Download as pdf or txt
Download as pdf or txt
You are on page 1of 20

Sensing and Imaging (2020) 21:51

https://doi.org/10.1007/s11220-020-00313-3

ORIGINAL PAPER

Estimation of Modulation Parameters of LPI Radar Using


Cyclostationary Method

Raja Kumari Chilukuri1,2 · Hari Kishore Kakarla1 · K. Subbarao3

Received: 29 May 2020 / Revised: 24 August 2020 / Accepted: 23 September 2020 /


Published online: 12 October 2020
© Springer Science+Business Media, LLC, part of Springer Nature 2020

Abstract
Low Probability of intercept (LPI) radars work on the principle of low peak-power
and wide bandwidth. To achieve this, the LPI radars use special type of modulated
waveforms which are difficult to intercept. The main tasks of intercept receivers are
to detect, estimate and classify LPI signals even in the presence of high noise. Accu-
rate estimation of the modulation parameters also provides information about threat
to radar so that necessary counter measures could be initiated against the enemy
radars. In the present work, Cyclo-stationary (CS) method is used to estimate the
parameters of Polytime coded LPI signals. CS method is efficient for the analysis
of periodic waveforms like LPI signals and finds applications in many fields such as
array signal processing, estimation of direction of arrival, signal detection. Cyclic
auto correlation function and the spectral correlation density function (SCD) which
are the basis for CS analysis are computed for all the four types of polytime codes.
Generation of all the four types of Polytime codes and computation of SCD coef-
ficients are carried out using MATLAB. From the contour plot of SCD function, the
parameters carrier frequency, bandwidth and code rate are measured. It is assumed
that the received signal is not corrupted with any noise. The simulation results are
compared with the values available in the literature. The maximum error of esti-
mation is well within ± 6% for all the codes in the proposed method whereas the
maximum error reported in the literature is 25.7%. Hence the proposed method is
better. Since CS method is sensitive to sampling frequency, the analysis is repeated
for three different sampling frequencies and in all cases, the estimation error is less
than 6%.

Keywords LPI radar · Polytime codes · Cyclo-stationary (CS) · Cyclic auto


correlation function (CACF) · Spectral correlation density (SCD) · Auto-correlation
function (ACF) · Fourier transform (FT) and code rate ­(Rc)

* Raja Kumari Chilukuri


Chrajakumari@gmail.com
Extended author information available on the last page of the article

13
Vol.:(0123456789)
51 Page 2 of 20 Sensing and Imaging (2020) 21:51

1 Introduction

A low probability of identification (LPID) and a low probability of intercept (LPI)


are two important tactical requirements of any radar. The property of LPI radar is
that, due to its low power, wide data transfer capacity, frequency fluctuation, or
other design aspects, makes it hard for it to be distinguished utilizing a passive inter-
cept receiver. A LPI radar uses a special emitted waveform intended to prevent a
non-cooperative intercept receiver from intercepting and detecting its emission. It
transmits a low power intra-pulse tweaked waveform so that the identified object
is resolved with a decent range resolution [1–4]. It utilizes separate transmitting
and receiving antennas that are co-mounted. Modulation of LPI radar signals also
increases the detection capability of the radar. In many applications like naviga-
tion, surveillance, tactical airborne targeting and altimeters, the interception of the
radar signals can rapidly prompt electronic attack or jamming if the parameters of
the emitter can be resolved. However, this is generally a difficult task because the
probability of intercepting and detecting LPI signals is low. Hence these signals are
also known as Low Probability of Detection (LPD). Since LPI radar works on the
principle of low power, the possibility of being destroyed by anti-radiation missiles
(ARMs) and precision-guided munitions is also less [5, 6].
Modern electronic intercept systems must perform the tasks of detection, estima-
tion and classification of LPI signals even in the presence of high noise interference
and multiple signals [7]. But it becomes more and more difficult for the non coop-
eration electronic war receivers to intercept the target because of LPI signals. The
characteristics of LPI signals are that the peak side lobe levels (PSL) and integrated
side lobe levels (ISL) in Auto-Correlation Function (ACF), at the output of the
matched filter must be low. While modulating the transmitted waveform, the order
for the phase or frequency must be chosen properly so that the magnitude of the
side lobes are very low [3, 8]. A low side lobe at the output of matched filter dem-
onstrates the capacity to recognize even weak targets in presence of solid targets.
On the other hand if any of the side lobe is high, it is possible that another target
which is present nearby may go undetected. In order to achieve these properties, LPI
signals are modulated and transmitted [4, 7, 9]. The types of modulation techniques
generally used are

1. Barker codes
2. Frequency modulated continuous wave (FMCW)
3. BPSK codes
4. Poly phase codes
5. Frank codes
6. Polytime codes

The importance of intrapulse modulation of LPI radars has been increased in


the field of electronic warfare (EW) and MIMO radar systems. Accurate estima-
tion of modulation parameters and identification of the type of modulation will pro-
vide information regarding the presence of threat to radar and hence an immediate

13
Sensing and Imaging (2020) 21:51 Page 3 of 20 51

counter action could be initiated by the Electronic support and/or Electronic attack
systems against the enemy. Estimation of modulation parameters also help to design
modern receivers which detect, intercept, analyse and classify signals [5]. Identifica-
tion of the parameters help to reguide and to retransmit to the transmitter without
any effect to our electronic systems. The parameters are very useful in the devel-
opment of many systems such as EA, ES and MIMO radar applications [3]. Also
accurate measurement of the parameters helps to detect two similar radars which are
operated at nearby frequencies [5]. It also provides required information regarding
the pulse compression waveform used which would help in the design of intercept
receivers [3].
Traditional signal processing methods are not suitable to analyse the complex LPI
radar waveforms. Many new algorithms called Time–Frequency (T–F) methods are
developed in the past two decades which will be able to analyse the signal in both
the domains simultaneously [7]. T–F methods are widely used for the analysis and
processing of non stationary signals [7, 8, 10–12].
The T-F methods generally used are

i. Choi–William distribution (CWD)


ii. Wigner–Ville distribution (WVD)
iii. Quadrature mirror filter bank (QMFB)
iv. Cyclo stationary (CS) method
v. Wavelet transform (WT) techniques
vi. S-transform/modified S-transform (MST)

Although several methods are reported in the literature, no single method is


suitable to analyse all the LPI radar waveforms. In other words, different meth-
ods analyse efficiently different types of modulations. Yongjian et al. [10] used
CWD method and radial integration method for detection of LPI radar signals.
The modulations detected are FMCW, BPSK and poly phase codes. The carrier
frequency of the LPI signals used is a low frequency (1.5 kHz). Ravi Kumar
et al. [3] classified different types of modulations (BPSK, LFM and polyphase)
using Feature extraction based on RAT and Fractional Autocorrelation. The esti-
mated error is less than 10% for most of the modulations. Spectral correlation
techniques and support vector machines are used to classify various LPI signals.
It is found that the percentage of correct classification varies from 60 to 100 as
the SNR is increased from − 5 to +10 dB [4]. Gulum et al. [11] used Pseudo
WVD (PWVD) method to detect the FMCW waveform and Hough transform for
the measurement of the modulation parameters (carrier frequency, band width
and code period) with a maximum error of 3.78% at + 6 dB SNR. The frequen-
cies used are very low which are in the range of 1.5–2.5 kHz. Also WVD method
introduces cross terms and filters are to be used to suppress them [7]. Singh et al.
[5] developed a digital receiver based ELINT system for detection and identifi-
cation of intra pulse modulated LPI radar signals and estimated the parameters
using All Phase FFT. The frequencies used are around 1.0 GHz. The ALL Phase
FFT is implemented in a high speed FPGA. Samsunder et al. [12] used Modified

13
51 Page 4 of 20 Sensing and Imaging (2020) 21:51

Simulated Annealing (MST) method and quadrature mirror filter bank method to
analyse polyphase sequences and reported that MST method is better. Samsun-
der et al. analysed only ­T1(n) codes using cyclostationary method. The maximum
estimation error is 25.7%. FFT accumulation method was used to compute the
Spectral correlation density (SCD) function [8].
In the present work, cyclostationary method is used to estimate the parameters
of Polytime coded LPI radar signals. The paper is organised as follows. Section 2
discusses cyclostationary techniques of analysis and Sect. 3 discusses the poly-
time codes and their generation. In Sect. 4, estimation of modulation parameters
and the simulation results are discussed. Conclusions are discussed in Sect. 5.

2 Cyclo Stationary (CS) Techniques for Estimation of Radar


Parameters

Cyclostationary signals are periodically stationary. By processing the signals as


CS, one can take advantage of periodic features of wave forms. The periodicities
arise from modulation, coding, keying and sampling. All LPI radar waveforms
are periodic [7, 8]. CS analysis transforms the signal into the cycle frequency-
frequency domain or bifrequency domain. It offers additional properties which
are not available in time–frequency domain. The main property of CS is that
they have spectral correlation with frequency shifted versions of itself at certain
frequency shifts. The spectral properties of the signal can be seen from the plot
of the spectral correlation density (SCD) function [13–15]. SCD function fully
characterizes the second order statistical behavior in the frequency domain of a
CS signal. Cyclostationary finds applications in many fields such as array signal
processing, estimation of direction of arrival, estimation of time of arrival, signal
detection and estimation of parameters. Bouillaut et al. [16] compared cyclosta-
tionary method and Bilinear method for early diagnosis of helicopter gear box
and reported that that the first method is preferable.
A signal is cyclostationary of order ‘n’ if and only if one can find some nth
order nonlinear change of the signal that will produce limited quality additive
sine wave segments that outcome in spectral lines [7, 8]. As an example for n = 2,
the auto correlation function (ACF) will generate spectral lines. Many character-
istics of LPI radar can be estimated using cyclic auto correlation function (CACF)
and the SCD function which are the basis for cyclostationary analysis.
Let x(t) be the LPI signal received by the intercept-receiver. The auto correla-
tion function(ACF) is given by
T∕
Rx (𝜏) ≜ lim
2 ( ) ( )
T→∞ �−T∕
𝜏 ∗ 𝜏
x t+ x t− dt (1)
2 2
2
The cyclic autocorrelation function (CACF) of x(t) is given by

13
Sensing and Imaging (2020) 21:51 Page 5 of 20 51

T∕ (
≜ lim
2 ) ( )
T→∞ T �−T∕
1 𝜏 ∗ 𝜏 −j2𝜋𝛼t
R𝛼x (𝜏) x t+ x t− e dt (2)
2 2
2
where α is the cycle frequency. It is well known that the power spectrum of the
signal is obtained by computing the Fourier transform (FT) of the ACF as given in
Eq. (3)

∫−∞ (3)
Sx (f ) = Rx (𝜏)e−j2𝜋f 𝜏 d𝜏

In a similar way, the SCD function is obtained by computing the FT of the


cyclic ACF as given in Eq. (4)

Sx𝛼 (f ) ≜
∞ ( ) ( )
�−∞
1 𝛼 ∗ 𝛼
R𝛼x (𝜏)e−j2𝜋f 𝜏 d𝜏 = lim XT f + XT f − (4)
T→∞ T 2 2

where
T∕
XT (f ) ≜
2
�−T∕
x(v)e−j2𝜋fv dv (5)
2
is the FT of the signal x(v). The factor α prompts a two-dimensional representation
Sx𝛼 (f ) which is the bi-frequency plane or (f, α) plane [7, 8]. Normally two methods
are used to compute the SCD function. They are (i) FFT accumulation method and
(ii) Direct frequency-smoothing (DFS) method. The second method is used here to
compute the SCD coefficients. The DFS algorithm is computed in two steps. In the
first step the spectral components of the data are computed and then the spectral cor-
relation operations are performed directly on the spectral components in the second
step. The DFS method is computationally efficient compared to Wigner-Ville Dis-
tribution. But DFS method is usually less efficient than a time-smoothing approach.
The discrete equivalent of Eq. (4) is given in Eq. (6.)
N−1 (
1 ∑ 𝛾 ) ∗( 𝛾)
SX𝛾 (n, k) = XN n, k + XN n, k − (6)
N N n=0 2 2

where
N−1

XN (n, k) = w(n)x(n)e−j2𝜋kn∕N (7)
n=0

Equation 7 represents the N-point FFT of the signal x(n) and w(n) is the win-
dow function, N is the total number of points of the FFT related to the total obser-
vation time, Δt. 𝛾 is the cycle frequency discrete equivalent of α.
The cycle frequency resolution of the algorithm is defined as

13
51 Page 6 of 20 Sensing and Imaging (2020) 21:51

Fs
𝛾res = (8)
N
and the frequency resolution is
MFs
kres = (9)
N
M is the Grenander’s Uncertainty condition and selected such that M ≫ 1. Fs
is the sampling frequency. From Eqs. (8) and (9), it is clear that the resolution is
better if ­Fs is small. The block diagram for the computation of SCD coefficients
using DFSM method is shown in Fig. 1.

3 Polytime Codes

Poly time codes are the counter part of poly phase codes. In polytime codes, the
phase steps vary as required to approximate an underlying RF or LFM waveform.
But the time spent at any phase state is constant. The minimum time spent at a
given phase state is a critical parameter that effects the BW and the sampling rate
[1, 8].
The main advantages of polytime codes are

(i) With different phase states, arbitrary time-bandwidth waveforms can be gener-
ated.
(ii) Polytime codes approximate the main compressed pulse of the underlying
waveform
(iii) As the number of phase states increases, the agreement in time side lobe
behavior improves.
(iv) Two spectral components that are mirror images of one another can be gener-
ated by a single code with two phase states.

Fig. 1  Direct frequency smoothing method of computing SCD

13
Sensing and Imaging (2020) 21:51 Page 7 of 20 51

3.1 Generation of Polytime Codes

3.1.1 T1(n) Code

T1(n) code is created utilizing the stepped-frequency waveform that is “zero-beat” at the
main section. It means that the first code fragment is at “zero” recurrence. The equa-
tion for the wrapped phase, φ(t), versus time for this code is given by
{ [ ] }
2𝜋 jn
𝜑(t) = MOD INT (kt − jT) , 2𝜋 (10)
n T

where n is the no. of phase states, k is the number of segments, j = 0, 1, 2, …, k − 1


is the segment number in the stepped RF waveform, T is the total duration of the
code and t is the time.

3.1.2 T2(n) Code

By approximating a stepped-frequency waveform that is zero-beat at the center fre-


quency produces the T ­ 2(n) code. If the waveform has an odd number of portions, the
frequency of the middle segment is the zero-beat frequency. If the number of sections
are even, the frequency halfway between the two center most sections gives the zero beat
frequency. For T­ 2(n) code, the relation between time and the wrapped phase is given by
{ [ ( ) ] }
2𝜋 2j − k + 1 n
𝜑(t) = MOD INT (kt − jT) , 2𝜋 (11)
n T 2

3.1.3 T3(n) Code

T3(n) code has a direct FM underlying waveform. The ­T3(n) is zero beats at its leading
edge. The ­T3(n) code is generated using
{ [ ] }
2𝜋 nΔFt2
𝜑(t) = MOD INT , 2𝜋 (12)
n 2T

where ∆F is the modulation bandwidth.

3.1.4 T4(n) Code

T4(n) code has additionally an underlying linear FM waveform to generate the signal.
Contrasted with T3(n), it has its zero–beat at the center frequency. The T
­ 4(n) code is
generated using
{ [ ] }
2𝜋 nΔFt2 nΔFt
𝜑(t) = MOD INT − , 2𝜋 (13)
n 2T 2

where all the variables are defined as above.

13
51 Page 8 of 20 Sensing and Imaging (2020) 21:51

The time spent at each phase state depends on the number of phases. On the off
chance that the number of phase states is expanded, at that point the time spent on
each phase state diminishes bringing about a signal which is hard to analyse. Pol-
ytime coded signals are analyzed using cyclostationary method and the following
modulation parameters are estimated.
Carrier Frequency ( FC),
Time Period (T),
Bandwidth (BW)
Code Rate ( RC)
Bit Duration ( tb)
Pulse Compression Ratio (PCR)
Figure 2 shows the pulse compression waveform and its ACF. The peak ACF
value is 13. The relation between time period and bit duration are also shown in the
figure. The parameter bit duration is very critical as it is related to BW [17].Band
width,
BW = 1∕tb . (14)

Time period,
T = 1∕Rc (15)
Pulse compression ratio is given by
T
PCR =
tb (16)

4 Estimation of Modulation Parameters

Estimation of modulation parameters plays a crucial role in the design of intercept


receivers and in the design of ES/ELINT systems. Correct estimation of carrier fre-
quency, FC , is very important as it provides jamming and other important counter
actions very effective [5]. The simulation work is carried for three different sampling

Fig. 2  a 13-bit pulse compres-


sion waveform. b Its ACF

13
Sensing and Imaging (2020) 21:51 Page 9 of 20 51

frequencies. The complete CS analysis including the generation of polytime coded


signals is carried out using MATLAB. It is assumed that the modulated signal is not
corrupted with any noise. The flow diagram for estimation of parameters using CS
method is shown in Fig. 3.

4.1 Analysis of ­T1(2) Code

T1(n) code is generated using Eq. 10. The number of phases is selected as n = 2 for
all the codes. SCD coefficients, SX𝛾 (n, k), are computed using Eq. 6 and its contour
N
plot (Cycle frequency vs. frequency) is shown in Fig. 4. The x-axis represents the
cycle frequency, y-axis represents frequency and the color of the plot depends on
SX𝛾 (n, k) values. The complete bi-frequency plot is shown in Fig. 4a. It may be
N
observed that the plot is spread in the form of four diamonds. The carrier frequency
is measured as follows. The difference from the origin to the centre of the selected
diamond on the cycle frequency axis gives 2FC as shown in the figure. The closer
approximation of the selected portion of Fig. 4a is shown in Fig. 4b. The 2FC point
can be seen more clearly in Fig. 4b. The x-coordinate of the center point gives 2FC
which is indicated as 3.982 GHz. It may be noted that for the 2FC point, the SCD
value is maximum and the color is red. Hence FC is measured as 1.991 GHz. The
actual carrier frequency used is 2.0 GHz. Band width and code rate are measured
from Fig. 4b. The bandwidth, BW, is measured as the difference from the center to
the end point of the diamond on x-axis as shown in the figure which is measured as
1758 MHz. The difference between any two adjacent spots on the cycle frequency
axis gives the code rate (­ Rc). It is measured as 62.0 MHz. The bit duration and time
period are calculated using Eqs. (15) and (16) respectively. The error for each of the
parameter is calculated using Eq. (17).
| true value − estimated value |
error = || |
| (17)
| true value |

The true values, the estimated values, the error in estimation and the litera-
ture values are shown in Table 1. The second column indicates the method used

Fig. 3  Flow diagram for measurement of T1 (n) Parameters

13
51 Page 10 of 20 Sensing and Imaging (2020) 21:51

Fig. 4  Contour plot of T1(n) code for ­Fs = 7Fc and ­Fc = 2GHz

and the third column shows the sampling frequency. The true value, estimated
value and the error for each of the three parameters are shown in columns 4
through 12. The analysis is repeated for three different sampling frequencies and
the simulation results are compared with the values reported in the literature [8].
One can find a good agreement with all the measured parameters and literature
values if ­Fs ≥ 5Fc. But when ­Fs = 3Fc, the proposed method is superior as the
error in the literature is 25.7% which is very high.

4.2 Analysis of ­T2(2) Codes

The ­T2(2) code is generated using Eq. 11 and its spectral correlation density is com-
puted as explained in Sect. 2 and its contour plot is shown in Fig. 5. It may be noted

13
Table 1  Comparison between true and estimated parameters for ­T1 code
Sensing and Imaging (2020) 21:51

S. No Method FS (GHz) FC (GHz) BW (MHz) RC (MHz)

True value Estimated %Error True value Estimated) %Error True value Estimated %Error
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)

1 Proposed 7 ­Fc 2 1.991 0.45 1750 1758 0.45 62 62 0


2 7 ­Fc 1 1 0 1750 1750 0 62 62.5 0.8
3 Literature values [8] 7 ­Fc 1 1 0 1750 1750 0 62.5 62.5 0
4 Proposed 5 ­Fc 3 2.991 0.02 1750 1750 0 62 64 3.22
5 5 ­Fc 1 0.997 0.3 1750 1760 0.57 62 62 0
6 Literature values [8] 5 ­Fc 1 1 0 1750 1750 0 62.5 62.5 0
7 Proposed 3 ­Fc 3 2.995 0.16 1750 1750 0 62 60 3.2
8 3 ­Fc 1 0.999 0.05 1750 1834 4.5 62 61 1.61
9 Literature values [8] 3 ­Fc 1 1 0 1750 1300 25.7 62.5 62 0.8
Page 11 of 20 51

13
51 Page 12 of 20 Sensing and Imaging (2020) 21:51

that this plot is also spread in the form of four diamonds. The carrier frequency, ­Fc
is estimated from Fig. 5a as explained for T ­ 1(2) and it is measured as 1.9910 GHz.
The closer approximation of the selected portion of Fig. 5a is shown in Fig. 5b. BW
and ­Rc are estimated as explained for ­T1(2). The measured values of BW and ­Rc are
1773 MHz and 62.0 MHz respectively.
Table 2 shows the comparison between actual, estimated parameters and error for
“T2 code” for three different sampling frequencies. To the best of the knowledge of
the authors, the estimated values are not available in the literature for ­T2(2), ­T3(2),
­T4(2) codes. As can be seen from the table, the maximum error is 4.83% for R ­ c for
­Fs = 5Fc. For this code also, the maximum estimation error is less than 5% for all the
parameters.

Fig. 5  Contour plot of T2(n) code for ­Fs = 7Fc with ­Fc = 2GHz

13
Sensing and Imaging (2020) 21:51

Table 2  Comparison between true and estimated parameters for ­T2 code
S.No FS GHz) FC (GHz) BW(MHz) RC(MHz)

True value Estimated %Error True value Estimated %Error True value Estimated %Error
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)

1 7 ­Fc 2 1.9910 0.45 1750 1773 1.29 62 62 0


2 7 ­Fc 1 0.995 0.5 1750 1766 0.90 62 62 0
4 5 ­Fc 2.5 2.493 0.28 1750 1750 0 62 65 4.83
5 5 ­Fc 1 0.9970 0.3 1750 1750 0 62 62 0
7 3 ­Fc 3 2.89 3.66 1750 1760 0.571 62 61.2 0.32
8 3 ­Fc 1 0.9995 0.05 1750 1835 4.63 62 61 1.61
Page 13 of 20 51

13
51 Page 14 of 20 Sensing and Imaging (2020) 21:51

4.3 Analysis of ­T3 (n) Codes

The ­T3(2) code is generated using Eq. 12 and the SCD coefficients are computed as
explained in Sect. 2. Its contour plot is shown in Fig. 6. It may be noted that this plot
is also spread in the form of four diamonds. The carrier frequency ­Fc is estimated
from Fig. 6a as explained for ­T1(2) and it is measured as 1.9910 GHz. The closer
approximation of the selected portion of Fig. 6a is shown in Fig. 6b. It may be noted
that for only T
­ 3(2) code, the band width is measured on y-axis. The bandwidth is
measured as the difference from the starting point to the end point of the diamond
on y-axis as shown in Fig. 6b. Hence BW is measured as 1835 MHz. The procedure
for measuring ­Rc is same as for ­T1(2) code and it is measured as 62.0 MHz.

Fig. 6  Contour plot of T3(n) code for ­Fs = 7Fc with ­Fc = 2GHz

13
Sensing and Imaging (2020) 21:51 Page 15 of 20 51

Table 3 shows the comparison between actual, estimated parameters and error
for “T3 code”. For this code, the maximum error is only 4.73% for band width when
­Fs ≥ 5Fc. But when ­Fs = 3Fc, the maximum error is 5.42% for BW which is more than
5%.

4.4 T4(n) Code

T4(2) code is generated using Eq. 13 and its spectral correlation density is computed
as explained in Sect. 2. Its contour plot is plotted in Fig. 7. For this code also, the
contour plot is spread in the form of four diamonds. But the patterns are different
for all the poly time codes. The procedure for measuring ­Fc, BW and ­Rc is same as
for ­T1(n) code and the measured values are 1.991 GHz, 1836 MHz and 62.0 MHz
respectively. Table 4 shows the comparison between actual, estimated parameters
and error for “T4 code” for three different sampling frequencies. From this table,
it may be observed that the maximum error is 4.68% for band width when F ­ s = 7Fc.
For this code also, the maximum error is less than 5% for all the parameters and for
all three cases.
From the above tables, it is observed that the maximum error is less than 5%
when the sampling frequency, Fs ≥ 5Fc. But when Fs = 3Fc, the maximum error is
5.42% for T3 code. The proposed method is better compared to the method reported
in the literature as the maximum error is less than 6% for all parameters. Selection of
sampling frequency is very critical. If Fs ≥ 5Fc, the error in estimation is less. But
the frequency resolution is poor as indicated in Eqs. 8 and 9. On the other hand, if
Fs = 3Fc, frequency resolution is good, but the estimation error is more.

5 Conclusions

Parameters of LPI radar, modulated with polytime codes, are estimated using
cyclostationary techniques. Accurate measurement of parameters is essential to
design modern receivers which help to take immediate counter action. Cyclostation-
ary method is efficient for the analysis of periodic signals like LPI signals. Genera-
tion of polytime coded signals and the complete analysis is carried out using MAT-
LAB. The estimated parameters are carrier frequency, band width and code rate. The
results are compared with the values reported in the literature. It is concluded that
the proposed method is superior as the maximum error is very much less compared
to the literature values. The simulation results indicate that the estimation error is
less than 5% when the sampling frequency, Fs ≥ 5Fc. But the frequency resolution
is poor. On the other hand, if Fs = 3FC , the estimation error is more than 5% and
the frequency resolution is good. It is also concluded that the minimum sampling

13
13
51 Page 16 of 20

Table 3  Comparison between true and estimated parameters for ­T3 code
S.No FS (GHz) FC(GHz) BW(MHz) RC(MHz)

True value Estimated %Error True value Estimated %Error True value Estimated %Error
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)

1 7 ­Fc 2 1.991 0.45 1750 1835 4.63 62 62 0


2 7 ­Fc 1 0.995 0.5 1750 1837 4.73 62 62 0
4 5 ­Fc 2 1.995 0.25 1750 1756 0.34 62 63.5 2.36
5 5 ­Fc 1 0.9970 0.3 1750 1750 0 62 62 0
7 3 ­Fc 3 2.89 3.6 1750 1760 0.57 62 61.2 1.29
8 3 ­Fc 1 0.9985 0.15 1750 1845 5.42 62 61 1.61
Sensing and Imaging (2020) 21:51
Sensing and Imaging (2020) 21:51 Page 17 of 20 51

Fig. 7  Contour plot of T4(n) code for ­Fs = 7Fc with ­Fc = 2GHz

frequency should be at least three times the carrier frequency to estimate the param-
eters within the acceptable error of 10%.

13
13
51 Page 18 of 20

Table 4  Comparison between true and estimated parameters for “T4 code”
S.No FS (GHz) FC (GHz) BW(MHz) RC(MHz)

True value Estimated %Error True value Estimated %Error True value Estimated %Error
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)

1 7 ­Fc 2 1.991 0.45 1750 1836 4.68 62 62 0


2 7 ­Fc 1 0.995 0.5 1750 1750 0 62 62 0
4 5 ­Fc 3 2.89 3.66 1750 1760 0.57 62 61.2 1.29
5 5 ­Fc 1 0.9985 0.15 1750 1835 4.63 62 61 1.61
7 3 ­Fc 3 2.89 3.66 1750 1760 0.56 62 61.2 0.81
8 3 ­Fc 1 0.9985 0.15 1750 1750 0 62 61 1.61
Sensing and Imaging (2020) 21:51
Sensing and Imaging (2020) 21:51 Page 19 of 20 51

Acknowledgements The authors express their sincere thanks to the management and principal,
VNRVJIET, Hyderabad for providing necessary facilities to carry out this work.

References
1. Fielding, J. E. (1999). Polytime coding as a means of pulse compression. IEEE Transaction on Aero-
space and Electronic Systems, 35(2), 716–721.
2. Phillip, E. P. (2009). Detecting and classifying low probability of intercept radars (2nd ed.). Nor-
wood: Artech House.
3. Kishore, T. R., & DeerghaRao, K. (2017). Automatic intrapulse modulation classification of
advanced LPI radar waveforms. IEEE Transaction on Aerospace and Electronic Systems, 53(2),
901–914.
4. Vanhoy, G., Schucker, T., & Bose, T. (2017). Classification of LPI radar signals using spectral corre-
lation and support vector machine. Analog Integrated Circuits and Signal Processing, 91, 305–313.
5. Singh, A. K., & SubbaRao, K. (2014). Digital receiver based electronic intelligent system configu-
ration for the detection and identification of intrapulse modulated radar signals. Defence Science
Journal, 64(2), 152–158.
6. Christer, P. (2009). Classification and analysis of low probability of intercept (LPI) radar signals
using image processing. Master’s Thesis, Naval Postgraduate School, Monterey.
7. Stephens, J. P. (1996). Advances in signal processing technology for electronics warfare. IEEE AES
Systems Magazine, 11, 31–38.
8. Shyamsunder, M., & Subbarao, K. (2015). Cyclostationary analysis of polytime coded signals for
LPI radars. International Journal of Research in Engineering and Technology, 04(06), 544–560.
9. Keerthi, Y., & Bhatt, T. D. (2015). LPI radar signal generation and detection. International Research
Journal of Engineering and Technology (IRJET), 02(07), 721–727.
10. Liu, Y., Xiao, P., Wu, H., & Xiao, W. (2015). LPI radar signal detection based on radial integration
of Choi-Willliams time-frequency image. IEEE Journal of Systems Engineering and Electronics,
26(5), 973–981.
11. Gulum, T. O., Erdogan, A. Y., Yildirim, T., & Ata, L. D. (2011). Parameter extraction of FMCW
modulated radar signals using Wigner–Hough transform. In 12th IEEE International Symposium on
Computational Intelligence and Informatics, Budapest, Hungary.
12. Shyamsunder, M. (2020). Classification and estimation of modulation parameters of LPI Radar sig-
nals. Ph.D Thesis, Osmania University, Hyderabad, India.
13. Taboada, F., Lima, A., Gau, J., Jarpa, P., & Pace, P. E. (2002). Intercept receiver signal processing
techniques to detect low probability of intercept radar signals. Center for Joint Services, Electronic
Warfare Naval Postgraduate School, Monterey, Canada.
14. Gardner, W. A. (1986). The spectral correlation theory of cyclostationary time-series. Signal Pro-
cessing, 11, 13–36.
15. Lima, A. F. Jr. (2002). Analysis of low probability of intercept (LPI) radar signals using cyclosta-
tionary processing. Master’s Thesis, Naval Postgraduate School, Monterey.
16. Bouillaut, L., & Sidahmed, M. (2001). Cyclostationary approach and bilinear approach: Compari-
sion, applications to early diagonosis for helicopter gear box and classification method based on
HOCS. Mechanical Systems and Signal Processing, 15(5), 923–943.
17. Skolnik, M. I. (2003). Introduction to radar systems (3rd ed.). New York: McGraw-Hill Education.

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published
maps and institutional affiliations.

13
51 Page 20 of 20 Sensing and Imaging (2020) 21:51

Affiliations

Raja Kumari Chilukuri1,2 · Hari Kishore Kakarla1 · K. Subbarao3


Hari Kishore Kakarla
kakarla.harikishore@kluniversity.in
K. Subbarao
kakarlasubbarao51@gmail.com
1
Department of Electronics and Communication Engineering, Koneru Lakshmaiah Education
Foundation, Guntur, Andhra Pradesh 522502, India
2
Department of ECE, VNRVJIET, Hyderabad 500090, India
3
Department of Electronics and Communication Engineering (Retd.), Osmania University,
Hyderabad 500007, India

13

You might also like