TEPZZ  5Z776B_T

(19)

(11) EP 2 250 776 B1

(12) EUROPEAN PATENT SPECIFICATION

(45) Date of publication and mention (51) Int Cl.:

of the grant of the patent: H04L 25/02 (2006.01)
26.12.2012 Bulletin 2012/52
(86) International application number:
(21) Application number: 09707254.0 PCT/EP2009/051018

(22) Date of filing: 29.01.2009 (87) International publication number:

WO 2009/098166 (13.08.2009 Gazette 2009/33)

(54) Whitening channel estimate by Cholesky factorization

Weissungskanalschätzung durch Cholesky-Faktorisierung

Procédé et appareil d’estimation de canal pour un traitement de signaux de communication

(84) Designated Contracting States: • LENARDI M ET AL: "A RAKE RECEIVER WITH
AT BE BG CH CY CZ DE DK EE ES FI FR GB GR INTRACELL INTERFERENCE CANCELLATION
HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL FOR A DS-CDMA SYNCHRONOUS DOWNLINK
PT RO SE SI SK TR WITH ORTHOGONAL CODES" VTC 2000-
SPRING. 2000 IEEE 51ST. VEHICULAR
(30) Priority: 07.02.2008 US 27492 TECHNOLOGY CONFERENCE PROCEEDINGS.
TOKYO, JAPAN, MAY 15-18, 2000; [IEEE
(43) Date of publication of application: VEHICULAR TECHNOLGY CONFERENCE], NEW
17.11.2010 Bulletin 2010/46 YORK, NY : IEEE, US, vol. CONF. 51, 15 May 2000
(2000-05-15), pages 430-434, XP000970655 ISBN:
(73) Proprietor: Telefonaktiebolaget LM Ericsson 978-0-7803-5719-8
(publ) • WANG HAI (ERICSSON (CHINA) COMPANY LTD)
164 83 Stockholm (SE) AND R RAMESH ET AL: "Approaches for fast,
adaptive, generalized rake reception"
(72) Inventor: CAIRNS, Douglas RESEARCH DISCLOSURE, MASON
Durham PUBLICATIONS, HAMPSHIRE, GB, vol. 475, no.
North Carolina 27705 (US) 41, 1 November 2003 (2003-11-01), XP007133089
ISSN: 0374-4353
(74) Representative: Åkerman, Mårten Lennart et al • GIRISH A REDEKAR ET AL: "Equivalence of the
Ericsson AB Linear Symbol-level MMSE and the Generalized
Patent Unit Mobile Platforms RAKE" TENCON 2005 2005 IEEE REGION 10,
Nya Vattentornet IEEE, PI, 1 November 2005 (2005-11-01), pages
221 83 Lund (SE) 1-3, XP031015743 ISBN: 978-0-7803-9311-0
• KAISER ET AL.: "Smart Antennas State of the Art"
(56) References cited: 1 January 2005 (2005-01-01), HINDAWI , USA ,
EP-A- 1 821 445 WO-A-01/61950 XP002558871 eqn (2.36) eqn (2.39) page 21 - page
WO-A-2006/067258 WO-A-2007/000620 22
• VENKATESAN S ET AL: "An iterative algorithm
• ESA TIIROLA, JUHA YLITALO: "Performance of for computing a spatial whitening filter" SIGNAL
Smart Antenna Receivers in WCDMA Uplink with PROCESSING ADVANCES IN WIRELESS
Spatially Coloured Interference" IST MOBILE COMMUNICATIONS, 2004 IEEE 5TH W ORKSHOP
COMMUNICATIONS SUMMIT 2001, [Online] 12 ON LISBON, PORTUGAL 11-14 JULY 2004,
EP 2 250 776 B1

September 2001 (2001-09-12), pages 1-15,
XP002558870 Retrieved from the Internet: URL:
http://www.telecomlab.oulu.fi/home/cou
rsematerial/MIMO_5_adapt_IRC.pdf> [retrieved
on 2009-12-03]
PISCATAWAY, NJ, USA,IEEE, 11 July 2004
(2004-07-11), pages 338-342, XP010806066 ISBN:
978-0-7803-8337-1

Note: Within nine months of the publication of the mention of the grant of the European patent in the European Patent
Bulletin, any person may give notice to the European Patent Office of opposition to that patent, in accordance with the
Implementing Regulations. Notice of opposition shall not be deemed to have been filed until the opposition fee has been
paid. (Art. 99(1) European Patent Convention).

Printed by Jouve, 75001 PARIS (FR) (Cont. next page)

 EP 2 250 776 B1

• BOCCUZZI J ET AL: "Adaptive antenna arrays

using sub-space techniques in a mobile radio
environment with flat fading and CCI",
VEHICULAR TECHNOLOGY CONFERENCE, 1999
IEEE 49TH HOUSTON, TX, USA 16-20 MAY 1999,
PISCATAWAY, NJ, USA,IEEE, US, vol. 1, 16 May
1999 (1999-05-16), pages 50-54, XP010341960,
DOI: 10.1109/VETEC.1999.778014 ISBN:
978-0-7803-5565-1

2
 EP 2 250 776 B1

Description

TECHNICAL FIELD

5 [0001] The present invention generally relates to communications signal processing, and particularly relates to channel
estimation in support of such processing.

BACKGROUND

10 [0002] Existing and future wireless communication standards focus on improving spectral efficiency and data through-
put. In terms of data throughput in particular, the developing wireless communication standards offer significantly im-
proved performance as compared to earlier standards, such as GSM, GPRS, IS-95, and CDMA2000. For example,
Release 7 of the Wideband Code Division Multiple Access (WCDMA) standards defines radio link peak data rates
approaching or exceeding 28.8 Mbits/s. Future radio standards are expected to continue increasing peak data rates.
15 [0003] Achieving anything close to the defined peak data rates in actual practice requires the use of high-quality signal
transmitters and receivers. For example, virtually all contemporary (and planned) wireless communication devices include
channel estimation processing, wherein "channel estimates" are generated to account for the distortion caused by
propagating an electromagnetic signal from the transmitter to the receiver. The quality of channel estimation at the
receiver directly affects its ability to recover transmitted information from the received signal with acceptably low error
20 rates.
[0004] Known efforts to improve channel estimation include the adoption of minimum mean square error estimation
techniques. For example, the co-pending and commonly assigned U.S. patent application, as filed on 6 Nov. 2007 and
assigned Application Ser. No. 11/935,604, discloses a minimum mean square error (MMSE) channel estimation process
that obtains the MMSE solution for medium channel coefficients as a function a medium coefficient correlation matrix,
25 an impairment correlation matrix, and (measured) net channel responses.
[0005] The document WO2007/000620 discloses a method for synchronization and channel estimation.

SUMMARY

30 [0006] Teachings presented herein present a "whitening" channel estimation method and apparatus that produce
high-quality net channel estimates for processing a received signal, such as a received CDMA signal.
[0007] The invention is as defined in the independent claims.

BRIEF DESCRIPTION OF THE DRAWINGS

35
[0008]

Fig. 1 is a block diagram of a channel estimator according to one or more embodiments taught herein, shown in the
context of an example communications receiver.
40 Fig. 2 is a logic flow diagram illustrating one embodiment of a method of channel estimation taught herein.
Fig. 3 is a logic flow diagram illustrating example details for the channel estimation method outlined in Fig. 2.
Fig. 4 is a block diagram of a wireless communication network base station and a corresponding mobile station,
either or both of which incorporate a channel estimator as taught herein.

45 DETAILED DESCRIPTION

[0009] Fig. 1 illustrates one embodiment of a receiver circuit 10 comprising a portion of a communications receiver
12. The receiver circuit 10 operates as an improved channel estimation circuit, and therefore is also referred to as the
"channel estimator 10." The word "improved" in this context denotes without limitation that the channel estimator 10
50 considers the effects of received signal interference in its channel estimation.
[0010] Broadly, the channel estimator 10 produces "whitened" channel estimates in which the effects of colored inter-
ference are substantially reduced. Such channel estimates are "improved" in quality with respect to non-whitened channel
estimates and these high-quality channel estimates improve the performance of the communications receiver 12 in
various areas of signal processing, such as received signal demodulation/decoding, received signal quality estimation,
55 etc. As a further, non-limiting point of improvement, the channel estimator 10 employs computationally efficient and
robust whitening techniques that have broad applicability in a range of communication receiver types.
[0011] To provide discussion context for the channel estimator 10 and the corresponding channel estimation method
implemented by it, Fig. 1 illustrates further functional receiver circuit elements. However, those skilled in the art will

3
 EP 2 250 776 B1

appreciate that the illustrated receiver circuit architecture is not limiting with respect to the channel estimator 10, and
that the channel estimator 10 (and/or its processing method) can be implemented in other receivers.
[0012] As for the illustrated example, one sees that the receiver 12 includes one or more receive antennas 14 for
receiving a communication signal from a remote transmitter (not shown). Generally, this antenna-received signal, e.g.,
5 a Wideband Code Division Multiple Access (WCDMA) signal, is a multipath signal and includes desired and interfering
signal components received on a plurality of propagation channels. As will be further understood, the antenna-received
signal may comprise a plurality of code-channelized signals, including pilot channel signals carrying pilot information
and traffic channel signals carrying traffic data.
[0013] The receiver 12 further includes a radiofrequency (RF) front-end circuit 16, that down-converts, digitizes, and
10 filters the antenna-received signal as needed or desired. A delay processor 18 processes the digitized signal, referred
to as the "received signal r" for purposes of this description, to identify the processing delays to be used for multipath
processing of the received signal. For example, the delay processor 18 may include a path searcher function that
correlates the received signal to a particular pilot or other correlation sequence, to identify correlation peaks at signal
delays corresponding to the strongest multipath components. As is known for Generalized Rake (G-Rake) and chip
15 equalization (CE) receivers, the delay processor 18 uses the identified path delays to determine the particular processing
delays to be used for multipath processing of the received signal, e.g., it calculates the G-Rake finger or CE filter tap
delays to be used for collecting desired signal and interfering signal energy from the received signal.
[0014] The receiver 12 further includes a demodulator/detector 20, which may be implemented as a G-Rake based
or CE based circuit, and which is operative to detect transmitted pilot and traffic data symbols from the received signal.
20 As is known in the art, G-Rake receivers include a number of "Rake fingers" (correlators) that are placed at selected
processing delays for recovering pilot and traffic data from the received signal. Some of these fingers are used to collect
desired signal energy and some are used to collect interfering signal energy. Finger output signals are combined using
combining weights that consider channel effects and interference effects. Similar processing is known for use in chip
equalization embodiments, wherein an equalization filter precedes a single correlator. The equalization filter includes a
25 series of delay stages, with each stage providing an output ("tap"). Channel equalization and interference suppression
are achieved in this context by combining selected ones of the filter taps using appropriately calculated combining
weights, as in the G-Rake case.
[0015] The receiver 12 further includes a decoder 22, which is operative to decode transmitted data from the demod-
ulated symbols output by the demodulator/detector 20. For example, traffic data and control signaling information encoded
30 on various code channels within the received signal are recovered for processing by one or more additional baseband
(BB) processing circuits 24.
[0016] Within this example context, the channel estimator 10 provides "improved" channel estimates. These improved
channel estimates are used, for example, in the demodulation process implemented in the demodulator/detector 20.
More particularly, the improved channel estimates produced by the-channel estimator 10 may be used in computing
35 Rake/CE combining weights, received signal SIR estimation, etc. In one or more embodiments, the channel estimator
10 comprises digital processing circuits, including a least squares (LS) processor 29 that includes an LS solver 30 and
a whitening term calculator 32, a data correlation processor 34, and a conversion matrix processor 36. These functional
circuits may be implemented in hardware, software, or any combination thereof.
[0017] In one embodiment, a general-purpose or special-purpose microprocessor (e.g., a DSP) executes stored pro-
40 gram instructions from a computer program held in a computer-readable medium, e.g., a memory device included in or
accessible by the receiver 12. In another embodiment, at least a portion of the channel estimator’s functionality is
implemented in dedicated signal processing hardware. In either case, it should be understood that the channel estimator
10 preferably comprises functional digital processing elements that can be integrated with other baseband signal process-
ing hardware and/or software.
45 [0018] In a more detailed discussion of the channel estimator 10, it may be helpful to begin with an overview of
terminology. Two types of channel coefficients will be referred to throughout the disclosure. The first type, "medium
coefficients," refers to the complex values associated with the (multipath) radio channel. The second type, "net coeffi-
cients," includes the effects of transmit filtering at the remote transmitter (not shown) and receiver filtering at the receiver
12, in addition to the medium coefficients. The medium coefficients are denoted by g, where g is a KLx1 vector of complex
50 values associated with L radio channel paths across K antennas. It is assumed that the coefficients corresponding to

each antenna are stacked according to where gk denotes the vector of medium coefficients

corresponding to the kth antenna. The net coefficients are denoted by h, where h is a KMx1 vector of complex values
55 (M ≥ L) associated with the M receiver fingers across K antennas. Similar to g, h is a composite vector formed by stacking

the individual antenna contributions One can convert from medium to net coefficients via

4
 EP 2 250 776 B1

5 where B is an KMxKL block-diagonal conversion matrix

10

15
The (i, j) element of Bk is given by

20

Here, RTX / RX (λ) is the convolution of the transmit and receive filters evaluated at λ, di is the delay of the ith receiver
finger, and τj is the jth path delay. The conversion from net to medium coefficients is given by

25

30 In one or more embodiments, the channel estimator 10 is configured to form a least squares estimation problem using,
e.g., data from a pilot channel in the received signal. The least squares problem is then transformed using a whitening
step, and the whitened least squares problem is then solved to give high-quality channel estimates. In this context,
"forming" a least squares problem connotes the channel estimator 10 setting up a least squares numerical computation
based on a model matrix having channel model element values, a coefficients matrix having (unknown) coefficients to
35 be solved for, and an observation matrix having observed pilot/channel element values.
[0019] To understand the channel estimation method taught herein, first consider the standard model for pilot despread
symbols, which is expressed as

40

Here, gpilot is a vector of medium channel coefficients scaled by the square root of the pilot symbol energy

45 is the ith known pilot symbol, and u(i) is a vector of impairment (interference plus noise)

samples for the ith symbol in a current transmission time period. For a WCDMA signal, that period may be, for example,
a defined transmission time slot within any given transmission time interval (TTI). To form a least squares problem here,
the channel estimator 10 is configured to "create" the following model and observation matrices, respectively, from the
50 Q pilot symbol intervals in the current slot

55

5
 EP 2 250 776 B1

10 and

15

20

Note that the pilot observation matrix b is a vector of observed pilot values for the received signal over a given slot or
other defined processing time period. Also, note that if spilot(0)=spilot(1)=··· =spilot(Q-1), then the channel estimator 10

25
could alternatively be formed from A = Bspilot(0) and

[0020] With the above formulation, the initial LS problem formed by the channel estimator 10 can be derived from or
represented according to the following expression,
30

where A is the channel model matrix, b is the observation matrix, and g is a vector of medium channel coefficients matrix
35 to be solved for via least squares processing. The solution for g minimizes the Euclidean norm squared of the residual
Ag-b.
[0021] However, rather than solving Eq. (7) for the medium channel coefficients g, the channel estimation method
taught herein applies a whitening transformation to A and b. The theoretically correct whitening transformation is the
inverse of the Cholesky factor of the impairment covariance matrix Ru, which can be estimated, for example, by taking
40 cross correlations between pilot values across different processing delays/antennas. However, there may be advantages
in approximating the impairment covariance matrix by a data correlation matrix. An estimate of the data correlation matrix
can be obtained using one or more channelization codes of the desired signal traffic channel in the received signal via

45

50
where is a vector of despread traffic symbols corresponding to the pth code for the kth symbol time. The

estimate of the data correlation matrix is Cholesky factorized

55

and the inverse of either the lower triangular or upper triangular factor is computed. The inverse Cholesky factor is

6
 EP 2 250 776 B1

applied to A and b to give

[0022] These transformed matrices can be used to compute high quality (whitened) medium channel estimates via
10

15
The corresponding high quality net channel estimates can then be computed from the whitened medium channel estimates
as

20

These high quality net channel coefficients can then be used by the receiver 12 in various aspects of received signal
processing, such as in demodulation, SIR estimation, etc.
25 [0023] Fig. 2 illustrates one embodiment of channel estimation according to the above processing operations. Such
processing can be implemented in the receiver circuit 10 via hardware, software, or any combination thereof. In general,
as noted, the receiver circuit 10 includes one or more processing circuits, e.g., the LS solver 30, the whitening term
calculator 32, etc. Whether the illustrated functional circuits are implemented or alternative arrangements are adopted,
the one or more processing circuits of the channel estimator 10 are configured according to Fig. 2 to produce high quality,
30 whitened net channel estimates according to the following processing steps.
[0024] The channel estimator 10 forms an initial least squares problem for medium channel estimates based on known
pilot values and corresponding pilot observations for the received signal (Step 100). Processing continues with calculating
a whitening transformation term based on a Cholesky factorization of a traffic data correlation matrix determined for the
received signal (Step 102), and with applying the whitening transformation term to the initial least squares problem to
35 obtain a transformed least squares problem (Step 104). Processing continues with solving the transformed least squares
problem to obtain whitened medium channel estimates for the received signal (Step 106), and converting the whitened
medium channel estimates to whitened net channel estimates for processing the received signal by applying a medium-
to-net conversion matrix to the whitened medium channel estimates (Step 108).
[0025] Forming the initial least squares problem for medium channel estimates based on known pilot values and
40 corresponding pilot observations for the received signal may include obtaining the pilot observations as pilot values
obtained from the received signal for all received signal processing delays of interest on each of one or more receive
antennas. More particularly, in a G-Rake embodiment, the pilot observations correspond to despread pilot symbols taken
at all processing delays of interest on each of one or more receive antennas in use.
[0026] Similarly, the traffic data correlation matrix may be estimated from despread traffic data values obtained for
45 one or more traffic channelization codes used in the received signal. With that, calculating the whitening transformation
term based on a Cholesky factorization of a traffic data correlation matrix determined for the received signal may include
estimating the traffic data correlation matrix from (despread) traffic data values obtained from the received signal and
determining a Cholesky factor of the traffic data correlation matrix.
[0027] In any case, it should be understood that in one or more embodiments the received signal comprises a CDMA
50 signal comprising pilot data and traffic data on different channelization codes. Therefore, forming an initial least squares
problem for medium channel estimates based on known pilot values and corresponding pilot observations for the received
signal may comprise forming a model matrix representing a channel model based on the conversion matrix and known
pilot values for the CDMA signal and forming an observation matrix based on despread pilot values obtained from the
received signal.
55 [0028] Also, as noted, the initial least squares problem of Step 100 in Fig. 2 includes a model matrix A representing
a channel model, an observation matrix b representing the pilot observations, and a coefficients matrix g for fitting the
model matrix to the observations. Thus, applying the whitening transformation term to the initial least squares problem
to obtain a transformed least squares problem comprises, in one or more embodiments, applying the Cholesky factor

7
 EP 2 250 776 B1

to the model and observation matrices. In at least one such embodiment, that step comprises applying the inverse
~
Cholesky factor to the model matrix to obtain a transformed model matrix A, and applying the Cholesky factor to the
~
observation matrix to obtain a transformed observation matrix b for use in the transformed least squares problem. In
another such embodiment, applying the Cholesky factor to the model and observation matrices comprises expressing
5 the transformed model matrix as a function of the Cholesky factor and the model matrix and solving for the transformed
model matrix, and expressing the transformed observation matrix as a function of the Cholesky factor and the observation
matrix and solving for the transformed observation matrix.
[0029] Still further, as was noted regarding Eq. (2), calculating the medium-to-net conversion matrix is, in one or more
embodiments, based on known transmit/receive filter characteristics and on differences between received signal process-
10 ing delays and received signal multipath delays. The transmit and receive filters-e.g., transmit and receive pulse-shaping
filters-may be the same or different. However, in either case, the filter characteristics are known to the channel estimator
10. For example, the channel estimator 10 may hold filter parameters or other values in memory representing the transmit/
receive filter characteristics and make these values accessible to the conversion matrix processor 36. Further, the
conversion matrix processor 36 is provided with multipath and processing delay values representing the multipath delays
15 of interest and the corresponding processing delays (e.g., Rake finger or CE tap delays) being used to process the
multipath received signal. In terms of converting the whitened medium channel estimates, one or more embodiments
obtain the whitened net channel estimates by calculating them as a product of the conversion matrix and the whitened
medium channel estimates.
[0030] Fig. 3 illustrates a more detailed processing flow, representing a preferred embodiment of the processing
20 introduced in Fig. 2, set in the context of a Rake receiver embodiment of the receiver 12. Processing includes forming
the medium-to-net conversion matrix B (Step 120), e.g., according to Eq. (2), obtaining despread pilot values for all
(Rake) fingers across all (active) receiver antennas 14 (Step 122), and forming the initial LS problem matrices (Step
124), e.g., according to Eq. (5) and Eq. (6). Processing continues with obtaining despread traffic values for all fingers
^
across all antennas for one or more codes (Step 126), forming an estimate of the traffic data correlation matrix Rd (Step
25 128), e.g., according to Eq. (8). Processing continues with Cholesky factorization of the estimated traffic data correlation
^ ~ ~
matrix Rd to obtain G (Step 130), and use of the Cholesky factor to obtain the transformed LS matrices A and b (Step
132). The transformation may be accomplished by inverting G, such as is done in Eq. (10). Alternatively, the channel
~ ~ ~ ~
estimator 10 may be configured to solve GA = A and Gb = b to obtain A and b. There may be an advantage to this latter
configuration because G is lower triangular.
30 [0031] Processing continues with solving the transformed LS problem (Step 134), e.g., according to Eq. (11), which
produces the whitened medium channel estimates gwhitened. Still further, processing continues with computing the whit-
ened net channel estimates (Step 136) hwhitened from gwhitened, e.g., according to Eq. (12).
[0032] Of course, except where indicated (or necessary), the teachings herein are not limited to the particular order
or sequence of processing steps illustrated in Figs. 2 or 3. Indeed, it should be understood that other processing step
35 orders are contemplated herein, and that one or more processing steps may be carried out in parallel. Indeed, the
processing steps of Figs. 2 and 3 may be carried out as part of or along with other received signal processing operations.
For example, the high-quality channel estimation taught herein may be carried out (and repeated) on a per-slot/per-TTI
basis, and may represent a part of overall received signal processing for received signal demodulation/decoding, etc.
[0033] For a further example context, Fig. 4 illustrates a wireless communication base station 40, such as a WCDMA
40 base station configured for use within a WCDMA-based cellular communication network. The example base station 40
includes a base station transmission system 42, which is configured to transmit one or more CDMA signals. Such signals
carry, for example, overhead and control signaling, common/dedicated pilot information, and traffic data signals for
potentially many remote mobile stations. The base station transmission system includes transmit processing circuits 44,
transmit filtering circuits 46, RF signal generation circuits 48, and transmit/receive antennas 50.
45 [0034] The base station 40 further includes a base station reception system 52, which is configured to receive uplink
signals on the transmit/receive antennas 50 from a plurality of remote mobile stations. The illustrated reception system
52 includes RF signal reception circuits 54, for front-end signal processing, RX filtering circuits 56, and RX processing
circuits 58. Those skilled in the art will appreciate that the base station 40 also includes other circuits/systems not
illustrated. For example, the base station 40 may include various core network interfaces, sidehaul interfaces to other
50 base stations, various call control/processing circuits, etc.
[0035] For purposes of this discussion, it will be understood that the illustrated base station architecture is exemplary
and that the previously described channel estimator 10 in any of its various embodiments may be implemented within
the RF reception circuits 54, or, more conveniently, within the (baseband) RX processing circuits 58. Further, the channel
estimator 10 may be duplicated in hardware and/or software for use in estimating channels for each in the plurality of
55 mobile stations being supported on the base station’s uplink. Such processing may yield particular improvements in
base station processing performance where uplink interference cancellation is practiced. That is, the whitened net
channel estimates produced by the channel estimator 10 are particularly useful in, for example, successive interference
cancellation (SIC) receivers. In SIC receivers, interfering signal components are regenerated and subtracted and that

8
 EP 2 250 776 B1

signal regeneration is based on the corresponding channel estimates. Thus, higher quality channel estimates yields
higher quality regenerated signals and more accurate signal subtraction.
[0036] The channel estimator 10 also may be additionally or alternatively implemented in the mobile station 60 illustrated
in Fig. 4. As with the base station 40, the mobile station 60 is, in a non-limiting example, a cellular radiotelephone or
5 other mobile communication device that is configured for operation within a cellular communication network. For example,
the mobile station 60 comprises a WCDMA device configured for operation in a WCDMA network that includes the base
station 40.
[0037] The illustrated embodiment of the mobile station 60 includes a transmitter 62, which includes a transmit process-
ing circuit 64, a transmit filtering circuit 66, RF signal generation circuits 68 for signal transmission, and associated TX/RX
10 antennas 70. The mobile station 60 further includes a receiver 72, which includes RF signal reception circuits 74, RX
filtering circuits 76, and RX processing circuits 78. As with the base station 40, the mobile station illustration is simplified,
and an actual mobile station 60 may include further processing circuits, user interface circuits, etc., depending upon its
intended use.
[0038] In any case, those skilled in the art will appreciate that the previously described channel estimator 10 can be
15 implemented, for example, in the (baseband) RX processing circuits 78 of the mobile station 60. Broadly, the channel
estimator 10 may be implemented in a variety of wireless communication devices systems, such as wireless communi-
cation network base stations for channel estimation supporting uplink signal processing, and in wireless communication
network mobile stations for channel estimation supporting downlink signal processing. Such mobile stations encompass
without limitation cellular telephones, wireless pagers, PDAs, network access cards or modules, wireless-enabled com-
20 puters, etc.
[0039] Further, those skilled in the art will appreciate the broad applicability of the channel estimator 10 (and its
embodied methods) to other signal processing devices and systems, both wireless and wired. As such, it should be
understood that the foregoing description and the accompanying drawings represent non-limiting examples of the meth-
ods, systems, and individual apparatuses taught herein. The present invention therefore is not limited by the foregoing
25 description and accompanying drawings. Instead, the present invention is limited only by the following claims and their
legal equivalents.

Claims
30
1. A method of channel estimation for a received signal characterized by:

forming an initial least squares problem for medium channel estimates based on known pilot values and corre-
sponding pilot observations for the received signal including forming a model matrix representing a channel
35 model based on a conversion matrix for medium-to-net conversion and known pilot values for the received
signal and forming an observation matrix based on pilot values obtained from the received signal;
calculating a whitening transformation term based on a Cholesky factorization of a traffic data correlation matrix
determined for the received signal;
applying the whitening transformation term to the initial least squares problem to obtain a transformed least
40 squares problem;
solving the transformed least squares problem to obtain whitened medium channel estimates for the received
signal; and
converting the whitened medium channel estimates to whitened net channel estimates for processing the re-
ceived signal by applying the medium-to-net conversion matrix to the whitened medium channel estimates,
45
wherein medium channel refers to radio channel and net channel refers to effects, in addition to medium channel,
of transmit filtering at a remote transmitter and receiver filtering at receiver.

2. The method of claim 1, wherein forming an initial least squares problem for medium channel estimates based on
50 known pilot values and corresponding pilot observations for the received signal includes obtaining the pilot obser-
vations as pilot values obtained from the received signal for all received signal processing delays of interest on each
of one or more receive antennas.

3. The method of claim 1, wherein calculating a whitening transformation term based on a Cholesky factorization of a
55 traffic data correlation matrix determined for the received signal includes estimating the traffic data correlation matrix
from traffic data values obtained from the received signal and determining a Cholesky factor of the traffic data
correlation matrix.

9
 EP 2 250 776 B1

4. The method of claim 3, wherein the initial least squares problem includes a model matrix representing a channel
model, an observation matrix representing the pilot observations, and a coefficients matrix for fitting the model matrix
to the observation, and wherein applying the whitening transformation term to the initial least squares problem to
obtain a transformed least squares problem comprises applying the Cholesky factor to the model and observation
5 matrices by applying the inverse Cholesky factor to the model matrix to obtain a transformed model matrix for use
in the transformed least squares problem, and applying the inverse Cholesky factor to the observation matrix to
obtain a transformed observation matrix for use in the transformed least squares problem.

5. The method of claim 4, wherein applying the Cholesky factor to the model and observation matrices comprises
10 expressing a transformed model matrix as a function of the Cholesky factor and the model matrix and solving for
the transformed model matrix for use in the transformed least squares problem, and expressing a transformed
observation matrix as a function of the Cholesky factor and the observation matrix and solving for the transformed
observation matrix for use in the transformed least squares problem.

15 6. The method of claim 1, further comprising calculating the medium-to-net conversion matrix based on known transmit/
receive filter characteristics and differences between received signal processing delays and received signal multipath
delays.

7. The method of claim 1, wherein the received signal comprises a Code Division Multiple Access (CDMA) signal
20 comprising pilot data and traffic data on different channelization codes, and wherein the forming of the model matrix
representing the channel model is based on the conversion matrix and known pilot values for the CDMA signal and
forming the observation matrix is based on despread pilot values obtained from the received signal.

8. The method of claim 7, wherein calculating a whitening transformation term based on a Cholesky factorization of a
25 traffic data correlation matrix determined for the received signal includes estimating the traffic data correlation matrix
from despread traffic data values obtained for one or more traffic channelization codes used in the received signal
and calculating the whitening transformation term as the Cholesky factor of the traffic data correlation matrix.

9. A receiver circuit (10) configured to generate channel estimates for a received signal received by an associated
30 communications receiver (12), said receiver circuit (10) characterized by one or more processing circuits (29, 30,
32, 34, 36) configured to:

form an initial least squares problem for medium channel estimates based on known pilot values and corre-
sponding pilot observations for the received signal including to form a model matrix representing a channel
35 model based on a conversion matrix for medium-to-net conversion and known pilot values for the received
signal and forming an observation matrix based on pilot values obtained from the received signal;
calculate a whitening transformation term based on a Cholesky factorization of a traffic data correlation matrix
determined for the received signal;
apply the whitening transformation term to the initial least squares problem to obtain a transformed least squares
40 problem;
solve the transformed least squares problem to obtain whitened medium channel estimates for the received
signal; and
convert the whitened medium channel estimates to whitened net channel estimates for processing the received
signal by applying the medium-to-net conversion matrix to the whitened medium channel estimates,
45
wherein medium channel refers to radio channel and net channel refers to effects, in addition to medium channel,
of transmit filtering at a remote transmitter and receiver filtering at receiver.

10. The receiver circuit (10) of claim 9, further characterized in that the receiver circuit (10) is configured to obtain the
50 pilot observations as pilot values obtained from the received signal for all received signal processing delays of
interest on each of one or more receive antennas (14) of the associated communications receiver (12).

11. The receiver circuit (10) of claim 9, further characterized in that the receiver circuit (10) is configured to estimate
the traffic data correlation matrix from traffic data values obtained from the received signal and determine a Cholesky
55 factor of the traffic data correlation matrix for use as the whitening transformation term, and the initial least squares
problem includes a model matrix representing a channel model, an observation matrix representing the pilot obser-
vations, and a coefficients matrix for fitting the model matrix to the observation, and the receiver circuit (10) is
configured to apply the Cholesky factor to the model and observation matrices to obtain the transformed least

10
 EP 2 250 776 B1

squares problem.

12. The receiver circuit (10) of claim 11, further characterized in that the receiver circuit (10) is configured to apply the
Cholesky factor to the model and observation matrices based on applying the inverse Cholesky factor to the model
5 matrix to obtain a transformed model matrix for use in the transformed least squares problem, and applying the
inverse Cholesky factor to the observation matrix to obtain a transformed observation matrix for use in the transformed
least squares problem.

13. The receiver circuit (10) of claim 11, further characterized in that the receiver circuit (10) is configured to apply the
10 Cholesky factor to the model and observation matrices based on expressing a transformed model matrix as a function
of the Cholesky factor and the model matrix and solving for the transformed model matrix, and expressing a trans-
formed observation matrix as a function of the Cholesky factor and the observation matrix.

14. The receiver circuit (10) of claim 9, further characterized in that the receiver circuit (10) is configured to calculate
15 the medium-to-net conversion matrix based on known transmit/receive filter characteristics and time differences
between received signal processing delays and received signal multipath delays, and wherein the received signal
processing delays comprise one of Rake finger correlation delays in a Rake receiver circuit (12) and equalization
filter tap delays in a chip equalization receiver circuit (12).

20 15. The receiver circuit (10) of claim 9, further characterized in that the receiver circuit (10) comprises all or part of a
Rake or chip equalization receiver circuit within the communications receiver (12) of a wireless communications
device (60), and wherein the one or more processing circuits (29, 30, 32, 34, 36) comprise one or more digital
processing circuits including a conversion matrix processor (36) configured to calculate the conversion matrix, a
least squares processor (29, 30, 32) configured to form the initial and transformed least squares problems and to
25 solve the transformed least squares problem for the whitened medium channel estimates, and a correlation processor
(34) configured to generate the data correlation matrix.

Patentansprüche
30
1. Verfahren zur Kanalschätzung für ein Empfangssignal, gekennzeichnet durch:

Bilden eines Ausgangsproblems der kleinsten Quadrate für Medienkanalschätzungen basierend auf bekannten
Pilotwerten und entsprechenden Pilotbeobachtungen für das Empfangssignal, umfassend das Bilden einer
35 Modellmatrix, die ein Kanalmodell darstellt, basierend auf einer Konversionsmatrix zur Medium-Netz-Konversion
und bekannten Pilotwerten für das Empfangssignal und Bilden einer Beobachtungsmatrix basierend auf Pilot-
werten, die aus dem Empfangssignalen erhalten werden;
Berechnen eines Weißungstransformationsterms basierend auf einer Cholesky-Faktorisierung einer Verkehrs-
daten-Korrelationsmatrix, die für das Empfangssignal bestimmt wird;
40 Anwenden des Weißungstransformationsterms auf das Ausgangsproblem der kleinsten Quadrate, um ein trans-
formiertes Problem der kleinsten Quadrate zu erhalten;
Lösen des transformierten Problems der kleinsten Quadrate, um geweißte Medienkanalschätzungen für das
Empfangssignal zu erhalten;
Konvertieren der geweißten Medienkanalschätzungen in geweißte Netzkanalschätzungen zum Verarbeiten des
45 Empfangssignals durch Anwenden der Medium-Netz-Konversionsmatrix auf die geweißten Medienkanalschät-
zungen,
wobei sich Medienkanal auf einen Funkkanal bezieht und Netzkanal auf Wirkungen, zusätzlich zu Medienkanal,
von Sendefilterung am einem Remotesender und Empfängerfilterung an einem Empfänger bezieht.

50 2. Verfahren nach Anspruch 1, wobei das Bilden eines Ausgangsproblems der kleinsten Quadrate für Medienkanal-
schätzungen basierend auf bekannten Pilotwerten und entsprechenden Pilotbeobachtungen für das Empfangssignal
ein Erhalten der Pilotbeobachtungen als Pilotwerte umfasst, die aus dem Empfangssignal für alle Empfangssignal-
Verarbeitungsverzögerungen von Interesse an jeder von einer oder mehreren Empfangsantennen erhalten werden.

55 3. Verfahren nach Anspruch 1, wobei das Berechnen eines Weißungstransformationsterms basierend auf einer Cho-
lesky-Faktorisierung einer Verkehrsdaten-Korrelationsmatrix, die für das Empfangssignal bestimmt wird, ein Schät-
zen der Verkehrsdaten-Korrelationsmatrix aus Verkehrsdatenwerten, die aus dem Empfangssignal erhalten werden,
und Bestimmen eines Cholesky-Faktors der Verkehrsdaten-Korrelationsmatrix umfasst.

11
 EP 2 250 776 B1

4. Verfahren nach Anspruch 3, wobei das Ausgangsproblem der kleinsten Quadrate eine Modellmatrix, die ein Kanal-
modell darstellt, eine Beobachtungsmatrix, welche die Pilotbeobachtungen darstellt, und eine Koeffizientenmatrix
zum Anpassen der Modellmatrix an die Beobachtung umfasst, und wobei das Anwenden des Weißungstransfor-
mationsterms auf das Ausgangsproblem der kleinsten Quadrate, um ein transformiertes Problem der kleinsten
5 Quadrate zu erhalten, ein Anwenden des Cholesky-Faktors auf die Modell- und Beobachtungsmatrizen durch An-
wenden des inversen Cholesky-Faktors auf die Modellmatrix, um eine transformierte Modellmatrix zur Verwendung
im transformierten Problem der kleinsten Quadrate zu erhalten, und Anwenden des inversen Cholesky-Faktors auf
die Beobachtungsmatrix umfasst, um eine transformierte Beobachtungsmatrix zur Verwendung im transformierten
Problem der kleinsten Quadrate zu erhalten.
10
5. Verfahren nach Anspruch 4, wobei das Anwenden des Cholesky-Faktors auf die Modell- und Beobachtungsmatrizen
ein Ausdrücken einer transformierten Modellmatrix als eine Funktion des Cholesky-Faktors und der Modellmatrix
und Lösen für die transformierte Modellmatrix zur Verwendung im transformierten Problem der kleinsten Quadrate
und ein Ausdrücken einer transformierten Beobachtungsmatrix als eine Funktion des Cholesky-Faktors und der
15 Beobachtungsmatrix und Lösen für die transformierte Beobachtungsmatrix zur Verwendung im transformierten
Problem der kleinsten Quadrate umfasst.

6. Verfahren nach Anspruch 1, ferner umfassend ein Berechnen der Medium-Netz-Konversionsmatrix basierend auf
bekannten Sende-/Empfangsfiltercharakteristiken und Differenzen zwischen Empfangssignal-Verarbeitungsverzö-
20 gerungen und Empfangssignal-Mehrwegverzögerungen.

7. Verfahren nach Anspruch 1, wobei das Empfangssignal ein Codemultiplexzugriffs (CDMA)-Signal umfasst, das
Pilotdaten und Verkehrsdaten auf verschiedenen Kanalbildungscodes umfasst, und wobei das Bilden der Modell-
matrix, die das Kanalmodell darstellt, auf der Konversionsmatrix und bekannten Pilotwerten für das CDMA-Signal
25 basiert, und das Bilden der Beobachtungsmatrix auf entspreizten Pilotwerten basiert, die aus dem Empfangssignal
erhalten werden.

8. Verfahren nach Anspruch 7, wobei das Berechnen eines Weißungstransformationsterms basierend auf einer Cho-
lesky-Faktorisierung einer Verkehrsdaten-Korrelationsmatrix, die für das Empfangssignal bestimmt wird, ein Schät-
30 zen der Verkehrsdaten-Korrelationsmatrix aus entspreizten Verkehrsdatenwerten, die für einen oder mehrere Ver-
kehrskanalbildungscodes erhalten werden, die im Empfangssignal verwendet werden, und Berechnen des Wei-
ßungstransformationsterms als den Cholesky-Faktors der Verkehrsdaten-Korrelationsmatrix umfasst.

9. Empfängerschaltung (10), die so konfiguriert ist, dass sie Kanalschätzungen für ein Empfangssignal erzeugt, das
35 durch einen zugehörigen Kommunikationsempfänger (12) empfangen wird, wobei die Empfängerschaltung (10)
durch eine oder mehrere Verarbeitungsschaltungen (29, 30, 32, 34, 36) gekennzeichnet ist, die konfiguriert sind zum:

Bilden eines Ausgangsproblems der kleinsten Quadrate für Medienkanalschätzungen basierend auf bekannten
Pilotwerten und entsprechenden Pilotbeobachtungen für das Empfangssignal, einschließlich zum Bilden einer
40 Modellmatrix, die ein Kanalmodell darstellt, basierend auf einer Konversionsmatrix zur Medium-Netz-Konversion
und bekannten Pilotwerten für das Empfangssignal und Bilden einer Beobachtungsmatrix basierend auf Pilot-
werten, die aus dem Empfangssignalen erhalten werden;
Berechnen eines Weißungstransformationsterms basierend auf einer Cholesky-Faktorisierung einer Verkehrs-
daten-Korrelationsmatrix, die für das Empfangssignal bestimmt wird;
45 Anwenden des Weißungstransformationsterms auf das Ausgangsproblem der kleinsten Quadrate, um ein trans-
formiertes Problem der kleinsten Quadrate zu erhalten;
Lösen des transformierten Problems der kleinsten Quadrate, um geweißte Medienkanalschätzungen für das
Empfangssignal zu erhalten;
Konvertieren der geweißten Medienkanalschätzungen in geweißte Netzkanalschätzungen zum Verarbeiten des
50 Empfangssignals durch Anwenden der Medium-Netz-Konversionsmatrix auf die geweißten Medienkanalschät-
zungen,
wobei sich Medienkanal auf einen Funkkanal bezieht und Netzkanal auf Wirkungen, zusätzlich zu Medienkanal,
von Sendefilterung am einem Remotesender und Empfängerfilterung an einem Empfänger bezieht.

55 10. Empfängerschaltung (10) nach Anspruch 9, ferner dadurch gekennzeichnet, dass die Empfängerschaltung (10)
so konfiguriert ist, dass sie die Pilotbeobachtungen als Pilotwerte erhält, die aus dem Empfangssignal für alle
Empfangssignal-Verarbeitungsverzögerungen von Interesse an jeder von einer oder mehreren Empfangsantennen
(14) des zugehörigen Kommunikationsempfängers (12) erhalten werden.

12
 EP 2 250 776 B1

11. Empfängerschaltung (10) nach Anspruch 9, ferner gekennzeichnet dadurch, dass die Empfängerschaltung (10)
so konfiguriert ist, dass sie die Verkehrsdaten-Korrelationsmatrix aus Verkehrsdatenwerten schätzt, die aus dem
Empfangssignal erhalten werden, und einen Cholesky-Faktor der Verkehrsdaten-Korrelationsmatrix zur Verwen-
dung als den Weißungstransformationsterm bestimmt, und das Ausgangsproblem der kleinsten Quadrate eine
5 Modellmatrix, die ein Kanalmodell darstellt, eine Beobachtungsmatrix, welche die Pilotbeobachtungen darstellt, und
eine Koeffizientenmatrix zum Anpassen der Modelmatrix an die Beobachtung umfasst, und die Empfängerschaltung
(10) so konfiguriert ist, dass sie den Cholesky-Faktors auf die Modell- und Beobachtungsmatrizen anwendet, um
das transformierte Problem der kleinsten Quadrate zu erhalten.

10 12. Empfängerschaltung (10) nach Anspruch 11, ferner gekennzeichnet dadurch, dass die Empfängerschaltung (10)
so konfiguriert ist, dass sie den Cholesky-Faktor auf die Modell- und Beobachtungsmatrizen basierend auf einem
Anwenden des inversen Cholesky-Faktors auf die Modellmatrix, um eine transformierte Modellmatrix zur Verwen-
dung im transformierten Problem der kleinsten Quadrate, und Anwenden des inversen Cholesky-Faktors auf die
Beobachtungsmatrix anwendet, um eine transformierte Beobachtungsmatrix zur Verwendung im transformierten
15 Problem der kleinsten Quadrate zu erhalten.

13. Empfängerschaltung (10) nach Anspruch 11, ferner gekennzeichnet dadurch, dass die Empfängerschaltung (10)
so konfiguriert ist, dass sie den Cholesky-Faktor auf die Modell- und Beobachtungsmatrizen basierend auf einem
Ausdrücken einer transformierten Modellmatrix als eine Funktion des Cholesky-Faktors und der Modellmatrix und
20 Lösen für die transformierte Modellmatrix und Ausdrücken einer transformierten Beobachtungsmatrix als eine Funk-
tion des Cholesky-Faktors und der Beobachtungsmatrix anwendet.

14. Empfängerschaltung (10) nach Anspruch 9, ferner gekennzeichnet dadurch, dass die Empfängerschaltung (10)
so konfiguriert ist, dass sie die Medium-Netz-Konversionsmatrix basierend auf bekannten Sende-/Empfangsfilter-
25 charakteristiken und Zeitdifferenzen zwischen Empfangssignal-Verarbeitungsverzögerungen und Empfangssignal-
Mehrwegverzögerungen berechnet, und wobei die Empfangssignal-Verarbeitungsverzögerungen eines von RAKE-
Finger-Korrelationsverzögerungen in einer RAK-Empfängerschaltung (12) und Entzerrungsfilter-Abgriffsverzöge-
rungen in einer Chipentzerrungs-Empfängerschaltung (12) umfassen.

30 15. Empfängerschaltung (10) nach Anspruch 9, ferner gekennzeichnet dadurch, dass die Empfängerschaltung (10)
einen Teil oder die Gesamtheit einer RAKE- oder Chipentzerrungs-Empfängerschaltung innerhalb des Kommuni-
kationsempfängers (12) einer drahtlosen Kommunikationsvorrichtung (60) umfasst, und wobei die eine oder die
mehreren Verarbeitungsschaltungen (29, 30, 32, 34, 36) eine oder mehrere digitale Verarbeitungsschaltungen
umfassen, die einen Konversionsmatrixprozessor (36), der so konfiguriert ist, dass er die Konversionsmatrix be-
35 rechnet, einen Prozessor (29, 30, 32) der kleinsten Quadrate, der so konfiguriert ist, dass er die Ausgangs- und
transformierten Probleme der kleinsten Quadrate für die geweißten Medienkanalschätzungen bildet, und einen
Korrelationsprozessor (34) umfassen, der so konfiguriert ist, dass er die Datenkorrelationsmatrix erzeugt.

40 Revendications

1. Procédé d’estimation de canal pour un signal reçu caractérisé par :

former un problème de moindres carrés initial pour les estimations de canal moyennes sur la base des valeurs
45 pilotes connues et des observations pilotes correspondantes pour le signal reçu incluant de former une matrice
de modèle représentant un modèle de canal sur la base d’une matrice de conversion pour une conversion
moyen à net et des valeurs pilotes connues pour le signal reçu et former une matrice d’observation sur la base
des valeurs pilotes obtenues d’après le signal reçu ;
calculer un terme de transformation par blanchissement sur la base d’une factorisation Cholesky d’une matrice
50 de corrélation de données de trafic déterminée pour le signal reçu ;
appliquer le terme de transformation par blanchissement au problème de moindres carrés initial afin d’obtenir
un problème de moindres carrés transformé ;
résoudre le problème de moindres carrés transformé afin d’obtenir des estimations de canal moyen blanchies
pour le signal reçu ; et
55 convertir les estimations de canal moyen blanchies en estimations de canal net blanchies en vue d’un traitement
du signal reçu en appliquant la matrice de conversion moyen à net aux estimations de canal moyen blanchies,
dans lequel le canal moyen fait référence à un canal radio et le canal net fait référence aux effets, en plus du
canal moyen, de filtrage d’émission au niveau d’un émetteur distant et de filtrage de réception au niveau d’un

13
 EP 2 250 776 B1

récepteur.

2. Procédé selon la revendication 1, dans lequel former un problème de moindres carrés initial pour les estimations
de canal moyen sur la base des valeurs pilotes connues et des observations pilotes connues pour le signal reçu
5 inclut d’obtenir les observations pilotes comme des valeurs pilotes obtenues d’après le signal reçu pour tous les
délais de traitement de signal reçu présentant un intérêt sur chacune de la ou des antennes de réception.

3. Procédé selon la revendication 1, dans lequel le calcul d’un terme de transformation par blanchissement sur la base
d’une factorisation Cholesky d’une matrice de corrélation de données de trafic déterminées pour le signal reçu inclut
10 d’estimer la matrice de corrélation de données de trafic d’après les valeurs de données de trafic obtenues d’après
le signal reçu et déterminer un facteur Cholesky de la matrice de corrélation de données de trafic.

4. Procédé selon la revendication 3, dans lequel le problèmes de moindres carrés initial inclut une matrice de modèle
représentant un modèle de canal, une matrice d’observation représentant les observations pilotes et une matrice
15 de coefficients pour ajuster la matrice de modèle à l’observation, et dans lequel l’application du terme de transfor-
mation par blanchissement au problème de moindres carrés initial afin d’obtenir un problème de moindres carrés
transformé comprend d’appliquer le facteur de Cholesky au modèle et aux matrices d’observation en appliquant le
facteur de Cholesky inverse à la matrice de modèle afin d’obtenir une matrice de modèle transformé à utiliser dans
le problème de moindres carrés transformé, et appliquer le facteur de Cholesky inverse à la matrice d’observation
20 afin d’obtenir une matrice d’observation transformée à utiliser dans le problème de moindres carrés transformé.

5. Procédé selon la revendication 4, dans lequel l’application du facteur de Cholesky au modèle et aux matrices
d’observation comprend d’exprimer une matrice de modèle transformé comme une fonction du facteur de Cholesky
et de la matrice de modèle et résoudre la matrice de modèle transformé à utiliser dans le problème de moindres
25 carrés transformé, et exprimer une matrice d’observation transformé comme une fonction du facteur de Cholesky
et de la matrice d’observation et résoudre la matrice d’observation transformée à utiliser dans le problème de
moindres carrés transformé.

6. Procédé selon la revendication 1, comprenant en outre de calculer la matrice de conversion moyen à net sur la
30 base des caractéristiques de filtre d’émission/réception connues et des différences entre les délais de traitement
de signal reçus et les délais de propagation par trajets multiples de signal reçu.

7. Procédé selon la revendication 1, dans lequel le signal reçu comprend un signal d’accès multiple par répartition de
code (CDMA) comprenant des données pilotes et des données de trafic sur différents codes de mise en canal, et
35 dans lequel la formation de la matrice de modèle représentant le modèle de canal est basé sur la matrice de
conversion et les valeurs pilotes connues pour le signal CDMA et la formation de la matrice d’observation est basée
sur les valeurs pilotes de désétalement obtenues d’après le signal reçu.

8. Procédé selon la revendication 7, dans lequel le calcul d’un terme de transformation par blanchissement sur la base
40 d’une factorisation Cholesky d’une matrice de corrélation de données de trafic déterminé pour le signal reçu inclut
d’estimer la matrice de corrélation de données de trafic d’après les valeurs de données de trafic de désétalement
obtenues pour un ou des codes de mise en canal de trafic utilisés dans le signal reçu et de calculer le terme de
transformation par blanchissement comme le facteur de Cholesky de la matrice de corrélation de données de trafic.

45 9. Circuit récepteur (10) configuré afin de générer des estimations de canal pour un signal reçu par un récepteur de
communications associé (12), ledit circuit récepteur (10) étant caractérisé par un ou plusieurs circuits de traitement
(29, 30, 32, 34, 36) configurés afin de :

former un problème de moindres carrés initial pour estimations de canal moyen sur la base des valeurs pilotes
50 connues et des observations pilotes correspondantes pour le signal reçu incluant de former une matrice de
modèle représentant un modèle de canal sur la base d’une matrice de conversion pour une conversion de
moyen à net et des valeurs pilotes connues pour le signal reçu et de former une matrice d’observation sur la
base des valeurs pilotes obtenues d’après le signal reçu ;
calculer un terme de transformation par blanchissement sur la base d’une factorisation Cholesky d’une matrice
55 de corrélation de données de trafic déterminée pour le signal reçu ;
appliquer le terme de transformation par blanchissement au problème de moindres carrés initial afin d’obtenir
un problème de moindres carrés transformé ;
résoudre le problème de moindres carrés transformé afin d’obtenir des estimations de canal moyen blanchies

14
 EP 2 250 776 B1

pour le signal reçu ; et

convertir les estimations de canal moyen blanchies en estimations de canal net blanchies en vue d’un traitement
du signal reçu en appliquant la matrice de conversion moyen à net aux estimations de canal moyen blanchies,
dans lequel le canal moyen fait référence à un canal radio et le canal net fait référence aux effets, en plus du
5 canal moyen, de filtrage d’émission au niveau d’un émetteur distant et de filtrage de réception au niveau d’un
récepteur.

10. Circuit récepteur (10) selon la revendication 9, caractérisé en outre en ce que le circuit récepteur (10) est configuré
afin d’obtenir les observations pilotes comme des valeurs pilotes obtenues d’ après le signal reçu pour tous les
10 délais de traitement de signal reçu présentant un intérêt sur chacune de la ou des antennes de réception (14) du
récepteur de communications associé (12).

11. Circuit récepteur (10) selon la revendication 9, caractérisé en outre en ce que le circuit récepteur (10) est configuré
afin d’estimer la matrice de corrélation de données de trafic d’après les valeurs de données de trafic obtenues
15 d’après le signal reçu et déterminer un facteur de Cholesky de la matrice de corrélation de données de trafic à
utiliser comme terme de transformation par blanchissement, et le problème de moindres carrés initial inclut une
matrice de modèle représentant un modèle de canal, une matrice d’observation représentant les observations
pilotes, et une matrice de coefficients pour ajuster la matrice de modèle à l’observation, et le circuit récepteur (10)
est configuré afin d’appliquer le facteur de Cholesky au modèle et aux matrices d’observation pour obtenir le problème
20 des moindres carrés transformé.

12. Circuit récepteur (10) selon la revendication 11, caractérisé en outre en ce que le circuit récepteur (10) est configuré
afin d’appliquer le facteur de Cholesky au modèle et aux matrices d’observation sur la base de l’application du
facteur de Cholesky inverse à la matrice de modèle pour obtenir une matrice de modèle transformé à utiliser dans
25 le problème de moindres carrés transformé, et de l’application du facteur de Cholesky inverse à la matrice d’obser-
vation pour obtenir une matrice d’observation transformé à utiliser dans le problème de moindres carrés transformé.

13. Circuit récepteur (10) selon la revendication 11, caractérisé en outre en ce que le circuit récepteur (10) est configuré
afin d’appliquer le facteur de Cholesky au modèle et aux matrices d’observation sur la base de l’expression d’une
30 matrice de modèle transformé comme une fonction du facteur de Cholesky et de la matrice de modèle et de la
résolution de la matrice de modèle transformée, et d’exprimer une matrice d’observation transformée comme une
fonction du facteur de Cholesky et de la matrice d’observation.

14. Circuit récepteur (10) selon la revendication 9, caractérisé en outre en ce que le circuit récepteur (10) est configuré
35 afin de calculer la matrice de conversion moyen à net sur la base des caractéristiques de filtre d’émission/réception
connues et des différences temporelles entre les délais de traitement de signal reçu et les délais de propagation
par trajets multiples de signal reçu, et dans lequel les délais de traitement de signal reçu comprennent un de délais
de corrélation de doigt Rake dans un circuit récepteur Rake (12) et délais de prise de filtre d’égalisation dans un
circuit récepteur d’égalisation d’élément (12).
40
15. Circuit récepteur (10) selon la revendication 9, caractérisé en ce que le circuit récepteur (10) comprend tout ou
une partie du circuit récepteur Rake ou d’égalisation d’élément à l’intérieur du récepteur de communication (12)
d’un dispositif de communication sans fil (60), et dans lequel le ou les circuits de traitement (29, 30, 32, 34, 36)
comprend un ou plusieurs circuits de traitement numérique incluant un processeur de matrice de conversion (36)
45 configuré afin de calculer la matrice de conversion, un processeur de moindres carrés (29, 30, 32) configuré afin
de former le problème de moindres carré initial et transformé et de résoudre le problème de moindres carrés
transformé pour les estimations de canal moyen blanchies, et un processeur de corrélation (34) configuré afin de
générer la matrice de corrélation de données.

50

55

15
EP 2 250 776 B1

16
EP 2 250 776 B1

17
EP 2 250 776 B1

18
 EP 2 250 776 B1

REFERENCES CITED IN THE DESCRIPTION

This list of references cited by the applicant is for the reader’s convenience only. It does not form part of the European
patent document. Even though great care has been taken in compiling the references, errors or omissions cannot be
excluded and the EPO disclaims all liability in this regard.

Patent documents cited in the description

• US 11935604 B [0004] • WO 2007000620 A [0005]

19