U.S. patent application number 10/953010 was filed with the patent office on 2006-03-30 for parameter estimate initialization using interpolation.
Invention is credited to Gregory Bottomley, Carmela Cozzo, Rajaram Ramesh.
Application Number | 20060067383 10/953010 |
Document ID | / |
Family ID | 35385585 |
Filed Date | 2006-03-30 |
United States Patent
Application |
20060067383 |
Kind Code |
A1 |
Cozzo; Carmela ; et
al. |
March 30, 2006 |
Parameter estimate initialization using interpolation
Abstract
A receiver is described herein that is capable of receiving and
processing a radio signal and further capable of using
interpolation to initialize receiver parameters when there is a
change in at least one delay associated with the received radio
signal or when there is at least one new correlator position. For
instance, the receiver parameters that can be initialized include:
(1) channel coefficients; (2) AFC parameters; (3) tracking
parameters; (4) noise statistics (noise correlations); (5) signal
statistics (channel coefficient correlations); (6) data statistics
(despread values or chip samples); or (7) combining weights.
Inventors: |
Cozzo; Carmela; (Cary,
NC) ; Bottomley; Gregory; (Cary, NC) ; Ramesh;
Rajaram; (Cary, NC) |
Correspondence
Address: |
ERICSSON INC.
6300 LEGACY DRIVE
M/S EVR C11
PLANO
TX
75024
US
|
Family ID: |
35385585 |
Appl. No.: |
10/953010 |
Filed: |
September 29, 2004 |
Current U.S.
Class: |
375/147 ;
375/E1.032 |
Current CPC
Class: |
H04B 2201/709727
20130101; H04B 1/7117 20130101; H04B 1/712 20130101 |
Class at
Publication: |
375/147 |
International
Class: |
H04B 1/707 20060101
H04B001/707; H04B 1/69 20060101 H04B001/69 |
Claims
1. A receiver that receives and processes a radio signal and then
uses interpolation to estimate an initial value of a parameter when
there is a change in at least one delay associated with the
received radio signal.
2. The receiver of claim 1, wherein said interpolation that
estimates the initial value of the parameter uses existing
parameter estimates.
3. The receiver of claim 1, wherein said interpolation that
estimates the initial value of the parameter uses searcher
information.
4. The receiver of claim 1, wherein said interpolation that
estimates the initial value of the parameter uses a combination of
existing parameter estimates associated with existing delays and
searcher information such as at least one of a complex/delay
profile (CDP) and a power/delay profile (PDP).
5. The receiver of claim 1, wherein said interpolation is linear
interpolation.
6. The receiver of claim 1, wherein said interpolation is Wiener
interpolation.
7. The receiver of claim 1, wherein said parameter is a channel
coefficient.
8. The receiver of claim 1, wherein said parameter is an automatic
frequency correction (AFC) parameter.
9. The receiver of claim 1, wherein said parameter is a tracking
parameter.
10. The receiver of claim 1, wherein said parameter is a noise
statistic.
11. The receiver of claim 1, wherein said parameter is a signal
statistic.
12. The receiver of claim 1, wherein said parameter is a combining
weight.
13. The receiver of claim 1, wherein said parameter is a data
statistic.
14. The receiver of claim 13, wherein said data statistic
corresponds to received sample data.
15. The receiver of claim 13, wherein said data statistic
corresponds to despread data.
16. The receiver of claim 1, wherein said receiver is a RAKE
receiver.
17. The receiver of claim 1, wherein said receiver is a chip
equalizer.
18. The receiver of claim 1, wherein said at least one delay
associated with the received radio signal is a new finger position
when said receiver is a RAKE receiver.
19. The receiver of claim 1, wherein said at least one delay
associated with the received radio signal is a new tap position
when said receiver is a chip equalizer.
20. A wireless communication system comprising: a transmitter
capable of transmitting a radio signal; and a receiver including:
an antenna capable of receiving the radio signal; a radio frequency
(RF) processor capable of processing the radio signal; and a
baseband processor capable of estimating at least one delay
associated with the processed radio signal and further capable of
using interpolation to initialize a parameter when there is a
change in the at least one delay associated with the processed
radio signal.
21. The wireless communication system of claim 20, wherein said
baseband processor includes a weight formation unit which uses
existing parameter estimates to interpolate an estimated initial
value of the parameter.
22. The wireless communication system of claim 20, wherein said
baseband processor includes a weight formation unit which uses
searcher information to interpolate an estimated initial value of
the parameter.
23. The wireless communication system of claim 20, wherein said
baseband processor includes a weight formation unit which
interpolates an estimated initial value of the parameter by using a
combination of existing parameter estimates and searcher
information.
24. The wireless communication system of claim 20, wherein said
baseband processor uses linear interpolation to initialize the
parameter when there is a change in the at least one delay
associated with the processed radio signal.
25. The wireless communication system of claim 20, wherein said
baseband processor uses Wiener interpolation to initialize the
parameter when there is a change in the at least one delay
associated with the processed radio signal.
26. The wireless communication system of claim 20, wherein said
parameter is at least one of: a channel coefficient; an automatic
frequency correction (AFC) parameter; a tracking parameter; a noise
statistic; a signal statistic; a data statistic; and a combining
weight.
27. A method for parameter initialization in a radio signal
receiver, said method comprising the steps of: estimating at least
one delay associated with the received radio signal; and using
interpolation to estimate an initial value of a parameter when
there is a change in at least one delay associated with the
received radio signal.
28. The method of claim 27, wherein said interpolation that
estimates the initial value of the parameter uses existing
parameter estimates.
29. The method of claim 27, wherein said interpolation that
estimates the initial value of the parameter uses searcher
information.
30. The method of claim 27, wherein said interpolation that
estimates the initial value of the parameter utilizes a combination
of existing parameter estimates and searcher information.
31. The method of claim 27, wherein said interpolation is linear
interpolation.
32. The method of claim 27, wherein said interpolation is Wiener
interpolation.
33. The method of claim 27, wherein said parameter is at least one
of: a channel coefficient; an automatic frequency correction (AFC)
parameter; a tracking parameter; a noise statistic; a signal
statistic; a data statistic; and a combining weight.
Description
BACKGROUND OF THE INVENTION
[0001] 1. Field of the Invention
[0002] The present invention relates in general to the wireless
telecommunications field and in one exemplary embodiment to a
receiver that receives a signal and estimates signal delays
associated with the received signal and when there is a new signal
delay then the receiver uses interpolation to estimate an initial
value of a parameter (e.g., channel coefficient, tracking
parameter) associated with the new signal delay.
[0003] 2. Description of Related Art
[0004] In direct-sequence code-division multiple-access (DS-CDMA)
systems, such as WCDMA and IS-2000, coherent RAKE receivers are
commonly used. This type of receiver often estimates a channel
response which consists of path delays and channel coefficients.
The receiver also uses despread values to search for signal paths.
Based on all of this information, the receiver positions its
correlators or "fingers" at certain delays. As these delays change
over time, the receiver needs to move the fingers to different
positions. Every time a new finger position is used, the receiver
needs to initialize a channel coefficient or some other parameter
like channel tracking parameters, automatic frequency correction
(AFC) quantities, noise correlations and fading correlations. The
last two parameters are needed for advanced receivers like G-RAKE
receivers or joint scaling receivers.
[0005] A traditional approach that can be used by a receiver to
initialize a channel coefficient after a finger is moved to a new
position is one that uses prior knowledge of the channel in a
nearby position. In this approach, the receiver considers the
distance between the new finger position provided and the closest
finger delay to the new position in a delay tracker. If the
distance is less than 1/4-chip period, it is assumed that the
finger has not moved much and the old channel coefficient is kept.
For larger distances, the channel coefficient in the new position
is set to zero, or its value is computed by scaling an initial,
noisy measured value. This approach is described in U.S. Pat. No.
6,560,273 entitled "Delay Searcher and Delay Trackers Interaction
for New Delays Assignment to RAKE Fingers". The contents of this
patent are incorporated herein.
[0006] Although this approach works well it does have a potential
shortcoming in that the accuracy of the initial channel coefficient
estimates may not be good enough to ensure a reliable channel
estimate and/or provide a fast enough convergence. And, if the
initial values of these initialized channel coefficient estimates
(or other parameters) are not good enough, then there can be a
transient loss in performance while the receiver develops better
estimates of the initial channel coefficients. Accordingly, there
is a need for a receiver that addresses the problem of
initialization of a channel coefficient (or other parameters) when
there is a new signal delay or when a finger is moved to a new
position. This need and other needs are satisfied by the receiver
of the present invention.
SUMMARY OF THE INVENTION
[0007] The present invention includes a receiver that is capable of
receiving and processing a radio signal and further capable of
using interpolation to initialize receiver parameters when there is
a change in at least one delay associated with the received radio
signal or when there is at least one new correlator position. For
instance, the receiver parameters that can be initialized include:
(1) channel coefficients; (2) AFC parameters; (3) tracking
parameters; (4) noise statistics (noise correlations); (5) signal
statistics (channel coefficient correlations); (6) data statistics
(despread or received sample correlations); or (7) combining
weights (for combining despread values or chip samples).
BRIEF DESCRIPTION OF THE DRAWINGS
[0008] A more complete understanding of the present invention may
be had by reference to the following detailed description when
taken in conjunction with the accompanying drawings wherein:
[0009] FIG. 1 is a block diagram of a wireless communication system
that includes a transmitter and a receiver which is configured in
accordance with the present invention;
[0010] FIG. 2 is a diagram that illustrates in greater detail the
components within a baseband processor in the receiver shown in
FIG. 1;
[0011] FIG. 3 is a diagram that illustrates in greater detail the
components within one embodiment of a weight formation unit in the
baseband processor shown in FIG. 2;
[0012] FIG. 4 (PRIOR ART) is a flowchart that shows the steps of a
traditional method that can be used to initialize a channel
coefficient within the weight formation unit shown in FIG. 3;
[0013] FIG. 5 is a flow chart that shows the steps of a method that
can be used within the weight formation unit shown in FIG. 3 to
initialize a channel coefficient for a new finger position by using
interpolated values from old or prior finger positions in
accordance with one embodiment of the present invention;
[0014] FIG. 6 is a diagram that illustrates an example used to help
describe the different types of interpolation methods that can be
used in the present invention;
[0015] FIG. 7 is a diagram that illustrates in greater detail the
components within one embodiment of a weight formation unit that
can be used in the baseband processor of a G-RAKE receiver in
accordance with the present invention;
[0016] FIG. 8 is a diagram that illustrates in greater detail the
components within one embodiment of a weight formation unit that
can be used in the baseband processor of a joint scaling RAKE
receiver in accordance with the present invention;
[0017] FIG. 9 is a diagram that illustrates in greater detail the
components within one embodiment of a finger placement unit that
can be used within the baseband processor shown in FIG. 2;
[0018] FIG. 10 is a flow chart that shows the steps of a method
where initialization of channel coefficients is performed by using
the most recent complex/delay profile (CDP) from a finger placement
unit in accordance with another embodiment of the present
invention;
[0019] FIG. 11 is a diagram that illustrates the components of a
finger placement unit that can be used within the baseband
processor of a G-RAKE receiver to initialize noise statistics in
accordance with another embodiment of the present invention;
[0020] FIG. 12 is a flowchart that shows the steps of a method on
how initialization of noise and channel statistics can be performed
within a joint scaling RAKE receiver by using interpolation in
accordance with yet another embodiment of the present
invention;
[0021] FIG. 13 is a diagram that illustrates the components within
an initialization unit shown in FIG. 3 that can use multiple CDP
measurements for channel tracking in accordance with another
embodiment of the present invention;
[0022] FIG. 14 is a diagram that illustrates the components within
a receiver that can be used to combine information from the finger
placement unit and the existing fingers in accordance with yet
another embodiment of the present invention; and
[0023] FIG. 15 is a diagram that illustrates an alternative
baseband processor that can be used in the receiver shown in FIG.
1.
DETAILED DESCRIPTION OF THE DRAWINGS
[0024] Referring to FIGS. 1-15, there is disclosed a receiver 100
that uses different types of interpolations to develop estimates of
initial receiver parameters like channel coefficients, channel
tracking parameters, AFC quantities, noise correlations, fading
correlations, data correlations and combining weights. Although the
receiver of the present invention is described as being in the form
of a RAKE receiver (RAKE, G-RAKE, joint scaling receiver) or chip
equalizer, it should be understood that the present invention
applies to any type of receiver that estimates signal delays.
Accordingly, the receiver 100 should not be construed in such a
limited manner.
[0025] FIG. 1 is a block diagram of a wireless communication system
101 that includes a transmitter 102 (only one shown) and a receiver
100 (only one shown) which is configured in accordance with the
present invention. As shown, the transmitter 102 transmits a radio
signal 114 that passes through a channel and is received by an
antenna 106 of the receiver 100. The receiver 100 also includes a
radio frequency (RF) processor 108 which processes the received
radio signal and a baseband processor 110 which converts the
processed received radio signal into a baseband signal 104 that is
further processed by an additional processor 112.
[0026] FIG. 2 is a diagram that illustrates in greater detail the
different components within the baseband processor 110. The
baseband processor 110 includes a correlation unit 202 (fingers
202), a finger placement unit 204 (searcher 204), a weight
formation unit 206 and a combiner 208. In DS-CDMA RAKE reception
(for example), the correlation unit 202 extracts different signal
images from the baseband signal 104 and performs a despreading
operation on the different signal images. To perform the
despreading operation, the correlation unit 202 uses delays or
finger positions that are provided by the finger placement unit
204. The finger placement unit 204 determines the delays or finger
positions for each signal image at a predetermined rate. And,
depending on the fading conditions the estimated delays made by the
finger placement unit 204 may change even if the true delays do
not. The combiner 208 combines the despread values received from
the correlation unit 202 by using combining weights provided by the
weight formation unit 206. In general, the combining weights depend
on estimates of the channel coefficients of the radio channel. The
combiner 204 then sends the combined values to the next processor
112. The present invention focuses on the weight formation unit
206.
[0027] FIG. 3 is a diagram that illustrates the components within
one embodiment of a weight formation unit 206' used within a RAKE
receiver 100. The weight formation unit 206' includes a correlation
unit 302, a channel tracking unit 304 and an initialization unit
306. The correlation unit 304 receives the delay values from the
finger placement unit 204 and correlates to the pilot signals in
the baseband signal 104 so it can send the despread values to the
channel tracking unit 304. Thereafter, the channel tracking unit
304 estimates the channel coefficients for each finger. If w
denotes the combining weights and c denotes the channel estimates
given by the channel tracking unit 304, then: w=c. Due to the
effect of noise and fading, as well as motion of the receiver 100,
the transmitter 102 and/or scattering objects, the positions of the
fingers change over time. Repositioning of fingers happens because
the channel actually changes and/or the channel is unchanged but
the finger placement unit 204 (searcher 204) assigns new positions.
When a finger is assigned a new position, the initialization unit
306 needs to determine and provide the initial value of the channel
coefficient.
[0028] FIG. 4 (PRIOR ART) is a flowchart that shows the steps of a
traditional method 400 that the initialization unit 306 can use to
initialize the channel coefficient. The process starts at step 402
and then at step 404 a new finger position is compared to the
closest old or existing finger. If the relative distance is less
than a specified value (for example 1/4-chip period), then at step
406 the channel tracking unit 304 is initialized to the channel
value of the old finger because it is assumed that the finger has
not moved. If the distance is larger then the specific value, then
at step 408 the channel tracking unit 304 is initialized to zero.
The process ends at step 412. For a more detailed description about
this traditional method 400 reference is made to the aforementioned
U.S. Pat. No. 6,560,273.
[0029] FIG. 5 illustrates a flow chart that shows the steps of a
method 500 that the initialization unit 306 can use to initialize
the channel coefficient for new fingers using interpolated values
from old or prior finger positions in accordance with one
embodiment of the present invention. The process starts at step 502
and then at step 504 the new finger positions are compared with the
old finger positions. If the position has not changed, then at step
506 the same channel coefficient is kept and tracking of the
channel continues. If the position is different, then at step 508 a
set of old positions is defined depending on the relative distance
between the old and new finger positions. A determination is then
made at step 510 as to whether the set is empty. If this set is
empty, then at step 512 the channel is initialized to zero or some
other form of prior art initialization may be performed. If the set
is nonzero, then at step 514 the channel is initialized to a value
obtained by the interpolation or extrapolation of the channel
coefficients of the old finger positions in the set, referred to
herein as the "interpolation set". The process ends at step
516.
[0030] Although the method 500 of the present invention can be used
with any finger placement strategy that is implemented by the
finger placement unit 204 (searcher 204), it is particularly useful
when the searcher 204 has the ability to use two strategies and can
switch from one strategy to the other depending on the channel
conditions, and/or system environment. For example, the searcher
204 could use a peak-based approach when in softer handoff and a
grid-based approach otherwise. It should be noted that regardless
of the finger placement strategies implemented by the searcher 204,
when the searcher 204 changes strategies some of the fingers are
going to be in new positions which means that one needs to
initialize the parameters for all of the new positions. Exemplary
finger placement strategies are described in U.S. patent
application Ser. No. 10/653,679 entitled "Method and Apparatus for
Finger Placement in a RAKE receiver". The contents of this document
are incorporated by reference herein.
[0031] FIG. 6 is a diagram that illustrates an example used to help
describe some of the different types of interpolation methods that
can be used in the present invention. In the example, the old
fingers positions are x1, x2 and x3, and the new finger positions
are y1 and y2. Clearly, the channel tracking unit 304 for the
channel in position y1 would be initialized to the estimate of
position x1 since the finger has not moved. For y2, the channel
tracking unit 304 is initialized with an interpolated value given
by the channel coefficients of the two adjacent old fingers (x2 and
x3) or of all three old fingers (x1, x2 and x3). In general, the
length of the region over which to perform interpolation is
predefined, for example it can be equal to one or two chip periods.
If a new finger position is farther apart than the specified
distance from any old finger position, the initial value can be set
to zero. In the example of FIG. 6, assume the interpolation over a
set of old finger positions x2 and x3 provides the initial value
for the channel coefficient of the new finger in position y2. As
such, the interpolation set includes the old finger positions x2
and x3 and the values of the channel coefficients for x2 and x3 are
h(x2) and h(x3), respectively.
[0032] One interpolation method that can be used to compute the
initial channel estimate for the new finger position y2 is linear
interpolation. In this case, the interpolated value of the new
finger is: h(y2)=(y2-x3)h(x2)/(x2-x3)+(y2-x2)h(x3)/(x3-x2).
[0033] Another interpolation method that can be used to compute the
initial channel estimate for the new finger position y2 is Wiener
interpolation. If h indicates the vector of the channel
coefficients of fingers belonging to the chosen set of old finger
positions (in this example h=[h(x1) h(x2)].sup.T), and G indicates
the vector of the filter coefficients, the channel coefficient of
the new finger y2 is: h(y2)=G.sup.Hh, with G=(R.sub.hh).sup.-1
r.sub.hh(y2), where r.sub.hh(y2) is the correlation vector between
the conjugate of the channel response at y2 and the set of channel
responses at the old finger positions, and R.sub.hh is the
covariance matrix of the set of channel coefficients of the old
finger positions. In this case, to obtain the coefficients of the
filter (entries in vector G), the correlation function of the
channel has to be estimated. And, then the correlation vector can
be obtained by interpolating values in the covariance matrix that
corresponds to nearby delays.
[0034] One way that the covariance matrix can be estimated is to
use channel coefficient estimates, as described in U.S. patent
application Ser. No. 10/672,127 entitled "Method and Apparatus for
RAKE Receiver Combining Weight Generation". The contents of this
document are incorporated by reference herein.
[0035] Another interpolation method is to use a simpler approach
that is based on estimating the "medium" response of the old finger
positions and then using the knowledge of the pulse shape "ringing"
to sum the responses of these signal images at the new position.
For instance, if h indicates the vector of the channel coefficients
of fingers belonging to the chosen set of old finger positions (in
this example h=[h(x1) h(x2)].sup.T), and B is a matrix that depends
only on the chip pulse shape, the medium response for the old
fingers is: g=B.sup.-1 h, where B is a square matrix depending on
how many fingers are in the interpolation set (length of g). If
p(t) indicates the chip pulse shape, then the elements of the B
matrix are approximated by elements of the chip pulse shape
autocorrelation function r.sub.p(xi-xj), where i and j are the
indices of the old fingers. As such, the channel response in a new
position, for example y2, is given by the contribution of paths
each of which are weighted by a coefficient that depends only on
the chip pulse shape autocorrelation (elements of the matrix A in
the next equation). The value of the channel in y2 is then:
h(y2)=Ag, where, like B, the elements of A are given by
r.sub.p(y2-xj) The A matrix has 1 row, and the number of columns
depends on the number of old finger positions used for
interpolation.
[0036] For a traditional RAKE receiver, the channel coefficient
estimates are used as the combining weights. Thus, interpolating
channel coefficient estimates is the same as interpolating
combining weights in this case. In FIG. 3, these combining weights
are formed by despreading pilot symbols and performing channel
tracking. As described in the incorporated U.S. Pat. No. 5,572,552,
it is possible to track the combining weights directly using an
adaptive filter, such as a least mean-squares (LMS) or recursive
least squares (RLS) filter. When new finger delays are introduced,
interpolation can be used to interpolate these combining weights to
obtain initial weight values at the new finger delays.
[0037] A description is provided next where noise statistics,
fading statistics and data statistics are estimated as well as
channel coefficients. For instance, a noise covariance matrix
and/or the channel covariance matrix will need to be estimated if
an advanced receiver 100 is used such as a G-RAKE receiver 100 (see
U.S. Pat. No. 6,363,104) or a joint scaling receiver 100 (see U.S.
patent application Ser. No. 10/672,127). The present invention
addresses the initialization of such covariance matrices.
[0038] An exemplary diagram of a weight formation unit 206'' that
can be used within a G-RAKE receiver 100 to estimate a noise
covariance matrix is shown in FIG. 7. The weight formation unit
206'' includes a correlation unit 702, a channel tracking unit 704,
a noise statistics unit 706, two initialization units 708 and 710
and a combiner 712. In this embodiment, if R.sub.n denotes the
noise covariance matrix, then the combining weights generated by
the combiner 712 correspond to the product of the inverse of the
noise covariance matrix and a vector of channel coefficient
estimates as follows: w=R.sub.n.sup.-1 c.
[0039] As described in U.S. Pat. No. 6,363,104 B1, the noise
covariance matrix can be replaced by a despread data correlation
matrix. Also, as described in an article by W. Hai et al.,
"Approaches for fast, adaptive, generalized RAKE reception,"
Research Disclosure Journal, No. 475041, Kenneth Mason Publications
Ltd., November 2003, the noise covariance matrix can be replaced by
a baseband sample or "chip" sample data correlation matrix. The
contents of both of these documents are incorporated by reference
herein.
[0040] An exemplary diagram of a weight formation unit 206''' that
can be used in a joint scaling RAKE receiver 100 to estimate the
noise and channel covariance matrices is shown in FIG. 8. The
weight formation unit 206''' includes a correlation unit 802, a
channel tracking unit 804, a noise statistics unit 806, a channel
statistics unit 808, three initialization units 810, 812 and 814
and a combiner 816. In this embodiment, if R.sub.c denotes the
channel covariance matrix, and R.sub.e denotes the covariance
matrix of the estimation error (usually computed by simply scaling
the noise covariance matrix R.sub.n), then the combining weights
generated by the combiner 816 have the following form:
w=[R.sub.n+R.sub.e(R.sub.c+R.sub.e).sup.-1 R.sub.c].sup.-1
R.sub.c(R.sub.c+R.sub.e).sup.-1c.* * It should be noted that a
covariance matrix R is referred to hereinafter as either the noise
or channel covariance matrix.
[0041] The estimation of the covariance matrices described above
can be done by smoothing multiple measurements over time. For
instance, if R(n-1) indicates the smoothed covariance matrix and
R.sub.i(n) is the instantaneous measurement, then the covariance
matrix can be updated as R(n)=.lamda.R(n-1)+(1-.lamda.) R.sub.i(n)
where .lamda. is the forgetting factor. And, when a new finger
position is used, a new row and column in R(n) is effectively
created. This row or column needs initial values that can be
initialized from elements in R(n) or R(n-1) corresponding to
existing finger positions. With the present invention, the initial
values are obtained by interpolating correlation values
corresponding to existing/old fingers. For the diagonal element,
interpolation using the diagonal elements of nearby fingers can be
used. Linear interpolation can be used. For the off-diagonal
elements, one can initialize the correlation of the new finger
position to position x1 using correlations of nearby fingers to
position x1. Again, this can be done by linear interpolation.
[0042] Sometimes, noise correlations between fingers are only a
function of relative finger delays, not absolute finger delays. In
this case, one can interpolate the noise correlation values that
correspond to nearby relative delays. For instance, consider the
exemplary scenario of FIG. 6 and indicate with R the covariance
matrix of the old fingers and indicate with {tilde over (R)}(0) the
initial estimate of the covariance matrix of the new finger
positions as follows: R = [ r 11 r 12 r 13 r 21 r 22 r 23 r 31 r 32
r 33 ] ##EQU1## R ~ .function. ( 0 ) = [ r ~ 11 r ~ 12 r ~ 21 r ~
22 ] ##EQU1.2## Then the elements of the R(0) matrix can be
computed as follows: [0043] since y1 is an old delay (y1=x1),
{tilde over (r)}.sub.11=r.sub.11; [0044] {tilde over (r)}.sub.22 is
given by the interpolation of r.sub.22 and r.sub.33, or by the
interpolation of r.sub.11, r.sub.22 and r.sub.33, depending on the
distance between y2 and the old fingers; and [0045] {tilde over
(r)}.sub.12 is given by the interpolation of the know correlation
between y1 and x2, and y1 and x3. Note that in this example these
correlation values are known because one of the fingers is an old
position finger, and the new position is in-between two old
fingers.
[0046] A description is provided next where AFC parameters are
estimated in accordance with the present invention. As described in
the incorporated U.S. patent application Ser. No. 09/678,901
entitled "Method and Apparatus for Automatic Frequency Control in a
CDMA Receiver", the AFC is estimated and possibly applied
separately to each path (finger location). In this patent
application, the estimation involves an initial value and
smoothing. With the present invention, the initial frequency offset
estimate for a new finger location can be obtained by interpolating
values from nearby finger locations. Linear interpolation can be
used. It may also help if the interpolated value was scaled down by
a factor between 0 and 1. This scaling may also be applied when
interpolating other quantities, such as channel estimates and
fading and noise statistics.
[0047] A description is provided next where tracking parameters are
estimated in accordance with the present invention. In the case for
channel tracking, AFC, and estimation of fading, noise, or data
statistics, there may be tracking parameters depending on the
tracking or estimation approach used. For example, the channel
tracking unit 304, 704 and 804 may have a step size if least-mean
square (LMS) tracking is used or a window size if a sliding window
average is used. The channel tracking unit 304, 704 and 804 may
also include smoothing filters that are designed based on power
spectral estimation as described in the incorporated U.S. patent
application Ser. No. 09/277,180 entitled "Smoothing Channel
Estimates by Spectral Estimation". While for AFC estimation, the
channel tracking unit 304, 704 and 804 may have phased-lock loop
step sizes. And for fading and noise statistics estimation, the
channel tracking unit 304, 704 and 804 could use smoothing
parameters that differ for different fingers or finger pairs. With
the present invention, all of these parameters or model estimates
can be interpolated from existing fingers. In addition,
interpolation with possible scaling may be used.
[0048] Up to this point, the present invention has been described
where the initialization of parameter estimates was performed by
using interpolation and existing finger locations. However, the
present invention has an alternative embodiment where
initialization can be performed by using interpolation and searcher
information. To illustrate this, reference is made to FIG. 9 which
shows an exemplary diagram of the components within one embodiment
of the finger placement unit 204' (searcher 204')(see FIG. 2). The
finger placement unit 204' includes a correlation unit 902, a
coherent smoothing unit 904, a noncoherent smoothing unit 906 and a
delay searching unit 908. In this example, the correlation unit 902
utilizes the delays and the baseband signal 104 to despread certain
symbols (pilots) at a range of delay values. Though not required,
the delay values are usually on an evenly spaced grid of values,
such as delays 1 or 1/2 chip apart. The coherent smoothing unit 904
processes the despread pilots to produce a complex/delay profile
(CDP). The noncoherent smoothing unit 906 then processes the CDP to
produce a power/delay profile (PDP). Thereafter, the delay
searching unit 908 processes the PDP and outputs delays to the
correlation unit 202 and the weight formation unit 206 (see FIG.
2).
[0049] In accordance with the present invention, the PDP and the
one or more CDPs are stored and used as part of initialization.
First, consider the case in which only the most recent CDP is
stored. For channel estimate initialization, an interpolation set
of nearby CDP delays is formed using, for example, the nearest 4
delays. The CDP values at those 4 delays form an interpolation set
that are interpolated using the aforementioned methods to determine
an initial channel estimate at the delay corresponding to the new
finger position. The interpolation set can be formed by CDP samples
that may or may not correspond to previous finger positions. If the
new finger position happens to correspond to a delay for which a
CDP value is available, then that single CDP value can be used as
the initial channel estimate. This is shown in FIG. 10.
[0050] FIG. 10 illustrates a flow chart that shows the steps of a
method 1000 that the weight control unit 206 can use to interpolate
and initialize the channel coefficient estimation using the most
recent CDP in accordance with the present invention. The process
starts at step 1002 and then at step 1004 it is assumed there is a
new delay at Di which corresponds to a new finger position. At step
1006, a determination is made at to whether a CDP value exists at
Di. If yes, then at step 1008 the channel coefficient is
initialized to CDP(Di). If no, then at step 1010 a set of CDP
values are chosen so they can be interpolated at step 1012.
Thereafter, at step 1014 the channel coefficient associated with
the interpolation value is initialized. The process ends at step
1016.
[0051] The searcher information can also be used to initialize
noise and/or fading statistics for an advanced receiver 100 like a
G-RAKE receiver 100 or a joint scaling receiver 100. FIG. 11 is a
block diagram illustrating the components within a finger placement
unit 204'' of a G-RAKE receiver 100 that can use interpolation to
initialize the signal statistics. The finger placement unit 204''
includes a PDP computer 1102, a noise power computer 1104, a
subtracter 1106 and a unit 1108 that initializes the diagonal
elements of R.sub.c. The PDP computer 1102 interpolates the PDP
values to obtain a PDP value for a new delay. The subtracter 1106
then subtracts a noise power estimate from this PDP value and
outputs a signal power estimate, which can be used by unit 1108 to
initialize the diagonal elements of R.sub.c. The off-diagonal
elements can be initialized to zero. It should be noted that
standard approaches can be used to estimate a noise power using the
PDP, such as taking the average. In certain scenarios, such as in
the downlink it may help to estimate noise power as a function of
delay and then interpolate such values.
[0052] FIG. 12 is a flowchart illustrating the steps of a method
1200 that can be used in a joint-scaling receiver 100 to
interpolate and initialize signal fading statistics. The process
starts at step 1202 and then at step 1204 it is assumed there is a
new delay at Di which corresponds to a new finger position. At step
1206, a determination is made at to whether a PDP exists at Di. If
yes, then at step 1208 the noise power is subtracted from PDP(Di)
and at step 1210 a diagonal element of R.sub.c is assigned. If no,
then at step 1212 a set of PDP values are chosen so they can be
interpolated at step 1214 to obtain PDP(Di). Thereafter, at step
1216 the noise power is subtracted from PDP(Di) and at step 1218 a
diagonal element of R.sub.c is assigned. The process ends at step
1220. It should be noted that one can use a weighted sum of a set
of PDP values to obtain the PDP at a certain delay even if the PDP
value exists at that position.
[0053] The case in which multiple CDPs are stored over time is
considered next. In this case, to initialize a channel estimate,
interpolation can be applied to each CDP to obtain channel
measurements at the new finger position over time. These
measurements are provided to a standard channel tracking algorithm
to produce a channel estimate or prediction for the present time.
This estimate can be the initial value used. Examples of channel
tracking algorithms that can be used include sliding window
averaging, Wiener filtering, Least-Mean Square (LMS), Normalized
Least-Mean Square (NLMS), Recursive Least Square (RLS) and Kalman
Least-Mean Square (KLMS) tracking. Any form of initialization of
these trackers can be used. The interpolated values can also be
used to estimate AFC quantities and tracking parameters such as
Doppler spread. In addition, channel tracking can be used with the
interpolated values to provide a track of the channel at that
delay. This track can be used to estimate diagonal and off-diagonal
elements of R.sub.c and R.sub.n with the aid of the approaches that
are described in the aforementioned U.S. patent application Ser.
No. 10/672,127.
[0054] FIG. 13 is a block diagram illustrating some of the
components within the initialization unit 306 (FIG. 3) that use
multiple CDP measurements for channel tracking. As can be seen,
CDPi and a new delay are input into an initial channel estimate
computer 1302 which is connected to a channel tracking unit 1304
that outputs an initial channel estimate.
[0055] In another embodiment of the present invention, the
initialization can be based on a mixture or combination of existing
finger information and searcher information. One approach is to
select information from one or the other where the selection
depends on what information is available from existing fingers.
Another approach is based on complexity where one can choose to use
the information that requires the least processing to obtain the
final initial channel estimate and/or channel statistics. In yet
another approach one can combine information from the searcher and
from the existing fingers. In this example, an initial channel
estimate can be obtained from both the searcher and the existing
fingers. And, then the weighted sum of these two can be used as the
initial channel estimate. The weights can sum to one and can be the
same (0.5 and 0.5) or different to account for which estimate is
believed to be noisier (the noisier estimate should have a weight
less than 0.5). FIG. 14 is a block diagram that illustrates an
exemplary set of components that can be used to combine information
in accordance with this approach of the present invention. As shown
in FIG. 14, the channel estimate from the searcher (see FIG. 13) is
input into a weighting unit 1402 and the channel estimate from
existing fingers (see FIG. 3) is input into another weighting unit
1404. The outputs from both weighting units 1402 and 1404 are input
to a combiner 1406 which is connected to an initialization of new
delay unit 1408 that outputs a final initial channel estimate.
[0056] It should be appreciated that combined information can also
be used for the initialization of the noise and channel statistics.
For example, one can use the information from the searcher 204 to
initialize the diagonal elements of the matrices of the noise and
fading statistics, and use the information from the existing
fingers to initialize the off-diagonal elements, as described in
detail above. Also, the weighted sum of information from both
sources can be used for the diagonal elements of the matrices. In
general, the preferred combination of the information depends on
the scenario under consideration and can be determined by using
some or all criteria mentioned above: availability, performance and
complexity. For instance, in high-speed scenarios, if searcher
information is available, one can simply use the CDP values for
channel coefficient initialization.
[0057] An alternative structure to the RAKE receiver is the chip
equalizer. With chip equalization, the combining weights are used
as equalization filter coefficients prior to despreading. FIG. 15
is a diagram that illustrates an alternative embodiment of a
baseband processor 110' based on chip equalization. The baseband
processor 110' includes an equalization filter 1502, a tap
placement unit 1504 (searcher 1504), a weight formation unit 1506,
and a correlator 1508. With one form of MMSE chip equalization (for
example), the equalization filter 1502 filters or combines baseband
samples. This filtering operation is performed using delays or tap
positions that are provided by the tap placement unit 1504. The tap
placement unit 1504 determines filter tap positions based on
traditional location of signal paths. The equalization filter 1502
also uses combining weights provided by weight formation unit 1506.
Similar to the RAKE receiver, the weight formation unit can produce
combining weights based on channel coefficient estimates. Like the
G-RAKE receiver, the weight formation unit may also use a baseband
sample data correlation matrix when forming the combining weights.
Direct adaptation of the weights is also possible, as described in
an article by F. Petre et al., "Pilot-aided adaptive chip equalizer
receiver for interference suppression in DS-CDMA forward link," in
Proc. IEEE Vehicular Technology Conference, Boston, Mass., Sep.
24-28, 2000. The contents of this document are incorporated by
reference herein.
[0058] With the chip equalizer structure, the interpolation
approaches described previously can also be applied. For example,
as tap locations change, interpolation can be used to determine
channel coefficient estimates for the new tap locations. For MMSE
chip equalization, interpolated data correlation values can also be
determined.
[0059] From the foregoing, it can be readily appreciated by those
skilled in the art that the receiver 100 of the present invention
uses interpolation to consider initialization of receiver
parameters when finger positions or channel response delay
estimates change over time. The receiver parameters that can be
estimated include: (1) channel coefficients; (2) AFC parameters;
(3) tracking parameters; (4) noise statistics (noise correlations);
(5) signal statistics (channel coefficient correlations); (6) data
statistics (despread values or chip samples); or (7) combining
weights. As described in detail above, the present invention
proposes a form of interpolation that uses existing information to
provide initial values for the new positions. One type of existing
information is the set of existing (old) parameter estimates
corresponding to the existing (old) finger positions. Another type
of existing information is information generated by a searcher
during the search process.
[0060] Following are some additional features, advantages and uses
of the present invention: [0061] The present invention can be used
in WCDMA base stations and terminals. [0062] The present invention
can be associated with technology governed in the WCDMA and IS-2000
standards. [0063] The present invention can be used in more
advanced receivers that support multiuser detection which requires
knowledge of each user's channel response. In this case, the
present invention could be used to maintain good channel
coefficient estimates for each user when delay estimates for
different users change. [0064] The present invention as described
herein focused on direct-sequence code-division multiple-access
(DS-CDMA) which is used in second and third generation digital
cellular systems such as WCDMA, CDMA2000, and IS-95. DS-CDMA is
also used in certain WLAN systems. In addition, it should be noted
that the present invention can be used to solve the problem of
channel tracking with changing delay estimates that can occur in
narrowband systems and OFDM systems as well. [0065] It should be
appreciated that many components and details associated with the
receiver 100 described above are well known in the industry.
Therefore, for clarity, the description provided above omitted
those well known components and details that are not necessary to
understand the present invention.
[0066] Although several embodiments of the present invention have
been illustrated in the accompanying Drawings and described in the
foregoing Detailed Description, it should be understood that the
invention is not limited to the embodiments disclosed, but is
capable of numerous rearrangements, modifications and substitutions
without departing from the spirit of the invention as set forth and
defined by the following claims.
* * * * *