See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/329972466

Automatic Modulation Recognition in Cognitive Radio Receivers using Multi-

Order Cumulants and Decision Trees

Article · November 2018

CITATIONS READS

8 341

2 authors:

M. Venkata Subbarao Samundiswary Punniakodi

Shri Vishnu Engineering College for Women Pondicherry University
27 PUBLICATIONS 73 CITATIONS 102 PUBLICATIONS 355 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Time-Frequency Analysis and its application to Signal Processing and Communications View project

Wireless communications View project

All content following this page was uploaded by M. Venkata Subbarao on 18 February 2019.

The user has requested enhancement of the downloaded file.

 International Journal of Recent Technology and Engineering (IJRTE)
ISSN: 2277-3878, Volume-7 Issue-4, November 2018

Automatic Modulation Recognition in Cognitive

Radio Receivers using Multi-Order Cumulants
and Decision Trees
M.Venkata Subbarao, P.Samundiswary

 poor performance under noisy conditions if proper signal

Abstract: Design of intelligent receiver is a major footstep in features are not provided for training [4], [5].
the implementation of Cognitive Radio (CR). Automatic
Modulation Recognition (AMR) of the received signal decides the
performance of the intelligent receiver. This paper proposes new
classification algorithms for AMR using supervised Decision Tree
(DT). DT Classifiers (DTC’s) are non-parametric classifiers
which provide high speed and low complex solutions in
classification. Fine Tree (FT), Medium Tree (MT) and Coarse
Tree (CT) classifiers are implemented in this paper which is
trained with multi-order cumulants to achieve optimum
classification accuracy. Performance of DTC’s is compared with Fig. 1 Block diagram of Intelligent Receiver
other classifiers stated in literature to prove their superiority in
modulation classification. To identify the specific modulation, Average and
Index Terms: Modulation Classification, Cognitive Radio,
Generalized Likelihood Ratio Tests (ALRT & GLRT)
Moments, Cumulants, Binary Trees compare received signal likelihood functions with various
available modulation functions [6], [7]. Higher order Statics
I. INTRODUCTION (HoS) or cumulants are used as features for PR classifiers [8],
[9]. Cumulants are the best features and the classification
Adaptive modulation and dynamic carrier selection are
accuracy is more even under fading conditions [10]- [12].
playing a major role in data security for military, commercial
Recently hybrid classification approaches are combined
and CR applications. Quality of Services can be provided by
cumulants and some other pattern recognition approaches like
altering the modulation technique dynamically based on the
Back Propagation Neural Network (BPNN), Genetic
channel characteristics. These techniques involve additional
Programming and KNN (GP-KNN) gives better classification
complex operations like spectrum sensing and AMR at
accuracy than Kolmogorov Smirnov (KS) and HoS
transmitter and receiver ends which results in existing
approaches [13]-[17]. In the last five decades, DT is used in
traditional receivers are inefficient. Intelligent receivers
various classification and regression applications [18].
capable of extracting the modulation information blindly may
Recently DTC‟s deals with many data mining, statistics,
improve transmission efficiency through reductions in
classification and prediction problems [19].
overhead or supplementary information on the modulation
This paper presents a new automatic recognition approach
type. The functional diagram of an intelligent receiver is
DTC‟s for MPSK (M=2, 4 & 8), 4QAM, 16QAM and
shown in Fig. 1.
64QAM signals through multi-order cumulants. The
Modulation classifiers are broadly categorized into
motivation behind to choose these particular classes is that
maximum likelihood and pattern recognition or feature based
most of the practical applications involves either MPSK or
classifiers [1]. Probability Density Function (PDF) of the
QAM modulation techniques. In literature, most of the
received signal waveform is used for classification in
researchers consider MASK, MFSK and limited MPSK and
Maximum Likelihood (ML) approaches [2], [3]. Pattern
QAM signals for classification. For MASK and MFSK
Recognition (PR) approach involves extraction of statistical
modulation classes, the proposed technique gives optimal
features, training the classifier and finally testing. The ML
classification accuracy similar to existing approaches in
classifiers are more accurate than PR classifiers, but they
literature. So in this paper those modulation classes are
require past knowledge of signal waveform characteristics
which is impractical. The PR classifiers are less complex in excluded for simulation. Simulation results shown, even with
design and these are signal independent. PR classifiers have more classes of modulation signals, the proposed approaches
achieve more accuracy than existing approaches.
The rest of this paper is organized as follow. Section II
Revised Version Manuscript Received on 30 November, 2018. describes system model and multi-order features for
M. Venkata Subbarao, Research Scholar, Department of Electronics modulation recognition.
Engineering, School of Engineering & Technology, Pondicherry University,
Pondicherry, India
P. Samundiswary, 2Assistant Professor, Department of Electronics
Engineering, School of Engineering & Technology, Pondicherry University,
Pondicherry, India

Published By:
Blue Eyes Intelligence Engineering
61 & Sciences Publication
 Automatic Modulation Recognition in Cognitive Radio Receivers using Multi-Order Cumulants and Decision Trees

Detailed description of proposed DTC‟s is presented in cumulants are calculated then pass these features through
Section III. Classification accuracy of DTC‟s are analyzed trained network for modulation identification.
through simulation results in Section IV and Section V
concludes the work. III. PROPOSED APPROACH
DT classifiers are non parametric supervised learning
II. SYSTEM MODEL classifiers. DTC‟s are simple to understand and fast in
The intelligent receiver system shown in Fig. 1 receives the classification or prediction. The complexity of these
noisy transmitted signal x(n), and it is given by classifiers is less so that they require less memory. The
accuracy of DTC‟s in classification is low under noisy
(1)
conditions when insufficient features are used for training.
where a(n) is the AWGN noise added in the channel and
Increase in depth of the tree leads to improve in the
r(n) is the received signal. In digital data transmission the
classification performance. DT classifiers are binary
transmitted signal x(n) is given by [20]
classifiers i.e. any node having two child nodes except leaf
nodes. To predict the accurate modulation format, the
decision starts from root to leaf node.
Based on the depth (number of leaves) of the tree, DTC‟s
(2) are sub categorized into Fine Tree Classifier (FTC), Medium
Here A is the amplitude, T is symbol time, θ_n is phase Tree Classifier (MTC) and Coarse Tree Classifier (CTC). The
jitter, x(k) is input data stream, c(.) is channel effect and ϵ is numbers of leaves in FTC are more, results classification
time shifts due to channel. accuracy is high. FTC is suitable for large class dataset. A
To recognize the exact modulation class of received signal CTC have minimum number of leaves, so that it provides
statistical features are extracted. Higher order statistical minimum accuracy among all DTC‟s. CTC is more robust and
features or cumulants are derived from moments. easy to interpret for small class problems. A MTC have
The moments are depends on the order of the statistical moderate number of leaves and it provides more accuracy
features and these are given by than CTC. In DTC‟s accuracy of a node depends on Gini's
(3) Diversity Index (GDI) and cross entropy. To split the nodes
twoing rule is used and by maximizing it, node accuracy will
Second order cumulants and are improve.
(4) The detailed DT classifiers algorithm is summarized as
follows.
(5)
Algorithm
Fourth order cumulants , and are
Classifier:
(6)
(7) Input: Set of Modulated Signals
(8) Output: Specific Modulation Class
Sixth order cumulants , and are
Phase 1: Training
(9)
Begin
(10)
Step 1: Define set of SNR values

Step 2: Define set of Modulation Classes

(11)
Step 3: Generate „M‟ signal copies of each modulated
class with each and every SNR

(12) Step 4: Extract higher order Moments and Cumulants

for all signal copies
Eight order cumulants and are
Step 5: Identify the best features for Training and
discard the remaining features
(13)
Step 6: Train the classifier with all SNR signal sets

End
(14)
These multi-order cumulants are useful for training the
network and testing the received signal. For modulation
identification of noisy received signal all multi-order

Published By:
Blue Eyes Intelligence Engineering
62 & Sciences Publication
 International Journal of Recent Technology and Engineering (IJRTE)
ISSN: 2277-3878, Volume-7 Issue-4, November 2018

Phase 2: Classification/ Testing

Begin

Step 1: Received signal preprocessing

Step 2: Extract the best features of received signal

Step 3: Apply extracted features to classifier for

modulation recognition

Step 4: Calculate the modulation accuracy and

misclassification rate

end

The DT classifier block diagram is shown in Fig. 2. It

involves training and testing phases. While training, set of
reference signals are considered for feature extraction. In this
paper , multi-order cumulants is considered as features. (a) Number of Observations

(b) Prediction accuracies

Fig. 3 Confusion Matrices of FN at SNR 20 dB
For better training, every modulated signal is generated 300
times for each SNR value, so that the total feature set size
becomes 3600*11. Among these futures, 90% of the features
are considered for training and remaining 10% features are
used for testing. M-ary PSK, 4QAM, 16 QAM and 64 QAM
Fig. 2 Decision Tree Classifier Model
signals are considered for comparison because most of the
IV. SIMULATION RESULTS existing approaches have a great reduction in accuracy at
lower SNR‟s. For testing, a 30*11 feature set is considered for
M-ary PSK (with M=2, 4 and 8), 4QAM, 16QAM and each modulation technique.
64QAM signals are considered for the simulation to verify the The confusion matrices of proposed FT in terms of
superiority of DTC‟s over existing approaches. To recognize observations and prediction rates at 20 dB SNR are shown in
exact modulation, Second, Fourth, Sixth and Eight order Fig. 3, and it's clear that for each modulation among 30
moments and cumulants are extracted for all classes of instances FT predicted 30 times accurately. Therefore, the
modulated signals under different noisy conditions (AWGN true positive rate of each modulation is 100%. Fig. 4 shows
Channel with SNR= 0, 5, 10, 15 and 20 dB‟s). From the the confusion matrices of FT, MT, and CT at SNR 0 dB. The
experimental tests, it is found that cumulants are enough for accurate modulation prediction rate of FT is better than MT
training to achieve optimal classification accuracy, so and CT for every modulation class even at lower SNR values.
moments are excluded in training phase to reduce the training
time.

Published By:
Blue Eyes Intelligence Engineering
63 & Sciences Publication
Automatic Modulation Recognition in Cognitive Radio Receivers using Multi-Order Cumulants and Decision Trees

(a) FT (b) MT

(c) CT
Fig. 4 Confusion matrices of DTC’s at SNR 0 dB
Table 1 The Performance Measure of Fine Tree
Predicted Class
SNR True
Overall % of
BPS QPS 16QA 64QA Accuracy
(dB) Class 8PSK 4QAM
K K M M

BPSK 100

QPSK 93 3 3

8PSK 3 80 7 10
0 83.9
4QAM 13 87

16QAM 7 10 70 13

64QAM 10 13 3 73

BPSK 100

QPSK 100

5 8PSK 97 3 88.9

4QAM 100

16QAM 20 67 13

Published By:
Blue Eyes Intelligence Engineering
64 & Sciences Publication
 International Journal of Recent Technology and Engineering (IJRTE)
ISSN: 2277-3878, Volume-7 Issue-4, November 2018

64QAM 7 3 20 70

BPSK 100

QPSK 100

8PSK 100
10 95.6
4QAM 100

16QAM 7 86 7

64QAM 13 87

BPSK 100

QPSK 100

8PSK 100
15 98.9
4QAM 100

16QAM 3 97

64QAM 3 97

BPSK 100

QPSK 100

8PSK 100
20 100
4QAM 100

16QAM 100

64QAM 100

respectively. From Tables 1 & Table 2, it is observed that FT

and MT almost have same performance apart from lower SNR

values. From Table 3, CT has more confusion in the

classification of 8 PSK, 16 QAM, and 64 QAM signals so that
the performance never reaches to 100%. At SNR values of 10
dB and 5 dB, the accuracy of FT is similar to MT, but there is
a variation in the classification in terms of prediction.
The recognition accuracy of FT for different modulation
techniques is shown in Fig. 5. The FT provides optimal
accuracy for MPSK and 4QAM signals even for lower SNR
values, but the performance is degraded for 16QAM and
64QAM signals at SNR 5dB and 0dB. Similarly, the
performance of MT for different modulation classes is shown
in Fig. 6. The performance of MT is almost similar to FT for
MPSK and 4QAM signals. The accuracy is slightly varying
for 16QAM and 64QAM signals.

Fig. 5 Recognition Accuracy of Fine Tree

The classification performance of Fine Tree, Medium Tree,
and Coarse Tree are shown in Table 1, Table 2 and Table 3

Published By:
Blue Eyes Intelligence Engineering
65 & Sciences Publication
Automatic Modulation Recognition in Cognitive Radio Receivers using Multi-Order Cumulants and Decision Trees

Table 2 The Performance Measure of Medium Tree

Predicted Class
SNR True
Overall % of
BPS QPS 16QA 64QA Accuracy
(dB) Class 8PSK 4QAM
K K M M

BPSK 100

QPSK 93 4 3

8PSK 3 77 7 13
0 81.7
4QAM 13 87

16QAM 10 10 57 23

64QAM 10 13 77

BPSK 100

QPSK 100

8PSK 97 3
5 88.9
4QAM 100

16QAM 20 67 13

64QAM 7 3 20 70

BPSK 100

QPSK 100

8PSK 100
10 95.6
4QAM 100

16QAM 7 86 7

64QAM 13 87

BPSK 100

QPSK 100

8PSK 100
15 98.9
4QAM 100

16QAM 3 97

64QAM 3 97

BPSK 100
20 100
QPSK 100

Published By:
Blue Eyes Intelligence Engineering
66 & Sciences Publication
 International Journal of Recent Technology and Engineering (IJRTE)
ISSN: 2277-3878, Volume-7 Issue-4, November 2018

8PSK 100

4QAM 100

16QAM 100

64QAM 100

Fig. 6 Recognition Accuracy of Medium Tree

Table 3 The Performance Measure of Coarse Tree
Predicted Class
SNR True
Overall % of
BPS QPS 16QA 64QA Accuracy
(dB) Class 8PSK 4QAM
K K M M

BPSK 100

QPSK 93 7

8PSK 7 13 80
0 62.8
4QAM 97 3

16QAM 17 7 77

64QAM 13 87

BPSK 100

QPSK 100

8PSK 83 17
5 78.3
4QAM 100

16QAM 37 0 63

64QAM 7 7 86

BPSK 100

10 QPSK 100 83.3

8PSK 100

Published By:
Blue Eyes Intelligence Engineering
67 & Sciences Publication
Automatic Modulation Recognition in Cognitive Radio Receivers using Multi-Order Cumulants and Decision Trees

4QAM 100

16QAM 7 93

64QAM 100

BPSK 100

QPSK 100

8PSK 100
15 83.3
4QAM 100

16QAM 3 97

64QAM 100

BPSK 100

QPSK 100

8PSK 100
20 83.3
4QAM 100

16QAM 100

64QAM 100

Fig. 7 represents the performance of fine, medium and

coarse trees at different SNR values. Among three FT and
MT provides better classification accuracy than CT because it
fails to distinguish between 16QAM and 64QAM, this reason
CT is not consider for final comparison. The performance
comparison of proposed FT and MT with some of the
available approaches in the literature is shown in Fig. 8. Most
of the existing approaches are consider limited MPSK,
MQAM signals along with MASK and MFSK signals. As the
number of modulation techniques increases, then the
performances of existing approaches are degraded. With all
classes of MPSK and QAM signals, the proposed decision
trees gave more classification accuracy for lower SNR values
too. From these simulation results, it is clear that decision
Fig. 7 Classification Accuracy of Decision Trees
trees can provide optimal classification accuracy in
modulation recognition under noisy channels.

V. CONCLUSION
This paper presented decision tree approaches for
automatic modulation recognition of digitally modulated
signals. Earlier, these decision trees are applied to the various
image and statistical data analysis. The limitation of decision
trees is accuracy which may vary based on characteristics of
the data. The simulation results proved that even under high
noisy conditions proposed,

Fig. 8 Performance Comparison of DTC’s with Existing

Approaches

Published By:
Blue Eyes Intelligence Engineering
68 & Sciences Publication
 International Journal of Recent Technology and Engineering (IJRTE)
ISSN: 2277-3878, Volume-7 Issue-4, November 2018

fine tree and medium trees give improved classification 19. Song Y, Lu Y, “Decision tree methods: applications for classification
and prediction”, Shanghai Archives of Psychiatry, vol. 27, no. 2, pp.
accuracy than existing approaches. The complexities of the 130-135, Feb. 2015.
proposed approaches are far lesser than traditional 20. Z. Shan, Z. Xin and W. Ying, "Improved modulation classification of
approaches due to their simple approaches in classification. In MPSK signals based on high order cumulants," 2nd International
future, these techniques are also applicable to various signal Conference on Future Computer and Communication, Wuhan, pp.
V2-444-V2-448, 2010.
classifications and communication signals under fading
conditions.
Acknowledgement: First Author expresses his deep sense
of gratitude to the management of Shri Vishnu Engineering
College for Women for encouraging him in Ph.D. research
work.

REFERENCES
1. A Dobre, A Abdi, Y Bar-Ness, W Su, “Survey of automatic modulation
classification techniques: classical approaches and new trends”, IET
Communications, vol. 1, no. 2, pp. 137–156, Apr. 2007.
2. W Wei, JM Mendel, “Maximum-likelihood classification for digital
amplitude-phase modulations”, IEEE Transactions on
Communications, vol. 48, no. 2, pp. 189–193, Feb. 2000.
3. V. Choqueuse, S. Azou, K. Yao, L. Collin, and G. Burel, “Blind
modulation recognition for MIMO systems”, MTA Rev., vol. 19, no. 2,
pp. 183–196, Jun. 2009.
4. W. C. Headley, J. D. Reed, and C. R. C. M. da Silva, “Distributed
cyclic spectrum feature-based modulation classification”, in
Proceedings of IEEE Wireless Communication Network Conference,
Las Vegas, NV, pp. 1200–1204, Apr. 2008.
5. O. A. Dobre, M. Oner, S. Rajan, and R. Inkol, “Cyclostationarity-based
robust algorithms for QAM signal identification”, IEEE
Communication Letters, vol.16, no. 1, pp. 12–15, Jan. 2012.
6. W. Wei and J. M. Mendel, “Maximum-likelihood classification for
digital amplitude-phase modulations,” IEEE Transaction on
Communication, vol. 48, no. 2, pp. 189–193, Feb. 2000.
7. P. Panagiotou, A. Anastasopoulos, and A. Polydoros, “Likelihood ratio
tests for modulation classification,” in Proceedings of IEEE Military
Communication Conference., Los Angeles, CA, vol. 2, pp. 670–674,
Oct. 2000.
8. H. C. Wu, M. Saquib, and Z. Yun, “Novel automatic modulation
classification using cumulant features for communications via
multipath channels,” IEEE Trans. Wireless Communication., vol. 7,
no. 8, pp. 3098–3105, Aug. 2008.
9. J. L. Barrera, F. E. Hernandez, “Classification of MPSK Signals
through Eighth- Order Statistical Signal Processing”, IEEE Latin
America Transactions, vol. 15, no. 9, pp. 1601-1607, Sep. 2017.
10. P. Liu and P. L. Shui, “A new cumulant estimator in multipath fading
channels for digital modulation classification,” IET Communication,
vol. 8, no. 16, pp. 2814–2824, Aug. 2014.
11. D. C. Chang and P. K. Shih, “Cumulants-based modulation
classification technique in multipath fading channels”, IET
Communication, vol. 9, no. 6, pp. 828–835, Apr. 2015.
12. Z. Wu, S. Zhou, Z. Yin, B. Ma and Z. Yang, “Robust Automatic
Modulation Classification Under Varying Noise Conditions,” in IEEE
Access, vol. 5, pp. 19733-19741, 2017.
13. Mohannad Abu-Romoh, Ahmed Aboutaleb, Zouheir Rezki,
“Automatic Modulation Classification using Moments and Likelihood
Maximization”, IEEE Communication Letters, vol. 22, no. 5, pp.
938-941, May 2018.
14. Zhenyu Zhang, Zhong Hua, Yingzhe Liu, “Modulation classification
in multipath fading channels using sixth-order cumulants and stacked
convolutional auto-encoders”, IET Communications, vol. 11, no. 6, pp.
910-915, June 2017.
15. M. W. Aslam, Z. Zhu and A. K. Nandi, "Automatic Modulation
Classification Using Combination of Genetic Programming and
KNN," IEEE Transactions on Wireless Communications, vol. 11, no.
8, pp. 2742-2750, August 2012.
16. Lubing H, Zan Li, OA Dobre, “Low complexity automatic modulation
classification based on order-statistics,” IEEE Transactions on
Wireless Communications, vol. 16, no. 1, pp. 400-411, Jan. 2017.
17. K. Zhang, E. L. Xu, Z. Feng and P. Zhang, "A Dictionary Learning
Based Automatic Modulation Classification Method," in IEEE Access,
vol. 6, pp. 5607-5617, 2018.
18. Loh WY, “Fifty years of classification and regression trees”,
International Statistical Review, vol. 82, no. 3, pp. 329–348, June
2014.

Published By:
Blue Eyes Intelligence Engineering
69 & Sciences Publication
View publication stats