2010 First International Conference on Pervasive Computing, Signal Processing and Applications

An Intelligent Anti-jamming Frequency Hopping System

Xiang-Yi Meng, Ran Tao, and Lin-Na Jia

Department of Electrical Engineering
Beijing Institute of Technology
Beijing, China
mxy0827@bit.edu.cn

Abstract—This paper proposes an integrated intelligent anti-

jamming scheme for frequency hopping (FH) system based on
the modern signal processing, including the algorithms for
jamming recognition and jamming suppression. Without
increasing the bandwidth and hopping rate, the proposed
scheme can improve the robustness of the FH system against
various jamming.

Keywords-frequency hopping system; jamming recognition;

jamming suppression

I. INTRODUCTION
Figure 1. Structure diagram of the intelligent anti-jamming FH system
The frequency hopping (FH) system has good anti-
jamming performance due to its hopping carrier frequency Finally, the effectiveness of the two algorithms and the
under the control of the pseudorandom code. So the system feasibility of the intelligent anti-jamming scheme are verified
can evade the narrow-band jamming to a certain extent. But by simulations in section č.
the anti-jamming performance will become worse when it
encounters the wide-band blocking jamming and the II. INTELLIGENT ANTI-JAMMING SCHEME
follower jamming. The structure of the intelligent anti-jamming FH system
In order to resist the follower jamming, American is shown in Fig. 1. At the receiver of the FH system, we
Sanders Company proposed the shortwave differential establish an intelligent anti-jamming module, including
frequency hopping (DFH) system [1] in 1995, which jamming recognition module, jamming suppression module,
introduced the correlation of the pre and post FH signals to jamming suppression effect evaluation module, and adaptive
increase the hopping rate. However, it may be attacked by frequency hopping (AFH) [2] module.
the narrow-band jamming due to the correlation among
hopping channels, and also face the influence of the wide- III. JAMMING RECOGNITION ALGORITHM
band blocking jamming simultaneously.
At present, the anti-jamming performance of the FH In this paper, the considered Narrow-band jamming
system is improved by the active method, which contains includes single tone jamming, noise amplitude modulation
the improvement of the modulation method and the increase (NAM) jamming, noise frequency modulation (NFM)
of both the FH bandwidth and the hopping rate. However, jamming, amplitude shift keying (ASK) jamming, frequency
the improvement of the modulation method causes the shift keying (FSK) jamming, and phase shift keying (PSK)
problem about the compatibility with the original system. jamming while the wide-band jamming includes wide-band
The increase of the FH bandwidth results in the increasing noise jamming, wide-band comb jamming, and wide-band
consumption of the pace spectrum resource inevitably. High sweep rate jamming.
hopping rate increases the complexity of the modulation and The proposed jamming recognition algorithm can be
demodulation of the system. divided into three steps. Firstly, we remove the FH signals
In this paper, we propose an integrated intelligent anti- in order to eliminate the influence of the FH signals on the
jamming FH system by jamming recognition and jamming signals. Secondly, we discriminate the narrow-
suppression based on the modern signal processing. The band jamming from wide-band jamming based on the
proposed algorithm can improve the anti-jamming bandwidth estimation. Finally, we identify the concrete type
performance without the increase of bandwidth and hopping of jamming to realize the jamming recognition.
rate. In section Ċ, the structure of the intelligent anti- A. Feature extraction for narrow-band jamming
jamming FH system is established. The algorithm for the The parameter Ra V a2 / P a2 [3] is used to distinguish the
jamming recognition and deterministic jamming suppression
are described in section ċ and section Č, respectively. constant envelope jamming from the non-constant envelope

978-0-7695-4180-8/10 $26.00 © 2010 IEEE 1054

1053
DOI 10.1109/PCSPA.2010.259
10.1109/PCSPA.2010.155
jamming, where ȝa and V a2 are the mean value and variance R fm max{PR (D )} / PR (D )
D
of the jamming envelop, respectively. The parameter
where PR(Į) is the accumulation of the Radon transform of
J max max | FFT [acn (m)]2 / N s | [4] is used to separate the NAM
the spectrogram. The values of Rfm has good effect on the
jamming from the ASK jamming, where Ns is sequence recognition of the wide-band sweep rate jamming with
Ns
length and acn (m) a j (m) / ma  1 , in which ma ¦ a ( m) / N
j s
higher chirp rate.
n 1

and aj(m) is the instantaneous amplitude of the jamming. C. Program for jamming recognition
We also introduce the parameter ȝf [5] defined by According to the above analysis, an efficient and rational
flow of jamming recognition is proposed as shown in Fig. 2.
Pf E{| J (k ) |4 } / {E[| J (k ) |2 ]}2
IV. JAMMING SUPPRESSION ALGORITHM
where J(k) is the instantaneous frequency of the jamming.
The parameter ȝf not only takes good effect on separating A. Suppression of Single Tone Jamming
the single tone jamming from the PSK jamming, but also The single tone jamming can not be rejected by the
discriminates the NFM jamming from the FSK jamming to filtering in the frequency domain as common due to the
a certain extent. But it can not deal with the case when the reason that the single tone jamming and the FH signals may
modulation index of the NFM jamming is 0.2 and the FSK be in coincidence in the frequency domain.
jamming has four spectrum peaks. To solve the problem at low JSR, we should firstly make
In order to solve the above problem, we propose a sure which hop channel is interfered by the single tone
parameter Rv ¦ J u (k ) /  mR  mL  1 where jamming. Then, we filter the single tone jamming in the
J u ( k ) t 0.5 frequency domain as common. After de-hopping, if the peak
J u (k ) | J ( k ) | / max(| J ( k ) |) , mL min (k ) , mR max (k ) . The of the spectrum for the jammed channel is low enough or the
J u ( k )t0.5 J u ( k )t0.5
kurtosis of the spectrum is low enough, the FH signals are
parameter Rv reflects the compactness of the spectrum peak filtered owing to the coincidence and furthermore the hop
and it can be utilized to separate the NFM jamming with a channel is used for the demodulation. If not, the FH signals
low modulation index from the FSK jamming. are not coincident with the jamming and the demodulation
Furthermore, we set Rp to separate the FSK jamming results are used directly. The realization diagram of the
from the PSK jamming, which can be written as Rp = Pm /Bw improved filtering algorithm is depicted in Fig. 3.
where Bw is the normalized 3dB bandwidth of the jamming The received FH signal

and Pm is the normalized peak of the average spectrum after FH signal elimination in the time-
frequency domain
the FH signal elimination. The value of Rp for the non- Bandwidth estimation Bw
frequency modulation jamming is almost larger than the Y N
Bw<TBw
ones for the frequency modulation jamming. Narrow-band jamming Wide-band jamming

B. Feature extraction for wide-band jamming FH signal elimination

in the time and
Y ( Fss ! TF ) &
(Dmax 90D )
N

frequency domain comb (Rf ! Tf 1)&

Y N
For the wide-band jamming, we use the parameter Y Ra ! Ta N
(Dmax z 90D )
Low (Rfm ! Tm1)&
N Y
Y J !T N sweep (Dmax z 90D )
Nt 1 max J
rate

¦ ( P (n)  P (n))
ASK NAM N (Rf !Tf 2)&
Y High
2
Fss s s / Nt Y Rp !Tp N (Rfm !Tm2) sweep
rate
unknown noise
n 0 Y Pf ! TP1 N
(P f  TP 2 )
Y N
ST PSK &( Rv ! Tv )

to extract the wide-band comb jamming , where Nt is the NFM FSK

L
length of Pqt(n) and Ps (n) Pqt (n)  ¦ Pqt (n  i ) / (2 L  1) , in Figure 2. Jamming recognition flow in the FH system
i L

which Pqt(n) is the average spectrum after the FH signal Use the hop
channal to
elimination and L is the width of moving average window. Yes
demodulate

In order to identify the wide-band sweep rate jamming, Jamming

recognition
Hop channel
judgement
Filtering in the
frequency domain
De-
hopping
Low peak or
low kurtosis

we introduce the parameter R f max{RT (u ,D )} / RT (u ,D ) No Demodulate

u ,D directly

where RT(u,Į) is the Radon transform of the spectrogram [6].

The parameter Rf can be used to extract the wide-band
Figure 3. Diagram of the improved filtering algorithm
sweep rate jamming from the received FH signals. But it
should be noted that the value of Rf becomes smaller with
the increase of the chirp rate due to the reason that the
higher chip rate results in more times repeat in the sampled
bandwidth and more peaks in the Radon transform domain.
In order to solve the above problem, we set
Figure 4. Diagram of the time domain elimination based on the FRFT

1054
1055
 recognition probability of different kinds of jamming can
almost approach to 95%. So the proposed algorithm is of
high robustness against different JSR and SNR.
According to the jamming suppression algorithm for the
deterministic jamming in the FH system, we obtained the
BER performance through the Monte Carlo realization for
50000 times at different SNR and JSR. Tables Ċ and ċ
present the BER of the FH system pre and post the
suppression against single tone jamming and wide-band
sweep rate jamming.
From the simulation results, we can find that the
proposed suppression algorithm for the deterministic
jamming is of good robustness against different JSR.
However, due to the incorrect jamming recognition, the BER
pre and post suppression at a low jam-to-signal ratio is close,
Figure 5. Deterministic jamming suppression flow in the FH system
even equal to each other.
Aiming at the side lobe leakage problem in the frequency
domain, a jamming suppression algorithm based on the VI. CONCLUSION
cancellation in the time domain is used. The amplitude of the In this paper, an intelligent anti-jamming module in the
jamming is derived from the value of the spectrum peak and FH system is proposed, including the jamming recognition
the initial phase of the original signal can be derived from algorithm and the deterministic jamming suppression
the bi-spectrum estimation of the new sinusoidal signal [7]. algorithm. In the jamming recognition process, three new
Furthermore, the jamming signals can be eliminated from the parameters are set besides the traditional parameters for
received FH signals using the reconstructed signals. further classification of different kinds of jamming. The
proposed algorithm can result in the recognition probability
B. Suppression of Wide-Band Sweep Rate Jamming
of different kinds of jamming approach to 95%. In the
As a generalization of the Fourier transform, the jamming suppression process, a novel method based on the
fractional Fourier transform (FRFT) is more suitable for time domain cancellation algorithm and the improved
processing the nonstationary signals [8]. Compared with the filtering in the transform domain is proposed for different
filtering in the Fourier domain, the filtering in the optimal JSR. In the future work, we can engage in the algorithm of
fractional Fourier domain (FRFD) can enable significant the nondeterministic jamming suppression in the FH system
reduction of the error for certain types of noise and and the realization of the jamming suppression effect
degradation [9], such as sweep rate jamming. Using the evaluation module and the AFH module.
filtering in the FRFD, the sweep rate jamming can be
suppressed effectively at low JSR. ACKNOWLEDGMENT
Similar to the single tone jamming, the wide-band sweep This work was supported by the Ministerial Natural
rate jamming in the FRFD also has large side lobe leakage at Science Foundation of China under Grant No.41001020201.
a high JSR. Thus, we propose a cancellation algorithm in the
time domain based on the FRFT and the bi-spectrum REFERENCES
estimation, which is depicted in Fig. 4. Above all, we can [1] D L Herrick, P K Lee. “CHESS: a new reliable high speed HF radio”,
describe the flow chart of the deterministic jamming Proceedings of IEEE MILCOM, 1996, pp. 684-690.
suppression in the FH system as shown in Fig. 5. [2] J Zander, G Malmgren. “Adaptive frequency hopping in HF
communications”, IEE Proceedings Commun, vol.142, Feb. 1995,
V. SIMULATIONS pp.99-105.
In the simulations, the working frequency band of the FH [3] Y T Chan, L G Gadbois, P Yansouni. “Identification of the
system is 208-224MHz, the FH rate is 12.5K/s, the number modulation type of a signal”, IEEE ICASSP. pp. 838-841, 1985.
of hopping frequency is 32, and the narrow-band modulation [4] A K Nandi, E E Azzouz. “Automatic analogue modulation
recognition”, Signal Processing, vol. 46 Feb. 1995, pp.211-222,.
is 4FSK. At the demodulation stage, the sampling frequency
[5] A K Nandi, E E Azzouz. “Algorithms for automatic modulation
is 16MHz, and every frame data contains six hopping recognition of communication signals”, IEEE Trans Commun, vol.46,
frequency. The types and corresponding parameters for Apr. 1998, pp.431-436.
different kinds of jamming are given in Table ĉ. [6] E Magli, G Olmo, L L Presti. “Pattern recognition by means of the
Based on the simulations at different SNR and different Radon transform and the continuous wavelet transform”, Signal
JSR, we set the optimum thresholds of the above Processing, vol.73, Mar. 1999, pp.277-289.
characteristic parameters as follows: TBw=0.06, Ta=0.19, TȖ= [7] C L Nikias, M R Raghuveer. “Bispectrum estimation: a digital signal
50, Tp=130, Tȝ1=1450, Tȝ2=200, Tv=0.09, TF=0.49, Tf1=10.5, processing framework”, IEEE Proceedings, vol. 75, July 1987,
pp.869-891.
Tm1=12, Tf2=3.5, and Tm2=3.5. The simulation results for
[8] R Tao, B Deng, Y Wang. “Research process of the fractional Fourier
jamming recognition over 500 Monte Carlo realizations are transform in signal processing”, Science in China: Series F, vol.49,
illustrated in Figs. 6 (a) and (b). We can find that the Jan. 2006, pp.1-25.

1055
1056
[9] M A Kutay, H M Ozaktas, O Ankan, L Onural. “Optimal filtering in [10] L Qi, R Tao, S Y Zhou, Y Wang. “Detection and parameter
fractional Fourier domains”, IEEE Trans Signal Process, vol.45, May estimation of multicomponent LFM signal based on the fractional
1997, pp.1129-1143. Fourier transform”, Science in China: Series F, vol.47, Feb. 2004,
pp.184-198.

TABLE I. LIST OF DIFFERENT KINDS OF JAMMING

Jamming types Jamming parameters
Narrow Single tone jamming {CW}
band Modulation jamming {NAMǃNFM}
jamming
Shift keying jamming {2ASKǃ4ASKǃ2FSKǃ4FSKǃBPSKǃQPSK}
Wide-band noise jamming Bandwidth {30%—100%} (FH bandwidth)
Wide Number of comb {2ˈ4ˈ6ˈ8ˈ10ˈ12ˈ14ˈ16}
band Wide-band comb jamming
jamming Comb modulation Any modulation of the narrow-band jamming
Wide-band sweep rate jamming Chirp rate {0.5ˈ1ˈ1.5ˈ2ˈ2.5ˈ3ˈ3.5ˈ4} (25/12K*FH bandwidth)
TABLE II. BER PRE AND POST SINGLE TONE JAMMING SUPPRESSION
BER Pre and JSR=0dB JSR=5dB JSR=10dB JSR=15dB JSR=20dB JSR=25dB JSR=30dB JSR=35dB JSR=40dB
Post Suppression
SNR= Pre 0.0118 0.0214 0.0210 0.0232 0.0227 0.0218 0.0237 0.0298 0.0452

-15dB Post 0.0118 0.0214 0.0099 0.0077 0.0063 0.0064 0.0056 0.0072 0.0071

SNR= Pre 0.0086 0.0143 0.0148 0.0142 0.0128 0.0150 0.0143 0.0178 0.0312

-10dB Post 0.0086 0.0027 0 0 0 0 0 0 0

SNR= Pre 0.0072 0.0161 0.0145 0.0158 0.0159 0.0154 0.0155 0.0179 0.0326

-5dB Post 0.0011 0 0 0 0 0 0 0 0

TABLE III. BER PRE AND POST WIDE-BAND SWEEP RATE JAMMING SUPPRESSION
BER Pre and Post JSR=0dB JSR=5dB JSR=10dB JSR=15dB JSR=20dB JSR=25dB JSR=30dB JSR=35dB JSR=40dB
Suppression
SNR= Pre 0.1483 0.1622 0.1738 0.1849 0.2047 0.2096 0.2199 0.2342 0.2793

-15dB Post 0.1476 0.1424 0.1368 0.1348 0.1467 0.1397 0.1398 0.1383 0.1433

SNR= Pre 0.0131 0.0350 0.0553 0.0712 0.0815 0.1031 0.1148 0.1308 0.1853

-10dB Post 0.0073 0.0058 0.0048 0.0059 0.0057 0.0083 0.0056 0.0069 0.0063

SNR= Pre 0.0008 0.0228 0.0476 0.0641 0.0726 0.0907 0.1027 0.1165 0.1627

-5dB Post 0 0 0 0 0 0 0 0 0

JSR=0dB SNR=0dB
1 1

0.99 0.99

0.98 0.98

0.97 0.97
recognition probability
recognition probability

0.96 0.96

0.95 0.95 ST
ST
NAM NAM
0.94 0.94
NFM NFM
ASK ASK
0.93 0.93
FSK FSK
0.92 PSK 0.92 PSK
noise noise
0.91 comb 0.91 comb
sweep sweep
0.9 0.9
-5 0 5 10 15 20 25 -5 0 5 10 15 20 25
SNR(dB) JSR(dB)

(a) (b)
Figure 6. Dependence of recognition probability for different kinds of jamming (a) versus JSR (b) versus SNR

1056
1057