Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
3, April 2012
ISSN 2225-7217
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
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)
1.2
DOWN-SAMPLING OPERATION
(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
ISSN 2225-7217
(4).
(5).
1.3
UP-SAMPLING OPERATION
(6).
1.4
DECIMATION
INTERPOLATION
AND
INTERPOLATION
229
ISSN 2225-7217
2.1.1
SCHEME USED FOR MULTIRATE
FILTER SYSTEM MODEL
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.
(7)
(8)
230
ISSN 2225-7217
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
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
PERFORMANCE ANALYSIS
231
ISSN 2225-7217
2.2
MULTIRATE
KALMAN
FILTERING FOR OPTIMAL SIGNAL
RECONSTRUCTION
FROM
NOISY
SUBBAND SYSTEM
2.2.1
FILTER
BANK
PROBLEM FORMULATION
a.
SYSTEMS
AND
232
ISSN 2225-7217
(17).
Now,
can
write
the
state
b.
MULTICHANNEL
OF SUBBAND SIGNALS
REPRESENTATION
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
that
STATE-SPACE
, where
x(-1)=x-1
+ v(n)
X(n+1)=A x(n)+B
r(n)=C x(n)+D
b.
(20).
, y(n) =
(19).
z(k+1)=
where f(n) =
as
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
(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
ISSN 2225-7217
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.
234
ISSN 2225-7217
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
235
ISSN 2225-7217
3. CONCLUSION
236
ISSN 2225-7217
REFERENCES
http://www.ejournalofscience.org
[1]
[2]
[3]
[4]
[5]
[6]
[7]
[8]
[9]
M.Simon
et
at,
Spread
Spectrum
Communications, vol.3, Computer Science
Press, Rockville, Maryland, 1985.
[10]
[11]
[12]
[13]
[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]
[16]
[17]
[18]
[19]
[20]
[21]
[22]
237