Upload
duongkien
View
215
Download
2
Embed Size (px)
Citation preview
Bi, J.; Rohling, H.: Complementary Binary Code Design based on MismatchedFilter. IEEE Transaction on Aerospace and Electronic Systems, Vol.48, No.2,January 2012.
© 2012 IEEE. Personal use of this material is permitted. Permission from IEEEmust be obtained for all other uses, in any current or future media, includingreprinting / republishing this material for advertising or promotional purposes,creating new collective works, for resale or redistribution to server or lists, orreuse of any copyrighted component of this work in other works.
Complementary Binary Code Design based onMismatched Filter
Complementary code (CC) pairs have the interesting
characteristic that the superposition of the two related
autocorrelation functions (ACFs) has zero sidelobes in range
direction. This is an important property for radar applications
to avoid range sidelobes. However, CC pairs exist only for
limited codeword lengths. Therefore, the general idea of CC
design is extended here by applying a mismatched filter (MMF)
procedure to the binary phase codeword pairs. There are two
objectives considered. The first objective is to design MMF
impulse responses to fulfill the zero sidelobe property. The second
objective is to find binary phase codeword pairs in cooperation
with the MMFs which have a high gain in signal-to-noise ratio
(SNR). It is shown that for all codeword lengths (even where the
classic CC pair does not exist) there exist binary phase codeword
pairs and the related MMF impulse response coefficients which
also have the zero sidelobe property. Furthermore, these codeword
pairs have high gains in SNR which are nearly the same as for the
classic matched filter technique.
I. INTRODUCTION
The mismatched filter (MMF) design procedurecan be considered as an extension of the classicmatched filter (MF) technique which has been appliedsuccessfully to periodic [1] and to aperiodic binarycodes [2, 3].A pair of aperiodic binary phase coded signals s1
and s2 is called a complementary code (CC) pair if thesuperimposed autocorrelation function (ACF) of thetwo codes has zero sidelobes [4] in range direction.Considering the ACF of an aperiodic signal meansimplicitly that the MF procedure is assumed andapplied in the receiver. The important zero sidelobeproperty of CC pairs can be described analytically inthe following equation.
A1(i) +A2(i) =
1Xn=¡1
s1(n) ¢ s1(n+ i)+1X
n=¡1s2(n) ¢ s2(n+ i)
=
½2N, i = 0
0, i 6= 0(1)
where A1,2(i) is the ACF of the signal s1,2 = fs1,2(n) j n= 0, : : : ,N ¡1g.
Manuscript received November 26, 2010; revised March 14, 2011;
released for publication July 5, 2011.
IEEE Log No. T-AES/48/2/943857.
Refereeing of this contribution was handled by C. Baker.
0018-9251/12/$26.00 c° 2012 IEEE
Equation (1) shows that any sidelobe interferences
can be avoided in stationary radar measurements and
radar applications where complementary binary phase
coded transmit signals are considered and the MF
procedure is applied in the receiver for each codeword
separately. The considered situation is demonstrated in
Fig. 1 for a CC pair example with a codeword length
of N = 10.
However the CC property of (1) is only fulfilled
for a very limited number of codeword lengths N.
Complementary binary code pairs are known for
codeword lengths N of the form:
N = 2a10b26c
where a, b, and c are nonnegative integer values [5].
In fact, for codeword lengths N · 100, only those CCpairs of length N = 1, 2, 4, 8, 10, 16, 20, 26, 32, 40,
52, 64, 80, 100 have been found [1].
To extend the theory of CCs in order to cover all
codeword lengths N, an alternative receiver structure
is considered in this paper. This means that the
application of the MF in the receiver is replaced by
an MMF design procedure for all codeword lengths N,
even where no CC pairs exist.
For an arbitrary binary phase codeword pair, two
digital MMF impulse responses, denoted by h1 and
h2, are calculated to have the same zero sidelobe
property as for classic CC pairs. This means that the
resulting cross-correlation functions (CCFs) will be
superimposed and should show zero sidelobes as
required for classic CC theory. These requirements
are described analytically by (2).
C1(i) +C2(i) =
1Xn=¡1
s1(n) ¢ h1(n+ i)+1X
n=¡1s2(n) ¢ h2(n+ i)
=
½2N, i= 0
0, i 6= 0(2)
where C1,2(i) is the CCF of the code s1,2 and the
related MMF impulse response h1,2 = fh1,2(n) j n=0, : : : ,N ¡ 1g. The MMF impulse response is requiredto have the same length N as the binary phase
code.
In contrast to the MF procedure and the resulting
ACF, the filter coefficients in an MMF procedure are
not necessarily binary and the CCF is not necessarily
symmetric. The MMF procedure offers more freedom
in the receive filter design and makes it possible to
design binary phase codeword pairs to have the same
zero sidelobe property as CC pairs (see (2)) for any
arbitrary length N.
II. MMF DESIGN PROCEDURE AND ZERO SIDELOBEPROPERTY
In this section a pair of aperiodic binary phase
coded signals of length N is considered where a
CORRESPONDENCE 1793
Fig. 1. ACF for each code of the CC pair and superposition of both for codeword length of N = 10.
classic CC pair does not exist. For an arbitrary pair of
binary phase codewords s1 and s2, the correspondingMMF impulse responses h1 and h2 are calculated
to fulfill the zero sidelobe condition described in
(2). It is required that all range sidelobes of the
superimposed CCF should be zero.
The binary phase codewords s1 and s2 as well as
the MMF impulse responses h1 and h2 are representedin a vector form of length N:
s1 = [s1(0),s1(1),s1(2), : : : ,s1(N ¡ 1)]s2 = [s2(0),s2(1),s2(2), : : : ,s2(N ¡ 1)]h1 = [h1(0),h1(1),h1(2), : : : ,h1(N ¡ 1)]h2 = [h2(0),h2(1),h2(2), : : : ,h2(N ¡ 1)]:
Equation (2) can be formulated as a system with
linear equations of dimension (2N ¡ 1)£ (2N):S ¢h= d (3)
where the vector h= [h1,h2]T contains the coefficients
of the two MMF impulse responses h1 and h2 in acascaded way. The vector
d= [0, : : : ,0,2N,0, : : : ,0]T (4)
denotes the desired filter output which describes the
zero sidelobe property.
For these linear equations the signal matrix S hasthe form of two adjoined Toeplitz matrices, as shown
in (5).
S= [S1,S2]
with S1,2 =
266666666666666666664
s1,2(N ¡ 1) 0 0 ¢ ¢ ¢ 0
s1,2(N ¡ 2) s1,2(N ¡ 1) 0 ¢ ¢ ¢ 0
s1,2(N ¡ 3) s1,2(N ¡ 2) s1,2(N ¡ 1) ¢ ¢ ¢ 0
.... . .
s1,2(0) s1,2(1) s1,2(2) ¢ ¢ ¢ s1,2(N ¡ 1)0 s1,2(0) s1,2(1) ¢ ¢ ¢ s1,2(N ¡ 2)0 0 s1,2(0) ¢ ¢ ¢ s1,2(N ¡ 3)...
. . .
0 0 0 ¢ ¢ ¢ s1,2(0)
377777777777777777775
:(5)
The maximum rank of this signal matrix S is2N ¡ 1
rankS· 2N ¡ 1: (6)
It has been validated by computer search that thereare many codeword pairs s1 and s2 for any arbitrarycodeword length N which have a signal matrix S ofmaximum rank 2N ¡ 1. For these codeword pairs,the signal matrix S has 2N ¡1 linear independentcolumns and there are 2N MMF impulse responsecoefficients. Therefore the MMF impulse responsecoefficients are linear dependent on a single parameter® 2R and are solved as follows:
h= ® ¢ a+b: (7)
The vector a is a single solution of the homogeneoussystem of linear equations (8) and b is a singlesolution of the system of linear equations (3).
S ¢ a= 0 (8)
S ¢b= d: (9)
Both vectors a and b are of dimension 2N and aredependent on the signal matrix S and the desiredMMF output d. The above equations have been solvedby using the Gauss-Jordan elimination technique.For a single codeword pair, all resulting MMF
impulse responses h of (7) have the zero sidelobeproperty for each value of parameter ®. This kind offreedom is used for an additional optimization stepwhere the receive signals are superimposed by someadditive noise. For a given codeword pair the
1794 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 48, NO. 2 APRIL 2012
Fig. 2. MMF coefficients for codeword length N = 14.
parameter ® is used to maximize the gain in
signal-to-noise ratio (SNR) for the MMF procedure,
denoted by SNRMMF in the following. It is calculated
in (10), for a binary phase coded input signal in the
presence of independent and identically distributed
(IID) white Gaussian noise.
SNRMMF =(2N)2PN¡1
n=0 h21(n) +
PN¡1n=0 h
22(n)
=(2N)2P2N¡1
j=0 (®a(j) + b(j))2
(10)
where a(j) and b(j) are the jth element of vectors aand b.Maximizing SNRMMF is equivalent to minimizing
the MMF energy E:
E =
N¡1Xn=0
h21(n) +
N¡1Xn=0
h22(n) =
2N¡1Xj=0
(®a(j)+ b(j))2: (11)
The optimal real valued ® is obtained, if the filter
energy E is minimized.
@E
@®= 0
! ®=¡P2N¡1j=0 a(j)b(j)P2N¡1j=0 a2(j)
:
(12)
The MMF impulse responses h1 and h2 for a
binary phase codeword pair of length N are finally
calculated, by inserting (12) into (7).
Example: As an example, an arbitrary binary
phase codeword pair of length N = 14 is considered
and the corresponding MMF impulse responses
have been calculated in accordance with the analysis
above (see Fig. 2). The superposition of the CCFs
fulfills the zero sidelobes property as shown in
Fig. 3.
CORRESPONDENCE 1795
TABLE I
CC Pairs for MMF Design with Maximum Normalized Gain G in SNR
Code Length G (%) Code Pair Example
4 100
5 73.64 [1 ¡1 1 1 1][1 1 ¡1 1 ¡1]
6 82.04 [1 ¡1 ¡1 1 1 1][1 1 1 ¡1 1 ¡1]
7 82.97 [1 ¡1 ¡1 ¡1 1 ¡1 1][1 ¡1 1 1 1 1 ¡1]
8 100
9 84.72 [1 ¡1 ¡1 1 ¡1 1 ¡1 ¡1 ¡1][1 1 1 ¡1 1 ¡1 ¡1 ¡1 1]
10 100
11 86.34 [1 ¡1 ¡1 1 1 1 ¡1 1 ¡1 ¡1 ¡1][1 1 1 ¡1 1 1 1 1 ¡1 ¡1 1]
12 95.15 [1 ¡1 1 1 ¡1 ¡1 ¡1 ¡1 1 ¡1 ¡1 ¡1][1 1 1 ¡1 1 ¡1 ¡1 ¡1 1 1 ¡1 1]
13 90.66 [1 ¡1 ¡1 ¡1 ¡1 1 1 ¡1 1 1 1 ¡1 1][1 ¡1 1 1 1 1 ¡1 1 1 1 1 1 ¡1]
14 92.10 [1 ¡1 ¡1 ¡1 1 ¡1 1 1 ¡1 ¡1 ¡1 ¡1 1 ¡1][1 ¡1 1 1 1 ¡1 ¡1 1 ¡1 ¡1 ¡1 ¡1 ¡1 1]
15 91.23 [1 ¡1 1 1 1 ¡1 1 ¡1 1 ¡1 ¡1 ¡1 ¡1 1 1][1 1 ¡1 ¡1 ¡1 ¡1 1 ¡1 ¡1 1 ¡1 ¡1 ¡11 ¡1]
16 100
17 91.21 [1 ¡1 1 1 ¡1 ¡1 ¡1 1 1 ¡1 1 1 1 1 1 ¡1 1][1 ¡1 1 1 1 1 ¡1 ¡1 ¡1 1 ¡1 1 1 ¡1 ¡1 1 ¡1]
18 93.42 [1 ¡1 ¡1 1 ¡1 ¡1 ¡1 1 1 1 1 1 ¡1 1 ¡1 1 1 1][1 1 1 ¡1 1 ¡1 1 1 ¡1 ¡1 ¡1 1 1 1 ¡1 1 1 ¡1]
19 91.25 [1 ¡1 ¡1 ¡1 ¡1 ¡1 ¡1 1 1 ¡1 1 1 1 ¡1 1 ¡1 1 1 ¡1][1 ¡1 ¡1 1 ¡1 1 ¡1 ¡1 ¡1 1 1 1 ¡1 ¡1 ¡1 ¡1 ¡1 ¡1 1]
24 95.15 [1 1 ¡1 1 1 1 1 ¡1 ¡1 1 ¡1 ¡1 ¡1 ¡1 ¡1 ¡1 1 1 ¡1 1 ¡1 ¡1 ¡1 1][1 ¡1 ¡1 ¡1 1 ¡1 1 1 ¡1 ¡1 ¡1 1 ¡1 1 ¡1 1 1 ¡1 ¡1 ¡1 ¡1 1 ¡1 ¡1]
Fig. 3. CCF for each code of CC pair based on MMF and superposition of both for codeword length of N = 14.
III. RESULTS
There are two objectives considered in this
paper. The first objective is to design MMF impulse
responses for binary phase codeword pairs in order
to fulfill the zero sidelobe property in target range
direction for all codeword length N. In this case
a high flexibility is given in the waveform design
procedure by selecting codeword pairs of different
length N. The second objective is to find binary phase
codeword pairs in cooperation with the resulting MMF
impulse responses which have a high gain in SNR.
The gain in SNR for the MF technique is
calculated as follows:
SNRMF =(2N)2PN¡1
n=0 s21(n) +
PN¡1n=0 s
22(n)
=(2N)2
N +N
= 2N: (13)
1796 IEEE TRANSACTIONS ON AEROSPACE AND ELECTRONIC SYSTEMS VOL. 48, NO. 2 APRIL 2012
Fig. 4. Normalized gain in SNR over codeword length.
The gain in SNR for the MF procedure cannot
be outperformed by the MMF scheme. However,
an optimum codeword pair and the resulting MMF
coefficients show nearly the same gain as the MF
technique. The normalized gain in SNR between
MMF and MF is calculated as follows:
G =SNRMMFSNRMF
· 1: (14)
The normalized gain is used to rank all possible
codeword pairs for a given length N.
All codeword pairs s1 and s2 are considered in
this paper which have a signal matrix S of maximum
rank 2N ¡ 1. For these codeword pairs the MMFimpulse responses are calculated according to (7). It
is furthermore required that in combination with the
MMF impulse responses these codeword pairs have
high normalized gains G in SNR.
Therefore, a computer search has been carried out
by investigating all possible binary phase codes for
each codeword length N. Some codeword pairs are
given in Table I. The resulting normalized gain G in
SNR is approximately 90% (see Fig. 4).
IV. CONCLUSION
It has been shown in this paper and analyzed by
computer search that for each arbitrary codeword
length N, there exist binary phase codeword pairs and
the related MMF impulse responses which have the
zero sidelobe property in range direction. In addition
the resulting gain in SNR for the MMF procedure is
nearly the same as the MF technique.
JINGYING BI
HERMANN ROHLING
Department of Telecommunications
Hamburg University of Technology
Eissendorfer Strasse 40
21073 Hamburg
Germany
E-mail: ([email protected])
REFERENCES
[1] Rohling, H. and Plagge, W.
Mismatched-filter design for periodic binary phased
signals.
IEEE Transactions on Aerospace and Electronic Systems,
25, 6 (1989), 890—897.
[2] Ackroyd, M. and Ghani, F.
Optimum mismatched filters for sidelobe suppression.
IEEE Transactions on Aerospace and Electronic Systems,
AES-9, 2 (1973), 214—218.
[3] Seidler, P.
Mismatched filtering for coded aperture imaging with
minimum sidelobes.
Electronics Letters, 17, 2 (1981), 96—97.
[4] Golay, M.
Complementary series.
IRE Transactions on Information Theory, IT-7, 2 (1961),
82—87.
[5] Turyn, R.
Hadamard matrices, Baumert-Hall units, four-symbol
sequences, pulse compression, and surface wave
encodings.
Journal of Combinatorial Theory, Series A, 16, 3 (1974),
313—333.
CORRESPONDENCE 1797