JP2007513564A - Method and apparatus for noise variance estimation for use in a wireless communication system - Google Patents

Abstract

  A method of noise variance estimation to be performed by a user equipment, the method of noise variance estimation to be executed by a user equipment, comprising: a training sequence and a noise vector transmitted via at least one propagation path; Receiving a signal vector comprising, estimating a channel impulse response of each propagation path according to the signal vector, constructing a channel impulse response matrix, and the channel impulse response during a dedicated time of the training sequence If the method is mainly invariant, a method is proposed comprising the step of calculating a noise variance of the signal vector according to the channel impulse response matrix and the signal vector.

Description

  The present invention relates generally to a noise variance estimation method and apparatus for use in a wireless communication system, and more particularly to a noise variance estimation method and apparatus by utilizing a training sequence.

  CDMA (Code Division Multiple Access) is a new radio communication technology developed after FDMA (Frequency Division Multiple Access) and TDMA (Time Division Multiple Access). In CDMA radio communication, different user equipments (UEs) are assigned different orthogonal spreading codes, and signals spread by different UEs having different spreading codes can be transmitted on the same frequency band.

A CDMA downlink transmission model has been proposed by A. Klein in a paper entitled “Data Detection Algorithms Specially Designed For The Downlink of CDMA Mobile Radio Systems” (VTC, 1997) as shown in FIG. . Signal vector ^{d (1), ···, d} (k), ···, d a ^(K) (where ^{d (k) (k = 1} ··· K) consists of N complex components) , UEk,..., UEK to transmit to UE1,..., UEk, respectively, first, the base station 200 first assigns the spreading code c _d ⁽¹⁾ assigned to UE1,. ^{_{, ···, c d (k)}} , ··· c d (K) signal vector d ⁽¹⁾ by utilizing, ···, ^d (k), the diffusion · · ·, ^d a ^(K) Then, the spread signal vector is combined into a signal vector s _d and transmitted to the corresponding UE 220 via the same channel 210. The signal vector s _d reaches UEk (k = 1... K) through a plurality of propagation paths, and the CIR (channel impulse response) of each propagation channel is h _{d (i)} ^(K) (i = 1 ^). 2..., The signal vector e _d ^(K) received by UEK can be expressed by the following equation (1).
_{^{e d (K) = H d}} (K) C d d + n d (K) = H d (K) s d + n d (K) (1)

Here, H _d ^(K) is a CIR matrix constructed by CIR h _{d (i)} ^(K) (i = 1, 2,...) Of each propagation channel, and C _d is a spreading code c. ^{_{d (1), ···, c}} d (k), how to build a ... a spreading code matrix constructed by ^{_{c d (K) (H d}} (K) and C _d is, a. Klein See the above paper by). d = ( d ^{(1) T} ,..., d ^{(k) T} ,..., d ^{(K) T} ) ^T , and [. ^T represents the transpose of the matrix, s ^d represents the signal vector obtained after d is diffused and combined, s _d = C _d d , and n _d ^(K) is the noise vector.

Equation (1) indicates that the received signal vector e _d ^(K) includes the signal vector and noise vector sent by the base station to other UEs along with the desired signal vector d ^(K) of UE ^k. Is shown.

For UEK to assist in obtaining a desired signal vector d ^(k) with a minimum error from the received signal vector e d ^_(K), a number of methods for signal reception have been proposed. “Iterative Multiuser Receiver / Decoders With Enhanced variance Estimation” by Kimmo Kettunen (VTC, 1999) and “Zero Forcing an Minimum Mean-Square-Error Equalization for Multiuser Detection in Code-Division multiple-access channels” by A. Klein ( See “IEEE Transactions on Vehicular Technology”, vol. 45, pp. 276-287, May 1996). However, all these methods for signal reception rely heavily on channel information, i.e. noise variance, to obtain the desired signal vector from the received signal vector, and thus the desired signal vector with minimal error. In order to obtain a signal vector, the noise variance needs to be accurately calculated.

  In order to obtain accurate noise variance, various noise estimation methods have been proposed. For example, a conventional variance estimation technique for use in the AWGN channel is described in “A novel variance estimator for turbo-code decoding” (Proc. Of ITC '97, pp. 173-178, 1997) by M. Reed and J. Asenstorfer. A Rake technique proposed in April) and proposed to reduce multipath interference is proposed in US patent application US200220110199 entitled “Method for Noise Energy Estimation in TDMA Systems”. In addition, there are several noise estimation methods in which the noise variance is calculated by convolving a training sequence. These noise estimation methods can meet the accuracy requirements of 2G wireless communication systems.

  However, in the 3G wireless communication system, more accurate noise distribution is required for signal reception. For example, both multi-user detection and turbo-code basic techniques have high requirements for accurate noise distribution. Current noise estimation methods cannot meet the accuracy requirements for noise variance in 3G wireless communication systems.

  It is an object of the present invention to provide a method and apparatus for noise variance estimation for use in a wireless communication system in which a training sequence is utilized to calculate the noise variance to obtain a more accurate noise variance. is there.

  The method of noise variance estimation proposed in the present invention for use in a wireless communication system receives a signal vector including a training sequence and a noise vector transmitted from a base station via at least one propagation path. Estimating a channel impulse response of each propagation path according to the signal vector and constructing a channel impulse response matrix, and the channel impulse response is mainly unchanged during a dedicated time of the training sequence. A noise variance of the signal vector according to the channel impulse response matrix and the signal vector.

  In order to describe the embodiment of the present invention in detail, TD-SCDMA is exemplified below.

  In TD-SCDMA, a base station transmits a signal vector to each UE in a corresponding time slot. According to the TD-SCDMA time slot format, the signal vector transmitted by the base station in the time slot corresponding to each UE consists of a training sequence and a spread user signal.

  For UEs assigned to the same time slot, the base station first combines the signal vector to be transmitted to each UE into a combined signal vector, and then combines the combined signal vector into the time slot. To each UE. The combined signal vector also consists of user signals and training sequences. Here, the user signal in the combined signal vector is obtained by combining the spread user signal in the signal vector to be transmitted to each UE, and the training sequence in the combined signal vector is transmitted to each UE. It is obtained by combining the training sequences in the signal vector to be transmitted.

  The training sequence assigned to each UE in the cell is obtained by performing different shift operations on the same basic training sequence. Therefore, the training sequence of the combined signal vector is considered as a basic training sequence. Since each UE acquires the basic training sequence used by the cell during the cell search process, the training sequence transmitted by the base station in the time slot is known in advance to each UE.

A training sequence included in a signal vector transmitted by the base station in a time slot reaches the UE through at least one propagation path, and the signal vector received by the UE in the time slot is r, Suppose that it consists of a training sequence and a noise vector n, and that the known value of the training sequence is s. From the equation (1), the signal vector r is expressed as follows.
r = Hs + n (2)
Here, H is a CIR matrix constructed by CIR of each propagation path between the UE and the base station.

は、以下のように表現されることができる。

ここで上付き文字^Ｈは、複素共役転置を表す。 Channel estimation method as described in “Low Cost Channel Estimation in the uplink receiver of CDMA mobile radio systems” by B. Steiner and PW Baier (Frequenz, Vol. 47, pp. 292-298, December 1993) According to the maximum likelihood estimated value of the training sequence contained in the signal vector r

Can be expressed as:

Here, the superscript ^H represents a complex conjugate transpose.

From equation (3), according to the known value s of the training sequence contained in the signal vector r, an estimate n ′ of the noise vector n is given by

ここで、Ｅ｛．｝は、期待値演算を示す。上述の式（５）の両辺の間に行列トレースの演算を実行することにより、ノイズベクトルｎの推定値

の平均分散

を計算する以下の式が容易に導かれる。

ここで、Ｎは前記トレーニングシーケンスのチップ幅（chip duration）であり、演算子ｔｒａｃｅ（）は行列のトレースの計算を意味し、σ^２は信号ベクトルｒのノイズ分散である。 Here, the covariance is as follows.

Here, E {. } Indicates an expected value calculation. Estimating the noise vector n by performing a matrix trace operation between both sides of equation (5) above

Mean variance

The following equation for calculating is easily derived.

Here, N is the chip duration of the training sequence, the operator trace () means the calculation of the matrix trace, and σ ² is the noise variance of the signal vector r.

が従来の方法を用いて計算される場合には非常に複雑になるであろう。実際には、分散

の計算は、１トレーニングシーケンスの時間に位置するノイズベクトルｎの推定値

について全ての要素の二乗平均値を算出することにより、当該時間においてチャネルが一定であるとみなされる場合には、近似されることができる。信号ベクトルｒのノイズ分散σ^２はこのとき以下のように導出されることができる。

Would be very complicated if calculated using conventional methods. In fact, distributed

Is an estimate of the noise vector n located at the time of one training sequence

Can be approximated by calculating the root mean square of all elements for, if the channel is considered constant at that time. The noise variance σ ² of the signal vector r can then be derived as follows:

In order to further improve the estimation performance, the noise variance σ ² calculated from the equation (7) in the time slot and the noise variance σ ² calculated from the equation (7) in each previous time slot are summed. It is also possible to average and take the average of the different σ _i ² as the noise variance σ ² of the signal vector r in the time slot.

  The above sections describe the principle of noise variance calculation by using the training sequence in the present invention.

  The following section describes in detail the proposed noise variance estimation method in conjunction with FIG.

  First, the UE receives a signal vector including a training sequence and a noise vector in a time slot transmitted through at least one propagation path from the base station (step S10).

  Second, the UE estimates the CIR of each propagation path according to the received signal vector, and constructs a CIR matrix by using the estimated CIR of each propagation path (step S20).

を推定する（ステップＳ３０）。 Third, the UE uses the equation (3) according to the signal vector and the CIR matrix to estimate the maximum likelihood of the training sequence included in the signal vector.

Is estimated (step S30).

及び前記トレーニングシーケンスの既知の値に従って、式（４）を利用することにより前記信号ベクトルに含まれるノイズベクトルの推定値ｎ’を計算する（ステップＳ４０）。ここで前記信号ベクトルに含まれるトレーニングシーケンスの既知の値は、セル検索処理において前記ＵＥによって取得される。 Fourth, the UE determines the maximum likelihood estimate (MLE) value of the training sequence included in the signal vector.

Then, according to the known value of the training sequence, the estimated value n ′ of the noise vector included in the signal vector is calculated by using Equation (4) (step S40). Here, a known value of the training sequence included in the signal vector is acquired by the UE in a cell search process.

Fifth, the UE calculates the noise variance σ ² of the signal vector by using Equation (7) according to the estimated value n ′ of the noise vector included in the signal vector and the CIR matrix H ( Step S50). Here, first, the power P _n ^{2 of} n ′ is calculated according to the formula p _n ² = (n ′) ^H (n ′), and then the trace cf of the matrix (H ^H H) is calculated. That is, cf = trace ((H ^H H) ⁻¹ ). Finally, the noise variance σ ² is calculated according to the equation σ ² = p _n ² / cf, ie equation (7).

Finally, the UE includes a noise variance sigma ² calculated from Equation (7) in said time slot, sums and averages the noise variance sigma ² calculated from Equation (7) in each preceding time slot, The average of the different σ _i ² is taken as the noise variance σ ² of the signal vector r in the time slot (step S60).

  A detailed description of the proposed noise variance estimation apparatus is given below in conjunction with FIGS.

  FIG. 3 is a block diagram showing a UE equipped with the proposed noise variance estimation apparatus. As shown in FIG. 3, in the cell search process before the UE communicates with the base station, the cell search means 40 acquires a basic training sequence used by the cell in which the UE exists. When the UE communicates with the base station, the UE antenna first transmits to the multiplexer 10 the signal vector Rx received in the time slot. The multiplexer 10 multiplexes the signal vector Rx received by the RF carrier generated by the VCO 20 and converts the signal vector Rx into a baseband signal vector. Next, the ADC 30 converts the baseband signal vector output from the multiplexer 10 into a digital baseband signal vector r. After that, the cell search means 40 synchronizes the digital baseband signal vector r output from the ADC 30, and the channel estimation means 50 uses the conventional channel estimation method for the synchronized digital baseband signal vector r. A CIR for each propagation channel is calculated and a CIR matrix is constructed using the calculated CIR for each propagation path. Next, the noise variance estimation means 60 performs digital base according to the CIR matrix calculated by the channel estimation means 50, the digital baseband signal vector r output by the ADC 30, and the basic training sequence obtained by the cell search means 40. The noise variance of the band signal vector r is calculated. Finally, the data detection means 70 uses a conventional data detection method, such as a multi-user detection method and turbo code decoding, so that the digital baseband signal is in accordance with the noise variance calculated by the noise variance estimation means 60. A desired user signal is obtained from the vector r.

を推定する、等化手段６０１と、
式（４）を用いて、等化手段６０１によって計算された前記ディジタルベースバンド信号ベクトルｒに含まれるトレーニングシーケンスのＭＬＥ値

と、基本トレーニングシーケンスｓ（即ち前記ディジタルベースバンド信号ベクトルｒに含まれるトレーニングシーケンスの既知の値）とに従って、前記ディジタルベースバンド信号ベクトルｒに含まれるノイズベクトルの推定値ｎ’を算出する、ノイズ推定手段６０２と、
式ｐ_ｎ ^２＝（ｎ’）^Ｈ（ｎ’）を用いて、ノイズ推定手段６０２によって計算された前記ディジタルベースバンド信号ベクトルｒに含まれるノイズベクトルの推定値ｎ’に従って、前記ノイズベクトルの推定値ｎ’のべき乗ｐ_ｎ ^２を算出する、ノイズべき乗算出手段６０３と、
行列（（Ｈ^ＨＨ）^−１）のトレースｃｆ、即ちｃｆ＝ｔｒａｃｅ（（Ｈ^ＨＨ）^−１）を計算する等化修正手段６０４と、
式σ^２＝ｐ_ｎ ^２／ｃｆを用いて、ノイズべき乗計算手段６０３により算出された前記ノイズベクトルの推定値ｎ’のべき乗ｐ_ｎ ^２と、等化修正手段６０４によって計算されたトレースｃｆとに従ってノイズ分散σ^２を算出する、ノイズべき乗修正手段６０５と、
を有する。 FIG. 4 is a block diagram showing the noise variance estimating means 60. Referring to FIG. 4, the noise variance estimation means 60
Using the equation (3), the MLE value of the training sequence included in the digital baseband signal vector r according to the CIR matrix H calculated by the channel estimation means 50 and the digital baseband signal vector r output from the ADC 30

Equalizing means 601 for estimating
The MLE value of the training sequence included in the digital baseband signal vector r calculated by the equalization means 601 using the equation (4)

And an estimated value n ′ of the noise vector included in the digital baseband signal vector r according to the basic training sequence s (that is, a known value of the training sequence included in the digital baseband signal vector r). Estimating means 602;
Using the expression p _n ² = (n ′) ^H (n ′), the noise vector estimation is performed according to the noise vector estimation value n ′ included in the digital baseband signal vector r calculated by the noise estimation unit 602. A noise power calculation means 603 for calculating a power p _n ² of the value n ′;
Equalization correction means 604 for calculating a trace cf of a matrix ((H ^H H) ⁻¹ ), ie, cf = trace ((H ^H H) ⁻¹ );
According to the formula σ ² = p _n ² / cf, the power _pn ^{2 of the} noise vector estimate n ′ calculated by the noise power calculator 603 and the trace cf calculated by the equalization corrector 604 Noise power correction means 605 for calculating noise variance σ ² ;
Have

  As mentioned above, in the proposed noise variance estimation method and apparatus for use in a wireless communication system, a training sequence is utilized to calculate the noise variance, so that the calculated noise variance is more accurate. Can meet the requirements for.

  It will be appreciated that the noise variance estimation method and apparatus for use in a wireless communication system disclosed in the present invention may be varied considerably without significantly departing from the spirit and scope of the present invention as defined by the appended claims. It should be understood by those skilled in the art.

2 shows a conventional CDMA downlink transmission model. It is a flowchart which shows the noise dispersion | distribution estimation method in this invention. It is a block diagram which shows the user apparatus (UE) provided with the noise dispersion | distribution estimation apparatus in the Example of this invention. It is a block diagram which shows the noise dispersion | distribution estimation apparatus in the Example of this invention.

Claims (13)

A noise variance estimation method to be performed by a user equipment, comprising:
(A) receiving a signal vector including a training sequence and a noise vector transmitted from the base station via at least one propagation path;
(B) estimating a channel impulse response of each propagation path according to the signal vector and constructing a channel impulse response matrix;
(C) calculating a noise variance of the signal vector according to the channel impulse response matrix and the signal vector if the channel impulse response is mainly unchanged during a dedicated time of the training sequence; ,
Having a method.   The method of claim 1, wherein the dedicated time is a duration of the training sequence. The step (c)
(C1) estimating a maximum likelihood estimate value of a training sequence included in the signal vector according to the channel impulse response matrix and the signal vector;
(C2) calculating an estimate value of a noise vector included in the signal vector according to a maximum likelihood estimate value of the training sequence and a known value of the training sequence;
(C3) calculating a noise variance of the signal vector according to the estimated value of the noise vector and the channel impulse response matrix;
The method of claim 2 comprising: The step (c3) has the formula:
σ ² ≈ (n ′ ^H n ′) / trace {(H ^H H) ⁻¹ }
To calculate the noise variance of the signal vector, where
σ ² is the noise variance of the signal vector,
n ′ is an estimate of the noise vector contained in the signal vector;
H is the channel impulse response matrix, the superscript ^H represents the complex conjugate transpose,
4. The method of claim 3, wherein trace {} represents a matrix trace calculation.   The noise variance of the signal vector and the noise variance calculated in the previous time slot are summed and then averaged, further comprising the step of setting the average noise variance as the noise variance of the signal vector. The method described. An apparatus for noise variance estimation,
Receiving means for receiving a signal vector including a training sequence and a noise vector transmitted from the base station via at least one propagation path;
Channel estimation means for estimating a channel impulse response of each propagation path according to the signal vector and constructing a channel impulse response matrix;
A calculation means for calculating a noise variance of the signal vector according to the channel impulse response matrix and the signal vector when the channel impulse response is mainly unchanged during a dedicated time of the training sequence;
Having a device.   The apparatus of claim 6, wherein the dedicated time is a duration of the training sequence. The calculating means is
Equalization means for estimating a maximum likelihood estimate of a training sequence included in the signal vector according to the channel impulse response matrix and the signal vector;
Noise estimating means for calculating an estimated value of a noise vector included in the signal vector according to a maximum likelihood estimated value of the training sequence and a known value of the training sequence;
Noise power calculating means for calculating the power of the estimated value of the noise vector according to the estimated value of the signal vector;
Noise power correction means for calculating a noise variance of the signal vector according to the power of the estimated value of the noise vector and the channel impulse response matrix;
The apparatus of claim 7, comprising: The noise power correction means has the formula:
σ ² ≈ (n ′ ^H n ′) / trace {(H ^H H) ⁻¹ }
To calculate the noise variance of the signal vector, where
σ ² is the noise variance of the signal vector,
n ′ is an estimated value of a noise vector included in the signal vector, n ′ ^H n ′ is a power of the estimated value of the noise vector,
H is the channel impulse response matrix, the superscript ^H represents the complex conjugate transpose,
9. The apparatus of claim 8, wherein trace {} represents a matrix trace calculation. Receiving means for receiving a signal vector including a training sequence and a noise vector transmitted from the base station via at least one propagation path;
Channel estimation means for estimating a channel impulse response of each propagation path according to the signal vector and constructing a channel impulse response matrix;
Noise variance estimation means for calculating a noise variance of the signal vector according to the channel impulse response matrix and the signal vector when the channel impulse response is mainly unchanged during a dedicated time of the training sequence; ,
A user device.   The user apparatus according to claim 10, wherein the dedicated time is a duration of the training sequence. The noise variance estimating means is
Equalization means for estimating a maximum likelihood estimate of a training sequence included in the signal vector according to the channel impulse response matrix and the signal vector;
Noise estimating means for calculating an estimated value of a noise vector included in the signal vector according to a maximum likelihood estimated value of the training sequence and a known value of the training sequence;
Noise power calculating means for calculating the power of the estimated value of the noise vector according to the estimated value of the signal vector;
Noise power correction means for calculating a noise variance of the signal vector according to the estimated value of the noise vector and the channel impulse response matrix;
The user apparatus according to claim 11, comprising: The noise power correction means has the formula:
σ ² ≈ (n ′ ^H n ′) / trace {(H ^H H) ⁻¹ }
To calculate the noise variance of the signal vector, where
σ ² is the noise variance of the signal vector,
n ′ is an estimated value of a noise vector included in the signal vector, n ′ ^H n ′ is a power of the estimated value of the noise vector,
H is the channel impulse response matrix, the superscript ^H represents the complex conjugate transpose,
13. The user equipment according to claim 12, wherein trace {} represents a matrix trace calculation.

Images