TYPE Original Research

PUBLISHED 21 August 2023

DOI 10.3389/frsip.2023.1197590

Detection of OFDM modulations

OPEN ACCESS based on the characterization in
EDITED BY
M. L. Dennis Wong,
Heriot-Watt University Malaysia, Malaysia
the phase diagram domain
REVIEWED BY
Mehdi Korki, Angela Digulescu 1,2*, Annamaria Sârbu 3, Denis Stanescu 1,3,
Swinburne University of Technology,
Australia
Dragoș Nastasiu 1,3, Cristina Despina-Stoian 3, Cornel Ioana 2,3 and
Milica Pejanovic-Djurisic, Ali Mansour 4
University of Montenegro, Montenegro
1
Telecommunications and Information Technology Department, Military Technical Academy “Ferdinand
*CORRESPONDENCE
I”, Bucharest, Romania, 2ALTRANS, Grenoble, France, 3GIPSA-Lab, UMR 5216 CNRS, Université Grenoble-
Angela Digulescu,
Alpes, Grenoble, France, 4ENSTA Bretagne—2 rue François VERNY—29806 BREST, Brest, France
[email protected]

RECEIVED 31 March 2023

ACCEPTED 17 July 2023
PUBLISHED 21 August 2023
Signal modulation identification is of high interest for applications in military
CITATION
Digulescu A, Sârbu A, Stanescu D, communications, but is not limited only to this specific field. Some possible
Nastasiu D, Despina-Stoian C, Ioana C applications are related to spectrum surveillance, electronic warfare, quality
and Mansour A (2023), Detection of services, and cognitive radio. Distinguishing between multi-carrier signals, such
OFDM modulations based on the
characterization in the phase as orthogonal frequency division multiplexing (OFDM) signals, and single-carrier
diagram domain. signals is very important in several applications. Conventional methods face a
Front. Sig. Proc. 3:1197590. stalemate in which the classification accuracy process is limited, and, therefore,
doi: 10.3389/frsip.2023.1197590
new descriptors are needed to complement the existing methods. Another
COPYRIGHT
drawback is that some features cannot be extracted using conventional feature
© 2023 Digulescu, Sârbu, Stanescu,
Nastasiu, Despina-Stoian, Ioana and extraction techniques in practical OFDM systems. This paper introduces a new
Mansour. This is an open-access article signal detection algorithm based on the phase diagram characterization. First, the
distributed under the terms of the proposed algorithm is described and implemented for simulated signals in
Creative Commons Attribution License
(CC BY). The use, distribution or MATLAB. Second, the algorithm performance is verified in an experimental
reproduction in other forums is scenario by using long-term evolution OFDM signals over a software-defined
permitted, provided the original author(s) radio (SDR) frequency testbed. Our findings suggest that the algorithm provides
and the copyright owner(s) are credited
and that the original publication in this good detection performance in realistic noisy environments.
journal is cited, in accordance with
accepted academic practice. No use,
KEYWORDS
distribution or reproduction is permitted
which does not comply with these terms. orthogonal frequency division multiplexing, phase diagram, signal recognition, cognitive
radio, electronic warfare

1 Introduction
The last few years have brought huge development in wireless communication systems.
With technology development, multiple-input and multiple-output (MIMO) systems have
become more widely used. The advantages of this technology are based on the increased data
capacity, which leads to reaching effective multiple parallel spatial data streams to increase
the number of users and enhance the transmission quality.
With the development of these systems, the types of modulation used in data
transmission have also become an important topic. Therefore, OFDM data transmission
represents one of the most important techniques in advanced wireless communications.
The limited availability of spectrum resources places constraints on the development
of wireless systems as the demand for wireless services continues to increase. To address
this challenge, wireless systems are moving toward incorporating more embedded
intelligence. The cognitive radio (CR) concept appears as a key solution to the
problem of making different systems coexist in the same frequency band. CR

Frontiers in Signal Processing 01 frontiersin.org

Digulescu et al. 10.3389/frsip.2023.1197590

FIGURE 1
BPSK signal (left) and its phase diagram representation (right).

terminals possess the capability to reconfigure themselves,

adjusting parameters such as modulation, carrier frequency, and
power based on the surrounding radio environment and spectrum
policy. Therefore, spectrum sensing and especially system
identification is a crucial step toward radio environment
awareness (H. Mahmoud et al., 2009). Thus, identification of
OFDM signals is highly important for adaptive receiver
algorithms and signal identification applications.
As a brief introduction, OFDM is based on the concept of the
inverse fast Fourier transform (IFFT), which requires the
simultaneous transmission of parallel data streams over mutually
orthogonal sub-carriers with overlapping frequency bands (Singh
et al., 2022). Each subcarrier may be modulated with a conventional
digital modulation scheme (such as binary phase-shift keying
(BPSK) and quadrature phase-shift keying (QPSK)), at a lower
symbol rate (Lin et al., 2002) than traditional single carrier
modulations. FIGURE 2
The documented automatic modulation classifier methods for BPSK signal (up) and the variation of the α angle (down).
OFDM systems include conventional spectral analysis (Jafar et al.,
2021), deep learning-based methods (Huynh-The et al., 2022), and
maximum a posteriori-based algorithms (Bahrani et al., 2016). The
automatic recognition of this type of modulation is rather proposed model required a longer input sample length and high
complicated with classical signal analysis techniques like spectral SNR value, and the accuracy result depends on the value of the
analysis, time-frequency analysis, wavelet analysis, or high-order thresholds used for FFT window length.
statistics (Hassan et al., 2012). A maximum a posteriori-based automatic modulation
In Park et al. (2021), the authors propose a deep learning-based classification for adaptive OFDM systems is presented in Haring
automatic modulation classification system to classify higher- et al. (2013), where the authors show that the joint probabilities of
order OFDM modulations, 64 OFDM to 512 OFDM, at an SNR the transmit and receive subcarrier bandwidth efficiencies must be
value of 20 dB. The results obtained are quite low in a classic deep precisely known to be able to use this approach. The numerical
learning manner, as the authors also showed, obtaining an results show that the approach can be used only at high values of
accuracy below 30%. For this reason, they combine this SNR. Above 20 dB, this method provides over 90% results for the
approach with a CNN model operating on the fast Fourier identification of OFDM. However, at a value of 10 dB, the
transformation window bank to extract the useful symbol accuracy is 0%.
length in OFDM, which represents the identification of each The integration of cyclostationary spectrum sensing detection
OFDM-based wireless communication technology, thus with an OFDM system is implemented and analyzed in Kumar et al.
managing to increase the accuracy by over 90%. However, the (2019). This approach, although simple from a computational point

Frontiers in Signal Processing 02 frontiersin.org

Digulescu et al. 10.3389/frsip.2023.1197590

Therefore, our proposal is to combine the iterative filter bank

decomposition with the phase diagram analysis, a data-driven
technique, applied in each frequency sub-band to address the detection
of the OFDM modulations. This approach allows us to identify the single
carrier in a given sub-band and, for all the signal’s bandwidth, to
determine the sub-bands containing each individual orthogonal carrier.
This paper is organized as follows: Section 2 presents the
theoretical aspects of the analyzed signal model, phase diagram,
and its characteristics used in the modulation recognition process.
In Section 3, we analyze the results obtained after the implementation
of the detection and characterization algorithm. Finally, in Section 4,
we present the conclusions and some future perspectives of our work.

2 Materials and methods

FIGURE 3 2.1 Theoretical aspects

OFDM signal and its subcarriers.
2.1.1 OFDM signal model
In general, an OFDM signal can be defined as follows:
N
of view, manages to highlight features necessary to be able to identify x(t) 
Ak · ej(2πfk t+φk (t)) , (1)
OFDM modulation. In the cases studied by the authors, the results k1

obtained are satisfactory and can identify the modulation at an SNR

where N represents the number of OFDM subcarriers and Ak , fk , φk
value of 5 dB with a probability of 90%.
are the amplitude, frequency, and phase of the kth subcarrier. For an
In Gorcin et al. (2015), an OFDM signal identification method
OFDM signal with BPSK subcarriers, the phase can be defined by:
that employs estimates of higher-order cumulants and their
covariance is proposed. Using this approach, the OFDM Mk

modulation can be identified with a probability of 80% at an φk 
bi,k · (u(t − (i − 1)Ts ) − u(t − iTs )) · π, (2)
i1
SNR value of 5 dB. However, the identification performance is
affected under low SNR because of the time-domain Gaussianity where Mk represents the total number of symbols transmitted
of the OFDM signals. through the kth subcarrier, Ts is the symbol period, u(t) stands
As we have seen, identifying the OFDM modulation is not an for the Heaviside step function, and bi,k  {0, 1} depending on the
easy task. There are different ways of approaching the problem, and transmitted bit.
no unanimously accepted method will provide the desired results. In
the case of most approaches, the biggest impediment is the low SNR 2.1.2 Phase diagram
value. Otherwise, the more complex the approach, the more factors The phase diagram is a way of analyzing nonlinear data based on
involved in the identification process that must be estimated. This a new representation domain of an analyzed time series. In this new
leads to the impossibility of using the blind identification of the representation space, different characteristics can be highlighted,
modulation, which is preferable in the case of real systems. bringing new information about the analyzed time series (Marwan

FIGURE 4
Analog OFDM signal spectrum (left) and spectrogram (right).

Frontiers in Signal Processing 03 frontiersin.org

Digulescu et al. 10.3389/frsip.2023.1197590

m
→ →
v[i] 
x[i + (k − 1)d] · ek , i  1, 2, . . . , M. (3)
k1

In the construction of the phase space vectors, we can observe the

introduction of two new parameters: the time delay d between the
samples of a signal’s vector and encapsulation dimension m—the
number of samples in the vectors; that is, the vectors’ projections on
the axis are the samples of the signal. Most often, these two
parameters can be determined through the mutual information
method and the false nearest neighbor method, respectively
(Digulescu et al., 2016).
One of the major advantages of this analysis method is that each
analyzed signal has a specific representation in the phase diagram.
Figure 1 shows the phase diagram representation of a BPSK signal,
defined as in Eq. 4:

FIGURE 5 xBPSK (t)  A0 · ej(2πf0 t+φ(t)) , (4)

Set of filters and the OFDM subcarriers (red).
where A0 , f0 , φ0 are the amplitude, frequency, and phase of the
signal.
Figure 1 shows that the phase diagram of a BPSK signal consists
et al., 2007). The transition to this new representation space of the of one ellipse with two distinct branches. These branches correspond
time series is performed based on the phase space vectors, according to the moments when the transition from one phase to another
to Eq. 3: happens in the analyzed signal.

FIGURE 6
Filtered signal: only one subcarrier present (left), two subcarriers present (middle), or no subcarrier present (left).

FIGURE 7
Variation of the angle α (blue) and the detected spikes (red) for the filtered signals: left—single carrier, middle—two subcarriers, and right—no
subcarrier.

Frontiers in Signal Processing 04 frontiersin.org

Digulescu et al. 10.3389/frsip.2023.1197590

FIGURE 8
α angle’s histograms for the filtered signals: left—single carrier, middle—two subcarriers, and right—no subcarrier.

2.1.3 The angle of the trajectory in the phase

diagram
In this paper, we use the angle extracted from the phase diagram
for the OFDM signal detection and its subcarrier characterizations.
The extraction of this angle is based on the analysis of the points that
build the trajectory of the series analyzed in the phase diagram
(Scripcaru et al., 2020).
Inspired by Figure 2, we consider any three points (A, B, and C)
of the representation. Thus, we aim to determine the angle between
the plane determined by these three points and the xOz plane based
on the following scalar product:
→ →
⎜ NABC · NxOz ⎞
α⎛
⎝    →⎟
⎠, (5)
N → · N 
 ABC   xOz 
→ → →
where NABC  AB×AC is the normal to the plane formed by the
→
three points and NxOz  (0, 1, 0). The lower image in Figure 2 shows
the variation of the α angle for the BPSK analyzed signal presented in
the upper image in Figure 2. As can be seen, the variation of the
extracted angle highlights the phase changes in the BPSK signal.
FIGURE 9
Block diagram of the proposed algorithm.

2.2 Phase-space-based OFDM recognition

Figure 4 shows the spectrum and spectrogram of the OFDM
In this subsection, we present the OFDM detection for two signal from Figure 3. The individual characteristics of the subcarriers
cases: with four and with eight subcarriers. The central idea of our are not visible in these representation domains, and we cannot find
work is the following: the detection process is based on the fact that any information on the type of modulation of the subcarriers.
each BPSK subcarrier of the OFDM signal has a specific To decompose the OFDM signal band, we use the set of orthogonal
characteristic in the field of the phase diagram. In this sense, bandpass Chebyshev type II filters, as shown in Figure 5. The pass band
using orthogonal filters, we decompose the analyzed frequency
band, and, with the phase diagram representation, we characterize of the filter is Δfi,k  f min + (k − 1) 2Bi−1 , f min + k 2Bi−1 , where
each signal extracted from its sub-band. Then, the characteristics B  f max − f min , i is the decomposition level, and k is the number
of the extracted signals are compared with the specific BPSK phase of sub-band frequencies.
diagram characteristics. Three situations can occur in the filtered frequency domain: the
presence of a single BPSK subcarrier, the presence of more than one
2.2.1 Analog OFDM signal with two BPSK BPSK carrier, or the absence of any subcarrier.
subcarriers To illustrate these examples, we have used the following filtering
We consider the case of an OFDM signal with a guard interval of bands: Δf2,1  [10.5 kHz, 11.5 kHz], Δf1,1  [9.5 kHz, 11.5 kHz],
20% of the total length of the signal shown in Figure 3 composed of and Δf2,0  [8.5 kHz, 9.5 kHz]. The signals obtained after applying
two orthogonal BPSK subcarriers with the following frequencies: the filters are presented in Figure 6, where each signal (from left to
f1  10 kHz, f2  11 kHz, and the sampling frequency of the right) represents the signal obtained after applying the bandpass
signal fe  55 kHz. filter at the ith level in the jth band.

Frontiers in Signal Processing 05 frontiersin.org

Digulescu et al. 10.3389/frsip.2023.1197590

FIGURE 10
PoS and PoA results obtained with white Gaussian noise added.

FIGURE 11
OFDM signal (left) and its corresponding amplitude spectrum (right).

TABLE 1 Results obtained with white Gaussian noise added.

Subcarrier scenario Parameter 2 BPSK subcarriers 4 BPSK subcarriers 8 BPSK subcarriers

SNR  10 dB SNR  5 dB SNR  10 dB SNR  5 dB SNR  10 dB SNR  5 dB

One subcarrier PoS1 0.37% 0.75% 1.26% 0.32% 0.24% 0.48%

PoA1 95.88% 94.01% 96.56% 87.38% 97.12% 97.84%

Multiple subcarriers PoSm 4.45% 1.12% 2.84% 7.89% 0.72% 18.71%

PoAm 84.64% 81.27% 42.90% 76.03% 70.74% 31.18%

No subcarrier PoS0 86.16% 4.87% 9.46% 19.56% 15.35% 8.63%

PoA0 86.16% 92.13% 41.32% 23.66% 40.77% 52.04%

On the filtered signals, we extract the angle from the phase samples, which are in the same bins as the spike’s samples, represent
diagram. Its variation can be observed in Figure 7. less than 3% of the total number of samples (Scripcaru et al., 2021). If
As it can be seen, depending on the case, the variation of the PoS is higher than 3%, we conclude that the filtered signal does not
angle α may highlight a different number of peaks. To be able to represent any BPSK subcarrier.
decide about the existence of the BPSK subcarrier, we quantify these The definition of the PoS is as follows:
peaks using the information provided by the angle variation
Nspike−bins
histogram, as shown in Figure 8. Therefore, we define the size(bini )
PoS  i1
· 100%, (6)
percentage of spikes (PoS) as in Eq. 6 to see if the number of N

Frontiers in Signal Processing 06 frontiersin.org

Digulescu et al. 10.3389/frsip.2023.1197590

FIGURE 12
Filtered signal with a single subcarrier present (left), the variation of the angle α (middle), and the histogram (right).

FIGURE 13
Filtered signal with two subcarriers present (left), the variation of the angle α (middle), and the histogram (right).

FIGURE 14
Filtered signal with no subcarrier present (left), the variation of the angle α (middle), and the histogram (right).

where N is the total number of samples and Nspike−bins is the number values of the samples are concentrated in a main bin or near a certain
of bins that contain at least one spike. value. In the case of two subcarriers, the values are distributed in
The PoS values for the three filtered signals analyzed are: multiple bins, and in the case of no carrier, false peaks are
PoS11  0.75%, PoSm  1.12%, and PoS0  88.39%. PoS11 introduced, and their distribution is similar to the previous case.
corresponds to a single subcarrier present, PoSm corresponds to Consequently, the algorithm must be completed with another
multiple subcarriers present, and PoS0 corresponds to the no condition.
subcarriers case. It is necessary to check if the number of samples from the two
Depending on the studied case, we have several possible most-populated bins represents more than 85% of the total number
interpretations of the histogram results. For a single carrier, the of samples. The two most-populated bins contain values close to the

Frontiers in Signal Processing 07 frontiersin.org

Digulescu et al. 10.3389/frsip.2023.1197590

FIGURE 15
Experimental setup and illustration of the OFDM signal recording.

FIGURE 16
Transmitted (left) and received (right) downlink LTE-OFDM signals.

TABLE 2 Parameters of the downlink LTE-OFDM signal. average value of the angle. We introduce the percentage of averages
(PoA) parameter defined by Eq. 7.
Parameter Value
size(bin1 + bin2 )
PoA  · 100%. (7)
Channel bandwidth 1.2 MHz N
Number of resource blocks 6 If PoA is smaller than 85%, then we decide that the filtered signal
Modulation QPSK is not a BPSK subcarrier (Scripcaru et al., 2021). The PoA values
for the three filtered signals analyzed are the following:
Subcarrier spacing 15 kHz
PoA11  89.89%, PoAm  54.68%, and PoA0  86.89%.
Sampling frequency 1.92 MHz Therefore, the steps of the presented algorithm for the OFDM
signal carrier detection and characterization are proposed in the
Occupied subcarrier number 72
diagram form shown in Figure 9.

Frontiers in Signal Processing 08 frontiersin.org

Digulescu et al. 10.3389/frsip.2023.1197590

FIGURE 17
Downlink LTE-OFDM filtered signal with one subcarrier (left), the phase diagram angle variation (middle), and its corresponding histogram (right).

FIGURE 18
Downlink LTE-OFDM filtered signal with no subcarrier (left), the phase diagram angle variation (middle), and its corresponding histogram (right).

FIGURE 19
Downlink LTE-OFDM filtered signal with two subcarriers (left), the phase diagram angle variation (middle), and its corresponding histogram (right).

Next, we aim to verify the efficiency of the proposed algorithm in 2.2.2 Digital OFDM signal with 4 BPSK subcarriers
the presence of noise. In addition, we have considered the cases with In this case, we start from the characteristics of the IEEE 802.11a
four BPSK subcarriers and eight BPSK subcarriers. The data are standard for WLAN [Wireless Local Access Network—(IEEE Std 802.11a-
analyzed by reporting at two levels of the signal-to-noise ratio 1999)] having a bandwidth B  20 MHz and the subcarrier spacing
(SNR): SNR  10 dB and SNR  5 dB. The obtained results are Δf  312.5 kHz, modulation BPSK, and four subcarriers (Figure 11).
shown in Table 1, Figure 10. Hereinafter, we present the evolution of the algorithm in the
We notice that the proposed algorithm successfully detects a three possible situations discussed previously (one carrier in the
single subcarrier of the multi-carrier signal for the two SNR values of filtered signal, more than one carrier in the filtered signal, or no
the additive white Gaussian noise. Thus, the noise presence in the carrier in the filtered signal). Figure 12 highlights the corresponding
analyzed signal still allows the detection of the OFDM signal. filtered signal for a single carrier present case.

Frontiers in Signal Processing 09 frontiersin.org

Digulescu et al. 10.3389/frsip.2023.1197590

FIGURE 20
PoS and PoA results obtained for the experimental setup.

TABLE 3 Results obtained for the experimental setup.

SNR [dB] 8 13 16 21 25

Subcarriers PoS PoA PoS PoA PoS PoA PoS PoA PoS PoA
0 17.2% 59.97% 22.29% 54.78% 19.45% 57.36% 15.74% 56.69% 23.35% 49.35%

1 0.3% 99.26% 0.21% 99.41% 0.18% 99.59% 0.05% 99.76% 0.02% 99.39%

2 1.09% 97.78% 0.82% 98.18% 0.64% 98.34% 0.7% 98.31% 0.8% 97.75%

3 1.68% 95.32% 1.46% 96.7% 1.4% 96.41% 1.91% 95.31% 1.16% 96.34%

4 1.4% 95.87% 2.43% 94.29% 2.19% 95.33% 2.47% 93.37% 1.91% 94%

5 2.75% 93.13% 3.27% 91.54% 2.73% 92.74% 2.25% 92.92% 2.78% 92.8%

6 3.55% 91.2% 3.67% 87.23% 3.41% 90.02% 2.93% 89.8% 3.88% 87.42%

The corresponding angle variation and histogram of this signal with shown in Figure 15. The distance between the two boards is 0.8 m,
a single subcarrier are displayed in Figure 12. The results obtained for and they are placed in an indoor location.
the two parameters required for detection are: PoS1  2.56% and The transmitted signal is generated based on the downlink LTE-
PoA1  95.27% for the bandwidth ΔW  [2.4106, 2.4109] MHz. OFDM standard (Rumney, 2013) according to the “R.4” reference
The results obtained for the case in which we have two subcarriers channel defined according to Annex A.3 of 3GPP TS 36.101 as
present are shown in Figure 13. The results obtained for the two presented in Table 2. Figure 16 presents the transmitted and received
parameters required for detection are: PoSm  15.42% and PoAm  signals.
46.28% for the bandwidth ΔW  [2.4106, 2.4112] MHz. Next, we have applied the same algorithm for the previously
The results obtained for the case in which no subcarrier is considered cases:
present are shown in Figure 14. The results obtained for the two
parameters required for detection are: PoS0  28.25% and PoA0  • No subcarrier is present: filter bandwidth is ΔW 
43.69% for the bandwidth ΔW  [2.4102, 2.4106]MHz. [2.5, 7.5]kHz.
Based on these results, we can observe that, in the case of a • One subcarrier is present: filter bandwidth is ΔW 
single carrier present, the proposed algorithm provides the results [22.5, 37.5]kHz.
in the imposed limits; see Figure 9. Furthermore, in the cases in • Multiple subcarriers are present:
which we do not have any subcarrier present or multiple carriers ○ For two subcarriers, the filter bandwidth is ΔW 
are present, the proposed algorithm corresponds to the chosen [22.5, 52.5]kHz.
restrictions. ○ For three subcarriers, the filter bandwidth is ΔW 
[22.5, 67.5]kHz.
○ For four subcarriers, the filter bandwidth is ΔW 
3 Results for experimental LTE OFDM [22.5, 82.5]kHz.
signals ○ For five subcarriers, the filter bandwidth is ΔW 
[22.5, 97.5]kHz.
In this section, we have performed an experimental setup using ○ For six subcarriers, the filter bandwidth is ΔW 
two B210 SDR boards connected with a VERT2450 antenna, as [22.5, 112.5]kHz.

Frontiers in Signal Processing 10 frontiersin.org

Digulescu et al. 10.3389/frsip.2023.1197590

We have performed the measurements for five different scenarios Data availability statement
depending on the gain of the receiver: SNR  {8, 13, 16, 21, 25} dB.
Figures 17–19 present the results for the SNR  8 dB. The original contributions presented in the study are included in
The recorded downlink LTE-OFDM signals are analyzed by the article/Supplementary Materials; further inquiries can be
reporting for all the noise scenarios and filtering cases. The obtained directed to the corresponding author.
results are shown in Table 3 and Figure 20.
From the aforementioned results, we can see that the restrictions
for the PoS and PoA previously imposed are available for the cases of Author contributions
no subcarrier present or one subcarrier present, but for the case of
multiple subcarriers present, the threshold should be higher than AD: conceptualization, data curation, formal analysis,
0.5% for PoS and 99% for PoA. methodology, and writing—original draft. CD, DN and DS:
conceptualization and validation. AD, CI, and AS: methodology
and validation. CI and AM: funding acquisition. AD and CI:
4 Conclusion conceptualization and supervision. All authors contributed to the
article and approved the submitted version.
This paper proposes a novel approach for the detection of the
OFDM modulation. First, using a set of orthogonal bandpass
filters, we perform the decomposition of an OFDM signal so that Funding
each sub-band is analyzed separately. Then, a phase diagram
representation analysis is performed, and the angle of the phase The project has been supported in part by the NATO grant,
diagram representation variation is obtained. The statistical “Implementation of a terahertz system dedicated to identification of
distribution of the angle variation is further used to recognize illicit substances”.
an OFDM signal and characterize its subcarriers.
Based on the defined statistical interpretation of the angle
variation, we implemented a subcarrier detection algorithm for Acknowledgments
BPSK-modulated subcarriers. This algorithm was successively
applied to Gaussian noise-corrupted signals, 802.11a OFDM The authors are grateful to AID-DGA (l’Agence de l’Innovation
signals, and in an experimental scenario using a software- de Défense à la Direction Générale de l’Armement—Minitère des
defined radio testbed to LTE signals. Our findings suggest that Armées) and ANR (Agence Nationale de le Recherche en France) for
the algorithm provides good results even in cases involving a low supporting our ANR-ASTRID—Project (ANR-19-ASTR-0005-03).
SNR level.
With this paper, we propose to present a new approach for
OFDM detection. In future work, we will improve these results by Conflict of interest
classifying different types of OFDM modulation (QPSK and
quadrature amplitude modulation (QAM)) in real scenarios Authors AD and CI were employed by the company ALTRANS.
using machine learning algorithms that make use of the The remaining authors declare that the research was conducted
present features. in the absence of any commercial or financial relationships that
Hereby, the work presented in this paper will represent a starting could be construed as a potential conflict of interest.
point for the further step of automatic digital modulation
classification. With our proposed algorithm, the separation of
each subcarrier in the OFDM transmission will be performed. Publisher’s note
Then, using several machine learning algorithms, such as in
Aslam et al. (2010), we will discriminate between each type of All claims expressed in this article are solely those of the authors
modulation on each subcarrier. and do not necessarily represent those of their affiliated
Furthermore, another next step is to use this approach for the organizations, or those of the publisher, the editors, and the
THz domain in order to enhance the identification of illicit reviewers. Any product that may be evaluated in this article, or
substances based on the different responses received on each claim that may be made by its manufacturer, is not guaranteed or
acquired subcarrier. endorsed by the publisher.

References
3 GPP TS 36.101, 2013 3 GPP TS 36.101 (2013). 3rd generation partnership project; Bahrani, S., Derakhtian, M., and Zolghadrasli, A. (2016). Performance analysis of
technical specification group radio access Network; evolved universal terrestrial radio a low-complexity MAP algorithm for automatic modulation classification in
access (E-UTRA); user equipment (UE) radio transmission and reception (release 11). adaptive OFDM systems. IET Commun. 10, 2363–2371. doi:10.1049/iet-com.
3GPP Organ. Partners (ARIB, ATIS, CCSA, ETSI, TTA, TTC). 2015.0768
Aslam, M. W., Zhu, Z., and Nandi, A. K. (2010). “Automatic digital modulation Digulescu, A., Murgan, I., Ioana, C., Candel, I., and Serbanescu, A. (2016).
classification using Genetic Programming with K-Nearest Neighbor,” in 2010 - MILCOM “Applications of Transient Signal Analysis using the concept of recurrence plot
2010 MILITARY COMMUNICATIONS CONFERENCE, San Jose, CA, USA, November analysis,” in Recurrence Plots and their quantifications: Expanding horizons sl.
2010, 1731–1736. Editors C. J. L. Webber, C. Ioana, and N. Marwan (Berlin: Springer), 19–38.

Frontiers in Signal Processing 11 frontiersin.org

Digulescu et al. 10.3389/frsip.2023.1197590

Gorcin, A., and Arslan, H. (2015). An OFDM signal identification method for wireless Lin, Y., Wang, H., and Chen, B. (2002). “Blind identification of OFDM channel using
communications systems. IEEE Trans. Veh. Technol. 64 (12), 5688–5700. doi:10.1109/ receiver diversity,” in 6th International Conference on Signal Processing, Beijing, China, 26-
TVT.2015.2388671 30 August 2002, 1332–1335.
Haring, L., and Kisters, C. “Map-based automatic modulation classification for wireless Mahmoud, H., Yucek, T., and Arslan, H. (2009). OFDM for cognitive radio: merits
adaptive OFDM systems,” in 2013 IEEE International Conference on Acoustics, Speech and challenges. IEEE Wirel. Commun. 16 (2), 6–15. doi:10.1109/MWC.2009.4907554
and Signal Processing, Vancouver, BC, Canada, May 2013 (IEEE), 5204–5208.
Marwan, N., Carmen Romano, M., Thiel, M., and Kurths, J. (2007). Recurrence plots
Hassan, K., Dayoub, I., Hamouda, W., Nzeza, C. N., and Berbineau, M. (2012). Blind for the analysis of complex systems. Phys. Rep. 438 (5-6), 237–329. doi:10.1016/j.
digital modulation identification for spatially-correlated MIMO systems. IEEE Trans. physrep.2006.11.001
Wirel. Commun. 11 (2), 683–693. doi:10.1109/twc.2011.122211.110236
Park, M. C., and Han, D. S. (2021). Deep learning-based automatic modulation
Huynh-The, T., Pham, Q.-V., Nguyen, T.-V., Da Costa, D. B., and Kim, D.-S. (2022). classification with blind OFDM parameter estimation. IEEE Access 9, 108305–108317.
“Automatic modulation classification with low-cost attention Network for impaired doi:10.1109/ACCESS.2021.3102223
OFDM signals,” in 2022 IEEE Wireless Communications and Networking Conference
Rumney (2013). LTE and the evolution to 4G wireless - design and measurement
(WCNC), Austin, Texas, 10-13 April 2022.
challenges. 2. New Jersey, United States: Wiley.
IEEE Standard for Telecommunications and Information Exchange Between Systems
Scripcaru, R., Nastasiu, D., Digulescu, A., Stanescu, D., Ioana, C., and Serbanescu, A.
(1999). IEEE standard for telecommunications and information exchange between
(2021). “On the potential of phase diagram analysis to identify OFDM modulation,” in
systems - LAN/MAN specific requirements - Part 11: wireless medium access control
Brest, Workshop on Security & Protection of Information (SPI21), Brest, France, 24-
(MAC) and physical layer (PHY) specifications: high speed physical layer in the 5 GHz
25 May 2023.
band. IEEE Std 802.11a-1999, 1–102. doi:10.1109/IEEESTD.1999.90606
Scripcaru, R., Nastasiu, D., Digulescu, A., Stănescu, D., Ioana, C., and
Jafar, N., Paeiz, A., and Farzaneh, A. (2021). Automatic modulation classification using
Şerbănescu, A. (2020). “On the potential of phase diagram analysis to identify
modulation fingerprint extraction. J. Syst. Eng. Electron. 32, 799–810. doi:10.23919/jsee.2021.000069
the wide band modulations,” in 13th International Conference on
Kumar, A., and Nandha kumar, P. (2019). OFDM system with cyclostationary feature Communications (COMM), Bucharest, Romania, June 24-27, 2013, 55–58.
detection spectrum sensing. ICT Express 5 (1), 21–25. doi:10.1016/j.icte.2018.01.007 doi:10.1109/COMM48946.2020.9141963

Frontiers in Signal Processing 12 frontiersin.org