Background technology
In a lot of communication systems, in order to improve transmission reliability, the channel impulse response of the channel that need obtain passing through in the signals transmission, i.e. channel estimating.For example, in the TD-SCDMA system, in order to realize joint-detection, wave beam forming or other measuring process, channel estimation process is essential.In a lot of communication systems, channel estimating often realizes through the mode of sending training sequence.In data detection process, can suppress noise and interference targetedly by channel information owing to knowing, improve data detection performance.
But the training itself that is used as channel estimating also can receive the influence of noise and interference, makes channel estimating not accurate enough, in addition, much adopts in the system of blind Channel Estimation, because channel estimation methods is limit, channel estimating is also very inaccurate.This will be to process deleterious impact such as the Sequence Detection of later use channel estimating, wave beam forming, measurements.
For this reason, after obtaining channel estimating, often need carry out some special noise reduction process, be called signal channel estimation denoising post-treatment.A kind of method relatively more commonly used is exactly the noise gate reprocessing.The main contents of noise gate reprocessing comprise: estimate the noise power in the channel estimating; Set a thresholding according to estimated noise power; The channel estimation value that is lower than this thresholding is thought noise, and only keeps the channel estimating tap (directly) greater than this thresholding.For multiaerial system, often each antenna independently calculates channel estimation results, and each antenna carries out this noise gate noise reduction process respectively; Perhaps at first utilize a plurality of antennas to estimate respectively to obtain averaging after the noise power; Obtain an average noise power; Then power and average by antenna is asked in each tap of the channel estimating that obtains on each antenna respectively, the result after average is carried out threshold processing.Describe in detail through mathematic(al) representation below:
If antenna element k
aThe channel impulse response that last estimation obtains can be written as:
Here L is a training sequence length.K
aBe the antenna element number.
At first, through the channel estimation results on each antenna, estimate to obtain the noise σ on each antenna
Ka 2, k
a=1...K
aAnd it asked on average obtain average noise:
Then to obtain on each antenna the channel estimating structure ask power with, promptly obtain:
According to sequence
each antenna in the formula (1) is carried out threshold processing then; Promptly set through
thresholding
in
less than the tap position of this thresholding, to tap zero setting corresponding in the formula (1).
Existing signal channel estimation denoising post-treatment method is under the high s/n ratio condition, and the estimation of noise is a kind of efficient ways more accurately under the condition.But under the low signal-to-noise ratio condition, some footpath will be submerged in the noise owing to power is less, is abandoned thereby be taken as noise, will cause channel estimating inaccurate like this.In addition, for the situation that many antennas receive, simple here power addition, the not noise reduction capability of good use smart antenna.
Embodiment
The present invention program's the noise reduction capability of to the effect that utilizing many antennas is carried out noise reduction process to channel estimation results, thereby finds useful footpath more accurately, the elevator system performance.The present invention program mainly is applicable to little spacing antenna array, like uniform circular array, and even linear array etc.
For making the object of the invention, technical scheme and advantage clearer, the present invention is done further to set forth in detail below in conjunction with specific embodiment.
The flow process of the signal channel estimation denoising post-treatment of embodiment of the invention realization multiaerial system is as shown in Figure 1, comprises the steps:
Step 101: try to achieve each antenna channel estimated result.The channel estimation methods that adopts can be a prior art, as carrying out channel estimating through training sequence or carrying out blind Channel Estimation etc.Channel estimation results is designated as:
Wherein
Represent the channel estimation results that each antenna element is total,
Expression is to k user's channel estimation results,
Represent k
aThe channel estimation results of individual antenna element,
Represent k
aIn the channel estimation results of individual antenna element, belong to k user's component,
Step 102: channel estimation results is carried out the noise reduction first time, and the noise-reduction method that is adopted here can be a prior art, like the method for mentioning in the background technology.If current each the antenna channel estimated result after the noise reduction process is:
And:
Wherein, h
k aRepresent k
aChannel estimation results behind the individual antenna element noise reduction, h
k k aFor belonging to k user's component in the channel estimation results behind this noise reduction.
Can each antenna channel estimated result of k user be written as:
Then have: h=[h
1, h
2..., h
K] (8)
Here k
aExpression antenna subscript, K representes total number of users.W representes that each user's window is long, and it has represented certain multidiameter delay.
Step 103: the figuration combined vector of trying to achieve each user's multi-antenna channel estimation
Here the figuration combined vector w of user k
kCan be according to the channel estimation results h in the formula (7)
kTry to achieve.The method for solving here is same as existing beamforming algorithm, promptly at first through (7), tries to achieve the space covariance matrix of user k
Then through finding the solution R
kEigenvalue of maximum characteristic of correspondence vector, said characteristic vector is said user's figuration combined vector w
k, this is commonly called eigen beam figuration method (EBB algorithm), perhaps passes through R
k(Direction Arrive Wave DOA) estimates, pass through arrival bearing's estimated result of user k then, and antenna obtains said user's figuration combined vector w corresponding to the array response vector of this direction to carry out arrival bearing user
k, this is commonly called fixed beam figuration method.
Owing to use the system of little spacing antenna array to tend to realize that DOA estimates and descending wave beam forming, thereby can directly take in the weight coefficient of wave beam forming and be used as the figuration combined vector w here
kRealize that like this amount of calculation does not increase.
Step 104: the channel estimation results
before the noise reduction carried out figuration and merged the first time that
step 101 is obtained, and was about to the channel estimation results that the figuration combined vector multiply by respective user; And it is vectorial that each user's figuration amalgamation result is lined up a row.Promptly have:
Theoretical according to smart antenna, the signal to noise ratio of the merging channel estimation results in the formula (10) is estimated with respect to each antenna channel in the formula (6)
Middle signal to noise ratio is wanted high about 10log (K
a) (dB), therefore, it is can Billy accurately a lot of with the channel estimation results in the formula (6) to adopt in the formula (10) channel estimation results to carry out noise reduction process.
Step 105: the row vector to obtaining after the figuration merging carries out noise reduction process according to the single antenna noise reduction algorithm; Channel estimation results
the noise reduction process here adopts existing noise-reduction method after obtaining noise reduction, repeats no more.
Step 106: each antenna channel estimated result before the noise reduction
is carried out noise reduction process according to
; The tap position of zero setting in promptly to
, the middle also zero setting of relevant position with
.In fact exactly will merge the tap zero setting of thinking noise in the channel estimating of back, can write down the zero setting position, the tap zero setting of correspondence position during original channel is estimated.Channel estimation results after the noise reduction process is as final channel estimation results.
Here mainly be the noise reduction and inhibition interference performance that has utilized smart antenna, promoted the signal to noise ratio of channel estimating, thereby promoted the right probability of useful grade, and then promoted systematic function.More serious for outer area interference, under the condition that perhaps signal to noise ratio is lower, the method is here compared existing method and is had improvement in performance clearly.
For step 102; Noise reduction mainly is to have promoted the accuracy that figuration merging weights calculate in the step 103 for the first time, if amount of calculation is too big, can skip; Directly utilize the original channel that obtains in the step 101 to estimate to carry out figuration and merge weights calculating, also can obtain preferable performance.
For in the step 103, the figuration combined vector can directly be taken the weight vector in down beam shaping.For wherein estimating earlier DOA, obtain the method for figuration weight vector again through DOA, then directly the system of getting is used to locate or the DOA estimated result of other functions calculates, thereby saves amount of calculation greatly.So, the present invention is actually with few amount of calculation lifting and has brought bigger systematic function to promote.
Formed a wave beam because the figuration here merges to be equivalent to like Fig. 2, wherein, disk representation space angle, from 0 ° of angle to ± 180 °, disk diameter to the size that gains of numeric representation.Curve representation is along footpath, the space of all directions, is main lobe 201 along the comparatively thick of 90 ° of directions wherein, and what other was less is secondary lobe 202.This wave beam has played certain space filtering effect, thereby has promoted signal to noise ratio.But main lobe 201 width are limited, thereby, when the environmental diffusion is very big, has a small amount of secondary lobe 202 and drop on beyond the main lobe scope, this moment, filtering also possibly given by space filtering in these useful footpaths.Thereby, can adopt following method to obtain the figuration combined vector in the step 103.Promptly calculate the broad beam figuration weight coefficient of all directions in advance; And be stored in the system; In step 103, obtain then after the DOA estimation of user k; According to arrival bearing's estimated result through user k, search the broad beam figuration weight coefficient of all directions of being preserved, get corresponding figuration weight coefficient as figuration combined vector w
k
Through above-mentioned processing, in fact be exactly the form that the wave beam among Fig. 2 is become Fig. 3, wherein main lobe 301 becomes more roomy, and secondary lobe 302 is then less comparatively speaking.Through adopting the method for broad beam, increased the scope of application of the present invention, make that it also can the obtained performance gain under the big environment in angular spread angle.Here beamwidth can be according to the residing environmental selection in base station.
In addition, under the bigger condition of angular spread, possibly have a plurality of arrival bearings corresponding to each user's signal, like this, for step 103, step 104 can also be carried out according to following realization:
Step 103: utilize each cell channel estimated result, obtain figuration combined vector corresponding to each each arrival bearing of user:
And the structure figuration merges matrix:
Here
All K user's k in the expression sub-district u
dIndividual arrival bearing, K
d kAll arrival bearings of k=1...K user of expression expression.
The channel estimation results that can obtain according to step 101 is found the solution and is obtained the figuration combined vector w of user k=1...K corresponding to each arrival bearing
(k) (k d )
The method for solving here mainly contains two kinds:
(1) method of characteristic is promptly at first tried to achieve the space covariance matrix of user k=1...K
Try to achieve R then
(k)All characteristic values
And characteristic vector
Characteristic value here and characteristic vector are corresponding one by one, and be theoretical according to smart antenna, the characteristic value here
In maximum some values be associated with several effective arrival bearings, several minimum values are associated with noise, thereby can set a thresholding Γ, and think greater than the characteristic value of this thresholding corresponding an arrival bearing, otherwise its correspondence noise.This several features value characteristic of correspondence vector is our figuration combined vector to be asked.
Get:
is as the figuration combined vector.In fact, the effective vectorial number when here is for the moment, promptly deteriorates to the figuration combined vector that eigen beam figuration method recited above (EBB algorithm) is tried to achieve.
(2) DOA method, this method are at first wanted effective arrival bearing of estimating user k=1...K.Be made as
Then the pairing array response vector of each effective arrival bearing is the figuration combined vector w corresponding to each arrival bearing
(k) (k d ), can be with figuration combined vector w according to formula (12)
(k) (k d )Merge, construct figuration and merge matrix A
(k)Here a plurality of effective arrival bearing who needs estimating user k=1...K needs the DOA method of estimation of high-resolution for the less environment of angular spread, also need try to achieve user's space covariance matrix as the one of which earlier
The DOA method of estimation here can be existing any DOA method of estimation, like the Bartlett spectral method, and the Capon least variance method; The MUSIC algorithm; ESPRIT algorithm etc., effectively the DOA number also can calculate through existent method, as the information standard (Akaike ' s information criterion; AIC) criterion or minimum description length (Rissanen ' s minimumdescription length, MDL) criterion etc.Obviously, when effective arrival bearing's number is one, then deteriorate to the figuration vector that fixed beam figuration method above-mentioned (GOB algorithm) obtains.
Step 104: the result
before the noise reduction first time of the channel estimating in the formula (7) is carried out figuration merge, and each user's amalgamation result is lined up a row vector.Promptly have:
The process that this figuration merges has just been accomplished a kind of effect of space filtering of multi-beam; Thereby the effect of having played noise reduction and having suppressed to disturb; The figuration merging process here is equivalent to and has utilized multi-beam figuration directional diagram as shown in Figure 4 to carry out space filtering (is example with two wave beams), and wherein main lobe 401 has two different directions.The method of formation broad beam shown in Figure 3 can be regarded as a special case of multi-beam figuration merging here.
The embodiment of the invention proposes a kind of signal channel estimation denoising post-treatment device that is applicable to multiaerial system, and its modular structure is as shown in Figure 5, comprising:
Channel estimation module 501 is used for each antenna to multiaerial system and carries out channel estimating and obtain channel estimation results
Wherein
Represent the channel estimation results that each antenna element is total,
Expression is to k user's channel estimation results,
Represent k
aThe channel estimation results of individual antenna element,
Represent k
aIn the channel estimation results of individual antenna element, belong to k user's component.
Preliminary noise reduction module 502 is used for the channel estimation results that said channel estimation module 501 obtains is carried out the noise reduction first time, and the channel estimation results h that incites somebody to action behind the noise reduction first time is input to figuration merging module 503.Also can omit this module in the practical application, make amount of calculation further reduce.
Figuration merges module 503; Be used to construct figuration combined vector or the figuration that each user's multi-antenna channel estimates and merge matrix; And carry out figuration with the channel estimation results
that said figuration combined vector or figuration merge each user that matrix obtains said channel estimation module 501 and merge, and each user's amalgamation result is lined up a row vector.
Noise reduction post-processing module 504; Be used for the row vector that said figuration merging module 503 obtains is carried out the noise reduction reprocessing; Channel estimation results
carries out noise reduction process according to
to each antenna channel estimated result before the noise reduction
after obtaining noise reduction; The tap position of zero setting in promptly to
, the middle also zero setting of relevant position with
.Channel estimation results after the noise reduction process is as final channel estimation results
In the TD-SCDMA system, utilize as above scheme to carry out emulation, simulated conditions is following:
1) CASE3 channel condition, single sub-district 12.2K is professional, single user and 6 users.
2) 120 degree sectors, even linear array, 6 antennas, 1000 data blocks.
3) utilize original estimation result to carry out user DOA and estimate that method of estimation adopts the BartLett spectral method, promptly tries to achieve the average Rxx corresponding to W space covariance matrix of the individual tap of W (W=128/Kcell) in certain user's the window.Try to achieve the Bartlett spectrum of Rxx subsequently, think that its spectrum maximum power place is the subscriber signal arrival bearing.Here trying to achieve Bartlett spectral corner degree step-length (resolution) is 2 degree.
4) user's incident footpath angular spread obedience
goes up the Laplace distributed random variable.
Here β=11.44.
The demodulation performance contrast that emulation obtains is as shown in Figure 6, and wherein ordinate representes to transmit Block Error Rate (BLER), and abscissa Ior/Ioc representes signal to noise ratio, and unit is a decibel (dB).It is thus clear that, utilize 6 antenna linear arrays, utilize algorithm of the present invention under amount of calculation increase situation seldom, under single user's condition, signal to noise ratio can obtain the 0.5dB gain, and under 6 user's conditions, signal to noise ratio can obtain the gain greater than 1dB.
The above is merely preferred embodiment of the present invention, not in order to restriction the present invention, all any modifications of within spirit of the present invention and principle, being done, is equal to and replaces and improvement etc., all should be included within protection scope of the present invention.