Symbol Based Modulation Classification using Combination of Fuzzy Clustering and Hierarchical Clustering

Negar Ahmadi & Reza Berangi
Signal Processing – An International Journal (SPIJ), Volume (4) : Issue (2) 123
Symbol Based Modulation Classification using Combination of
Fuzzy Clustering and Hierarchical Clustering
Negar Ahmadi negar.ahmadi670@gmail.com
Department of Computer Engineering
Iran University of Science and Technology
Narmak, Tehran, Post Code: 1684613114, Iran
Reza Berangi rberangi@iust.ac.ir
Department of Computer Engineering
Iran University of Science and Technology
Narmak, Tehran, Post Code: 1684613114, Iran
Abstract
Most of approaches for recognition and classification of modulation have been
founded on modulated signal’s components. In this paper, we develop an
algorithm using fuzzy clustering and consequently hierarchical clustering
algorithms considering the constellation of the received signal to identify the
modulation types of the communication signals automatically. The simulation that
has been conducted shows high capability of this method for recognition of
modulation levels in the presence of noise and also, this method is applicable to
digital modulations of arbitrary size and dimensionality. In addition this
classification finds the decision boundary of the signal which is critical information
for bit detection.
Keywords: Fuzzy C-means, AMR, Modulation Classification, Hierarchical Clustering.
1. INTRODUCTION
Modulation classification (MC) is an important subject in both military and commercial
communication applications. This is a challenging problem, especially in non-cooperative
environments, where no prior knowledge on the incoming signal is available. The aim of
modulation classification (MC) [1-5] is to identify the modulation type [6] of a Communication
signal. It plays an important role in many communication applications such as software radio,
intelligent modem, and electronic surveillance system [7].
In many military communication situations such as reconnaissance, surveillance and electronic
warfare, an indispensable step is to classify automatically the modulation types of the received
signals. Automatic modulation recognition plays an important role in civil fields such as
interference identification and spectrum management.
Modulation recognition is an intermediate step on the path to full message recovery. As such, it
lies somewhere between low level energy detection and a full fledged demodulation. Therefore,
correct recovery of the message per se is not an objective, or even a requirement [8, 9]. The
existing methods for modulation classification span four main approaches. Statistical pattern
recognition, decision theoretic (Maximum Likelihood), M-th law non-linearity and filtering and ad
hoc [10, 11].

Early on it was recognized that modulation classification is, first and foremost, a classification
problem well suited to pattern recognition algorithms. A successful statistical classification
requires the right set of features extracted from the unknown signal. There have been many
attempts to extract such optimal feature. Histograms derived from functions like amplitude,
instantaneous phase, frequency or combinations of the above have been used as feature vectors
for classification, Jondral [12], Dominguez et al. [13], Liedtke [14]. Also of interest is the work of
Aisbett [15] which considers cases with very poor SNR.
The current state of the art in modulation classification is the decision theoretic approach using
appropriate likelihood functional or approximations thereof. Polydoros and Kim [16] derive a
quasi-log-likelihood functional for classification between BPSK and QPSK modulations. In a later
publication, Huang and Polydoros [17] introduce a more general likelihood functional to classify
among arbitrary MPSK signals. They point out that the S-classifier of Liedtke, based on an ad hoc
phase-difference histogram, can be realized as a noncoherent, synchronous version of their
qLLR. Statistical Moment-Based Classifier (SMBC) of Solimon and Hsue [18] are also identified
as special coherent version of qLLR. Wei and Mendel [19] formulate another likelihood-based
approach to modulation classification that is not limited to any particular modulation class. Their
approach is the closest to a constellation-based modulation classification advocated here
although they have not made it the central thesis of their work. Carrier phase and clock recovery
issues are also not addressed. Chugg et al [20] use an approximation of log-ALF to handle more
than two modulations and apply it to classification between OQPSK/BPSK/QPSK. Lin and Kuo
[21] propose a sequential probability ratio test in the context of hypothesis testing to classify
among several QAM signals. Their approach is novel in the sense that new data continuously
updates the evidence.
Past work on modulation recognition has primarily used signal properties in time and/or frequency
domain to identify the underlying modulation. One of the typical analysis methods for the
modulated signal is the extraction of In-Phase  I and Quad-Phase  Q components. According
to these components, we can see the signal as a vector in the I Q plane which is referred to as
the constellation diagram. With the use of modulated signal constellation, modulation
classification can be investigated as pattern recognition problem and well known pattern
recognition algorithms can be used. For example in a recent publication Ahmadi and Berangi [22]
introduce a method to classify modulation by using Genetic Algorithm and template matching
Based on constellation diagram.
2. CLASSIFICATION OF QAM AND PSK MODULATIONS USING FUZZY
CLUSTERING
As mentioned, constellation diagram, which consists of In-phase and Quad-phase components,
can be used for modulation classification. Since the constellation is symmetric with respect to its
axes, in order to reduce complexity, we can map all the received symbols into the first quadrant in
the constellation diagram. After obtaining number and location of clusters in the first quadrant,
centroids of the clusters could be extended to the whole constellation, symmetrically.
The proposed technique has been designed so that it would be capable of recognizing the types:
4-PSK, 8-PSK, 16-PSK, 4-QAM, 16-QAM and 64-QAM. So the initial number of clusters has been
set to 16 in the first quadrant. This technique consists of two stages of FCM (Fuzzy C-means)
clustering and post processing. Therefore the initial clusters can be defined as a vector of 16
elements in which, each element is a point in the first constellation quadrant. Initial value of
centroids of the clusters is used for obtaining initial elements of the membership matrix. In order
to reduce processing, calculation is done in the first quadrant, absolute value of signal’s I Q
components is calculated and stored in a 2D matrix and used in the future processing.
2.1. An Overview of Fuzzy C-mans Algorithm

Fuzzy C-means algorithm executed as follows [23, 24, and 25]:
Initial membership matrix constructed according to equation (1).
2
1
1
1
ij
m
C i j
k
i k
u
x c
x c



 
 
 
 
 (1)
where m is the Fuzzy exponent, ix are the input symbols and jc denote the centroids.
2. New centroids shall be calculated using membership function as equation (2).
1
1
N m
ij ii
j N m
iji
u x
c
u






(2)
Values of the old membership matrix are stored for the future comparisons.
3. New membership matrix is calculated using equation (1) and the objective function is
calculated using equation (3).









  
 
mcxuJJ
N
i
C
j
ji
m
ij
N
i
im 1
1 1
2
1
(3)
In fact the aim of clustering is assigning symbols to clusters so that the value of the objective
function is minimized. The algorithm convergence and termination condition are evaluated as
equation (4).
   
 1
max k k
ij ij iju u 
  (4)
If the above condition is satisfied or the number of iterations reaches its end, then the algorithm is
terminated, otherwise it returns to the step 2. Through this stage the number of centroids is
constant. Hence to reduce the number of centroids, a post-processing step is necessary.
2.2. Post-Processing Stage
In this stage the centroids resulted from fuzzy clustering have been merged such that they result
in the actual number and the location of centroids. The number of clusters in one quadrant could
be 1, 2, 4, 8 or 16. The foundation of this step is based on hierarchical clustering [26]. Therefore,
upon the completion of the fuzzy algorithm when centroids are obtained naturally, we use
hierarchical clustering. To do this, first we calculate the Euclidean distance of the clusters pair-
wise and construct the hierarchical tree. After constructing the tree, all the possibilities of the
merged centroids are calculated based on the number of the centroids. For each possibility the
fitness value of the merged centroids with signal’s symbols is calculated. This fitness value is
calculated using hard C-means clustering [26]. Since clustering is performed to assign symbols to
clusters, in a sense that the objective function (fitness value) is minimized. The number of
optimum clusters corresponds to the objective function that is minimum between all possibilities.
It must be mentioned that, since the value of the objective function increases naturally as the
number of clusters increases, which in turn causes an error in recognition of the actual number of
cluster, the value of the objective function is multiplied by a weight proportional to the number of
clusters to prevent producing error.

After multiplication of the weights, the number of clusters corresponding to the objective function
with minimum value is chosen as the final number of cluster. After the final clusters on the first
quadrant are obtained, these clusters are extended to the whole constellation diagram. For
example, consider Figure (1) which corresponds to post processing of 16-QAM. It can be seen
that the value of the objective function for 4 clusters in one quadrant is minimum with respect to
the function’s value for other possibilities; therefore 4 clusters are chosen for the first quadrant,
and for the whole constellation diagram, 16 clusters are chosen (equal to the number of points of
the 16-QAM constellation).
Figure 1: Values of the objective functions for possible clusters in the first quadrant (16-QAM).
Also in Figure (2) which corresponds to 8-PSK modulation, objective function for two clusters in
the first quadrant has the minimum value, so the modulation levels in the whole constellation
diagram are equal to 8, which is equal to points of 8-PSK constellation. In this way, for all
modulations of QAM and PSK families, the post-processing step can lead to finding the actual
number of clusters from the 16 clusters resulted by the fuzzy algorithm.

Figure 2: Values of the objective functions for possible clusters in the first quadrant (8-PSK).
3. EVALUTION AND SIMULATION RESULTS
In order to evaluate the performance of the proposed method, simulation has been performed for
various SNR values and various types of QAM and PSK modulations. Channel model, applied in
this work, has been assumed to be an AWGN channel, and it is also assumed that there is no
time and/or frequency synchronization error. Simulation results show that this method has an
efficient performance and high accuracy for the recognition of modulation. The performances of
this method for 16-QAM and 16-PSK have been investigated. Figure (3) and Figure (4) show the
centroids in the first quadrant obtained from fuzzy algorithm.
After completion of fuzzy algorithm, post-processing is performed to achieve the actual number of
clusters in the first quadrant and consequently modulation type is recognized. For both 16-QAM
and 16-PSK modulations, the value of objective function for 4 clusters in one quadrant will be
minimum. Therefore closest clusters are combined so that 4 clusters are obtained in the first
quadrant. Figure (5) and Figure (6) show extended clusters and signal symbols together, for 16-
QAM and 16-PSK modulations, respectively.
Accuracy of this method for different modulations and different values of SNR is evaluated. The
accuracy percentages have been obtained by executing the algorithm enough number of times
and by calculating the ratio between correct recognition and total number of execution. Figure (7)
shows the accuracy of recognition versus SNR ratio for different modulation types of QAM family;
Also, this figure shows the results for the PSK family. As it can be seen in Figure (7) and (8), as
modulation level increases, higher SNR is needed for the correct recognition of modulation type.

Figure 3: Centroids in the first quadrant obtained from fuzzy clustering (16-QAM).
Figure 4: Centroids in the first quadrant obtained from fuzzy clustering (16-PSK).

Figure 5: Extended clusters and their centroids in whole constellation diagram after post-processing
step (16-QAM).
Figure 6: Extended clusters and their centroids in whole constellation diagram after post-processing
step (16-PSK).

Figure 7: Accuracy of QAM modulation recognition versus SNR.
Figure 8: Accuracy of PSK modulation recognition versus SNR.

Having analyzed the obtained results from the fuzzy clustering method, we can imply the
following cases:
 This method can recognize all the 4-QAM modulations with 100% accuracy and its recognition
truth does not depend on the signal to the noise, and it can also identify this modulation type with
SNR=0.
 This method can identify all the 16-QAM modulations with SNR=8 with a 95% accuracy,
whereas it can also recognize the signal to the noise ratio lower than 8 with a less exactness and
higher than 8 with a more exactness; it is noteworthy to say that in lower SNR, the more the
sample is increased, the more efficient is the recognition of the modulation type.
 Accuracy of recognition of 64-QAM for the signal of the noise 10 and higher than that is 100%
and for a level lower than it is less. Here, for the lower SNR, as the number of specimens
increases, the accuracy would also increase.
 This research aims to recognize the modulation of QAM modulation family, but can be applied
practically to recognize and separate the PSK modulation family. The proposed method to
recognize PSK modulation can recognize all the QPSK modulations with 100% accuracy.
Exactitude of this method to recognize the 8-PSK modulation with the signal to noise 8 and higher
is 100% and for the lower level it is less; moreover, this method is able to identify 16PSK
modulation with SNR=15 and higher than that with 100% accuracy and for a level lower than this
it is less.
In this method fuzzy clustering and hierarchical clustering was used to classify different
modulation types of QAM and PSK families, using the constellation diagram of the received
signal, which is the noisy version of transmitted signal.
As can be seen in the simulation section, the proposed method shows a good performance for
recognition even in low SNR condition, of course it must be mentioned that the performance
could be increased with a higher number of data symbols. Another advantage of this method is
calculating final centroids of clusters and determining the location of these centroids in
constellation diagram. Also this approach could be extended and modified to recognize other
types of digital modulation. Figures (9), (10), (11), (12) show some more simulation results of this
method.
4. CONCLUSION
In this paper fuzzy clustering and hierarchical clustering were used to classify different modulation
types of QAM and PSK families, using the constellation diagram of the received signal, which is
the noisy version of the transmitted signal. By using fuzzy clustering algorithm, the clustering of
input samples in the I Q plane is done with a high accuracy, and then in the post-processing
stage, by using hierarchical clustering the number of real clusters that are representative of levels
of modulation is recognized. As can be seen in simulation section, the proposed method shows a
good recognition performance even in low SNR condition, and adapt to engineering applications.
of course, it must be mentioned that the performance could be improved with higher number of
data symbols. Another advantage of this method is calculating final centroids of the clusters and
determining the location of these centroids in constellation diagram. In addition this approach
could be extended and modified to recognize other types of digital modulations.
The method that has been used can be expanded and uses them for modulation recognition of
any PAM signals. These signals have one-dimensional constellation while in this research we
study the signals with two-dimensional constellations, which are more complicated. Thus with a
little change we can use them for recognition of PAM signals. From these signals the MFSK and
MASK modulations can be referred.
With little changes in the proposed method it can be used in the recognition of modulations that
have non-standard one dimensional or two-dimensional constellation. By rotating the

constellation diagram of PAM signals for 45
in the I Q plane, without any change in the
proposed method, it can recognize modulations.
(a) (b)
(c) (d)
Figure 9: Recognition of 4-QAM with SNR=0dB and 1000 samples.
a) Centroids in the first quadrant obtained from fuzzy clustering,
b) values of the objective functions for possible clusters in first quadrant,
c) centroids of the clusters in whole constellation diagram after post-processing step and
d) extended clusters and their centroids in whole constellation diagram after post processing step.

(a) (b)
(c)
Figure 10: Recognition of 64-QAM with SNR=13dB and 1000 samples.
a) centroids in the first quadrant obtained from fuzzy clustering,
b) centroids of the clusters in whole constellation diagram after post processing-step and
c) extended clusters and their centroids in whole constellation diagram after post processing step.

(a) (b)
(c) (d)
Figure 11: Recognition of 8-PSK with SNR=7dB and 2000 samples.
a) centroids in the first quadrant obtained from Fuzzy clustering,
c) centroids of the clusters in whole constellation diagram after post processing-step and
d) extended clusters and their centroids in whole constellation diagram after post processing-step.

(a) (b)
(c) (d)
Figure 12: Recognition of 16-PSK with SNR=13dB and 1000 samples.
a) centroids in the first quadrant obtained from fuzzy clustering,
c) centroids of the clusters in whole constellation diagram after post processing-step and
d) extended clusters and their centroids in whole constellation diagram after post processing step.
REFERENCES
[1] A. Swami, B. M. Sadler, “Hierarchical Digital Modulation Classification Using Cumulants”,
IEEE Trans. Communications, Vol. 48(3), pp.416-429, 2000
[2] A.K. Nandi, E.E. Azzouz, “Algorithms for Automatic Modulation Recognition of Communication
Signals”, IEEE Trans. Communications ,Vol. 46(4), pp. 431-436, April 1998
[3] E.E. Azzouz, A.K.Nandi., “Automatic Modulation Recognition of Communication Signals”,
Kluwer Academic Publisher, Norwell, MA, 1996

[4] Wen Wei, Jerry M.Mendel, “A Fuzzy Logic Method for Modulation Classification in Nonideal
Environments”, IEEE Trans. Fuzzy Systems, Vol.7 (3), pp.333-344, June 1999
[5] J. Lopatka, M.Pedzisz, “Automatic Modulation Classification Using Statistical Moments and a
Fuzzy Classifier”, in Proceedings of ICSP2000, pp 1500-1506, 2000
[6] S. G. Wilson, “Digital Modulation and Coding”, New York, Prentice-Hall, Inc., Ch.3, (1996)
[7] Daniel Boudreau, Christian Dubuc, Francois Patenaude et al., “A Fast Automatic Modulation
Recognition Algorithm and Its Implementation in a Spectrum Monitoring Application”,
MILCOM2000, Los Angeles, California, Oct. 22-25, 2000
[8] J. Reichert, “Automatic Classification of Communication Signals using Higher Order Statistics”,
ICASSP 92, pp.221-224, 1992
[9] R. Schalkoff, “Pattern Recognition: Statistical, Structural and Neural Approach” , John Wiley,
(1992)
[10] Bijan G. Mobaseri, “Constellation shape as a robust signature for digital modulation
recognition”, Military Communications Conference Proceedings, MILCOM IEEE, Volume 1, Issue,
pp. 442-446, 1999
[11] Bijan G. Mobasseri, “Digital Modulation Classification using Constellation Shape”, Signal
Processing, Vol. 80, No. 2, pp.251-277, 2000
[12] F. Jondral, “Automatic Classification of High Frequency Signals”, Signal Processing, Vol. 9,
No. 3, pp.177-190, 1985
[13] L. Dominguez, J. Borrallo, J. Garcia, “A General Approach to the Automatic Classification of
Radiocommunication Signals”, Signal Processing, Vol. 22, No. 3, pp.239-250, 1991
[14] F.F. Liedtke, “Computer Simulation of an Automatic Classification Procedure for Digitally
Modulated Communication Signals with Unknown Parameters”, Signal Processing, Vol. 6,
pp.311.323, 1984
[15] J. Aisbett, “Automatic Modulation Recognition using Time-Domain Parameters”, Signal
Processing, Vol.13, No. 3, pp.323-329, 1987
[16] A. Polydoros, K. Kim, “On the Detection and Classification of Quadrature Digital Modulation
in Broad-Band Noise”, IEEE Transactions on Communications, Vol. 38, No. 8, pp. 1199-121,
1990
[17] C. Huang, A. Polydoros, “Likehood Method for MPSK Modulation Classification”, IEEE
Transaction on Communications, Vol. 43, No. 2/3/4, pp.1493-1503, 1995
[18] S. Soliman, S. Hsue, “Signal classification using statistical moments”, IEEE Transactions on
Communications, Vol. 40, No. 5, pp. 908-915, 1992
[19] W. Wei, J. Mendel, “A New Maximum Likelihood for Modulation Classification,” Asilomar-29,
pp. 1132-1138, 1996
[20] K. Chugg, et al, “Combined Likelihood Power Estimation and Multiple Hypothesis Modulation
Classification”, Asilomar-29, pp. 1137-114, 1996

[21] Y.Lin, C.C. Kuo, “Classification of Quadrature Amplitude Modulated (QAM) Signals via
Sequential Probability Ratio Test (SPRT)”, Report of CRASP, University of Southern California,
July 15, 1996
[22] Negar Ahmadi, Reza Berangi, “A Template Matching Approach to Classification of QAM
Modulation using Genetic Algorithm”, Signal Processing: An International Journal, Vol.3, Issue 5,
pp: 95-109, 2009
[23] Krishna K. Chintalapudi and Moshe Kam, “A Noise-Resistant Fuzzy C Means Algorithm for
Clustering”, Fuzzy systems proceedings, IEEE international Conference, Vol. 2, pp. 1458-1463,
1998
[24] Sadaaki Miyamoto, “An Overview and New Methods in Fuzzy Clustering”, Knowledge-Based
Intelligent Electronic Systems, Second International Conference, Vol. 1, , pp. 33-40, 1998.
[25] Frank Chung-Hoon Rhee and Cheul Hwang, “A Type-2 Fuzzy C-Means Clustering
Algorithm”, 20 th NAFIPS international conference, Vol. 4, pp. 1926-1929, 2001
[26] E. Gose, R. Johnsonbaugh, S. Jost, “Pattern Recognition and Image Analysis”, Prentice Hall
PTR, (1996)

Symbol Based Modulation Classification using Combination of Fuzzy Clustering and Hierarchical Clustering

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Symbol Based Modulation Classification using Combination of Fuzzy Clustering and Hierarchical Clustering (20)

Recently uploaded (20)

Symbol Based Modulation Classification using Combination of Fuzzy Clustering and Hierarchical Clustering