Method and Arrangement for Improved G-RAKE Scaling Parameter Estimation

Abstract

The present invention discloses a method of improved impairment covariance matrix estimation for a received signal in a Generalized Rake receiver arrangement, providing SIO a plurality N of despread pilot symbols representative of the signal, determining S20 an estimate of an impairment covariance matrix R for the received signal. Subsequently, providing S30 an estimate of scaling parameters a, β for the estimate R by means of a weighted least squares estimate based on the color of a residual noise vector for the impairment covariance matrix estimate R, and forming S40 an improved estimate of the impairment covariance matrix based on the weighted least squares estimate.

Description

TECHNICAL FIELD

The present invention relates to telecommunication systems in general, specifically to an improved G-Rake method and arrangements in such a system.

BACKGROUND

Telecommunication systems utilizing Wideband Code Division Multiple Access (WCDMA) continues to evolve to support high-bit rate applications. As the demand for higher data rates increases, however, greater self-interference from the dispersive radio channel limits performance. Consequently, advanced receivers for WCDMA terminal platforms and base stations have been and are continuously being developed.

Signals transmitted in a wireless communication system such as a Code Division Multiple Access (CDMA) or Wideband CDMA (WCDMA) system are subjected to multiple sources of interference and noise as they propagate via radio channels. The interference and noise components that affect signal transmission and reception in a wireless communication system are broadly referred to as impairments. Certain types of impairments may be correlated. That is, two seemingly independent signal impairments may in fact be related, and thus are said to be correlated. Some conventional receiver types such as the Generalized-RAKE (G-RAKE) receiver, see e.g. [1]-[3], and its Chips Equalizer (CEQ) counterpart use knowledge of impairment correlations to improve received signal quality. G-Rake receivers and CEQs also use an estimate of a multipath fading channel response in their received signal processing.

For example, a G-Rake receiver includes various signal “fingers” where each finger has an assigned path delay for receiving a particular image of a multipath signal and a correlator for despreading the received image. In combination, the signal fingers de-spread multiple signal images of a received multipath signal, this utilizing the multipath channel dispersion phenomenon. Additional “probing” fingers may be placed off path delays for capturing impairment correlations information. The finger outputs are weighted and coherently combined to improve received signal demodulation and/or received signal quality reception estimation, e.g. signal-to-interference (plus noise) (SIR) estimation. The processing weights assigned to the finger outputs are conventionally a function of the channel response and impairment correlations. As such, knowledge of signal impairments may be used to improve received signal processing. In a similar manner, CEQs utilize impairment correlations information for improving received signal processing where the selection of equalization filter taps in a CEQ is comparable to the placement of fingers in a G-Rake receiver and the generation of equalization filter coefficients is comparable to the generation of G-Rake combining weights.

Parametric G-Rake receivers estimate impairment correlations using a modeling approach. The model employs parameters, sometimes referred to as fitting parameters that can be estimated in a number of ways such as least-squares fittings. The parametric impairment correlations modeling process depends on corresponding model fitting parameters and on estimates of the channel response. However, signal impairments affect the channel response estimation process, particularly when the impairments are severe. As such, impairment correlation estimation and channel response estimation may be interdependent, particularly when interference is severe.

One specific type of receiver that has been developed is the so-called Rake receiver and the subsequently evolved Generalized-Rake or G-Rake receiver. In a Rake receiver signal energy is collected from different delayed versions of a transmitted signal. The channel response generates multiple images of the transmitted signal (that is the dispersive, multi-path channel gives rise to different versions). The “fingers” of the Rake receiver extract signal energy from delayed signal images by despreading and combining them. The Rake receiver coherently combines the finger outputs using complex conjugates of estimated channel coefficients to estimate the modulation symbol. Each despread value consists of a signal component, an interference component, and a noise component. When combining the values the Rake receiver aligns the signal components so that they add to one another, creating a larger signal component.

A G-Rake receiver operates in a similar, but slightly different manner. The G-Rake receiver uses fingers and combining techniques to estimate a symbol. However, the G-Rake uses extra interference fingers to collect information about interference on the signal fingers. This interference might result from other symbols of interest (self-interference) or symbols intended for other users in the cell (own-cell interference) or symbols intended for other users in other cells (other-cell interference). The extra fingers capture information about the interference. This is used to cancel interference on the signal fingers. In addition, a separate procedure is used to form combining weights. Rake receivers use a weighted sum of despread values to estimate symbols. Despread values are thus combined using combining weights. Besides estimating the channel, the G-Rake estimates the correlations between the impairment (interference plus noise) on different fingers. The correlation captures the “color” of the impairment. This information can be used to suppress interference. Channel estimates and impairment correlation estimates are used to form the combining weights. As a result, the combining process collects signal energy and suppresses interference. The G-Rake receiver combines two despread values to cancel interference and increase the signal component. By contrast, the Rake receiver solely maximizes the signal component.

In order to remain competitive on the market, WCDMA systems are constantly evolving and striving for higher bit rates. In order to achieve this, concepts like higher order modulation and MIMO are considered. However, to be able to benefit from all these new features better signal to noise ratio (SNR) conditions are required. Consequently, it is common for telecommunications systems to operate at significantly higher E_c/N₀ regions, which make the interference situation more severe. Hence, in order to achieve the desired targets a good G-Rake becomes essential. Unfortunately, when moving to these higher E_c/N₀ regions the implementation of the G-Rake becomes more sensitive, and the current G-Rake algorithm is in general not good enough. In particular, we need to improve the channel estimation procedure and especially the estimation of the covariance matrix required in the G-Rake.

SUMMARY

According to a basic aspect, the present invention enables an improved telecommunication system.

According to a further aspect, the present invention enables an improved receiver in a telecommunication system.

According to a further aspect, the present invention enables an improved G-Rake receiver.

According to a further aspect, the present invention enables a G-Rake receiver with improved interference suppression.

In general, the present invention comprises a method of improved estimation of an impairment covariance matrix of a received signal in a G-Rake receiver arrangement. The method comprises the steps of providing S10 a plurality of despread pilot symbols of the received signal and determining S20 an initial estimate of the impairment covariance matrix for the signal. Subsequently, providing S30 scaling parameter estimates α, β for the determined estimate by a weighted least squares estimate based on the residual noise vector of the estimate, and finally forming an improved estimate of the impairment covariance matrix based on the weighted least squares estimate of the scaling parameters.

According to a further embodiment, the weighted least squares estimate further utilizes information relating to the color of the residual noise vector.

Advantages of the present invention include

• • a G-Rake with improved performance at high SNR;
  • a G-Rake with improved scaling parameter estimation.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention, together with further objects and advantages thereof, may best be understood by referring to the following description taken together with the accompanying drawings, in which:

FIG. 1 is a general communication system;

FIG. 2 is a schematic flow chart of an embodiment of a method according to the present invention;

FIG. 3 is a schematic flow chart of a further embodiment of the present invention;

FIG. 4 is a schematic illustration of an embodiment of a receiver arrangement according to the present invention.

ABBREVIATIONS

CPICH Common Pilot Channel

CQI Channel Quality Indicator

CRB Cramér-Rao Bound

DPCCH Dedicated Physical Control Channel

G-Rake Generalized Rake

HOM Higher Order Modulation

HSPA High Speed Packet Access

MIMO Multiple Input Multiple Output

ML Maximum Likelihood

MMSE Minimum Mean Square Error

SINR Signal-to-Interference-plus-Noise Ratio

SNR Signal-to-Noise Ratio

TPC Transmit Power Control

WCDMA Wideband Code Division Multiple Access

Mathematical notations:

• • E is the expectation operator;
  • T is the transpose;
  • H is the Hermitian transpose;
  • cov denotes the covariance matrix.

DETAILED DESCRIPTION

Embodiments of the present invention will be described in the context of a general telecommunication system utilizing Wideband Code Division Multiple Access (WCDMA). Such a system is illustrated in FIG. 1. A mobile terminal or user equipment (UE) communicates with one or several Node Bs. The term Node B refers to a logical node, responsible for physical layer processing such as error detection coding, modulation and spreading, as well as conversion from baseband to the radio-frequency signal transmitted from the antenna. A Node B is handling transmission and reception in one or several cells. Every Node B is connected to a Radio Network Controller (RNC). The RNC controls one or normally several Node Bs. The RNC, among other things, is in charge of call setup, quality-of-service handling, and management of the radio resources in the cells for which it is responsible. Most of the “intelligence” in a radio-access network is located in the RNC, while the Node Bs mainly act as modems. For high speed packet access (HSPA) the Node B further deals with channel dependent scheduling and link adaptation. Finally, the RNC is connected to the Internet and the traditional wired telephony network through the core network.

In order to enable an in-depth understanding of the merits of the present invention, a brief description of the concept of G-Rake receivers and particularly parametric G-Rake receivers follows below.

Parametric G-Rake receivers estimate impairment correlations using a modeling approach. The model employs parameters, sometimes referred to as fitting parameters that can be estimated in a number of ways such as least squares fittings. The parametric impairment correlations modeling process depends on corresponding model fitting parameters and on estimates of the channel response. However, signal impairment affects the channel response estimation process, particularly when the impairments are severe. As such, impairment correlation estimation and channel response estimation may be interdependent, particularly when interference is severe.

The G-Rake receiver is extensively used in communication systems, for example WCDMA, as an important means to provide interference suppression at moderate to high data rates. As such, the G-Rake tries to balance interference suppression and channel equalization by taking the so-called color of the impairment covariance matrix into account when forming the combining weights. To be more specific, the G-Rake combining weights w are found by solving the following system of linear equations

Rw=h  (1)

where R is the total impairment covariance matrix and h is the net channel response vector, see [2]. For the basic parametric G-Rake it holds that R can be modeled as follows:

R=αR _I(g)+βR_n  (2)

where g is the medium channel response, R_I spawns from interference, R_n corresponds to a white Gaussian noise source (used to model the aggregate of all un-modeled signal sources), and α, β are the so-called scaling/fitting parameters. The scaling parameters are in general unknown and these values can also change quite rapidly over time. Consequently, in practice the scaling parameters α, β need to be estimated. This estimation problem, which needs to be accurate to achieve good G-Rake performance, suffers from several weaknesses.

In the baseline G-Rake, α, β are estimated by using a least squares methodology according to the relation of Equation 3

{circumflex over (R)} _m =αR _I({circumflex over (g)})+βR_n  (3)

where ĝ is an estimate of the medium channel response, and {circumflex over (R)}_m is an estimate of the impairment covariance matrix. The latter is commonly obtained using N despread pilot symbols according to Equation 4.

$\begin{array}{cc}{\hat{R}}_m=\frac{1}{N-1}\sum _{k=1}^N\left(y_k-\hat{h}\right){\left(y_k-\hat{h}\right)}^H& \left(4\right)\end{array}$

where ĥ is obtained via averaging of the N despread pilot symbols

$\begin{array}{cc}\hat{h}=\frac{1}{N}\sum _{k=1}^Ny_k& \left(5\right)\end{array}$

Notice also that the scaled medium channel g can be obtained from the estimate of h by observing that the despread pilot channel (e.g. dedicated physical control channel (DPCCH) or common pilot channel (CPICH)) can be written as

y _k =h+n _k=√{square root over (E_p)}Ag+n_k  (6)

where n_k is assumed to be white Gaussian noise with covariance matrix

R=E _c R _l(g)+N₀R_n  (7)

Here E_c is the total chip energy, E_p denotes the pilot symbol energy, N₀ represents the variance of the white noise source, and A is a matrix that depends on the pulse shape autocorrelation function, the path delays and the finger delays. Notice also that it is common to consider the scaled medium channel g=√{square root over (E_p)}g when discussing the G-Rake parameter estimation problem.

Obviously, there exist different variations of the G-Rake, in which different approaches to obtain the required G-Rake components are employed. For example, the medium channel ĝ can be obtained using least squares, maximum likelihood, or minimum mean square error criterions.

Two related and very important problems when discussing the G-Rake and combining weight estimation are channel quality indicator (CQI) and transmit power control (TPC). Both these concepts need an estimate of the SNR or signal to interference plus noise ratio (SINR). For the G-Rake a formulation of the SINR can be written as

SINR=h^HR⁻¹h  (8)

which indicates that an accurate estimate of R is essential for good CQI and TPC performance.

With reference to FIG. 2, the present invention basically comprises providing S10 a plurality N of pilot symbols of a received signal and determining S20 an initial estimate {circumflex over (R)}_m of an impairment covariance matrix R for the received signal. Subsequently, scaling parameters α and β for the estimated {circumflex over (R)}_m impairment covariance matrix are estimated utilizing a weighted least squares estimate S30 based on a residual noise vector ε for the estimate {circumflex over (R)}_m. Finally, an improved estimate of the impairment covariance matrix is provided S40 based on the weighted least squares method.

The main idea of the present invention is thus to extend a current G-Rake algorithm structure to include a more accurate estimate of the covariance matrix R at the cost of a moderate increase in computational complexity. In particular, a solution to the problem of estimating the scaling parameters α and β is further developed. Consequently, the present invention discloses an algorithm that significantly increases the accuracy of α and β estimates, particularly at high E_c/N₀ regions. In order to keep the computational complexity within reason, the disclosed embodiments all utilize a least squares framework. However, it is evident that also other estimates can be utilized with minor modifications of the present invention.

In essence, the improved estimate of the scaling parameters α and β, according to the present invention, all rely on the same principle as a G-Rake. Similarly as the G-Rake takes the color of the impairment covariance matrix into account, the disclosed scheme for estimating α and β will benefit from knowledge about the color of the residual error in the least squares regression model.

In order to ease the presentation, a single antenna scenario and a real valued medium channel g are assumed. Consequently, all involved quantities of the disclosure are real valued in the current embodiment. However, it is equally possible to implement the present invention on multiple antenna scenarios and complex valued medium channels.

Initially the relation for a parametric G-Rake model according to

R=αR _I(g)+βR_n  (9)

is further explored. This expression can be written in the classic regression form according to

$\begin{array}{cc}r=\mathrm{\Phi \theta }\mathrm{where}& \left(10\right)\\ r=\left[\begin{array}{c}{e}_1^T{\mathrm{Re}}_1\\ ⋮\\ e_i^T{\mathrm{Re}}_j\end{array}\right],\Phi =\left[\begin{array}{cc}{e}_1^TR_I\left(g\right)e_1& e_1^TR_ne_1\\ ⋮& ⋮\\ e_i^TR_I\left(g\right)e_j& e_i^TR_ne_j\end{array}\right],\mathrm{and}\theta =\left[\begin{array}{c}\alpha \\ \beta \end{array}\right].& \left(11\right)\end{array}$

Here e_i is the zero vector with a 1 at the i:th position, meaning that e_i^TXe_j represents the (i,j) element of the matrix X. Consequently, it can be seen that in general not all equation elements from (9) need to be used when forming (10). Exactly which equation elements to use is typically implementation specific. However, it is common to include the main diagonal and possibly some of the off-diagonal elements. Furthermore, it is assumed that the medium channel g includes a scaling factor, i.e. g=√{square root over (E_p)}g (see the discussion following Equation (6)). This means that the true scaling parameters are given by

α=E_c/E_p and β=N₀  (12)

By using estimated quantities of R and g the following expression is obtained

$\begin{array}{cc}\hat{r}=\hat{\Phi }\theta +\varepsilon \mathrm{where}& \left(13\right)\\ \hat{r}=\left[\begin{array}{c}{e}_1^T{\hat{R}}_me_1\\ ⋮\\ e_i^T{\hat{R}}_me_j\end{array}\right],\hat{\Phi }=\left[\begin{array}{cc}{e}_1^TR_I\left(\hat{g}\right)e_1& e_1^TR_ne_1\\ ⋮& ⋮\\ e_i^TR_I\left(\hat{g}\right)e_j& e_i^TR_ne_j\end{array}\right],\varepsilon =\left(\hat{r}-r\right)-\left(\hat{\Phi }-\Phi \right)\theta & \left(14\right)\end{array}$

and ε is the residual noise vector for the model estimate.

In particular, assume that {circumflex over (R)}_m is given by Equation 4. A known baseline least squares estimate of the scaling parameters is then given by

{circumflex over (θ)}_LS=({circumflex over (Φ)}^T{circumflex over (Φ)})⁻¹{circumflex over (Φ)}^T{circumflex over (r)}  (15)

However, as identified by the inventors, the residual noise vector ε in the model of Equation 13 is neither white nor has zero mean. In fact, the color becomes severe at high E_c/N₀ regions. Consequently, the baseline least squares approach may suffer from poor performance at high E_c/N₀ regions

Consequently, the inventors have realized that in order to obtain an improved estimate of the impairment covariance matrix of the received signal at high E_c/N₀ regions it is necessary to take also the color of the residual noise vector into consideration.

According to embodiments of the present invention, with specific reference to FIG. 3, as an alternative to the baseline least square estimate, the color of the residual noise vector will be taken into account. Therefore, according to an embodiment of the present invention, a parameter or matrix Σ representative of the color of the residual noise vector is determined S31 or estimated. In mathematical terms the parameter or matrix Σ represents an estimate of the second order momentum of the residual vector ε. In principal, the parameter or matrix τ describe the correlation between the elements of the observation vectors {circumflex over (r)} and {circumflex over (Φ)} of Equation 13, or rather the dependencies between the elements or observations of the model of Equation 10.

An estimate of θ is thus obtained S30 by a weighted least squares approach

{circumflex over (θ)}_WLS=({circumflex over (Φ)}^TΣ⁻¹{circumflex over (Φ)})⁻¹{circumflex over (Φ)}^TΣ⁻¹{circumflex over (r)}  (16)

where Σ represents the correlation or interdependency between the observations according to {circumflex over (r)} and {circumflex over (Φ)}, where Σ=E{εε^T}. On a mathematical level this implies that the under certain assumptions this choice of Σ minimizes the MSE ({circumflex over (θ)}). However, from a more technical viewpoint Equation 16 represents a least squares method where the observations have been pre-filtered (or whitened). One way of describing this filter is by the so-called Cholesky-factorization of Σ.

According to a further embodiment of the present invention, instead of determining the actual Σ, which is impossible or at least very difficult, an approximate Σ is used according to Equation 17

Σ≈cov({circumflex over (r)})  (17)

A number of things are worth to point out. First, by using proper approximations it can be shown that this choice of Σ will minimize the MSE of {circumflex over (θ)}. Second, an expression for cov({circumflex over (r)}) is obviously needed to successfully implement Equation 16. However, from Equations (14) and (17), it follows that

$\left(18\right)$ $\sum \approx \left[\begin{array}{ccc}{\mathrm{Ee}}_1^T\left({\hat{R}}_m-R\right)e_1e_1^T\left({\hat{R}}_m-R\right)e_I& \cdots & {\mathrm{Ee}}_1^T\left({\hat{R}}_m-R\right)e_1e_i^T\left({\hat{R}}_m-R\right)e_j\\ ⋮& \ddots & ⋮\\ {\mathrm{Ee}}_i^T\left({\hat{R}}_m-R\right)e_je_1^T\left({\hat{R}}_m-R\right)e_1& \cdots & {\mathrm{Ee}}_i^T\left({\hat{R}}_m-R\right)e_je_i^T\left({\hat{R}}_m-R\right)e_j\end{array}\right]$

where, under the given assumptions, it holds that

$\begin{array}{cc}{\mathrm{Ee}}_i^T\left({\hat{R}}_m-R\right)e_je_k^T\left({\hat{R}}_m-R\right)e_I=\frac{1}{N-1}\left(e_i^T{\mathrm{Re}}_qe_j^T{\mathrm{Re}}_r+e_i^T{\mathrm{Re}}_re_q^T{\mathrm{Re}}_j\right)& \left(19\right)\end{array}$

A further comment is that Σ needs to be invertible. This should of course be checked in the implementation, and also be taken into account when choosing which matrix elements to include from (10) when forming (11). Nevertheless, given R it is now straightforward to construct the required matrices and obtain the weighted least squares estimate of the scaling parameters α and β.

The next and final problem is that Σ is a function of R (Σ=Σ(R)) and that R is unknown (after all, the overall goal is to obtain an accurate estimate of R). However, this problem can be solved by using a two step approach. The algorithm can be summarized as follows, see also FIG. 3:

1) Estimate the (scaled) medium channel g, for instance via an estimate of the net channel h. Also, form S20 the non-parametric estimate (4) of R.2) Using the estimate in step 1) estimate the scaling parameters S21 using the baseline least squares algorithm in Equation 15. By using the estimated scaling parameters build S22 a parametric estimate of R.3) Use the baseline parametric estimate of R (from step 2) to form Σ. Then the weighted least square estimate is used to provide improved estimates S30 of the scaling parameter. Finally, rebuild S40 R using the improved scaling parameters.

Notice that step 2) is optional and can be avoided by using the non-parametric estimate in Equation 4 of R when forming Σ in step 3. In this case, the weighted least squares method S30 is used directly to obtain the scaling parameters.

In conclusion:

• • The proposed weighted least squares scheme provides significantly improved estimates of the scaling parameters, especially at regions with high E_c/N₀, which are of importance when considering higher order modulation (HOM) or MIMO. Experience indicates that an improvement of the scaling parameters of the order of several magnitudes can be obtained. In fact, results indicate that the method according to the present invention yields results close to the Cramér-Rao Bound (CRB).
  • The increased performance comes at cost of a modest increase in computational power, since it is necessary to build and invert Σ. Also, two estimation steps are used even though the second step is optional and can be avoided.
  • The method according to the invention can be combined with other methods, and different embodiments can be envisioned. For example, the estimate of the channel can be obtained using maximum likelihood (ML) or minimum mean square error (MMSE). In addition, the improved estimate of the scaling parameters, which result in an improved estimate of the impairment covariance matrix R can be used to improve SINR estimates or TPC performance.
  • The method can be extended to handle complex valued channels (Rayleigh distributed) as well as to multi antenna scenarios, or even MIMO systems.

The framework disclosed in the present invention is in principle applicable to other receivers as well, e.g. GRAKE+ etc.

The above disclosed embodiments are preferably implemented in a GRAKE in a receiver in a telecommunication system. Consequently, it can be implemented in a user equipment such as a mobile phone or other mobile user equipment, or in a receiver in a Node B.

A receiver arrangement according to the present invention will be described below with reference to FIG. 4.

An embodiment of a receiver arrangement, e.g. a G-Rake receiver arrangement, is provided with a unit 10 for receiving and providing multiple despread pilot symbols representative of a signal received at the receiver arrangement. The receiver also comprises a unit 20 for estimating an impairment covariance matrix R for the received signal. In addition, the receiver comprises a unit 30 for estimating scaling parameters α, β for the previously estimated R by means of a weighted least squares estimate based on the color of a residual noise vector for the impairment covariance matrix estimate R, and a unit 40 forming an improved estimate of the impairment covariance matrix based on at least the weighted least squares estimate.

According to a further embodiment of the present invention, the arrangement also comprises a unit 31, indicated by the dotted box 31, for determining a parameter Σ representative of the color of the residual noise error vector of the estimated impairment covariance matrix. The color parameter is based on correlations between elements of the covariance matrix estimate. Consequently, the providing means 30 are further adapted to provide the weighted least squares estimate of the scaling parameters α, β for the impairment covariance matrix estimate R based on the determined color parameter.

The unit 20 for determining the initial estimation of the impairment covariance matrix R can be further adapted to determine the impairment covariance matrix R by forming a non-parametric estimate of said impairment covariance matrix R. In addition, the receiver arrangement comprises a unit 21 for calculating a baseline least squares estimate of the scaling parameters α, β based on an estimate of the medium channel and the non-parametric estimate of the impairment covariance matrix; and a unit 22 for determining a parametric estimate of the impairment covariance matrix based on the estimated scaling parameters. Consequently the providing unit 30 is further adapted to provide an improved estimate of scaling parameters α, β for the estimate R by means of the weighted least squares (WLS) estimate, and on the parametric estimate, and the forming unit 40 is adapted to form an improved estimate of the impairment covariance matrix based on the improved estimate for the scaling parameters.

The units 21 and 22 can be implemented as separate units or included in the unit 20 for estimating the initial estimate of the impairment matrix R, as indicated by the dotted box surrounding the units 20, 21, 22.

Advantages of the present invention comprise:

• • The purpose of the current invention is to increase the performance of the G-Rake at high SNR regions, which is essential in order to reach the desired targets when introducing concepts such as HOM and MIMO. In particular, we focus on the impairment covariance matrix and especially on how the estimate of the scaling parameters can be improved. This can have a profound impact, not only on the G-Rake combining weight calculation, but also when considering CQI estimates and TPC. The improved parameter quality comes at a modest cost of a slight increase in computational complexity.

It will be understood by those skilled in the art that various modifications and changes may be made to the present invention without departure from the scope thereof, which is defined by the appended claims.

REFERENCES

• [1] C. Cozzo et. al., “Method and Apparatus for Scaling Parameter Estimation in Parametric Generalized Rake Receivers”, Pub. No. US 2006/0007990 A1, 2006.
• [2] G. E. Bottomley and T. Ottosson and E. Wang, “A Generalized RAKE Receiver for Interference Suppression”, IEEE Journal on selected areas in communications, num 18(8): 1536-1545, 2000.
• [3] D. Cairns et. al., “Method and Apparatus for Parameter Estimation in a Generalized Rake Receiver”, Pub. No. US 2005/0201447 A1, 2005.

Claims

1. A method of improved impairment covariance matrix estimation for a received signal in a Generalized Rake receiver arrangement, said method comprising: providing a plurality N of despread pilot symbols representative of said signal;determining an estimate of an impairment covariance matrix R for said received signal;providing an estimate of scaling parameters α, β for said estimate R according to a weighted least squares estimate based on the color of a residual noise vector for said impairment covariance matrix estimate R, andforming an improved estimate of said impairment covariance matrix based on at least said weighted least squares estimate.

2. The method according to claim 1, wherein said step of providing the estimate of scaling parameters includes: determining a parameter Σ representative of the color of the residual noise error vector of said estimated impairment covariance matrix, said color being based on correlations between elements of the covariance matrix estimate; andproviding the weighted least squares estimate of said scaling parameters α, β for said impairment covariance matrix estimate R based on said determined parameter.

3. The method according to claim 2, wherein said step of determining the estimate of the impairment covariance matrix R comprises forming a non-parametric estimate of said impairment covariance matrix R.

4. The method according to claim 3, wherein said step of determining the estimate of the impairment covariance matrix R includes: forming a baseline least squares estimate of said scaling parameters α, β based on an estimate of the medium channel and said nonparametric estimate of the impairment covariance matrix; anddetermining a parametric estimate of the impairment covariance matrix based on the estimated scaling parameters; andproviding an improved estimate of scaling parameters α, β for said estimate R according to said weighted least squares estimate, and on said parametric estimate, andforming an improved estimate of said impairment covariance matrix based on said improved estimate for said scaling parameters.

5. The method according to claim 2, further comprising determining a matrix Σ such that it minimizes the weighted least squares estimate of the scaling parameters.

6. The method according to claim 2, further comprising estimating said color parameter Σ as the covariance value of the estimated impairment covariance matrix.

7. The method according to claim 1, wherein said weighted least squares estimate {circumflex over (θ)}WLS of scaling parameters α, β is determined according to the expression: {circumflex over (θ)}WLS=({circumflex over (Φ)}T{circumflex over (Σ)}−1{circumflex over (Φ)})−1{circumflex over (Φ)}T{circumflex over (Σ)}−1{circumflex over (r)}

8. The method according to claim 2, further comprising determining the color parameter {circumflex over (Σ)} as set forth below: Σ=E{εεT}≈cov({circumflex over (r)})

9. A receiver arrangement for improved impairment covariance matrix estimation for a received signal in a Generalized Rake receiver arrangement, said receiver arrangement comprising one or more processing circuits including: a despreading unit for providing a plurality N of despread pilot symbols representative of said signal;a first covariance estimation unit for determining an estimate of an impairment covariance matrix R for said received signal;a scaling parameter estimation unit for providing an estimate of scaling parameters α, β for said estimate R according to a weighted least squares estimate based on the color of a residual noise vector for said impairment covariance matrix estimate R, anda second covariance estimation unit for forming an improved estimate of said impairment covariance matrix based on at least said weighted least squares estimate.

10. The arrangement according to claim 9, further comprising color parameter estimation unit for determining a parameter Σ representative of the color of the residual noise error vector of said estimated impairment covariance matrix, said color being based on correlations between elements of the covariance matrix estimate, and wherein said scaling parameter estimation unit is configured to provide the weighted least squares estimate of said scaling parameters α, β for said impairment covariance matrix estimate R based on said determined parameter.

11. The arrangement according to claim 10, said first covariance estimation unit is configured to determine the impairment covariance matrix R by forming a non-parametric estimate of said impairment covariance matrix R.

12. The arrangement according to claim 11, further comprising: a calculation unit for forming a baseline least squares estimate of said scaling parameters α, β, based on an estimate of the medium channel and said non-parametric estimate of the impairment covariance matrix; anda parametric estimation unit for determining a parametric estimate of the impairment covariance matrix based on the estimated scaling parameters; and whereinsaid scaling parameter estimation unit is further adapted to provide an improved estimate of scaling parameters α, β for said estimate R according to said weighted least squares estimate and said parametric estimate; andsaid second covariance estimation unit is adapted to form an improved estimate of said impairment covariance matrix based on said improved estimate for said scaling parameters.