments under different SNRs are synthesized into a dataset. A network model based
on ResNext is then established, the residual blocks in the traditional ResNet are replaced
with a residual block of the same topology stacked in parallel and an attention layer CBAM
is introduced, connected behind each convolution block of ResNext. The impaired I/Q
signals are used directly as input, by increasing the dimension of ‘cardinality’ to extract
greater signal features and improve the accuracy of modulation recognition. Simulation
results show that the established network outperforms other neural networks.
The remainder of this paper is organized as follows. Section 2 describes the construc-
tion of the signal recognition model. Section 3 discusses the experimental results. Section 4
provides the conclusions.

2. System Models and Scenatios

AMR is an intermediate process that occurs between signal detection and receiver
demodulation. Compared with the traditional method, the structure of the proposed AMR
method in this paper is shown in Figure 1. The preprocessing in Figure 1 refers to the
sampling and quantization of the IF signal. The traditional AMR process, as shown in the
dashed box, contains the extraction, selection, and classifier of expert features, which is
replaced by the CNN method in this paper. Moreover, provided that the SNR range of
the communication channel is known, CNN can learn the characteristics adapted to the
corresponding conditions. This property means that the method proposed is unrelated to
SNR estimation.
Electronics 2022, 11, 2100 3 of 13

Received signal

Preprocessing

Feature
extraction
established method

ResNeXt+CBAM

Traditional AMR
This paper

methods
Feature SNR
selection estimation

Classifier

Demodulation

Figure 1. Automatic modulation recognition model comparison.

2.1. Signal Model

Modulation recognition can be expressed as a classification problem with N modula-
tions. This paper focuses on the modulated signal affected by the complex electromagnetic
environment; the received signal r (t) is described in Equation (1):

r (t) = s(t) + g(t) (1)

where g(t) is the additive white Gaussian noise (AWGN), s(t) is the transmit signals of
different modulation types, and SNR is defined as Ps /Pn (Ps is the signal power, Pn is the
noise power). The commonly used modulation methods are as follows:
When the transmit signal is a PSK or FSK signal, s(t) can be expressed in Equation (2)
as follows:
s(t) = [ Am ∑ an n(t − nTs )] cos(2π ( f c + f m )t + Ψ0 + Ψm ) (2)
n

where Am and an are the modulation amplitude and symbol sequence, respectively, n(t)
is the signal pulse, and Ts denotes the symbol period. f c and f m denote the carrier fre-
quency and modulation frequency. Ψ0 and Ψm denote the initial phase and modulation
phase, respectively.
When the transmit signals are M-QAM signals, which is slightly different from PSK
and FSK signals in that there are two quadrature carriers and the two carriers are modulated
by an and bn , respectively, s(t) can be expressed in Equation (3):

s(t) = [ Am ∑ an n(t − nTs )] cos(2π f c t + Ψ0 )

n
(3)
+[ Am ∑ bn n(t − nTs )] sin(2π f c t + Ψ0 )
n
Electronics 2022, 11, 2100 4 of 13

After determining the transmitting signal s(t), for the actual radio wave propagation
channel, the electromagnetic waves will be transmitted from different paths to the receiver
through reflections from multiple objects, creating a multipath effect. However, since there
are different transmission paths with different time delays, each propagation path will
change with time, and the interrelationship between the component fields involved in the
interference will also change with time, causing random changes in the synthetic wave-field,
and thus causing the fading of the total received field. In a multipath propagation scenario
with a strong path, the received signal is a statistical model of a multipath channel whose
impulse response amplitude follows Rician fading α(t).
A clock offset is caused by inaccurate internal time sources of the transmitter and
receiver. Clock offset causes the center frequency (used to down-convert the signal to
baseband) and the digital-to-analog converter (DAC) sample rate to be different from the
ideal value. Therefore, it is necessary to perform frequency offset f 0 and phase offset θ0 on
the signal based on the clock offset factor and the center frequency.
To simulate the complex electromagnetic environment, it is necessary to add Rician
multipath fading, frequency offset, and phase offset to the channel. The received sampled
signal r (t) can be re-expressed in Equation (4).

r (t) =α(t)e j(2π f0 t+θ0 (t)) s(t) + g(t) (4)

The purpose of modulation recognition is to determine the P(s(t) ∈ N (i )|r (t) ) after re-
ceiving the signal r (t), where N (i ) denotes the i-th modulation, and the goal is to recognize
the modulation type i from the received signal r (t). For simplicity, the received signal is
usually represented by its in-phase and quadrature I/Q components, which represent the
r (t) real and imaginary parts, respectively. For this purpose, the ResNext network, based
on the attention mechanism, is used to learn recognition, first processing the dataset to set
the network parameters and then calculating the recognition accuracy on the test dataset.

2.2. Network Model

Theoretically, the more layers of a deep learning model network there are, the better
the performance should be. In practice, as the number of network layers increases, the
gradient disappears and the gradient explodes, resulting in poor recognition performance.
However, due to the topology of the sub-modules, the ResNext structure can improve
accuracy without increasing parameter complexity, while also reducing the number of
hyper-parameters. As shown in Figure 2, with parallel stacking of blocks of the same
topology, instead of the three-layer convolution block of the original ResNet, the accuracy
of the model is improved without significantly increasing the parameter order. At the same
time, due to the same topology, the hyper-parameters are also reduced, which is convenient
for model porting.
To further improve the performance of the ResNext network model and select the
most discriminative features, an attention mechanism is introduced into the ResNext
network model to explore the dependencies between features. The attention mechanism
is a common data processing method in deep learning and is widely used in various
deep learning tasks, such as natural language processing, image recognition, and speech
recognition. Assembling features by assigning larger weights to some ‘significant’ features
not only reduces the parameters of the network but also improves the discriminative power
of the features.
A convolutional block attention module (CBAM) is an attention module that can
be inserted into convolutional neural networks. A CBAM will infer the attention map
along two independent dimensions (channel and spatial) accordingly, and then multiply
the attention map with the input feature map to perform adaptive feature optimization.
As shown in Figure 3, the output result of the convolution layer is first weighted by the
channel attention module, and then passed through the spatial attention module to obtain
the final result.
Electronics 2022, 11, 2100 5 of 13

256-d in
256-d in

256,1  1,64 256,1  1,4 256,1  1,4 256,1  1,4

Total
32
paths
4,3  3,4 4,3  3,4 4,3  3,4
64,3  3,64 ……

4,1  1,256 4,1  1,256 4,1  1,256

64,1  1,256

256-d out
256-d out

Figure 2. ResNet residual block (left) vs. ResNext residual block (right).

Channel Spatial
Attention Attention
Module Module

Input X X Refined
Feature Feature
Figure 3. CBAM structure diagram.

The channel attention module, as shown in Figure 4a, compresses the feature map
in the spatial dimension and then operates after obtaining a one-dimensional vector. The
input feature maps are compressed by the MaxPooling layer and the MeanPooling layer,
and then sent to the shared fully connected layer. Then the results of the shared fully
connected layer are summed and activated by the activation function sigmoid to obtain the
final channel attention weights MC ( F ), which can be expressed in Equation (5).
c c
MC ( F ) = σ (W1 (W0 ( Favg )) + (W1 (W0 ( Fmax ))) (5)

where F is the input feature mapping, and W0 and W1 represent the weight matrices of the
hidden layer and the fully connected layer, respectively.
The spatial attention module in Figure 4b can be regarded as channel compression,
performing MeanPooling and MaxPooling on the feature maps of the channel dimension.
The previously obtained feature maps (the number of channels is equal to 1) are merged to
obtain a two-channel feature map, and to obtain the spatial attention weight Ms ( F ), which
can be expressed in Equation (6).

MS ( F ) = σ ( f 7×7 ([ Favg
s s
; Fmax ])) (6)

where σ is the sigmoid activation operation and f 7×7 represents the kernel size of the convolution.
Electronics 2022, 11, 2100 6 of 13

MaxPool
Input
Feature Channel
F Shared MLP Attention
AvgPool
MC(F)
(a)

Conv layer

F [MaxPool,AvgPool] Spatial Attention

MS (F)
(b)
Figure 4. CBAM channel, spatial attention module.

2.3. CNN Signal Recognition Model Based on Attention Mechanism

The ResNext network introduces residual connections between convolutional blocks
and adds the output to the input of the convolutional blocks to optimize the training
process, overcoming the degradation problem of deep neural networks, and achieving
better recognition. The basic structure of the residual unit is shown in Figure 5, where x
represents the input of the first layer and the expectation output function is H ( x ), i.e., H ( x )
is the expected complex potential mapping; however, such a model would be very difficult
to train. Therefore, the learning objective is transformed into the learning of the identity
map; that is, the input x is approximated to the output H ( x ) to keep the accuracy in the
later layers without loss. Through shortcut connection, the input x is passed directly to
the output as the initial result, and the output is H ( x ) = F ( x ) + x. When F ( x ) = 0, then
H ( x ) = x, which is the identity mapping. Thus, the learning objective shifts from the
original learning complete output to the difference between the learning target value H ( x )
and x, which is the residual function F ( x ) = H ( x ) − x. Therefore, the training objective is
to approximate the residual function to 0 such that the accuracy does not decrease as the
network deepens.

Weight layers

F(x) Relu
x
Weight layers

F(x)+x

Relu

Figure 5. Residual unit.

Figure 6 shows the network structure designed for automatic feature extraction of
10 types of modulated signals. Based on ResNext, the CBAM module is introduced and
connected after each convolutional block of ResNext. It consists of four convolutional
modules and two fully connected layers, where each convolutional block consists of a down-
sampling layer and two ResNext residual blocks, and a MaxPooling layer. In each ResNext
Electronics 2022, 11, 2100 7 of 13

residual block, to avoid the gradient disappearance and slow network convergence caused
by the internal covariate shift during the training process, it is necessary to batch normalize
the activation values at the end. The original radio signal first enters the convolution
module through the input layer, and the convolution module inputs the extracted features
to the attention layer, weights the features in the attention layer, and finally inputs to the
next layer after the cascade processing of nodes. The final features are fed to fully connected
layers for subsequent classification. The first fully connected layer uses the Selu scaling
exponential linear unit activation function, and the second fully connected layer uses the
Softmax activation function with an output size of 10.
During the network training process, the cross-entropy loss function in Equation (7) is
chosen to evaluate the network.
N
1
Loss = −
N ∑ log(oM(i)) (7)
i =1

where N represents the number of training samples and oM (i ) represents the prediction
probability that the i-th sample belongs to class M (i ). The training process uses Adam
optimizer back-propagation to update all network parameters (including convolutional,
attention and fully connected layers).
input

Input stem
Down sampling

ResNext
Residual

ResNext Block 1 CBAM

Residual
Attention weights
Max Pooling

Block2 CBAM

Attention weights
...
256,11,4 256,11,4 256,11,4
Block 3 CBAM
Cardinality=16

4,33,4 4,33,4 4,33,4

Attention weights

4,11,256 4,11,256 4,11,256

Block 4 CBAM

Attention weights

Fully connected
layers,Selu
Relu

Batch Fully connected

normalization layers,Softmax

output
Figure 6. ResNext network model based on attention mechanism.
Electronics 2022, 11, 2100 8 of 13

3. Experiment and Analysis

3.1. Dataset
The realistically complex electromagnetic environment has many effects on the trans-
mitted signals, which makes it very difficult to simulate. Although the use of simulated and
synthetic data sets for learning is not favored in deep learning, the field of radio communi-
cation is a special case. Automatic modulation recognition models require large amounts
of labeled data for training, and, with a complex electromagnetic environment, this makes
it difficult to capture and label radio signals. As the realistically complex electromagnetic
environment is quantified as much as possible, introducing unknown scales, translations,
inflation and noise into our models, the gap between synthetic and real datasets will narrow
as the simulation methods become better. The use of simulated and synthetic datasets will
become an increasingly important tool when real data is difficult to capture or label.
The establishment of the dataset is shown in Figure 7. Bits are modulated and passed
through the channel, including AWGN, Rician fading and clock offset, to synthesize 10 types
of transmission-impaired IQ signals under various SNRs.To generate well-characterized
data sets, ten widely used modulations are selected: eight digital modulations and two
analog modulations. These include 8-PSK, BPSK, CPFSK, GFSK, 4-PAM, 16-QAM, 64-QAM,
QPSK for digital modulation, and WB-FM and AM-DSB for analog modulation.
Modulation

I/Q Impaired
2128
Bits signal Signal I/Q signalI 0 6 14 7 7 14 8 ... 8 0
Q 7 0 3 5 3 7 0 ... 0 1
Channel
AWGN

fading

Offset
Rician

Clock

Figure 7. Dataset Generation Process.

The dataset parameters are also set as close as possible to the realistically complex
electromagnetic environment. The sampling frequency f s affects the classification perfor-
mance only in the fading channel, and the Rician channel is modeled as a flat channel when
f s = 200 kHz. In the flat channel, the multipath structure of the channel allows the spectral
characteristics of the transmitted signal to be preserved on the receiver side. Taking into
account that the length of the signal sequence may have an impact on the results, the model
shown in Figure 6 is used to train on datasets with sequence lengths of 32, 128, and 256,
respectively. The over-sampling rate is 4. Other signal parameters are set as follows: the
SNR range is [−6, 4], and the step size is 2 dB. The center frequencies of digital and analog
modulation types are 902 MHz and 100 MHz, respectively. A total of 10,000 data points are
generated for each modulation type at each SNR, of which 80% are used for training, 10%
for validation, and 10% for testing. Figure 8 shows the effect of different sequence lengths
on the recognition results, so the dataset with sequence length L = 128 was chosen as the
input to the modulation recognition model.

3.2. Performance Comparison Analysis

To demonstrate the advantages of the model established in this paper, seven different
models of CNN [21], DenseNet [22], CLDNN [23], CNN1 [24], CNN2 [24] and ResNext
were trained using the same training set. Then the best model and the parameters after
training were saved, and the test set was used to obtain the recognition accuracy under
different SNRs.
Electronics 2022, 11, 2100 9 of 13


 




$FFXUDF\ 











     

615G%

Figure 8. Recognition accuracy for different sequence lengths.

From the comparative performance plots of various neural networks given in Figure 9,
it is clear that the recognition accuracy of all five networks increased with increase in
SNR due to there being less noise at high SNR. Under high SNR, ResNext + CBAM and
ResNext were able to achieve more than 90% recognition accuracy, which was the best
result among all models, due to the residual connection method and the topology of
ResNext’s group convolution. The ResNext model, with introduction of a CBAM attention
layer, was about 3% higher than the ordinary ResNext model, reflecting the benefit of
introducing a CBAM attention mechanism, which can help to extract effective features from
noise-contaminated data.




$FFXUDF\



5HV1H[W

 &11
'HQVH1HW
&/'11
5HV1H[W&%$0

 &11
&11

      

615G%

Figure 9. Average recognition accuracy under different SNRs.

Electronics 2022, 11, 2100 10 of 13

3.3. Analysis of Results

The training set was used to train the network shown in Figure 6, using the overall
accuracy (OA) to evaluate the recognition results of the built model. A total of 40 indepen-
dent experiments were conducted to complete the recognition of 10 types of modulated
signals with impaired transmission; the average OA of the classification results was 80.57%.
For further confirmation, the recognition results of the test set under several SNRs were
selected and presented in the form of confusion matrices.
The network parameters of this design are set as follows: the cross-entropy loss
function is used to view the change in the loss value during the training process, and the
optimizer selects Adam, where the learning rate is 0.001. The learning rate becomes 0 after
each parameter update, and the updated exponential decay rate is β 1 = 0.9, β 2 = 0.999.
The batch_size = 512 in the training process; after each iteration, the training data is re-
shuffled. When the training loss value is reached for 10 consecutive iterations without
reducing, the training is stopped, the trained model saved and the weight parameters are
used in testing.
The variation in the training accuracy function (dashed line) and the validation ac-
curacy function with the number of iterations is shown in Figure 10. It can be seen that
as the number of training iterations increases, the values of both the training accuracy
function and the validation accuracy function increase and finally reach a stable value at
40 iterations.

0.85
0.80
0.75
0.70
Accuracy

0.65
0.60
0.55
0.50
Validation accuracy
0.45 Training accuracy
0 10 20 30 40
Epochs
Figure 10. Training, validation accuracy iterative process.

The confusion matrix tested at SNRs ranging from −6 dB to 4 dB with a step size of
2 dB is shown in Figure 11; the numbers in the percentage column indicate the percentage
of the left signal type that were a misrecognition of the right signal type. Figure 11a shows
that SNR = −6 dB and OA = 54.86%. Figure 11b shows that SNR = −4 dB and OA = 69.87%.
Figure 11c shows that SNR = −2 dB and OA = 82.95%. Figure 11d shows that SNR = 0 dB
and OA = 89.19%. Figure 11e shows that SNR = 2 dB and OA = 92.78%. Figure 11f shows
that SNR = 4 dB and OA = 92.17%. It can be seen that the established model can accurately
identify most signal types, while for AM-DSB and WBFM, 16-QAM and 64-QAM have
poor recognition results. This is because, when the data set is generated, the observation
window is small, the information rate is low, and the correlation between the information
is small, so it is difficult to distinguish AM-DSB and WBFM. For 16-QAM and 64-QAM,
Electronics 2022, 11, 2100 11 of 13

due to the amplitude characteristics and phase information of the QAM signal, the I/Q
signals of these two signals are similar, and it is difficult to classify and recognize them.

(a) SNR = −6 (b) SNR = −4

(c) SNR = −2 (d) SNR = 0

(e) SNR = 2 (f) SNR = 4

Figure 11. Confusion matrix under different SNRs.

4. Conclusions
Due to the complexity of the wireless communication environment, it is difficult to
obtain high recognition accuracy in practical applications. In this paper, an attention-
mechanism-based ResNext model is established, and Matlab simulation is used to generate
a modulated signal with impaired transmission in a complex electromagnetic environment
Electronics 2022, 11, 2100 12 of 13

as the dataset. The final results show that more than 90% recognition accuracy was able to
be achieved when the SNR is greater than 0 dB.
The established network model outperforms the CNN, DenseNet, CLDNN, and
ordinary ResNext models. By introducing an attention mechanism, the recognition accuracy
is improved by about 3%, which enriches the types of signal feature extraction and avoids
the occurrence of overfitting and gradient disappearance phenomena. High accuracy is
achieved for modulation recognition in complex electromagnetic environments.

Author Contributions: Conceptualization, F.T. and M.X.; software, L.W.; validation, F.T., M.X. and
L.W.; investigation, F.T.; data curation, L.W.; writing—original draft preparation, L.W.; writing—
review and editing, F.T. and M.X. All authors have read and agreed to the published version of
the manuscript.
Funding: This research received no external funding.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Zhu, Z.; Nandi, A.K. Automatic Modulation Classification (Principles, Algorithms and Applications); John Wiley & Sons: Hoboken, NJ,
USA, 2014. [CrossRef]
2. Dobre, O.A.; Abdi, A.; Bar-Ness, Y.; Su, W. Survey of automatic modulation classification techniques: Classical approaches and
new trends. IET Commun. 2007, 1, 137–156. [CrossRef]
3. Huang, S.; Yao, Y.; Wei, Z.; Zhang, P. Automatic Modulation Classification of Overlapped Sources Using Multiple Cumulants.
IEEE Trans. Veh. Technol. 2017, 66, 6089–6101. [CrossRef]
4. Tian, J.; Pei, Y.; Huang, Y.D.; Liang, Y.C. Modulation-Constrained Clustering Approach to Blind Modulation Classification for
MIMO Systems. IEEE Trans. Cogn. Commun. Netw. 2018, 4, 894–907. [CrossRef]
5. Polydoros, A.; Kim, K. On the Detection and Classification of Quadrature Digital Modulations in Broad-Band Noise. IEEE Trans.
Commun. 1990, 38, 1199–1211. [CrossRef]
6. Beidas, B.F.; Weber, C.L. Asynchronous classification of MFSK signals using the higher order correlation domain. IEEE Trans.
Commun. 1998, 46, 480–493. [CrossRef]
7. Prokopios, P. Likelihood ratio tests for modulation classification. In Proceedings of the MILCOM 2000 Proceedings, 21st Century
Military Communications, Architectures and Technologies for Information Superiority (Cat. No.00CH37155), Los Angeles, CA,
USA, 22–25 October 2000.
8. Wen, W.; Mendel, J.M. Maximum-likelihood classification for digital amplitude-phase modulations. IEEE Trans. Commun. 2000,
48, 189–193. [CrossRef]
9. Majhi, S.; Gupta, R.; Xiang, W.; Glisic, S. Hierarchical Hypothesis and Feature based Blind Modulation Classification for Linearly
Modulated Signals. IEEE Trans. Veh. Technol. 2017, 66, 11057–11069. [CrossRef]
10. Swami, A.; Sadler, B.M. Hierarchical digital modulation classification using cumulants. IEEE Trans. Commun. 2000, 48, 416–429.
[CrossRef]
11. Soliman, S.S.; Hsue, S.Z. Signal Classification Using Statistical Moments. IEEE Trans. Commun. 1992, 40, 908–916. [CrossRef]
12. Mobasseri, B.G. Digital modulation classification using constellation shape. Signal Process. 2000, 80, 251–277. [CrossRef]
13. Grimaldi, D.; Rapuano, S.; De Vito, L. An Automatic Digital Modulation Classifier for Measurement on Telecommunication
Networks. IEEE Trans. Instrum. Meas. 2007, 56, 1711–1720. [CrossRef]
14. Lecun, Y.; Bengio, Y.; Hinton, G. Deep learning. Nature 2015, 521, 436. [CrossRef] [PubMed]
15. Fenn, P. The Deep Learning Revolution; Sejnowski, T.J., Ed.; The MIT Press: Cambridge, MA, USA, 2018; Volume 352, p. 24;
ISBN 978-0-262-03803-4.
16. Liu, W.; Wang, Z.; Liu, X.; Zeng, N.; Liu, Y.; Alsaadi, F.E. A survey of deep neural network architectures and their applications.
Neurocomputing 2017, 234, 11–26. [CrossRef]
17. Hinton, G.; Deng, L.; Yu, D.; Dahl, G.E.; Mohamed, A.R.; Jaitly, N.; Senior, A.; Vanhoucke, V.; Nguyen, P.; Sainath, T.N.; et al. Deep
Neural Networks for Acoustic Modeling in Speech Recognition: The Shared Views of Four Research Groups. IEEE Signal Process.
Mag. 2012, 29, 82–97. [CrossRef]
18. Le, L.; Zheng, Y.; Carneiro, G. Deep Learning and Convolutional Neural Networks for Medical Image Computing; Springer: Cham,
Switzerland, 2017.
19. Al-Ayyoub, M.; Nuseir, A.; Alsmearat, K.; Jararweh, Y.; Gupta, B. Deep learning for Arabic NLP: A survey. J. Comput. Sci. 2017,
26, 522–531. [CrossRef]
20. O’Shea, T.; Hoydis, J. An Introduction to Deep Learning for the Physical Layer. IEEE Trans. Cogn. Commun. Netw. 2017, 3, 563–575.
[CrossRef]
21. O’Shea, T.J.; Corgan, J.; Clancy, T.C. Convolutional Radio Modulation Recognition Networks. In Proceedings of the International
Conference on Engineering Applications of Neural Networks, Aberdeen, UK, 2–5 September 2016.
Electronics 2022, 11, 2100 13 of 13

22. Huang, G.; Liu, Z.; Laurens, V.; Weinberger, K.Q. Densely Connected Convolutional Networks. IEEE Comput. Soc. 2016.
[CrossRef]
23. Sainath, T.N.; Vinyals, O.; Senior, A.; Sak, H. Convolutional, Long Short-Term Memory, fully connected Deep Neural Networks.
In Proceedings of the ICASSP 2015—2015 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP),
South Brisbane, QLD, Australia, 19–24 April 2015.
24. Zheng, S.; Qi, P.; Chen, S.; Yang, X. Fusion Methods for CNN-Based Automatic Modulation Classification. IEEE Access 2019, 7,
66496–66504. [CrossRef]