VOL. 2, NO.

3, April 2012

ISSN 2225-7217

ARPN Journal of Science and Technology

2011-2012. All rights reserved.
http://www.ejournalofscience.org

Multirate Filtering For Digital Signal Processing and

Its Applications
1

Suverna Sengar and 2 Partha Pratim Bhattacharya

Department of Electronics and Communication Engineering

Faculty of Engineering and Technology
Mody Institute of Technology & Science (Deemed University)
Lakshmangarh, Dist. Sikar, Rajasthan,
Pin 332311, India
1
suvernasengarmrt@gmail.com, 2 hereispartha@gmail.com

ABSTRACT
Multirate Filtering techniques are used when conventional method becomes extremely costly and this technique is widely
used in both sampling rate conversion system and in constructing filters with equal input and output rates. The basic
concepts and building blocks in multirate digital signal processing are discussed here which includes down-sampler, upsampler and analysis/synthesis representation. Applications of multirate digital filters in DS/CDMA code acquisition,
Kalman filtering for optimal signal reconstruction from noisy subband system and lossy compression approach to
transmultiplexed images are also reviewed.
Keywords: Multirate filters, Decimation, Interpolation, Multistage system and Analysis/Synthesis filter.

1.

INTRODUCTION

A multirate filter is a digital filter that changes

the sampling rate of the input signal into another desired
one. These filters are of essential importance in
communications, image processing, digital audio, and
multimedia. Unlike the singlerate system, the sample
spacing in the multirate system can vary from point to
point [1]-[2]. This often result in more efficient
processing of signals because the sampling rates at
various internal points can be kept as small as possible,
but this also results in the introduction of a new type of
error, i.e., aliasing.

1.1
TIME-DOMAIN
REPRESENTATION
OF
DOWNSAMPLING AND UP-SAMPLING
Two discrete signals with different sampling
rates can be used to convey the same information. For
example, a band limited continuous signal xc(t) might be
represented by two different discrete signals {x[n]} and
{y[n]} obtained by the uniform sampling of the original
signal xc(t) with two different sampling frequencies Ft
and Ft.
x[n]=xc(nT) and y[n]=xc(nT')

(1)

where T= 1/Ft and T=1/ Ft are the corresponding

sampling intervals. When the sampling frequencies Ft
and Ft are chosen in such a way that each of them
exceeds at least two times the highest frequency in the
spectrum of xc(t), the original signal xc(t) can be

reconstructed from either {x[n]} or {y[n]}. Hence, the

two signals operating at two different sampling rates are
carrying the same information. By using the discretetime operations, signal {x[n]} can be converted to
{y[n]}, or vice versa, with minimal signal distortions.
The two basic operations in sampling rate
alteration process are the down-sampler and up-sampl er.
These two operators can perform the sampling rate
alteration: a down-sampler used for decreasing the
sampling rate, and an up-sampler used for increasing the
sampling rate.

1.2

DOWN-SAMPLING OPERATION

The down-sampling operation with a downsampling factor M, where M is a positive integer, is

implemented by discharging M1 consecutive samples
and retaining every Mth sample. Applying the downsampling operation to the discrete signal {x[n]},
produces the down-sampled signal
y[m]= x[ mM]

(2).

The down-sampling can be imagined as a twostep operation. In the first step, the original signal {x[n]}
is multiplied with the sampling function {sM[n]} defined
by,

sM[n]=

(3).

228

VOL. 2, NO. 3, April 2012

ISSN 2225-7217

ARPN Journal of Science and Technology

2011-2012. All rights reserved.
http://www.ejournalofscience.org

Multiplying the sequence {x[n]} by the

sampling function {sM[n]} results in the intermediate
signal {ys[m]},
ys[m]=x[n]sM[n]=

(4).

This operation is called a discrete sampling. In

the second step, the zero valued samples in {ys[m]} are
omitted resulting in the down-sampled sequence {y[m]},
y[m]=ys[mM]=x[ mM]

(5).

The down-sampling operation is sometimes

called the signal compression, and the down-sampler is
also known as a compressor. A block diagram
representing the down-sampling operation is shown in
Figure 1 [3].

Figure 1: Block diagram representation of a

down-sampler

1.3

UP-SAMPLING OPERATION

called interpolation. Two devices, the down-sampler and

the up-sampler, are elements that change the sampling
rate of the signal. The drawback of the down-sampling is
the aliasing effect, where as the up-sampling produces
the unwanted spectra in the frequency band. Decimation
has used to avoid the effects of aliasing, which occurs
when the highest frequency in the spectrum of a downsampled signal H exceeds the value /M. In
interpolation, the L-1 images caused by inserting L-1
zeros between the samples should be removed.
1.4.1 DECIMATION
Decimation requires that aliasing should be
prevented. So before, down sampling with the factor of
M, the original signal has to be band limited to /M. This
means that the factor-of-M decimation has to be
implemented in two steps:i) Band limiting of the original signal to /M
ii) Down-sampling by the factor-of-M.

Figure 3 [4] shows the block diagram of a

decimator implemented as a cascade of the decimation
Filter H(z), also called the anti-aliasing filter, and the
factor-of-M down-sampler. The performance of a
decimator is mainly determined by the filter
characteristics.

The up-sampling by an integer factor L is

performed by inserting L-1 zeros between two
consecutive samples. Applying the up-sampling
operation to the discrete signal {x[n]}, produces the upsampled signal {y[m]} is defined as,
y[m]=

(6).

A block diagram presentation of equation-6 is

given in Figure 2 [3].

Figure 2: Converting the sampling rate with an upsampler

The up-sampling is sometimes called the
sequence expansion, and the term expander is sometimes
used for the device.

1.4
DECIMATION
INTERPOLATION

AND

The process of sampling rate decrease is called

decimation, and the process of sampling rate increase is

Figure 3: Block diagram representation of

decimator
1.4.2

INTERPOLATION

Interpolation requires the removal of the

images. This means that the factor-of-L interpolation has
to be implemented in two steps:
i) Up-sampling of the original signal by inserting
L-1 zero-valued samples between two
consecutive samples.
ii) Removal of the L-1 images from the spectrum
of the up-sampled signal.
Figure 4 [4] shows the block diagram of an
interpolator implemented as a cascade of a factor-of-L
up-sampler and a lowpass filter, frequently called the
anti-imaging filter. The cut-off frequency of the filter is
/L. The anti-imaging (interpolation [5]) filter H (z) is
used to remove images from the spectrum of the upsampled signal. Removal of images from the spectrum of
the signal causes the interpolation of the sample values
in time domain. The zero-valued samples in the upsampled signal {xu[m]} are filled with the interpolated
values.

229

VOL. 2, NO. 3, April 2012

ISSN 2225-7217

ARPN Journal of Science and Technology

2011-2012. All rights reserved.
http://www.ejournalofscience.org

2.1.1
SCHEME USED FOR MULTIRATE
FILTER SYSTEM MODEL

Figure 4: Block diagram representation of an

interpolator

2.
APPLICATIONS OF MULTIRATE
FILTERS
A multirate filter is used to change the sampling
rate of the input signal into another desired one and
multirate filtering [6] can be used in various fields, like it
is used in multirate adaptive filtering for DS/CDMA
code acquisition, Multirate kalman filtering for optimal
signal reconstruction from noisy subband system,
multirate filters use in lossy compression approach to
transmultiplexed image and many more. Different
applications of multirate filters are described in the
subsequent sections.
2.1
MULTIRATE ADAPTIVE FILTERING
FOR DS/CDMA CODE ACQUISITION
In direct sequence code division multiple access
(DS/CDMA) system [7], to recover the transmitted
information, the received signal should first despreaded
using a locally generated pseudo noise (PN) code
sequence. The code acquisition performs initial code
timing alignment between the locally generated and
received signals. Code acquisition method is usually
conducted using a correlator to serially search the code
phase, which is related to the channel propagation delay.
This approach performs well for the adaptive filtering
approaches [8] have high acquisition-based capacity and
drawback of these schemes is high computational
complexity. Due to down-sampling operations, the
computational complexity of the adaptive filters can be
effectively reduced.
Code synchronization is a very essential and
important part of any spread spectrum (SS) system in
order to remove the spreading effect induced by the
transmitter and to exploit the processing gain of the
spread signal. The receiver must be able to estimate the
delay offset between the spreading code in the received
signal and the locally generated replica of the code
before data demodulation
is
started.
Code
synchronization is usually developed over two steps,
namely acquisition and tracking [9]-[10].In this
application of multirate processing, not only the
computational complexity, but also the mean acquisition
time can be effectively reduced. Here the basic idea of
the scheme used in multirate adaptive filtering for
DS/CDMA code acquisition is discussed.

There are K active users and each user is

specified with a pseudo-noise (PN) sequence with length
L, where L is the processing gain. Consider an additive
white Gaussian noise (AWGN) channel. The user-k
transmitted signal is given by
Xk(n)=

(7)

where dk(j) {1, -1} is the j-th BPSK(binary phase

shifting keying) symbol of user- k, Ck(l)
is the
l-th chip of user-ks PN sequence, and p(n) is the
rectangular pulse having a unit amplitude and a chipduration, Tc . Now, assume that user-1 is the desired
user. The received signal can be written as
r(n) =
=

(8)

where w(n) is AWGN (additive white Gaussian noise)

and i(n) is the summation of interference and noise. Let
the mean of i(n) be zero and the variance be . Assume
that , k = 1, ... , K are integer multiples of Tc, and
uniformly distributed over {0, 1. . . L - l}, The carrier
synchronization is established before code acquisition
and no data are transmitted during acquisition (i.e. d1(j)
= 1). The main objective is to estimate the delay of the
desired user,
from the composite received signal,
{r(n)}. Here proposed to use an adaptive filter (AF) to
estimate the delay of the desired user from the tapweight vector of the filter. The block diagram of the AF
based DS/SS code acquisition system is shown in Figure
5 [8].

Figure 5: Block diagram for the adaptive acquisition

system

230

VOL. 2, NO. 3, April 2012

ISSN 2225-7217

ARPN Journal of Science and Technology

2011-2012. All rights reserved.
http://www.ejournalofscience.org

THE SCHEME
Figure 6 [11] shows the Structure of the
proposed scheme, which consists of two adaptive filters.
The input of the first filter (wc(.)) is the down-sampled
locally generated PN sequence; the down-sampling
factor is D. The reference signal is the down-sampled
received signal. Here, the lowpass filter is an averaging
where is an
filter with length D. Let
integer,
and
is a fractional
delay, -D/2 <
. To determine the fractional
delay, then need the second filter (wf (.)). The filter
appearing in front of wf (.) is named the delay tuning
filter (DTF); its output is just a delay version of the
input. If copy the delay identified by the first filter to the
DTF, the second filter can identify the fractional delay.
Feedback the peak position of wf(.) to the filter in front
of wc(.), which is namely the phase tuning filter (PTF).
The fractional delay will vanish. The convergent
weights of the first filter only have a non-zero weight. In
noisy environment, this will greatly reduce acquisition
error. The cross adaptation of these two filters will
effectively determine the code phase.

where N is the number of chips for tap-weight

converging. Then consider the probability of acquisition
error. An acquisition error may occur due to the first
filter, the second filter, or both. The probability of
acquisition error is then
(10)

where
and
are the correct taps of the first and
the second filters, respectively;
and
denote the
probabilities of correct acquisition of the first and the
second filters, respectively. Finally, consider the mean
acquisition time. The propose scheme gives an estimate
after N chips elapses. Figure 7 [11] shows the state
transition diagram, whose transfer function is

where TP is the penalty time, z is a delay operator and P e

is the probability of acquisition error.

Figure 7: The state transition diagram of the

proposed acquisition system
2.1.3
Figure 6: The Proposed Scheme
2.1.2

PERFORMANCE ANALYSIS

Here, first consider the computational

complexity. Now, only consider the required
multiplications per chip duration. The second filter has
2D-1 taps and each tap needs two multiplications for
LMS weight-updating. The first filter has Mp taps and its
tap-weights updating rate is D times slower than the
second filter. Thus, its computational complexity per
chip duration, Cp, is given by

EXISTING SIMULATION RESULTS

The simulation results are demonstrating the

effectiveness of the proposed scheme. Here, K = 20
(SINR = -13dB), L = 128, D = 4, Mp = 32, and Tp =
(chips) are considered. Both filters use a same step size
p. For a 32-tap conventional adaptive acquisition
method, it needs 64 multiplications per chip, but the
proposed scheme only requires 30 multiplications per
chip. Figure 8 shows the theoretical and simulated
acquisition error probabilities Pe in various step sizes.

231

VOL. 2, NO. 3, April 2012

ISSN 2225-7217

ARPN Journal of Science and Technology

2011-2012. All rights reserved.
http://www.ejournalofscience.org

2.2
MULTIRATE
KALMAN
FILTERING FOR OPTIMAL SIGNAL
RECONSTRUCTION
FROM
NOISY
SUBBAND SYSTEM

Figure 8: Theoretical and simulated probabilities of

acquisition error P, versus step sizes p for the proposed
scheme
The probability Pe was evaluated after N chips
using
samples. For P = 2
samples were
observed. Now, N is selected to be four times of the time
. These results are
constant of the first filter
substituted into equation (13) for theoretical and
simulated mean acquisition time calculation. The results
are shown in Figure 9.
As shown in Figure 8, as step size increases the
probability of error also increases and from Figure 9 it
can be seen that as step size increases the mean
acquisition time decreases. So, not only the
computational complexity is low, but also the mean
acquisition time is low.

The multirate signal processing is to decompose

the original signal into complementary frequency bands
and then process them separately in each subband. In
this application use decimation and interpolation filters
or analysis/synthesis filter banks that allow perfect
reconstruction [12]. The perfect reconstruction (PR)
filter bank systems are based on the assumption that the
subband components are free of noise. While, in
practical systems, the subband components are always
contaminated by noises due to the effect of quantization
[13], round-off [14] and other distortion, therefore, the
perfect reconstruction is no longer possible. For
improving the applicability of filter bank systems use the
multirate kalman synthesis filter [15] in place of
conventional synthesis filters to achieve optimal
reconstruction of the input signal in noisy filter bank
systems Figure 11 [16].
The input signal embedded in the state vector,
the multichannel representation of subband signals is
combined with the statistical model of input signal to
derive the multirate state-space model for the filter bank
systems and the subband noises are assumed to be
additive ones. The multirate Kalman filter can be
constructed to provide the minimum variance estimation
of the input signal based on observations of noisy
subband components. Here going to discuss first the
basics idea of multirate kalman synthesis filtering and to
explain the state-space model and also mention the
multirate state-space model for 2-D kalman filtering.

Figure10: Noisy M-band filter bank system equipped

with a multirate Kalman synthesis filter
Figure 9: Theoretical and simulated mean acquisition
time versus step sizes of the proposed scheme

2.2.1
FILTER
BANK
PROBLEM FORMULATION
a.

SYSTEMS

AND

FILTER BANK SYSTEMS

Let {Hk (Z), Gk(Z): k = 0,1,... , M - 1} denotes

the M band filter bank systems. The bank of filters {Hk
(Z)} constitutes the analysis filters. Each filter output is

232

VOL. 2, NO. 3, April 2012

ISSN 2225-7217

ARPN Journal of Science and Technology

2011-2012. All rights reserved.
http://www.ejournalofscience.org

down-sampled and transmitted to the receiver. Where

they are up-sampled and fed into the bank of synthesis
filters {Gk (Z)} for signal reconstruction Figure 11 [16].

By substituting k = k-, we can rewrite equation (14) as

a causal form
y(n)=
x(n)=
then

(17).
Now,

can

write

the

state

b.
MULTICHANNEL
OF SUBBAND SIGNALS

REPRESENTATION

Let { hi , gi ; i=0,1} denote the impulse

response of 2-band analysis/synthesis filter bank, the
equivalent multichannel representation of subband
.
signals
(14)

y(n) =
=

(15)

H(k) =

The
above are the so-called
polyphase components of
respectively.
If additive noisy corruption are included in the subband
components, the received subband signal r(n) can be
expressed as follows:
r(n) = y(n)+ v(n)

(16)
in (16) is the additive-noise

v(n) =
disturbance vector.
2.2.2
MULTIRATE
FOR 1-D SIGNAL
a.

MODEL

THE BASIC SIGNAL MODEL

Let p =

that

STATE-SPACE

, where

are chosen such

.

x(-1)=x-1
+ v(n)

X(n+1)=A x(n)+B
r(n)=C x(n)+D
b.

(20).

THE STATISTICAL MODEL

The state space model describing the statistical

characteristic of the input signal, which has the form
f(k) = H z(k)
(21)

, y(n) =

and H(k) is the multichannel impulse response matrix of

the form

(19).

Considering the effect of noisy disturbance

vector v(n), now formulate the problem as the form of
state-space model

z(k+1)=
where f(n) =

as

Here further denote

, the subband output is

y(n) = C x(n)+D
Figure 11: M-band filter bank system with clean
subband paths

(18).

X(n+1)=A x(n)+B

and

vector

where z(k) and u(k) are the state vector and driving
source, respectively. Let z(n) = z[2(n+)], the
equivalent block generation model is given below
z(n+1)=
f(n+)=
f[2(n+)]= H z(n)
(22)
c.
THE
MODEL
below
w(n+1)

MULTIRATE

STATE-SPACE

The multirate state space model will be given

=
r(n)=

(23)

where w(n) =
, is the state vector of the
system model. Similarly, the multirate state-space model
for 2-D signal [15] can also be written.
2.2.3 NUMERICAL RESULTS
Simulation is carried out to show the feasibility
and effectiveness of the proposed 2-D Kalman filtering
for optimal 2-D signal reconstruction from noisy

233

VOL. 2, NO. 3, April 2012

ISSN 2225-7217

ARPN Journal of Science and Technology

2011-2012. All rights reserved.
http://www.ejournalofscience.org

subband systems. Here compared the performance of the

proposed 2-D multirate Kalman filter versus the
conventional perfect reconstruction (PR) filters for 2-D
signal reconstruction under different noise levels. The
simulation was implemented with Matlab. Now adopted
a set of 2-band quadrature mirror filter (QMF) PR filter
bank {h (n)}of length 8 in the simulation [17]. The 2x2
subband decomposition of test image is implemented
through two 1-D separate analysis filter bank
{
i=0,1. with

Two quantitative measures, SNRim,n and SNRr ,

are input noise level and reconstruction performance,
respectively. The test image Hillside is of size 160 160
8 (Figure 12 (a) [15]). The image is converted to zero
mean prior to 2 x 2 subband decomposition and its
autocorrelation function. White additive noise at
different SNR levels was added to all the four subband
images. Figure 13 [15] demonstrates the reconstruction
performance comparison with both the proposed
multirate Kalman filtering and the conventional PR
synthesis filters under different SNR level. It is observed
that the improvement in reconstruction SNR with the
proposed 2-D multirate Kalman filtering is considerable.

Figure 12: (a) Original image Hillside of size 160

160;
(b) 2 2 subband decomposition of Hillside;
(c) noise-corrupted subband image at SNRi = 0
dB;
(d) reconstructed image with conventional PR
filter banks at SNRi = 0 dB;
(e) reconstructed image with 2-D Kalman
filtering at SNRi = 0 dB;
(f) Reconstructed image with conventional PR
filter banks at SNRi = 8 dB;
(g) reconstructed image with 2-D Kalman
filtering at SNRi = 8 dB

Figure 13: Reconstruction SNR versus additive noise

SNR level
Figure 12(e) [15] show that, even in the
extremely low SNR, case, the main structure of the
original image is still distinguishable on the
reconstructed image with 2-D multirate Kalman filtering.

2.3
LOSSY
COMPRESSION
APPROACH TO TRANSMULTIPLEXED
IMAGES
Multimedia content is more and more popular
in many different types of telecommunications. Thats
why use a new and efficient method for sending several
images through a single transmission line is needed. The
transmultiplexer is a structure that combines up-sampled
and filtered signals for the transmission over a single
transmission line. Transmultiplexing [18] is easy to
apply because it needs only simple digital processing:
up-sampling, filtering and summing. All these operations
used in transmultiplexer are linear and time-invariant.
Figure 14 [20] shows the structure of 4-channel (M=4)
transmultiplexer.

Figure 14: Scheme of 4-channel image transmultiplexer

The main problem in transmultiplexers is
preventing image distortion caused by the change of
amplitude and phase as well as image leakage from one
channel into another. The main motivation of this
application, lossy compression is to achieve a perfect
image reconstruction in the receiver by using appropriate
filters [19].
In compression algorithm the number of bits
needed to represent the signal (image) or its spectrum is

234

VOL. 2, NO. 3, April 2012

ISSN 2225-7217

ARPN Journal of Science and Technology

2011-2012. All rights reserved.
http://www.ejournalofscience.org

minimized. The fundamental concept of compression is

to split up the frequency band of a signal (image) and the
quantized each subband using a bit rate accurately
matched to compromise between the two opposite
criteria: minimize distortions and maximize the
compression rate.

highpass filters which gives useful transformations and

allow recovering the transmitted images.

Over the past several years, the wavelet

methods have gained widespread acceptance in signal
processing and in the signal compression. Multirate
processing is related to signal transformation using
wavelets. Wavelet packets are a way to analyze a signal
using base functions which are well localized both in
time and in frequency. The properties of wavelet
transform make it useful in compression. The main
advantage of the wavelet transform is used in lossy
compression which reduces the signal (image) spectrum
by eliminating redundant
information.
Lossy
compression is generally used where a loss of a certain
amount of information will not be detected by the users.
2.3.1

IMAGE TRANSMULTIPLEXING

Figure 14 [20] shows the structure of the fourchannel (M=4) image transmultiplexer. The input images
are up-sampled and filtered vertically and summed to
obtain two combined images. These combined images
are then up-sampled and filtered horizontally and
summed to obtain the final version of combined image
[19]. In presented system the combined image consists
of four times more pixels than each input image. At the
receiver end, the signal is relayed first two channels of
the detransmultiplexing, where the signals are filtered
and down-sampled horizontally. Then these signals are
relayed to four channels where images are filtered and
down-sampled vertically to recover the original image.
2.3.2COMPRESSION
Wavelet packet algorithm generates a set of
orthogonal sub-images that are derived from a single
combined image. The wavelet spectra are produced by
cascading filtering and down-sampling operations in a
tree-structure. Wavelet packets were introduced for
splitting images into its frequency components so that to
compress the image by non-uniform quantization.
The 2-D wavelet packet transform (WPT) can
be viewed as a decomposition system in Figure 15 [20]
for three levels. The basis data are the coefficients of
wavelet series of the original image. The next level
results of one step of the 2-D wavelet discrete transform
(WDT). Subsequent levels are constructed by recursively
applying the 2-D wavelet transform to the low and high
frequency sub-bands of the previous wavelet transform.
The higher level component has two times narrower
frequency bands when compare with the subsequent
lower frequency components. Multirate processing
involves the application of filtering and down-sampling.
The main objective is the design of lowpass and

Figure 15: 2-D wavelet packet transforms structure

2.3.3 EXAMPLE
Four test images (boats, F-16, Lena and
512 pixels resolution in 256
baboon) with 512
grayscale levels were selected for the analysis. The
1024 pixel resolution. Its
combined image has 1024
luminance values are integer numbers from -1088 to
1884. The combining filters
and the separation
filters
were designed by algorithm presented in [21].
Linearity property of the transmultiplexer
system enables to split the combined image into its
frequency
components. The periodicity of spectrum of the
combined image is shown in Figure 16 [20]. The
sampling densities in the frequency domain for all levels
are the same but the amounts of samples are different.
The spectrum of the up-sampled signal consists of the
original spectrum and its components.
The three level 2-D wavelet packet
decomposition was provided by the discrete Meyer
wavelets. The absolute values of coefficients of wavelet
packet decomposition are shown in Figure 17 [20]. The
main part of energy (96.6%) of the signal is localized in
last four sub-bands of the network AAA (2.5%), HAA
(13.3%), VAA (10.4%) and DAA (70.4%). The
compression with factor 16 was obtained by omitting
information from other bands. The reconstructed output
images are shown in Figure 18 [20]. The large part of the
signal energy was sent and the compression was not very

235

VOL. 2, NO. 3, April 2012

ISSN 2225-7217

ARPN Journal of Science and Technology

2011-2012. All rights reserved.
http://www.ejournalofscience.org

high, some distortions are visible. In the presented

example data from bands with the energy higher than
0.2% of the whole energy was transmitted.

Another possible solution is the preliminary

compression of the input images, e.g. using JPEG
algorithm [22], and Figure of JPEG for lossy
compression in transmultipexer system [20] and
applying the 1-D transmultiplexer after converting of the
bit streams into integer 1-D signal. The compression
with factor 16 using the JPEG algorithm on the each of
input images allows to gain the better quality of images
than in case of the compression of the combined image.

Figure 18: Comparison of input and output images

3. CONCLUSION

Figure 16: Amplitude [dB] of spectrum of

combined image

Figure 17: 2-D wavelet packet transform coefficients

The fundamentals of multirate filters and its

applications in various fields like multirate adaptive
filtering for DS/CDMA code acquisition, Kalman
filtering for optimal signal reconstruction from noisy
sub-band system and lossy compression approach to
transmultiplexed images are discussed in this paper. In
DS/CDMA code acquisition, it is proposed that a low
complexity is achieved using fast acquisition adaptive
filtering using the multirate signal processing technique.
Unlike the conventional methods, the proposed scheme
does not divide the uncertainty region into cells and
search the code phase cell by cell. Instead, it uses a twostep procedure for searching. Firstly, it searches the code
phase in the whole region with a lower resolution.
Secondly, it searches the code phase in a small region
with a high resolution. Due to this multirate processing
and searching strategy the computational complexity and
mean acquisition time can be effectively reduced.
Multirate Kalman filtering combines the multichannel
representation of sub-band signal with the statistical
model of input signal to derive the multirate state-space
model for noisy filter bank systems. The signal
reconstruction can be formulated as the optimal state
estimation with multirate Kalman filtering. Lastly lossy
compression approach to transmultiplexed images uses
the transmultiplexer for sending the several images over
a single transmission line. The transmultiplexer is a
structure that combines suitably up-sampled and filtered
signals for the transmission by a single channel that uses
different algorithms to reduce the high frequency
components which causes major error in output images
of the transmultiplexer.

236

VOL. 2, NO. 3, April 2012

ISSN 2225-7217

ARPN Journal of Science and Technology

2011-2012. All rights reserved.

REFERENCES

http://www.ejournalofscience.org

[1]

R.W.Schafer and L.R.Rabiner, A digital signal

processing
approach
to
interpolation,
Proc,IEEE,vol.61, pp. 692-702, June 1973.

[2]

G.Oetken, T.W.Parks,and H.W.Schussler, New

results in the design of digital interpolators,
IEEE Trans. Acoust. Speech signal Proc.,vol.23,
pp. 301-309,June 1975.

[3]

Ljiljana Mili, University of Belgrade, Serbia,

Multirate Filtering For Digital Signal
Processing: MATLAB Applications, 2009.

[4]

Ljiljana Mili and Tapio Saram

ki and Robert
Bregovic, Multirate Filters: An Overview.

[5]

Nkwachukwu Chukwuchekwa, Interpolation

Techniques in Multirate Digital Signal Processing
for
efficient
communication
system,vol.3.no.2.March, 2011.

[6]

L. D., Milic and M.D. Lutovac, "Efficient

multirate filtering," in Gordana Jovanovi &Dolecek, (ed.), Multirate Systems: Design &
Applications, Hershey, PA: Idea Group
Publishing, pp. 105-142, 2002.

[7]

Seokho Yoon, Lickho Song, So Ryoung Park,

Sun Youg Kim and Jooshik Lee, A Code
Acquisition Scheme for DS/CDMA systems,
Hallym University, 1998.

[8]

Mohamed G. El-Tarhuni and Asrar U. Sheikh,

"An Adaptive Filtering PN Code Acquisition
Scheme with Improved Acquisition Based
Capacity in
DS/CDMA,"
IEEE
Ninth
International Symposium on PIMRC, vol.3, pp.
1486-1490, 1998.

[9]

M.Simon
et
at,
Spread
Spectrum
Communications, vol.3, Computer Science
Press, Rockville, Maryland, 1985.

[10]

R.Dixon, Spread Spectrum System, John Wilay

and Sons, New York, 1984.

[11]

Hua-Lung Yang, Wen-Rong Wu, Multirate

Adaptive Filtering for DS/CDMA Code
Acquisition, 2003.

[12]

P.P. Vaidyanathan, Multirate Digital Filters,

Filter Banks, Polyphase Networks, and
Applications: A Tutorial, Proc. of The IEEE,
vo1.78, pp.56-93, Jan. 1990.

[13]

R. A. Haddad and N. Uzun, Modeling, analysis

and compensation of quantization effects in Mband subband codes, in Proc. ICASSP 93, pp.
III-173-176.

[14]

P. P. Vaidyanathan, On coefficient-quantization
and computational roundoff effects in lossless
multirate filter banks, IEEE Trans. Signal
Processing, vol. 39, no. 4, pp. 1006-1008, Apr.
1991.

[15]

J.Q.Ni, K.L. Ho, K.W. Tse, and M.H.Shen,

Multirate Kalman Filtering Approach for
Optimal Two-Dimensional Signal Reconstruction
from Noisy Subband Systems, IEEE, vol.1, 06
August 2002.

[16]

B. Chen, Optimal Signal Reconstruction in

Noisy Filter Bank Systems: Multirate Kalman
Synthesis Filtering Approach, IEEE Trans.
Signal Processing, vo1.43, pp.2496-2504, Nov.,
1995.

[17]

A.Delopoulos, Optimal Filter Banks for Signal

Reconstruction
from
Noisy
Subband
Components, IEEE Trans. signal processing, vol.
44, pp212-224, Feb., 1996.

[18]

A. N. Akansu, P. Duhamel, X. Lin, & M.

Courville, Orthogonal Transmultiplexers in
Communications: A Review, IEEE Trans. on
Signal Processing, vol. 46(4), pp. 979-995, 2001.

[19]

P. Sypka, B. Ziolko, and M. Ziolko, "Integer-tointeger filters in image transmultiplexers", IEEE

International Symposium on Communications,
Control and Signals, Marrakech, Morocco, March
13-15. 2006.

[20]

Przemyslaw Sypka , Mariusz Ziolko and Bartosz

Ziolko, Lossy Compression Approach to
Transmultiplexed Images, AGH University of
Science and Technology al. Mickiewicza 30, 30059 Krakow, Poland, 07-09 June 2006.

[21]

P. Sypka, M. Zio6ko, and B. Zio6ko, Lossless

JPEG-Base Compression of Transmultiplexed
Images, IEEE 14-th European Signal Processing
Conference, Florence 2006.

[22]

ISO/JEC IS 10918-1 / ITU-T Rec. T.81,

Information technology- digital compression and
coding of continuous still images, Requirements
and guideline.

237