CN1710899A - Interactive parameter estimation method of multi-input and multi-output system - Google Patents

Abstract

Considering multiple input/ output communication system (MIMO) as composed of multiple sub single input/ output systems (SISO), the invention resolves issue of estimating parameters of MIMO system as estimating parameters of SISO system with interference. The disclosed method includes steps: approach of iteration interference cancellation is adopted for estimating parameters of SISO system with interference, i.e. after channel estimation and frequency deviation estimation, interference estimation is carried out first; then, using the approach of iteration interference cancellation carries out estimation for channel gain and frequency deviation, according to presetting number, iteration is carried out repeatedly so as to obtain final estimated value of channel gain and frequency deviation. Comparing the existed methods, the disclosed method possesses features of low complexity, high transmission efficiency, and low cost.

Description

The interactive parameter estimation method of multiple-input-multiple-output communication system

Technical field

The invention belongs to communication technical field, relate to wireless communication signals and handle, is a kind of method of estimating multi-input multi-output system channel gain and frequency departure.

Technical background

Over past ten years, ICT (information and communication technology) and application system have obtained developing rapidly, present unprecedented prosperity scene.Mobile communication, radio communication, MMS (Multimedia Message Service) and development of Internet can be carried out the information exchange of any kind of at any time and any place for realizing anyone, have represented fine prospect.The twentieth century end, the extensive use of Internet technology and people more and more rely on the development of impacting the 3G (Third Generation) Moblie technology to data transmission service.And about future mobile communications, people have proposed the notion of super 3G or 4G, and a plurality of International Standards Organization and forum are also actively developing the research of future mobile communications.International Telecommunications Union goes out in the document to the future development of 3G and super 3G system: the ability of 3G land wave point expanded to nearly 30Mbps in 2005; Imagination under high-speed mobile condition, will support the peak rate of about 100Mbps in the new system of super 3G about 2010, under the low speed mobile condition, will support the peak rate of about 1Gbps.

The multiple-input and multiple-output mimo system is a kind of communication system with a plurality of transmitting antennas and a plurality of reception antennas, comprise and transmit and receive two parts, wherein radiating portion comprises modulator, frequency mixer, radio-frequency power amplifier and many transmit antennas, receiving unit comprises many reception antennas, frequency mixer, filter, D/A and baseband digital signal processor, as shown in Figure 1.This system can improve the channel capacity under the wireless channel exponentially, according to information-theoretical achievement in research, if different transmit receive antennas between decline separate, under identical transmitting power and bandwidth, the channel capacity that system that has n transmitting antenna and a m reception antenna can reach is the min (n of present a single aerial system, m) doubly, thus present other technology capacity potentiality that are beyond one's reach are provided.So mimo system is considered to one of key technology that realizes future mobile communications.

Because wireless signal when the channel, can be subjected to the influence of channel parameter,, and then utilize the estimated value of channel parameter to do the demodulation of data so at receiving terminal, at first will utilizing makes a start sends known training sequence and estimate channel parameter.The channel parameter here is meant channel gain and carrier frequency offset, and channel gain is meant that wireless signal through power attenuation and phase change behind the channel, is a complex values; Carrier frequency offset is to cause owing to Doppler frequency shift or the two differences of sending out frequency oscillators of transmitting-receiving.

At present, existing method for parameter estimation includes the method for parameter estimation under the single antenna and the method for parameter estimation of mimo system, and the method for parameter estimation under the conventional single-antenna can not be in operate as normal under the mimo system, and the method for parameter estimation of mimo system mainly contains:

1. think that the carrier frequency offset between receiver and the transmitter has only one, carries out parameter estimation method again.The shortcoming of this method is not have generality, because mimo system is to utilize the difference of different antennae decline to improve the capacity of channel, the arrival angle that is wireless signal arrival different antennae is different, when transmitter or receiver motion, because different arrival angles can cause different Doppler frequency shifts, perhaps produce different frequency departures owing to the frequency oscillator on each antenna is different, these all are ubiquitous phenomenons in wireless communication procedure, obviously only think that the method for parameter estimation that has only a carrier frequency offset between receiver and the transmitter is also improper in this case, so general model be think the different dual-mode antennas of mimo system between frequency departure all inequality, promptly for a mimo system that has n transmitting antenna and m reception antenna, receiver must be estimated gain and nm carrier frequency offset of channel earlier before demodulating data.

2. existing method of estimation to above-mentioned universal model has only the maximum likelihood of employing ML to estimate, it is optimum that this method of estimation can reach, it is the lower bound that its estimated mean-square can reach estimation, be Crmer-Rao circle, yet this estimation does not have the form of sealing, be the non-linear search problem of a n dimension, computation complexity is very high, only only limits to theoretical research.

3. feasible using method is that the training sequence that sends of making a start is done special design at present, promptly an antenna transmission training sequence time, other antennas do not send, each antenna sends training sequence in such a manner successively, on all antennas, send data-signal more simultaneously after distributing training sequence, parameter Estimation with a mimo system has just become the parameter Estimation of doing a single aerial system on different time sections like this, uses traditional single antenna method for parameter estimation.The shortcoming of this method is:

(1) gross power of all transmitting antennas of general mimo system is constant, and the transmitted power on the individual antenna is the 1/n of gross power, and n is the number of transmitting antenna.If the method that other antennas are not sent out when adopting an antenna transmission training sequence, increase n when then the transmitted power on the individual antenna is than transmission of data signals doubly, produced the excessive problem of peak-average power ratio like this, promptly very high to the requirement meeting of the radio-frequency power amplifier on the antenna, and radio-frequency power amplifier is the core of antenna, has increased the cost of system greatly;

(2) training sequence is all known data of transmitter and receiver, its effect only is that the parameter of estimating channel comprises channel gain and frequency departure, not the actual Useful Information data of transmission, so it is oversize that training sequence can not account in frame data, and said method is in an antenna transmission, and other antennas do not send, i.e. the idle time of transmission for other antennas, particularly when number of transmit antennas is many, increased the expense of training sequence greatly.

Summary of the invention

The objective of the invention is to avoid the deficiency of above-mentioned prior art, a kind of interactive parameter estimation method of mimo system is provided, with complexity that solves the existing channel method for parameter estimation and the problem that cost requirement is too high and efficiency of transmission is low excessively.

The object of the present invention is achieved like this:

MIMO communication system as shown in Figure 1 is generally at transmitting terminal, at first data modulated, after the mixing, power amplification, goes out through a plurality of transmission antennas transmit again.At receiving terminal, after the signal process mixing and filtering that receive on a plurality of reception antennas, enter analog/digital converter again model signals is become digital signal, then the digital signal on a plurality of antennas is entered the baseband digital signal processor and carry out channel parameter estimation and data demodulates, the present invention realizes the estimation to channel gain and frequency departure parameter just on the baseband digital signal processor.Its basic ideas are: for a reception antenna number is m, number of transmit antennas is the mimo system of n, can with wherein any one antenna between regard the single output of a single input SISO subsystem as, can see this mimo system as by mn SISO subsystem and form.For the parameter Estimation problem of mimo system, the parameter that need estimate is this mn their channel gain and frequency departure separately of SISO subsystem just.For certain reception antenna, its received signal is the summation of n SISO subsystem, offer an explanation out the parameter of each SISO subsystem in this summation signals, and this n SISO subsystem can be thought the phase mutual interference.So just can be the parameter Estimation problem of the SISO system of parameter Estimation the question resolves itself into band of mimo system interference.

Technical scheme of the present invention is: the parameter Estimation of the SISO system that band is disturbed adopts the method for parameter estimation of iterative interference cancellation, promptly after channel estimating and frequency departure are estimated, at first carries out Interference Estimation; Adopt the method for Interference Cancellation that channel gain and frequency departure are carried out the iteration estimation then, detailed process is as follows:

1. the parameter of the single output of single input SISO subsystem is estimated, the received signal y of the SISO subsystem between j transmitting antenna and k the reception antenna _{K, j}(t) be expressed as:

$y_{k,j}\left(t\right)=h_{k,j}{e}^{\mathrm{jt}{\mathrm{\ω}}_{k,j}}{x}_j\left(t\right)+I_{k,j}\left(t\right)+z_k\left(t\right)$

In the formula: h _{K, l}And ω _{K, l}Channel gain and the carrier frequency offset of representing this subsystem

T is the discrete time coordinate, t=1, and 2,3 ...

x _j(t) training sequence that sends on j transmitting antenna of expression

z _k(t) noise on the expression reception antenna is a Gaussian random variable

I _{K, j}(t) expression can be approximately then I of a Gaussian random variable from the interference signal of other SISO subsystem _{K, j}(t)+z _k(t) be Gaussian random variable;

2. the method for parameter estimation that utilizes traditional SISO is to h _{K, j}And ω _{K, j}Estimate, obtain the channel gain estimated value

With the carrier frequency offset estimated value

I=0 wherein, 1,2,3 ..., I, expression is through the estimated value after the i time iteration, and I need altogether to represent the iterations that carries out;

3. utilize the channel gain estimated value

With the carrier frequency offset estimated value To interference signal from other SISO subsystem

Estimate, obtain the estimated value of interference signal

Promptly

${\hat{I}}_{k,j}^{\left(i\right)}=\underset{l=1,l\≠j}{\overset{T}{\mathrm{\Σ}}}{\hat{h}}_{k,l}{e}^{\mathrm{jt}{\widehat{\mathrm{\ω}}}_{k,l}}{x}_l\left(t\right)$

In the formula: x _j(t) represent the training sequence that sends on other antenna:

4. use the estimated value of interference signal

Offset, obtain the received signal behind the Interference Cancellation

Promptly

$y_{k,j}^{(i+1)}\left(t\right)=y_{k,j}\left(t\right)-{\hat{I}}_{k,j}^{\left(i\right)}\left(t\right);$

5. will be through the received signal of Interference Cancellation Returned for (1)～(4) step, after channel gain, carrier frequency offset and interference signal are reappraised, carry out Interference Cancellation again, finish iteration for the first time;

6. carry out above-mentioned (1)～(4) step repeatedly according to the iterations I that sets, obtain final channel gain And carrier frequency offset

Estimated value.

The present invention compared with prior art, the present invention has following advantage:

(1) the present invention is owing to allow all antennas to send training sequence simultaneously, and the through-put power on each antenna is equal, thereby has reduced the radio-frequency power amplifier requirement of antenna end, thereby has reduced the cost of system.

(2) the present invention is owing to adopt a kind of general model, think all dual-mode antenna between carrier frequency offset all different, like this when realizing mimo system, not need all adopt on each antenna identical high accuracy frequency oscillator guarantee each antenna between have identical frequency departure, saved the cost of system.

(3) mountain of the present invention has made full use of the time of transmitting in adopting all antennas to transmit the method for training sequence simultaneously, has increased the efficiency of transmission of mimo system, has saved the bandwidth of communication.

(4) the present invention has versatility, is fit to the various situations of wireless communication transmissions; And complexity reduces greatly compared with the method for estimation of maximum likelihood, is very beneficial for the realization of actual wireless communication mimo system, and low simultaneously complicated method of estimation has reduced the requirement to the core digital signal processor of mimo system.

Description of drawings

Fig. 1 is a MIMO communication system block diagram

Fig. 2 is an interactive parameter estimation method flow chart of the present invention

Fig. 3 is that frequency departure estimated mean-square of the present invention is to the signal to noise ratio curve chart

Fig. 4 is that channel gain estimated mean-square of the present invention is to the signal to noise ratio curve chart

Embodiment

The present invention is by selecting 4 transmit antennas for use, and the embodiment of the mimo system of 1 reception antenna illustrates the detailed process of its channel parameter estimation.There are 4 frequency departure ω in this system _1,1, ω _1,2, ω _1,3, ω _1,4Be respectively: 0.02,0.03,0.01 and 0.015; The frequency departure of this system is a relative frequency deviation, i.e. the ratio of absolute frequency deviation and system data rates; The channel of this system is the white Gaussian noise channel, i.e. 4 channel gain h _1,1, h _1,2, h _1,3, h _1,4All be 1; The length of training sequence is 32, and signal power to noise power ratio is 30dB.

With any one dual-mode antenna in this mimo system between can be regarded as the single output of a single input SISO subsystem, so this mimo system can be thought to be made up of 4 SISO subsystems.And, need the parameter estimated their channel gain and frequency departures separately of these 4 SISO subsystems just for the parameter Estimation problem of this mimo system.And reception antenna hereto, its received signal is the summation of 4 SISO subsystems, the purpose of parameter Estimation is exactly to offer an explanation out the parameter of each SISO subsystem in this summation signals, so these 4 SISO subsystems can be thought the phase mutual interference.So just the parameter Estimation problem of the SISO system of 4 band interference of parameter Estimation the question resolves itself into of this mimo system.Because the process that 4 SISO subsystems in this mimo system are estimated is all identical, so only need provide the estimation procedure of the SISO subsystem between the 1st transmitting antenna and the 1st reception antenna here.

With reference to Fig. 2, parameter Estimation flow process of the present invention is as follows:

(1) with the received signal y of the SISO subsystem between the 1st transmitting antenna and the 1st reception antenna _1,1(t) can be expressed as:

${\mathrm{<figure-callout\; id="1"\; label="y"\; filenames="A20051004273500101.png,A20051004273500111.png"\; state="\{\{state\}\}">y</figure-callout>}}_{\mathrm{1,1}}\left(t\right)=h_{\mathrm{1,1}}{e}^{\mathrm{jt}{\mathrm{\&omega;}}_{\mathrm{1,1}}}{x}_1\left(t\right)+I_{\mathrm{1,1}}\left(t\right)+z_1\left(t\right)$

Here h _1,1=1 and ω _1,1=0.02, represent the channel gain and the carrier frequency frequency difference of this subsystem; T is the discrete time coordinate, because the length of training sequence is 32, thus t=1,2,3 ..., 32.x ₁(t) training sequence that sends on the 1st transmitting antenna of expression is known to receiver; z ₁(t) noise on the 1st reception antenna of expression is a Gaussian random variable, I _1,1(t) expression can be similar to and think Gaussian random variable, then an I from the interference signal of other SISO subsystem _1,1(t)+z ₁(t) also be Gaussian random variable;

(2) because with I _1,1(t) approximately think a Gaussian random variable, so will estimate h _1,1And ω _1,1Value, be the parameter Estimation problem of a traditional SISO, therefore utilize traditional SISO method of estimation just can obtain the estimated value of this SISO subsystem ${\hat{h}}_{\mathrm{1,1}}^{\left(0\right)}=0.9942+0.054i$ With ${\widehat{\mathrm{\&omega;}}}_{\mathrm{1,1}}^{\left(0\right)}=0.129,$ With actual value h _1,1=1 sharp ω _1,1=0.02 relatively, can see that the frequency departure estimated value has bigger gap.

(3) all 4 SISO subsystems are estimated simultaneously after, can estimate the channel gain of 4 SISO subsystems: h _1,1, h _1,2, h _1,3, h _1,4, frequency departure: ω _1,1, ω _1,2, ω _1,3, ω _1,4This obviously estimation is more coarse, reason be exactly this estimation be under noisy situation, to do, and for the SISO subsystem of the 1st transmitting antenna and the 1st reception antenna, I _1,1(t) being interference signal, promptly is the summation by other antenna transmission signals, because the estimates of parameters of all SISO subsystems all obtains, so I _1,1(t) do not need to think the Gaussian random variable an of the unknown, can use the estimates of parameters of these subsystems to estimate I _1,1(t), promptly

${\hat{I}}_{\mathrm{1,1}}^{\left(0\right)}\left(t\right)=\underset{l=2}{\overset{4}{\mathrm{\&Sigma;}}}{\hat{h}}_{1,l}^{\left(0\right)}{e}^{\mathrm{jt}{\widehat{\mathrm{\&omega;}}}_{1,l}^{\left(0\right)}}{x}_l\left(t\right)$

(4) interference signal I _1,1(t) estimated value is  _1,1 ⁽⁰⁾(t), do Interference Cancellation this moment, and promptly the received signal of the SISO subsystem of the 1st transmitting antenna and the 1st reception antenna deducts interference signal  _1,1 ⁽⁰⁾(t), then be through the new received signal behind the Interference Cancellation

$y_{\mathrm{1,1}}^{\left(1\right)}=y_{\mathrm{1,1}}\left(t\right)-{\hat{I}}_{\mathrm{1,1}}^{\left(0\right)}\left(t\right)$

(5) will be through the received signal y of Interference Cancellation _1,1 ⁽¹⁾(t) return (1) step replacement y again _1,1(t), do the SISO parameter Estimation in (2) step again, can obtain carrying out the estimated value behind the Interference Cancellation the 1st time ${\hat{h}}_{\mathrm{1,1}}^{\left(1\right)}=0.9993+0.012i$ With ${\widehat{\mathrm{\&omega;}}}_{\mathrm{1,1}}^{\left(1\right)}=0.01987,$ Compare as seen with the 1st estimated value with the estimated value behind this Interference Cancellation, through estimated value and the actual value h behind 1 Interference Cancellation _1,1=1 and ω _1,1=0.02 gap is littler.Continue to utilize new estimated value to do

(3) Bu Interference Estimation:

${\hat{I}}_{\mathrm{1,1}}^{\left(1\right)}\left(t\right)=\underset{l=2}{\overset{4}{\mathrm{\&Sigma;}}}{\hat{h}}_{1,l}^{\left(1\right)}{e}^{\mathrm{jt}{\widehat{\mathrm{\&omega;}}}_{1,l}^{\left(1\right)}}{x}_l\left(t\right)$

Utilize new Interference Estimation value to do Interference Cancellation again:

${\mathrm{<figure-callout\; id="1"\; label="y"\; filenames="A20051004273500101.png,A20051004273500111.png"\; state="\{\{state\}\}">y</figure-callout>}}_{\mathrm{1,1}}^{\left(2\right)}\left(t\right)={\mathrm{<figure-callout\; id="1"\; label="y"\; filenames="A20051004273500101.png,A20051004273500111.png"\; state="\{\{state\}\}">y</figure-callout>}}_{\mathrm{1,1}}\left(t\right)-{\hat{I}}_{\mathrm{1,1}}^{\left(1\right)}\left(t\right)$

Received signal y with the 2nd Interference Cancellation _1,1 ⁽²⁾(t) return (1) step replacement y again _1,1(t) do the parameter Estimation of SISO in (2) step, can obtain carrying out the estimated value behind the Interference Cancellation the 2nd time ${\hat{h}}_{1,1}^{\left(2\right)}=0.9993-0.0007i$ With ${\widehat{\mathrm{\&omega;}}}_{\mathrm{1,1}}^{\left(2\right)}=0.020002,$ Same process can obtain the estimated value behind the Interference Cancellation the 3rd time ${\hat{h}}_{\mathrm{1,1}}^{\left(3\right)}=0.9999-0.0000i$ With ${\widehat{\mathrm{\&omega;}}}_{\mathrm{1,1}}^{\left(3\right)}=0.020001,$ The estimated value that obtains after each as can be seen iteration all can be approached actual value more.

In order to embody overall performance of the present invention, we use signal power to noise power ratio SNR that the curve of mean square error MSE is estimated performance of the present invention.The mean square error is here represented the average departure degree of estimated value to actual value, is defined as:

MSE=ask on average | estimated value-actual value | ²}

Fig. 3 is the curve of the mean square error of frequency departure estimated value to signal power and signal to noise ratio, under the situation of low signal-to-noise ratio, be below the 10dB, the mean square error of estimated value is gradually to the boundary of optimal estimation, be that Cramer-Rao circle is approaching, under the situation of middle and high signal to noise ratio, i.e. 10dB～40dB, estimated mean-square and Cramer-Rao circle lean on very closely, in figure

With Article three, shown in the curve.

Fig. 4 is the curve of the mean square error of channel gain estimated value to signal to noise ratio, and the mean square error of its estimated value almost overlaps with Cramer-Rao circle, in figure

With

Article three, shown in the curve.

By Fig. 3 and Fig. 4 as can be seen, the non-iterative situation among the figure shown in the dotted line is exactly the situation that the single antenna estimation approach is used in mimo system, and along with the increase of signal to noise ratio, the MSE performance can not obtain any improvement in this case.And the present invention whenever carries out iteration one time, the MSE performance will be improved in a nearly step, performance after the 1st iteration is compared with the performance of iteration not and is improved significantly, the performance of the 2nd iteration is compared with the performance of the 1st iteration, improve significantly again when signal to noise ratio is 30dB～40dB, and the performance curve of the 2nd iteration and the 3rd iteration almost overlaps, convergence rate of this explanation iteration is very fast, only need 2 iteration just can realize as can be seen near optimum systematic function, in two figure

With

Article three, shown in the curve.

The foregoing description does not constitute any restriction to invention, and obviously, described those skilled in the art can be without any creative work, and utilizes technical conceive of the present invention, makes the iterations and the method that are not limited only to the embodiment of the invention.

Claims (1)

1. the method for parameter estimation among the multiple-input-multiple-output communication system MIMO is after channel estimating and frequency departure are estimated, at first carries out Interference Estimation; Adopt the method for Interference Cancellation that channel gain and frequency departure are carried out the iteration estimation then, detailed process is as follows:

(1) parameter of the single output of single input SISO subsystem is estimated, the received signal y of the SISO subsystem between j transmitting antenna and k the reception antenna _{K, j}(t) be expressed as:

$y_{k,j}\left(t\right)=h_{k,j}{e}^{\mathrm{jt}{\mathrm{\&omega;}}_{k,j}}{x}_j\left(t\right)+I_{k,j}\left(t\right)+z_k\left(t\right)$

In the formula: h _{K, l}And ω _{K, l}Channel gain and the carrier frequency offset of representing this subsystem

T is the discrete time coordinate, t=1, and 2,3 ...

x _j(t) training sequence that sends on j transmitting antenna of expression

z _k(t) noise on the expression reception antenna is a Gaussian random variable

I _{K, j}(t) expression can be approximately Gaussian random variable, then an I from the interference signal of other SISO subsystem _{K, j}(t)+z _k(t) be Gaussian random variable;

(2) utilize the method for parameter estimation of traditional SISO to h _{K, j}And ω _{K, j}Estimate, obtain the channel gain estimated value With the carrier frequency offset estimated value I=0 wherein, 1 ..., I, expression is through the estimated value after the i time iteration, and I need altogether to represent the iterations that carries out;

(3) utilize the channel gain estimated value With the carrier frequency offset estimated value

To interference signal I from other SISO subsystem _{K, j}(t) estimate, obtain the estimated value of interference signal

Promptly

${\hat{I}}_{k,j}^{\left(i\right)}\left(t\right)=\underset{l=1,l\&NotEqual;j}{\overset{T}{\mathrm{\&Sigma;}}}{\hat{h}}_{k,l}{e}^{\mathrm{jt}{\widehat{\mathrm{\&omega;}}}_{k,l}}{x}_j\left(t\right)$

In the formula: x _j(t) represent the training sequence that sends on other antenna;

(4) estimated value of usefulness interference signal

Offset, obtain the received signal y behind the Interference Cancellation _{K, j} ⁽ⁱ⁺¹⁾(t), promptly

$y_{k,j}^{(i+1)}\left(t\right)=y_{k,j}\left(t\right)-{\hat{I}}_{k,j}^{\left(i\right)}\left(t\right);$

(5) will be through the received signal y of Interference Cancellation _{K, j} ⁽ⁱ⁺¹⁾(t) returned for (1)～(4) step, after channel gain, carrier frequency offset and interference signal are reappraised, carry out Interference Cancellation again, finish iteration for the first time;

(6) carry out above-mentioned (1)～(4) step repeatedly according to the iterations I that sets, obtain final channel gain And carrier frequency offset Estimated value.

Images