CN1625075A - Noise variance estionating method and device for radio communication system - Google Patents

Abstract

A noise variance estimation method and device used in radio communication system. This method includes the following processes: receive the signal vector that is from the basic station and transmitted by at least one transmitting path and contains training sequence and noise vector; estimate channel impulsion response of every transmitting path according to the signal vector to compose a channel impulsion response matrix; if the channel impulsion response remains the same in the given time range of the training sequence, calculate the noise variance of the signal vector according to the channel impulsion response matrix and the signal vector.

Description

A kind of noise variance evaluation method and device that is used for wireless communication systems

Technical field

The present invention relates to a kind of noise variance evaluation method and device that is used for wireless communication systems, relate in particular to a kind of method and device that utilizes training sequence to carry out the noise variance estimation.

Technical background

CDMA (code division multiple access) is a kind of novel wireless communication technology that grows up afterwards at FDMA (frequency division multiple access) and TDMA (time division multiple access).In the cdma wireless communication technology, different user terminals is assigned with different mutually orthogonal spreading codes, and the signal of the different spreading code spread processing of employing that sent by different user terminals can transmit on identical frequency band.

The paper of being write by A.Klein on VTC in 1997 (vehicle technology meeting) periodical " Data Detection Algorithms Specially Designed For The Downlink ofCDMA Mobile Radio Systems (being in particular the Data Detection Algorithm of the down link design of CDMA mobile wireless system) " has proposed a kind of mode of CDMA down link.The mode of this CDMA down link as shown in Figure 1.As seen from Figure 1, for signal vector d ⁽¹⁾..., d ^(k)..., d ^(k)(wherein, signal vector d ^(k)(k=1...K) constitute by N plural elements) send to respectively user terminal 1 ..., k ..., K, base station 200 at first use distribute to user 1 ..., k ..., the spreading code of K c _d ⁽¹⁾..., c _d ^(k)..., c _d ^(K)Respectively to signal vector d ⁽¹⁾..., d ^(k)..., d ^(K)Carry out spread spectrum, then will be through the signal vector of spread spectrum d ⁽¹⁾..., d ^(k)..., d ^(K)Be merged into signal vector s _dAnd send to each relevant user terminals 220 simultaneously via identical channel 210.Suppose signal vector s _dArrive user terminal k (k=1...K) through a plurality of transmission paths, and the channel impulse response of each transmission path is respectively h _{D (i)} ^(k)(i=1,2 ...), the signal vector received of user terminal k then e _d ^(k)Can describe by equation (1):

e _d ^(k)＝ H _d ^(k) C _d d+ n _d ^(k)＝ H _d ^(k) s _d+ n _d ^(k) (1)

In following formula, H _d ^(k)It is channel impulse response by each transmission path h _{D (i)} ^(k)(i=1,2 ...) the channel impulse response matrix that constitutes, C _dBe by spreading code c _d ⁽¹⁾..., c _d ^(k)..., c _d ^(K)The spreading code matrix that structure obtains ( H _d ^(k)With C _dConcrete building method referring to the above-mentioned paper of writing by A.Klein), d=( d ^{(1) T}..., d ^{(k) T}..., d ^{(K) T}) ^T, [.] ^TThe representing matrix transposition, s _dIt is right to represent dCarry out the signal vector that obtains after spread spectrum merges and s _d= C _d d, n _d ^(k)It is noise vector.

By equation (1) as can be seen, received signal vector e _d ^(k)In not only comprise user terminal k and want the signal vector that receives d ^(k), comprise that also the base station sends to the signal vector and the noise vector of other user terminals.

In order to make the user terminal k can be from received signal vector e _d ^(k)In obtain the base station with minimal error and send to its signal vector d ^(k)The paper that people have proposed many signal acceptance methods. is delivered by Kimmo Kettunen on VTC (vehicle technology meeting) periodical in 1999 " Iterative Multiuser Receiver/Decoders With Enhanced VarianceEstimation (having the iteraction multiple users receiver/decoder that strengthens the variance estimation) ", and the paper of being delivered by A.Klein on the IEEE Transaction in May, 1996 on Vehicular Technology the 45th phase 276-287 page or leaf " Zero Forcing and MininumMean-Square-Error Equalization for Multiuser Detection inCode-Division multiple-access channels (ZF and the Minimum Mean Square Error that are used for CDMA channel are balanced) " is described these signal acceptance methods. From the description of these signal acceptance methods as can be known, they rely on channel information (that is:noise variance) all and obtain the signal vector that desire receives from received signal vectors, therefore for the signal vector that desire receives that obtains of minimal error, need to calculate accurate noise variance (noise variance).

In order to obtain more accurate noise variance, people have proposed various noise removing methods.Such as, on in April, 1997 Proc.Of ITC ' 97 173-178 pages or leaves by the paper " a novel variance estimator for turbo-codedecoding (a kind of new variance estimation device that is used for the decoding of turbo sign indicating number) " of M.Reed and J.Asenstorfer co-present, a kind of traditional variance estimating techniques that are used for awgn channel that propose, the American National patent publication No. is US20020110199, Rake (RAKE) technology that is used to eliminate the multipath interference that denomination of invention proposes for " Method for Noise Energy Estimation inTDMA Systems (method that is used for the noise energy estimation in the TDMA system) ", the process of convolution that also has some to use training sequence is come the noise removing method of calculating noise variance in addition.Described these noise removing methods can both satisfy the required precision of second generation wireless communication systems to noise variance.

But, in third generation wireless communication systems, signal acceptance method requires accurate more noise variance, such as, the key technology multi-user test method and the turbo sign indicating number coding/decoding method that are used for third generation wireless communication systems just require point-device noise variance, and existing noise removing method can not satisfy the required precision of third generation wireless communication systems to noise variance.

Summary of the invention

An object of the present invention is to provide a kind of noise variance evaluation method and device that is used for wireless communication systems.In this noise variance evaluation method and device, use training sequence to come calculating noise variance, to obtain the very high noise variance of precision.

According to a kind of noise variance evaluation method of in a user terminal, carrying out of the present invention, comprise step: receive the signal vector that comprises training sequence and noise vector via at least one transmission path from the base station; According to this signal vector, estimate the channel impulse response of each transmission path, to constitute a channel impulse response matrix; If this channel impulse response remains unchanged substantially,, calculate the noise variance of this signal vector then according to this channel impulse response matrix and this signal vector in the special time length of this training sequence.

The accompanying drawing summary

Fig. 1 is the mode of conventional CDMA down link;

Fig. 2 is the flow chart of noise variance evaluation method of the present invention;

Fig. 3 is the block diagram of user terminal of using the noise variance estimating device of one embodiment of the invention;

Fig. 4 is the block diagram of the noise variance estimating device of one embodiment of the invention.

Detailed Description Of The Invention

Be example below with TD-SCDMA, describe one embodiment of the present of invention in detail.

In TD-SCDMA, the base station in corresponding time slot to each user terminal transmission signal vector.According to the time slot format of TD-SCDMA, the signal vector that the base station sends to each user terminal in time slot is made of the subscriber signal of training sequence and process spread spectrum.

For the user terminal that is distributed in identical time slot, the base station at first merges the signal vector that sends to each user terminal and obtains the combined signal vector, in this time slot this combined signal vector is sent to each user terminal then.Described combined signal vector also is made of subscriber signal and training sequence two parts, wherein, the subscriber signal of this combined signal vector is obtained by the subscriber signal merging through spread spectrum in the signal vector that sends to each user terminal, and the training sequence of this combined signal vector is obtained by the merging of the training sequence in the signal vector that sends to each user terminal.

Reside in a training sequence that each user terminal distributed in the sub-district and obtain through different displacements, so the training sequence of this combined signal vector can be regarded as basic training sequences by identical basic training sequences.Again because each user terminal has just obtained the basic training sequences that use this sub-district in cell search process, so the training sequence that the base station sends in time slot is known to each user terminal.

Suppose that the training sequence that signal vector that the base station is sent comprises arrives a user terminal via at least one transmission path in a time slot, the signal vector that is made of described training sequence and noise vector n that this user terminal is received at this time slot is r, and the given value of described training sequence is s, then according to equation (1), signal vector r can be expressed from the next:

r＝Hs+n (2)

Wherein, H is the channel impulse response matrix that the channel impulse response by each transmission path between this user terminal and base station constitutes.

According to the channel estimation methods that the paper of being write by B.Steiner and P.W.Baier on 1993 11/12 month Frequenz the 47th phase 292-298 page or leaf " Low Cost Channel Estimation in theuplink receiver of CDMA mobile radio systems (the low-cost channel estimating in the up-link receiver of CDMA mobile radio system) " is proposed, the maximum likelihood estimator of the training sequence that signal vector r is comprised (the maximumlikehood estimate)  can be described by following formula:

＝(H ^HH) ^-1H ^Hr＝s+(H ^HH) ^-1H ^Hn＝s+n′ (3)

In following formula, subscript ^HThe expression complex-conjugate transpose.

By equation (3),, can be easy to the estimated value n ' that signal calculated vector r comprises noise vector n according to the given value s of signal vector training sequence that r comprises:

n′＝-s＝(H ^HH) ^-1H ^Hn (4)

The covariance matrix that signal vector r comprises the estimated value n ' of noise vector n is:

E{n′n′ ^H}＝E{H ^HH) ^-1H ^Hn·n ^HH(H ^HH) ^-1}

＝(H ^HH) ^-1H ^HE(nn ^H)H(H ^HH) ^-1}

＝σ ²(H ^HH) ^-1 (5)

Wherein, E{.} is the computing of carry out desired value.The operation of asking trace of a matrix (trace) is carried out on the both sides of equation (5), just can be obtained the average variance σ of the estimated value n ' of the noise vector n that signal vector r comprised _{N '} ²:

${\stackrel{\‾}{\mathrm{\σ}}}_{n^{\′}}^2={\mathrm{\σ}}^2\·\mathrm{trace}\left\{{\left(H^{H}H\right)}^{-1}\right\}/N---\left(6\right)$

In following formula, N is the chip lengths of the training sequence that comprised of signal vector r, and operator trace () is the operation of asking trace of a matrix, σ ²It is the noise variance that needs estimated signals vector r.

Signal vector r comprises the average variance σ of the estimated value n ' of noise vector n _{N '} ²It is cumbersome calculating according to the method for routine.In fact, if the channel in the time span of signal vector training sequence that r comprises can be considered to indeclinable, the average variance σ of the estimated value n ' of noise vector n then _{N '} ²Can be approximated to be the mean-square value of the middle all elements of estimated value n ' of noise vector n, therefore, need the noise variance σ of estimated signals vector r ²Can represent by equation (7):

σ ²≈(n′ ^Hn′)/trace{(H ^HH) ^-1} (7)

In order further to improve estimated performance, this time slot can be utilized the noise variance σ of the signal vector r that equation (7) calculates ²Utilized equation (7) to calculate noise variance σ with former each time slot ²Carry out addition and ask on average, and obtaining the noise variance σ of average noise variance as this time slot signal vector r ²

Above-mentioned is exactly the principle of use training sequence calculating noise variance of the present invention.

Below in conjunction with Fig. 2, describe noise variance evaluation method of the present invention in detail.

At first, user terminal receives the signal vector that comprises training sequence and noise vector (step S10) via at least one transmission path from the base station in a time slot.

Secondly, user terminal is estimated the channel impulse response of each transmission path according to described signal vector, and the channel impulse response of each transmission path that is obtained by estimation constitutes a channel impulse response matrix H (step S20).

Then, user terminal uses the maximum likelihood estimator  (step S30) of equation (3) the estimation training sequence that described signal vector comprised according to described signal vector and described channel impulse response matrix.

Then, user terminal uses equation (4) to calculate the estimated value n ' (step S40) of noise vector that described signal vector comprises according to the maximum likelihood estimator  of training sequence that described signal vector comprises and the given value of training sequence.Wherein, the given value of training sequence that described signal vector comprises is obtained in cell search process by user terminal.

Afterwards, user terminal uses equation (7) to calculate the noise variance σ of described signal vector according to the estimated value n ' and the described channel impulse response matrix H of noise vector that described signal vector comprises ²(step S50).Wherein, can use equation p earlier according to the estimated value n ' of this noise vector that calculates _n ²=(n ') ^HThe power p of the estimated value n ' of (n ') calculating noise vector _n ²Compute matrix ((H again ^HH) ^-1) mark cf, i.e. cf=trace ((H ^HH) ^-1); At last according to the power p of the estimated value n ' of this noise vector _n ²And the mark cf that calculates, use equation σ ²=p _n ²/ cf, that is: equation (7), calculating noise variance σ ²

At last, user terminal utilizes this time slot the noise variance σ of the described signal vector that equation (7) calculates ²The noise variance σ that utilizes equation (7) to calculate with former each time slot ²Carry out addition and ask on average, and obtaining the noise variance σ of average noise variance as the described signal vector of this time slot ²(step S60).

Below in conjunction with Fig. 3 and Fig. 4, describe noise variance estimating device of the present invention in detail.

Fig. 3 is a kind of block diagram of using the user terminal of noise variance estimating device of the present invention.As shown in Figure 3, in the cell search process before user terminal and base station communicate, cell search unit 40 is obtained the basic training sequences s that this user terminal institute persistent district uses.When user terminal and base station communicate, the signal vector Rx that the antenna of user terminal is received in a time slot at first delivers to multiplier 10, this multiplier 10 multiplies each other the radio-frequency carrier that the signal vector Rx that receives and voltage-controlled oscillator (VCO) 20 generate, so that this signal vector Rx is transformed into the baseband signal vector; Then, AD conversion unit (ADC) 30 converts the baseband signal vector of multiplier 10 outputs to digital baseband signal vector r; Then, the digital baseband signal vector r of 40 pairs of AD conversion unit of cell search unit, 30 outputs carries out Synchronous Processing, the digital baseband signal vector r that 50 pairs of processes of channel estimation unit are synchronous, use conventional channel estimation, calculate the channel impulse response of each transmission path, and constitute the channel impulse response matrix by the channel impulse response of the transmission path that calculates; Next, the digital baseband signal vector r of the noise variance evaluation unit 60 channel impulse response matrix that 50 estimations obtain according to channel estimation unit, AD conversion unit 30 outputs and the basic training sequences s that cell search unit 40 is obtained, the noise variance of calculating digital baseband signal vector r; At last, the noise variance that Data Detection unit 70 calculates according to noise variance evaluation unit 60, use conventional Data Detection method,, from digital baseband signal vector r, obtain the subscriber signal that needs such as multi-user test method and turbo sign indicating number coding/decoding method etc.

Fig. 4 is the block diagram of noise variance evaluation unit 60.As shown in Figure 4, noise variance evaluation unit 60 comprises:

Balanced unit 601, be used for the channel impulse response matrix H that calculates according to channel estimation unit 50, the digital baseband signal vector r of AD conversion unit 30 outputs, use equation (3) to calculate the maximum likelihood estimator  of described digital baseband signal vector training sequence that r comprises;

Noise estimation unit 602, the maximum likelihood estimator  and the basic training sequences s that are used for the described digital baseband signal vector training sequence that r comprises that calculates according to balanced unit 601, that is: the given value of described digital baseband signal vector training sequence that r comprises uses equation (4) to calculate the estimated value n ' of described digital baseband signal vector noise vector that r comprises;

Noise power calculation unit 603 is used for the estimated value n ' of the described digital baseband signal vector noise vector that r comprises that calculates according to noise estimation unit 602, use equation p _n ²=(n ') ^HThe power p of the estimated value n ' of (n ') calculating noise vector _n ²

Isostatic correction unit 604 is used for compute matrix ((H ^HH) ^-1) mark cf, also be cf=trace ((H ^HH) ^-1);

Noise power correcting unit 605 is used for the power p of the estimated value n ' of the described digital baseband signal vector noise vector that r comprises that calculates according to noise power calculation unit 603 _n ²And the mark cf that calculates of isostatic correction unit 604, use equation σ ²=p _n ²/ cf calculating noise variance σ ²

Beneficial effect

In sum, owing to estimate at the noise variance for wireless communication systems provided by the invention In calculation method and the device, come calculating noise variance with training sequence, what therefore calculate makes an uproar The sound variance can satisfy the more application of high-precision requirement.

It will be appreciated by those skilled in the art that disclosed in this invention for wireless communication systems Noise variance evaluation method and device can be made each on the basis that does not break away from content of the present invention Plant and improve. Therefore, protection scope of the present invention should be true by the content of appending claims Fixed.

Claims (13)

1, a kind of noise variance evaluation method of carrying out in a user terminal comprises step:

(a) reception is from the signal vector that comprises training sequence and noise vector via at least one transmission path of base station;

(b), estimate the channel impulse response of each transmission path, to constitute a channel impulse response matrix according to this signal vector;

(c),, calculate the noise variance of this signal vector then according to this channel impulse response matrix and this signal vector if this channel impulse response remains unchanged substantially in the special time length of this training sequence.

2, the method for claim 1, wherein said special time length are the time spans of described training sequence.

3, method as claimed in claim 2, wherein, step (c) comprising:

(c1), estimate the maximum likelihood estimator of training sequence that described signal vector comprises according to described channel impulse response matrix and described signal vector;

(c2), calculate the estimated value of noise vector that described signal vector comprises according to the given value of the maximum likelihood estimator and the described training sequence of this training sequence;

(c3), calculate the noise variance of described signal vector according to the estimated value and the described channel impulse response matrix of this noise vector.

4, method as claimed in claim 3, wherein step (c3) is calculated the noise variance of described signal vector according to following formula:

σ ²≈(n’ ^Hn’)/trace{(H ^HH) ^-1}

Wherein:

σ ²It is the noise variance of described signal vector;

N ' is the estimated value of noise vector that described signal vector comprises;

H is described channel impulse response matrix, subscript ^HThe expression complex-conjugate transpose;

Trace{} represents to ask matrix trace.

5, as claim 3 or 4 described methods, wherein, also comprise step:

The noise variance that the noise variance of described signal vector and previous time slot are calculated carries out addition to be asked on average, and with the average noise variance that the obtains noise variance as described signal vector.

6, a kind of noise variance estimating device comprises:

A receiving element is used to receive the signal vector that comprises training sequence and noise vector via at least one transmission path from the base station;

A channel estimation unit is used for estimating the channel impulse response of each transmission path according to this signal vector, to constitute a channel impulse response matrix;

A computing unit is used for if this channel impulse response remains unchanged in the special time length of this training sequence substantially, then according to this channel impulse response matrix and this signal vector, calculates the noise variance of this signal vector.

7, device as claimed in claim 6, wherein said special time length are the time spans of described training sequence.

8, device as claimed in claim 7, wherein said computing unit comprises:

A balanced unit is used for estimating the maximum likelihood estimator of training sequence that described signal vector comprises according to described channel impulse response matrix and described signal vector;

A noise estimation unit is used for the given value according to the maximum likelihood estimator and the described training sequence of this training sequence, calculates the estimated value of noise vector that described signal vector comprises;

A noise power calculation unit is used for the estimated value according to this noise vector, calculates the power of the estimated value of this noise vector;

A noise power correcting unit is used for the power according to the estimated value of the noise vector of this calculating, and described channel impulse response matrix, calculates the noise variance of described signal vector.

9, device as claimed in claim 8, wherein said noise power correcting unit calculates the noise variance of described signal vector according to following formula:

σ ²≈(n’ ^Hn’)/trace{(H ^HH) ^-1}

Wherein:

σ ²It is the noise variance of described signal vector;

N ' is the estimated value of noise vector that described signal vector comprises, n ' ^HN ' is the power of the estimated value of described noise vector;

H is described channel impulse response matrix, subscript ^HThe expression complex-conjugate transpose;

Trace{} represents to ask matrix trace.

10, a kind of user terminal comprises:

A receiving element is used to receive the signal vector that comprises training sequence and noise vector via at least one transmission path from the base station;

A channel estimation unit is used for estimating the channel impulse response of each transmission path according to this signal vector, to constitute a channel impulse response matrix;

A noise variance evaluation unit is used for if this channel impulse response remains unchanged in the special time length of this training sequence substantially, then according to this channel impulse response matrix and this signal vector, calculates the noise variance of this signal vector;

A data detecting unit is used for the noise variance according to the signal vector of this calculating, the signal vector that is received is detected, with the signal that needing to obtain.

11, user terminal as claimed in claim 10, wherein said special time length are the time spans of described training sequence.

12, user terminal as claimed in claim 11, wherein said noise variance evaluation unit comprises:

A balanced unit is used for estimating the maximum likelihood estimator of training sequence that described signal vector comprises according to described channel impulse response matrix and described signal vector;

A noise estimation unit is used for the given value according to the maximum likelihood estimator and the described training sequence of this training sequence, calculates the estimated value of noise vector that described signal vector comprises;

A noise power calculation unit is used for the estimated value according to this noise vector, calculates the power of the estimated value of this noise vector;

A noise power correcting unit is used for the power according to the estimated value of the noise vector of this calculating, and described channel impulse response matrix, calculates the noise variance of described signal vector.

13, user terminal as claimed in claim 12, wherein said noise power correcting unit calculates the noise variance of described signal vector according to following formula:

σ ²≈(n’ ^Hn’)/trace{(H ^HH) ^-1}

Wherein:

σ ²It is the noise variance of described signal vector;

N ' is the estimated value of noise vector that described signal vector comprises, n ' ^HN ' is the power of the estimated value of described noise vector;

H is described channel impulse response matrix, subscript ^HThe expression complex-conjugate transpose;

Trace{} represents to ask matrix trace.