You are on page 1of 9

724

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 48, NO. 3, MAY 1999

Effects of Cell Correlations in a Matched-Filter


PN Code Acquisition for Direct-Sequence
Spread-Spectrum Systems
Wern-Ho Sheen, Member, IEEE, Jiun-Kai Tzeng, and Ching-Kae Tzou

Abstract Rapid pseudonoise (PN) code acquisition with


matched-filter correlators has been very popular in directsequence (DS) spread-spectrum systems. Conventionally, the
analysis of this acquisition method is based on the assumption
that the detections among cells are independent. However, there
may be strong correlations among cell detections for the case
that the cell size is less than a chip duration. In this paper, the
mean acquisition time performance of the acquisition method
is analyzed with the cell correlations being taken into account.
Numerical results show that depending on the threshold value
and other system parameters, the effect of cell correlations may
be over 20% of the mean acquisition time for signal-to-noise
ratios (SNRs) of practical interest. The analytical results are
substantiated by computer simulations.
Index Terms DS spread spectrum, matched filter, PN code
acquisition.

I. INTRODUCTION

N DIRECT-SEQUENCE (DS) spread-spectrum systems,


pseudonoise (PN) code acquisition aligns the received and
locally generated PN codes to within a small range of timing
error before a code tracking loop can finetune and maintain a
precise timing synchronism of the two PN codes. (The range
of permissible timing error is often determined by the pull-in
range of the code tracking loop.) Rapid PN code acquisition
has been the most challenging work faced by a DS spreadspectrum receiver [1][3].
PN code acquisition for DS spread-spectrum systems has
been drawing much research attention in the past [1][20].
Acquisition methods such as multiple-dwell serial search acquisition [1][6], serial search acquisition with sequential
detection [1][3], [7][9], serial (or parallel) search acquisition
based on matched-filter (MF) correlators [10][15], and acquisition with recursion-aided sequential estimation [16], [17]
have been investigated extensively in the literature. Of these,
the method based on MF correlators has a faster acquisition
speed at the expense of a larger system complexity for a
large PN code period, as opposed to other types of acquisition
Manuscript received July 7, 1996; revised February 25, 1997. This work
was supported in part by the National Science Council, Taiwan, R.O.C., under
NSC Grant 85-2213-E-194-001.
W.-H. Sheen is with the Department of Electrical Engineering, National
Chung Cheng University, Chia-Yi 621, Taiwan, R.O.C.
J.-K. Tzeng is with the National Space Program Office, Science-Based
Industrial Park, Hsin-Chu 300, Taiwan, R.O.C.
C.-K. Tzou is with the TranSwitch Corporation, Taipei 13F-3, Taiwan,
R.O.C.
Publisher Item Identifier S 0018-9545(99)01043-9.

methods. In this study, we are concerned with the performance


analysis of the serial search PN code acquisition based on MF
correlators.
In a serial search PN code acquisition, the time uncertainty
, is divided into cells,
of the incoming code, denoted as
and the cells are searched on a cell-by-cell basis until true
code acquisition is accomplished (timing error is within the
permissible range).1 Quite often, the cell size is taken as a
fraction of a chip duration. Therefore, there are more than one
cell at which true acquisition is possible. These types of cells
cells, and the set of all
cells will be
will be denoted as
region. All other cells for which the true
denoted as the
region,
acquisition is not possible will be denoted as the
cells. For the
and the cells in the region be denoted as the
acquisition method based on MF correlators, since the decision
whether a particular cell is a correct cell (true acquisition) can
be made in a duration of cell size, the acquisition speed is much
faster than other types of acquisition methods, for example, the
method of multiple-dwell acquisition [2], [3].
Conventionally, the performance analysis of the PN code
acquisition based on MF correlators assumes that the detections between cells are independent. Then, a transfer function
and/or a time domain approach can be employed to obtain
the mean and variance and/or the probability density function
(pdf) of the acquisition time [1][3], [10], [18][20]. However,
as to be shown, the assumption of independency between cell
detections is only valid for the case when the cell size is equal
to the chip duration. For cell sizes less than one chip duration,
the detections between cells may be strongly correlated. In
this paper, the mean acquisition time performance of the
acquisition method is analyzed with the cell correlations being
taken into account. The results show that, depending on the
threshold values and other system parameters, the effect of
cell correlations may be over 20% of the mean acquisition
time for signal-to-noise ratios (SNRs) of practical interest.
Computer simulations have been used to substantiate the
analytical results.
The remainder of this paper is organized as follows.
Section II describes the serial search PN code acquisition
based on MF correlators. Section III describes a state
transition diagram with which the mean acquisition time
can be evaluated. In Section IV, the expression for the mean
acquisition time is derived, and some numerical examples
1 For

simplicity, the frequency uncertainty will not be considered.

00189545/99$10.00 1999 IEEE

SHEEN et al.: EFFECTS OF CELL CORRELATIONS IN PN CODE ACQUISITION

725

Fig. 1. A typical matched-filter detector for DS PN code acquisitions.

are shown in Section V. Finally, conclusions are given in


Section VI.
II. SYSTEM DESCRIPTION
A typical structure for DS PN code acquisition based on
MF correlators is shown in Fig. 1. The received signal is
(1)
and
are the carrier power, frequency and
where
is a maximum length sequence with
phase, respectively,
nonreturn-to-zero shaping function, is the timing information
is additive white Gaussian
needed to be acquired, and
noise (AWGN) with one-sided power spectra density (PSD) of
watts/hertz. For simplicity, the effects of data modulation
and frequency shift will not be considered. The received
signal is first downconverted to the inphase and quadrature
s, the inphase and
components. After integration by
quadrature signals are sampled and correlated, respectively,
with an MF correlator as shown in Fig. 2, where
and
denotes the chip duration. The samples
s at the instants of
are taken every
with
The MF correlators perform the correlation
between the incoming and locally generated PN codes. Each
-stage tapped delay line with a tap
MF uses an
s, where is the period of the maximum length
spacing of
is the locally
sequence (full period correlation).
generated PN sequence to be defined later. The MF correlator
is used to compare
outputs are squared, and the sum
For one-dwell systems, if
exceeds
with a threshold
then the cell being tested is considered as
the threshold
a correct cell and the synchronization process is transferred
to the code tracking. For two-dwell systems, however, the
acquisition enters the second dwell (verification mode) after
.2 As an example, an active correlator with
the excess of
is assumed for the
a second threshold comparison
verification mode, although some other schemes may also be
employed, for example, the coincidence detector [10]. If the
is exceeded again, then the cell is considered
threshold
as a correct cell. Otherwise, the cell will be considered as
an incorrect one, and a new cell will be searched and tested
[1][3]. This means that the first sample in the tap delay lines
2 The extension of the proposed method to more than two-dwell systems is
straightforward.

Fig. 2. The matched-filter correlator.

will be shifted out and the new coming one will be moved in,
and the correlation and threshold comparisons repeat. Many
strategies can be used for searching a new cell [2], [3]. Here,
only the straight line search strategy will be used as an example
for evaluating the mean acquisition time.
The case of returning false alarm will be considered which
means that the tracking loop can always detect the false
alarm after some time, called the penalty time, and return the
synchronization to the code acquisition. The penalty time will
with
be modeled as a fixed value and denoted by
This model is a bit simplified. Nevertheless, it is commonly
employed in the study of acquisition systems.
and
By using the chip synchronous model [8], i.e.,
, the inphase and quadrature outputs at the
assume that
MF correlators are given by

(2)
and

(3)
respectively, where
is the maximum length sequence,

is the integer

726

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 48, NO. 3, MAY 1999

part of , and
and
independent Gaussian noise with

are base-band mutually


and

for a large
Since
are Gaussian random variables,
uncorrelation implies independency. Therefore, the decision
variables

for
denotes the expectation value of the random variable
From (2) and (3), we have

(13)

(4)
and
(5)

III. STATE TRANSITION DIAGRAM

where
(6)
and
(7)
Note that

otherwise
(8)
In (8), we have used the property of the periodic autocorrelation of a maximum length sequence. From (4) and (5), it can
be shown that

(9)
and

(10)
otherwise
or
where
correlation coefficient

depend only on
Conventionally, the performance analysis of the rapid acquisition using MF correlators is based on the assumption of
independency between cell detections. As shown in (12), this
In the following, a new method is
is not the case for
proposed to analyze the performance of mean acquisition time
that takes the correlations between cell detections into account.

Therefore, from (8) and (10), the

(11)
is approximated by

A state transition diagram will be first devised to describe


the acquisition process. Then, the mean acquisition time will
be evaluated by using a time-domain approach. In the state
states, which are
transition diagram, there are total
to
, respectively. Each state is associated
denoted as
, except that the state
with a cell in the time uncertainty
(or ACQ) is used to denote the acquisition state. The
region contains
cells which are denoted as
to
The state
is the first
cell. As shown in (12),
depends on the detections
the detection at the cell (state)
To take this dependency
at the cells
substates,
into account, each state is divided into
with each substate defined by

(14)
,
, and
are to
where
and
when
denote the events of
respectively. In (14), is the decimal
detecting the cell
representation of the state content where the left-most bit is
the most significant bit. The state transition diagram for the
is shown in Fig. 3. As can be seen, there are
case of
cells, and for each substate in the
region, there is
five
Note that in the
a transition to the acquisition state
region, two transitions have been associated with the substate
and
pairs
This is attributed to that if
at
the first dwell, then two events could happen. One is that
is exceeded, and the system enters
the second threshold
the tracking mode. The other is that the second threshold
is not exceeded, and a new cell is searched and tested. The
acquisition process can be initiated at any state, say , with
As to be shown, it is appropriate to use the
probability of
as the starting substate.
substate
The next step is to define transition gains in the state
transition diagram such that the mean acquisition time can
be evaluated. To this end, we define the transition gains as
the transition gains for
follows. For
in the
region are defined
the substate pair
is exceeded)
by (the event that

(12)
(15)

SHEEN et al.: EFFECTS OF CELL CORRELATIONS IN PN CODE ACQUISITION

and
(16)
respectively, and for

727

TABLE I
GENERAL EXPRESSIONS FOR THE Pd ji

( ) AND
l

f ji

IN

FIG. 3

the transition gain is


(17)

In (15)(17),
is the false alarm
where
is the originating
probability for the first dwell given that
and
are the false
substate of the transition, and
alarm probability and dwell time of the second dwell. Since
the channel is stationary, the false alarm probabilities
and
are independent of the cell index For the substate
in the
region, the transition gain is
pair
defined as
(18)
for

and is defined as
(19)

respectively. In the above,


is
for
the detection probability for the first dwell given that
is the originating substate of the transition, and
is
the detection probability for the second dwell when detecting
For the
region, the detection probability is a
the cell
Recall that for each substate
function of the tested cell
region, there is a transition to the acquisition state
in the
(or ACQ). The transition gains for these transitions are
given by
(20)
Since the transition gains are not a function of time, the state
transition diagram describes a stationary Markov chain. Table I
and
for the
shows the general expressions of
In the table, and are to denote the source
case of
and destination substates of a transition, respectively. For the
and
considered active correlator, the probabilities
are given by the general expression of
(21)
is the decision variable for the verification mode.
where
for some in the substate
Note that if
, i.e., the acquisition process enters
the tracking mode for one-dwell systems or enters the verification mode for two-dwell systems when detecting the cell
the detection of the cell
is independent of the detections
for
At the very beginning of the
of the cell
acquisition process, since there are no detections preceding
as
the starting cell, it is appropriate to use the substate
the starting substate as shown in Fig. 3.
From Fig. 3 and Table I, we can see that the substates
and
can be combined as one for the case of
This
type of combination is always possible for
and will largely reduce the complexity of the state transition
diagram. For example, only 5 rather than 16 substates are
The simplified state
required for the case of
is shown in
transition diagram for the case of
for
Fig. 4, where the substates have been renamed as
with

Fig. 3. The state transition diagram with

1=1 3

= :

IV. MEAN ACQUISITION TIME


With the defined transition diagram, the expression for the
mean acquisition time can now be derived. A time domain
approach will be used for the derivation. As evident in Fig. 4,
one needs to know the probabilities and elapsed time for every
in
visit of the substates
order to calculate the exact mean acquisition time. However,
this is extremely difficult if not impossible even for
Instead, if the correlations from the
region to the
region
can be neglected, then, as to be shown, the calculation of the
is
mean acquisition time will be largely simplified. Since
usually quite large in practical applications, the effect due to
this simplification will be very small (see Section V). Once
as an example. Then, without considering
again take
to
regions, the state transition
the correlations from the
diagram Fig. 4 becomes the one shown in Fig. 5. Note that
and
have degenerated to one and two
the cells
substates, respectively. In addition, the general expressions for
and
are different from those defined
in Fig. 4. The new expressions are given in Table II.

728

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 48, NO. 3, MAY 1999

TABLE II
GENERAL EXPRESSIONS FOR THE Pd ji M

( 0 5) AND

Fig. 4. The simplified state transition diagram with

d ji

( 0 4) IN FIG. 4
M

The average dwell time from state


to
given that acquisition is reached during the
state
region. Let
be the
next pass of the
dwell time for the particular state transition sequence
Then

1=1 3

(23)

= :

The average dwell time from the starting state


to the state
For
(the
region)
(24)
where

(25)
and

Fig. 5. The simplified state transition diagram for no correlations from the
H0 to H1 region with
= :

1=13

To derive the mean acquisition time, the following definitions are useful.
The detection probability for one passes through the

region. Let
denote a sequence of state tranand the final state
and
sitions with the initial state
denote the probability of this particular sequence
of transitions. Then, the overall detection probability is
given by
(22)
denote the set of all possible sequences
where
and final state
of state transitions with the initial state

(26)

and
with
In (25) and (26),
and
denote the transition probability and dwell time for the
transitions associated with the substate pair
respectively. Recall that there may be two transitions
and that
associated with the substate pair
and
are independent of with
For
(the
region)
(27)

SHEEN et al.: EFFECTS OF CELL CORRELATIONS IN PN CODE ACQUISITION

729

For

where

(28)
is the average dwell time from

to

and

(35)

(29)

and
, where
with
the last term is due to the fact that it is possible to acquire
region when
the code phase without searching through the
Consequently, the overall mean
is given by
acquisition time

The average dwell time


to the state itself (no
from the state
region).
acquisition occurs when passing through
It can be shown that

(36)

is used in calculating

(30)
where

is given in the Appendix to


A simpler example with
explain the derivation of (33) more clearly.
The final step to complete the analysis is to obtain the
and
From
probabilities
Tables II and III, we know that
and
take the
general forms of
(37)
or

(31)
is the average dwell time for missing the

region and

(38)
where
known that

is given in (13). In (37), it is well


is chi-square distributed with the pdf given by

(32)

is used in calculating
With these definitions, the mean acquisition time for the
is given as follows. For
case that the starting cell is

(39)
is the zeroth-order modified Bessel function of
where
, and
is defined
the first kind,
in (9). Therefore, (37) can be evaluated as

(40)
where
(41)
is the Marcum generalized function, which can be evaluated
very efficiently by the saddle point integration method [21].
then (40) becomes
For
(42)
(33)
From (33), it is easy to see that

(34)

and
take the
For active correlators, the probabilities
form of (37) and can be evaluated similarly as in (40). Unfortunately, there is no efficient method to calculate (38) due
to that
are correlated central or noncentral
chi-square random variables. To the best of our knowledge,
for central chi-square cases, one way to do it is to express the
joint pdf of
in a convergent infinite series

730

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 48, NO. 3, MAY 1999

Fig. 6. Effects of the cell correlations on the performance of mean acquisition time.

and then the probability of the form

(43)
or

1=13

= 127

=0

V. NUMERICAL EXAMPLES

can be evaluated numerically [22], [23]. However, for noncentral chi-square cases, no such convergent infinite series
have been found in the open literature. Moreover, since a
numerical integration is still needed in the above infinite series
method, the method would not be easier than the Monte Carlo
integration employed in this study. Note that all conditional
probabilities for the thresholds of interest can be found by
using one integration. This is important in the process of
finding the optimum threshold. Let

where

Fig. 7. Comparisons of analytical and simulation results for the cases of


= ; K
; and b
dB.

and

Since

In this section, numerical examples are used to show the


effects of cell correlations. For simplicity, only one-dwell
is employed for all
systems will be considered.3
numerical examples in this section. Monte-Carlo integration
with 10 samples have been used to obtain the probability
(38). The mean acquisition time is evaluated only for the case
, i.e., the worst case.
of
Fig. 6 shows typical examples of the effects of cell correis the time uncertainty of the incoming
lations. Recall that
is defined by
code. In the figures, the SNR
(47)

and
are approximately
It can be shown that
for a large
and, hence,
uncorrelated, if
they are independent Gaussian variables. Equation (45) can be
used to efficiently generate the random variables
Taking
as an example, the Monte-Carlo
integration can be used to calculate the probability

As expected, the effects of the cell correlations are generally


Also, at the expense of a
more significant for a smaller
larger system complexity, a better performance is obtained
with a smaller (larger ) due to the fact that there are more
region. However, as evident in the figure, the
cells in the
less than
performance improvement becomes smaller with
In
1/3. Similar results are observed for the cases of
Fig. 7, computer simulations are used to verify the analytical
As can be seen, analytical
results for the case of
results agree very well with the simulation ones, and, therefore,
to
regions
it is verified that the correlations from the
can be safely neglected. In our simulations, the sequence is
, and
generated by the primitive polynomial
500 acquisitions are averaged to obtain the mean acquisition
time.
Fig. 8 shows the relative acquisition time error for the case
and
with various SNRs and s. The
of
relative acquisition time error is defined as

(46)

(48)

we have
(44)
By using (44), (4) and (5) become

(45)

by generating the random variables


ing to (45) and (13).

and

accord-

3 For two-dwell systems, since the verification process provides independent


samples, the effects of cell correlations may be smaller than those presented
here.

SHEEN et al.: EFFECTS OF CELL CORRELATIONS IN PN CODE ACQUISITION

731

Fig. 9. The state transition diagram with

1 = 1=2 (H1 region).

parameters. As can be seen, for an SNR larger than 0 dB,


and
over 10% and 20% of errors are observed for
respectively. For
5%8% of errors are
observed for the same range of SNRs. In addition, the effects
of cell correlation are not sensitive to the values of
Fig. 8. Example relative acquisition time errors.
TABLE III
THE OPTIMUM THRESHOLD, THE MINIMUM MEAN ACQUISITION TIME, AND THE
ASSOCIATED RELATIVE ERROR FOR VARIOUS SYSTEM PARAMETERS

VI. CONCLUSIONS
In this paper, a new method has been proposed to analyze
the matched-filter PN code acquisition for DS spread-spectrum
systems. By using a time-domain approach along with a state
transition diagram, the mean acquisition time performance
can be evaluated with the cell correlations being taken into
account (the effects of the cell correlations have been neglected
in the previous analyses). Numerical results show that: 1)
the effects of the cell correlations are more significant for
a smaller cell size and/or a larger SNR; 2) for SNRs of
practical interest, over 10% and 20% acquisition time error
and
(at the optimum thresholds) are observed for
, respectively; and 3) the effects of cell correlations
at the optimum thresholds are quite insensitive to the value
of penalty time.

APPENDIX
DERIVATION OF (33)
is used
In this Appendix, a simple example with
to explain the derivation of (33) more clearly. Fig. 9 depicts
is used to denote
the state transition diagram, where
the transition gain for easy presentation. Using our notation,
for the first pass, the average acquisition time (if acquired) is
region)
given by ( in the
where
is the mean acquisition time without
considering the cell correlations. As can be seen, the larger
the SNR, the larger the relative acquisition time error. Also,
over 20% of errors are observed. Note that the largest relative
error does not necessarily occur at the optimum threshold.
(The optimum thresholds in the figure can be read from
Table III.) Once again, the result shows that the effects of
the cell correlations are more significant for a smaller
Table III shows the optimum threshold, minimum mean
for various system
acquisition time, and the associated

where

and
For the second pass, the

732

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 48, NO. 3, MAY 1999

average acquisition time (if acquired) is

Similarly, the average acquisition time for the third pass is

and so on. Summing all passes is exactly (33).


REFERENCES
[1] J. K. Holmes, Coherent Spread Spectrum Systems. New York: Wiley,
1982.
[2] M. K. Simon, J. K. Omura, R. A. Scholtz, and B. K. Levitt, Spread
Spectrum Communications, vols. IIII. MD: Computer Science, 1985.
[3] R. L. Peterson, R. E. Ziemer, and D. E. Borth, Introduction to SpreadSpectrum Communications. Englewood Cliffs, NJ: Prentice-Hall,
1995.
[4] G. F. Sage, Serial synchronization of pseudonoise systems, IEEE
Trans. Commun., vol. COM-12, pp. 123 and 127, Dec. 1964.
[5] J. K. Holmes and C. C. Chen, Acquisition time performance of PN
spread-spectrum systems, IEEE Trans. Commun., vol. COM-25, pp.
778783, Aug. 1977.
[6] D. M. DiCarlo and C. L. Weber, Multiple dwell serial search: Performance and application to direct sequence code acquisition, IEEE Trans.
Commun., vol. COM-31, pp. 650659, May 1983.
[7] Y. H. Lee and S. Tantaratana, Sequential acquisition of PN sequences
for DS/SS communications: design and performance, IEEE J. Select.
Areas Commun., vol. 10, pp. 750759, May 1992.
[8] K. K. Chawla and D. V. Sarwate, Acquisition on PN sequences in chip
synchronous DS/SS systems using a random sequence model and the
SPRT, IEEE Trans. Commun., vol. 42, pp. 23252334, June 1994.
[9] S. Tantaratana, A. W. Lam, and P. J. Vincent, Noncoherent sequential
acquisition of PN sequences for DS/SS communication with/without
channel fading, IEEE Trans. Commun., vol. 43, pp. 17381746,
Feb./Mar./Apr. 1995
[10] A. Polydoros and C. L. Weber, A unified approach to serial search
spread-spectrum code acquisitionPart II: A matched filter receiver,
IEEE Trans. Commun., vol. COM-32, pp. 550560, May 1984.
[11] Y. T. Su, Rapid code acquisition algorithms employing PN matched
filters, IEEE Trans. Commun., vol. 36, pp. 724733, June 1988.
[12] U. Cheng, Performance of a class of parallel spread spectrum code
acquisition schemes in the presence of data modulation, IEEE Trans.
Commun., vol. 36, pp. 596604, May 1988.
[13] E. Sourour and S. C. Gupta, Direct-sequence spread-spectrum parallel
acquisition in a fading mobile channel, IEEE Trans. Commun., vol. 38,
pp. 992998, July 1990.
[14] B. B. Ibrahim and A. H. Aghvami, Direct sequence spread spectrum matched filter acquisition in frequency-selective Rayleigh fading
channels, IEEE J. Select. Areas Commun., vol. 12, pp. 885890, June
1994.
[15] V. M. Jovanovic and E. S. Sousa, Analysis of noncoherent correlation
in DS/BPSK spread spectrum acquisition, IEEE Trans. Commun., vol.
43, pp. 565573, Feb./Mar./Apr. 1995.

[16] R. B. Ward, Acquisition of pseudonoise signals by sequential estimation, IEEE Trans. Commun., vol. COM-13, pp. 475483, Dec.
1965.
[17] R. B. Ward and K. P. Yiu, Acquisition of PN signals by recursion
aided sequential estimation, IEEE Trans. Commun., vol. COM-25, pp.
784794, Aug. 1977.
[18] H. Meyr and G. Poltzer, Performance analysis for general PN spread
spectrum acquisition techniques, IEEE Trans. Commun., vol. COM-31,
pp. 13171319, Dec. 1983.
[19] V. M. Jovanovic, Analysis of strategies for serial-search spread spectrum code acquisitiondirect approach, IEEE Trans. Commun., vol. 36,
pp. 12081220, Nov. 1988.
[20] S. M. Pan, D. E. Dodds, and S. Kumar, Acquisition time distribution
for spread spectrum receivers, IEEE J. Select. Areas Commun., vol. 8,
pp. 800807, June 1990.
[21] C. W. Helstrom, Approximate evaluation of detection probabilities in
radar and optical communications, IEEE Trans. Aerosp. Electron. Syst.,
vol. AES-14, pp. 630640, July 1978.
[22] A. S. Krishnamoorthy and M. Parthasarathy, A multivariate Gammatype distribution, Ann. Math. Statist., vol. 22, pp. 549557, 1951.
[23] T. Royen, Expansions for the multivariate Chi-square distribution, J.
Multivariate Anal., vol. 38, pp. 213232, 1991.

Wern-Ho Sheen (S89M91) received the B.S.E.E.


degree from the National Taiwan Institute of Technology, Taiwan, R.O.C., in 1982, the M.S.E.E. degree from the National Chiao Tung University,
Taiwan, in 1984, and the Ph.D. degree from the
Georgia Institute of Technology, Atlanta, in 1991.
From 1984 to 1993, he was with Telecommunication Laboratories, Taiwan, where he was mainly
involved in the projects of personal communications
and basic rate ISDN. Since 1993, he has been an
Associate Professor in the Department of Electrical
Engineering, National Chung Cheng University, Taiwan. His research interests
include adaptive signal processing, spread-spectrum communications, and
personal and mobile radio systems.

Jiun-Kai Tzeng was born in Taipei, Taiwan,


R.O.C., on August 12, 1972. He received the
B.S. degree in 1994 from the National ChiaoTung University, Taiwan, and the M.S. degree in
1996 from the National Chung Cheng University,
Chia-Yi, Taiwan.
Since October 1996, he has been with the National Space Program Office, Hsin-Chu, Taiwan. His
current research interests include spread-spectrum
communications and satellite communications.

Ching-Kae Tzou was born in Hsin-Chu, Taiwan,


R.O.C. He received the B.S.E.E. and M.S.E.E. degrees from the National Chiao-Tung University,
Taiwan, in 1984 and 1986, respectively, and the
Ph.D. degree in system electrical engineering from
the University of South California, Los Angeles.
During 19931996, he was with the Department
of Communication Technology at Computer and
Communication Laboratories, Industry of Technology Research Institute, Taiwan, where he worked
on personal wireless communication system design.
In 1997, he became a Senior Engineer in the VLSI Group, TranSwitch
Corporation, Taipei, Taiwan, and was promoted to Principle Engineer in
May of the same year, working on ASIC design for telecommunications. His
research interests include wireless communication system design, digital signal
processing, and ASIC design for applications in communication systems.
Dr. Tzou received the Phi Tau Phi Award from the Phi Tau Phi Scholastic
Society, Taiwan, in 1986.

You might also like