CN1063600C - CDMA communication system - Google Patents

Abstract

The invention relates to a CDMA communication system comprising at least a transmitter in which a data sequence is spread with the code sequence and comprising at least a receiver in which the data sequences are recovered by a detector, estimate sequences for the values of the transmitted data sequences being computed for the purpose of the detection while a description of the communication channel between the transmitter or transmitters and a receiver is incorporated. For the detection of data sequences there is provided to compute joint probability distributions for the transmitted data sequences by computing the associated moments and determine therefrom the estimate sequences for the transmitted data sequences.

Description

The wave detector that is used for the code division multiple address communication system receiver

The present invention relates to cdma communication system, it comprises that at least one data sequence is therein by the transmitter of sign indicating number sequence institute spread spectrum, also comprise the receiver that at least one data sequence is detected device therein and is recovered, calculate the estimation sequence that is launched data sequence for the purpose of detection, introduce description simultaneously the channel between transmitter or a plurality of transmitter and receiver.The invention still further relates to the receiver of cdma communication system, specifically, relate to the wave detector of this receiver.

In the communication system based on code division multiple access (CDMA), various subscriber signals use common carrier frequency side by side to be launched on common frequency band.Code division multiple access system is based on the spread bandwidth technology, that is to say, the signal that be launched is expanded in fact than the wide frequency range of the required the narrowest frequency range that is used to transmit at one.The consequence of this spread bandwidth is code division multiple access system influence without interruption on the whole.

For the band spreading in the digital communication system, for example, each bit that be launched is taken advantage of by the mutual consensus code word of transmitter and receiver.When adopting mutually orthogonal code word, the phase mutual interference that causes from each user's signal has been excluded substantially.But, no longer observe these actual specifications for ripple transmission (for example causing the multipath transmission that orthogonality takes place) at the earth's surface.In addition, if be removed by the synchronism of each user capture so that realize the visit of greatly simplifying to common frequency band, if perhaps allow different bit rates, then subscriber signal can only use the cost of more Circuits System and Geng Gao or reduce quality and be detected, because signal can be no longer mutually orthogonal.

" the linear multi-user wave detector that is used for the S-CDMA channel " that information theory journal the 35th volume the 1st phase 123-126 page or leaf that in January, 1989, IEEE published is shown by Ruxandra Lupas and Sergio Verd ú is that a kind of each user therein is by the known wave detector that is used for cdma receiver of while detection.By each user of while detection, the detection quality can be improved by increasing Circuits System, and the reciprocation in the aforementioned circuit system between the signal section of different user is eliminated by means of linear image (image) or reduces.Thereby the interference that is caused by receiver noise also is exaggerated.As for other, this wave detector is a suboptimum with respect to bit error rate.

The information theory journal IT-32 that in January, 1986, IEEE published rolls up the 1st phase 85-96 page or leaf and has described the nonlinear detector that is bordering on the best with respect to bit error rate by " minimum error probability of asynchronous Gauss's access channel " of Sergio Verd ú work.But the cost of this desirable nonlinear detector increases exponentially along with cdma communication system user's number.

The object of the present invention is to provide the wave detector that is used for cdma communication system, its performance with respect to linear detector be improved and its be embodied as and originally remain on below the best wave detector cost.

This purpose is achieved in that the joint probability distribution of transmission of data sequences is to calculate by calculating adjoint matrix and it being derived from the estimation sequence of the data sequence of having launched.

In communication process, new data is constantly launched and is made for the judgement of launching data value by the receipts machine.Thereby joint probability distribution constantly changes so that the joint probability distribution sequence will be calculated.In addition, joint probability distribution can constantly be gone into to cry received signal to improve.When calculating joint probability distribution, the information of transmitting channel and noise characteristic can both be taken in the lump.

In addition, probability distribution automatically provides the quality information of relevant estimation.This quality information can be by favourable estimation in the decoder in circuit downstream.

The advantage that adopts square to describe probability distribution is to utilize square to obtain very simple probability distribution and describes.

Joint probability distribution only adopts the incompatible calculating of restriction set of square in another embodiment of the present invention.

The limitations set of square only provides the roughly description of joint probability distribution.Yet the limitations set of use square can reduce the cost of wave detector under the situation of not bringing tangible detection mass loss.

The first and second rank squares that only calculate joint probability distribution in another embodiment of the present invention are so that calculating joint probability distribution roughly.

Verified for best detection, only calculate first and second squares of joint probability distribution, thereby calculating mean value and covariance are just enough.

In a specific embodiments of the present invention, use Kalman filter as wave detector.

Employing is commonly called the improved Kalman filter of soft decision feedback equalizer at J.Tielecke, Proceedings ICASSP 90 (INTERNATIONAL CONFERENCE ON ACOUSTICS, SPEECH AND SIGNAL PROCESSING; APRIL 3-6,1990) " the new nonlinear equalizer that is used for mobile radio channel " of 1667-1670 page or leaf done discussion, and it is used for conventional tdma communication system to eliminate intersymbol interference.In tdma system, the time distributed channel mainly causes the generation of intersymbol interference.(under the situation of intersymbol interference, the bit of the mutual time shift of same subscriber overlaps each other.) in cdma system, this intersymbol interference compare with the interference that causes by other users only play a part very little.Concerning the technology skilful person of this area, described research does not produce any prompting of adopting this Kalman filter technique at cdma system and their various types of problems.

Traditionally, utilize Kalman filter not follow the estimation vector sum that calculates wrong covariance matrix to give probability distribution.Yet as is known, estimation vector can be considered to the average vector (first square of joint probability distribution) of joint probability distribution and wrong covariance matrix can be counted as the covariance matrix (second square of joint probability distribution) of joint probability distribution.

In another embodiment of the present invention, make about the judgement of the value of data sequence and with these decision-feedback according to estimation and to go back to wave detector inside.

This result of judgement through feedback is, the fair of detection process can be improved widely, and this is because considered the discrete nature of institute's transmission of data sequences.Specifically this feedback of Kalman filter is not taken place.

In another embodiment of the present invention, during on duty being fed, the value that has been fed is considered estimation quality.

By avoiding hard decision, the quality of detection process can further be improved.Can handle the decoder that quality grading is judged (soft decision) if add one in wave detector in circuit, then communication reliability is obviously strengthened.

Embodiment shown in the present invention incites somebody to action with reference to the accompanying drawings further explains following making, wherein:

Fig. 1 illustrates the cdma communication system that comprises a N mobile radio station and a base station;

Fig. 2 illustrates the receiver of this cdma system;

Fig. 3 illustrates the wave detector functional block;

Fig. 4 illustrates the time diagram of a plurality of users' that represent cdma communication system transmission of data sequences;

Fig. 5 illustrates the state diagram of the improved Kalman filter that is used to estimate transmitted data signal;

Fig. 6 illustrates the circuit block diagram in the feedback step in the state diagram shown in Figure 5;

Fig. 7 illustrates the state diagram of the estimation of covariance matrix;

Fig. 8 illustrates the circuit block diagram in the feedback step in the state diagram as shown in Figure 7.

Emission The data binary data sequence b in CDMA data communication system shown in Figure 1 _iIn an embodiment, binary data sequence b _iBe resumed from speech signal by suitable coding.These binary data sequences b _iBy at each mobile wireless phone with cover between the base station BS in wireless mobile telephone set zone just within it and exchange.In response to the number N of launching mobile radio station simultaneously, can use N data source b _i, the subscript i=1 on the accompanying drawing wherein ... N is to distinguish each data source.Be spread data sequence, each data sequence b _iBy sign indicating number sequence s _iMultiply each other.Following, a bit of sign indicating number sequence is known as " sheet " (chip) so that distinguish mutually with the bit of the data sequence of launching.

Want separated opening owing to transmit at receiver side, each yard sequence is selected discriminatively.For this reason, for example mobile radio station is selected a sign indicating number sequence randomly and this selection is sent to the base station, has at this moment just set up to be connected between mobile radio station and base station.

The diverse location of each mobile radio station makes to transmit and arrives the base station by the different radio channel.By these wireless channels, transmit for example to be exposed to and transmit caused distortion by reflection and multipath.The signal of these distortion overlaps each other to form continuous received signal y (t) at the antenna of base station BS, and this received signal comprises noise signal n (t).

In the receiver of base station BS, receiving data sequence b _iBe at the transmission of data sequences b among the received signal y (t) _iAnd estimation.For example, be distributed in from the reception data of the recovery of voice signal of these input data and each receiving port that for example arrives that this is not shown by sending communication network to.

Fig. 2 illustrates the basic structure according to receiver of the present invention.The signal y (t) that receives is at first through HF preamplifier state 21 preposition amplification and bandpass filterings.By means of from HF oscillator 22, in response to the HF signal that is used in the carrier frequency in the transmitter, the received signal of bandpass filtering in frequency mixer 23a, 23b with HF signal self and the signal mixing mutually that has quadrature phase with respect to the HF signal.Consequently, real number input signal y (t) is converted directly to base band, produces the complex base band signal that comprises real part and imaginary part in this operating process.In case these two signals are low pass filtering (this filtering is not shown in Fig. 2), they are just taken a sample by sampler 24a, 24b at equidistant kT constantly.Thereby will consider sampling theorem, that is sampling frequency should be the twice at least of base-band signal frequency.In the present embodiment, this be done in such a way that promptly take a sample with the section speed twice finish.

Sampling value is transformed to digital value sequence y (k) subsequently in analog to digital converter 25a, 25b.This digital value sequence y (k) is added to digital signal processor 26, and this processor 26 comprises read-only memory 27 for example, be used to estimate the program of the binary data sequence of emission.The intermediate object program that estimation produces is cached in the read-write memory 28 by signal processor 26.According to the number of users that will be detected, the calculated capacity of individual signals processor is no longer enough under special circumstances.To provide more processor to disperse calculated capacity in this case.

If Fig. 3 illustrates the function that signal estimation need be realized respectively by single or multiple processors with the form of circuit diagram.In order in wave detector 30, to finish signal estimation, need carry out channel description to each user, this explanation is produced by channel-estimator 31.Should be noted that this point: because the position difference of mobile radio station, mobile radio station is different with each wireless channel between the base station.Often use so-called channel impulse response to describe as channel.In order to determine whether the channel description is possible, for example, in transmitting, insert the training data sequence, can be from this sequence by the impulse response of correlator calculating channel receiver.In addition, sign indicating number sequencer 32 needs, and it produces the value of a sheet at sampling moment k to each unique user.As will be further shown, will be detected device 30 by the channel impulse response that radio channel impulse responds and yard sequence forms is advantageously used.

In order to calculate the sequence of joint probability distribution square, advantageously in state space, adopt the description of channel.For this reason, each user is combined into vectorial b (k) in the data that sampling moment k influences discrete time received signal y (k).Can cause the phase mutual interference owing to be not only in the channel of time diffusion, so vectorial b (k) does not include only every user's one bit and spread according to the time and also to comprise the many bits of every user from different users' bit and from a user's successive bits.Suppose noise superimposed and cause that the linear channel of distortion done simplification that following experimental equation can be used at the sampling input signal y (k) that detects at receiver of k constantly:

y(k)=h ^T(k)b(k)+n(k)

Vector h (k) has described the emission bit and how to have caused mutual interference.This vector has been considered filter and the frequency expansion sequence in actual wireless channel, the telecommunication circuit.

In order to understand the influence of sign indicating number sequence to h (k), Fig. 4 shows a very simple embodiment.In this embodiment, cdma communication system has three users.Modulator approach is that phase shift keying (Phase Shift Keying) and each user's wireless channel is formed (no multipath transmission, nothing decline, undamped) by direct link, and the white Gauss noise that this link only is applied (AWGN) is disturbed.At moment k ₀Obtain following vectorial b (k ₀), this vector comprises the emission bit that causes mutual interference: b (k ₀)=[+1 ,+1 ,-1] ^T

Impulse response vector h (k ₀) be that only to be used to the sign indicating number sequence of spread spectrum determined under this simple scenario, as can be seen from Figure 4:

h(k ₀)=[+1，-1，+1] ^T

Because sheet (chip) value continuously changes, new impulse response vector h (k) in fact constantly all will be calculated in each new sampling, and this vector has been considered the currency of sheet.

It is favourable introducing a sign indicating number sequence in impulse response vector h (k), because consequently this vector can have the structure of high flexibility: for example, can allow yard sequence period not in response to used cycle data.In addition, can allow the user to have different and variable data rate.Embodiment shown in Figure 4 expresses binary code sequence is incorporated into impulse response vector h (k) is very simple, because symbol only changes according to sheet in pulse vector h (k).

This experimental equation can be finished by state transition equation, and this state transition equation has been described being combined in from sampling moment k of vectorial b (k) and how to have been changed when k+1 changes constantly to sampling:

b(k+1)=A(k) ^*b(k)+b _a(k+1)

By means of matrix A (k) and since data moment k+1 no longer to received signal y (k+1) work, from vectorial b (k) or b (k+1), remove when describing them formal.By means of vectorial b △ (k+1), data are added to vectorial b (k) or b (k+1) respectively, and above-mentioned data just are launched at moment k+1, thereby for the first time the signal that receives are exerted an influence.

The state equation of this form only is used to generally how to be added to vectorial b (K+1) by means of the formula data of description or to be read out from this vector.When calculating b (K+1) in signal processor, this computing is not preferably finished by means of matrix multiplication or addition of matrices, but by special storage operation.For not having the situation that new data are launched at moment k+1, vectorial b (k) changes into b (k+1) on the turn.

By means of quantity of being introduced by the description of state space (magnitude) and description, calculating operation is how to be performed with the estimated data's in the wave detector with being described in.In order to simplify, suppose that for preferred embodiment binary value is used as data transmission and baseband receiving signals is a real-valued signal.Under the situation of complex values baseband signal, the real part of plural in a preferred embodiment sampling value y (k) and imaginary part as two independently real-valued signal be received continuously.Even if obtaining slightly in Kalman filter, the superiority of quality does not need intensifier circuit system and cost yet.

The first and second rank squares are recursive calculation in a preferred embodiment, that is mean vector b (k) and b ₊(k) and follow covariance matrix P (k) and P ₄(k).These squares can be endowed joint probability distribution, and this distribution is similar to the real number probability distribution.Fig. 5 illustrates the part of state diagram, takes this and with illustrated form the calculation procedure that signal processor 26 is finished is shown.Wave detector itself also can be looked at as improved Kalman filter, and the feedback of its judgement (soft decision) by having relevant quality is improved.For the calculating of the necessary covariance matrix of estimated data's sequence provides in the legend mode in state diagram shown in Figure 7.By signal processor 26 finish to the feedback calculating with state diagram form respectively shown in Fig. 6 and Fig. 8.In order to understand better, can note in Fig. 5 and Fig. 6 marking with 10 and in Fig. 7 and Fig. 8 with the square frame of the 11 expressions feedback character that mark.

But be to use according to Kalman filter of the present invention, compare the judgement of nothing feedback with the suboptimum wave detector of prior art and also show significantly improved characteristic, particularly aspect flexibility.

Usually at this moment transmitting training data when data transmission begins connect and have set up.Thereby for the user of other increase, the initialization of wave detector is simple, because only suitable data enter mean vector.In covariance matrix, follow covariance value will be changed to 0.During this mode, at moment k mean vector ${\underset{-}{\overset{^}{b}}}_{+}\left(k\right)$ And follow covariance matrix P ₊(k) be known.Thereby no longer estimate the signal value y (k) of reception at moment k.Improvement estimation or the mean vector of computing mode vector b (k) subsequently $\underset{-}{\overset{^}{b}}\left(k\right)$ So that the basic colonel at reception value y (k) estimates vector ${\underset{-}{\overset{^}{b}}}_{+}\left(k\right)$ 。This is to be finished by the filter equation of Kalman filter, and this equation is based on the experimental equation of state space description.Provide the mean vector of correction below:

$\underset{-}{\overset{^}{b}}\left(k\right)={\underset{-}{\overset{^}{b}}}_{+}\left(k\right)-g\left(k\right)e\left(k\right)$ Correction term can be calculated by kalman gain vector:

$g\left(k\right)=\frac{{\underset{-}{P}}_{.}\left(k\right)+h\left(k\right)}{h^T\left(k\right){\underset{-}{P}}_{+}\left(k\right)h\left(k\right)+{\mathrm{\σ}}_n^2}$ And calculate estimation error e (k) by received signal y (k): $e\left(k\right)=y\left(k\right)-{\underset{-}{h}}^T\left(k\right){\underset{-}{\overset{^}{b}}}_{+}\left(k\right).$

In kalman gain vector g (k), will receive the power σ of noise n (k) ² _nTake into account.This power can be easily estimated in receiver for example in the scope of channel identification.Needn't estimate accurately that because as is known, Kalman filter is tough and tensile (robust) to the noise power error.

Just be because proofreaied and correct mean vector, follow covariance matrix so must proofread and correct.Corresponding Kalman filter equation is: $\underset{-}{P}\left(k\right)={\underset{-}{P}}_{.}\left(k\right)-g\left(k\right){\underset{-}{h}}^T\left(k\right){\underset{-}{P}}_{+}\left(k\right),$ Wherein reused kalman gain vector g (k).Square $\underset{-}{\overset{^}{b}}\left(k\right)$ With P(k) be the first and second rank squares, it is considered the sampling value y (k) of current reception.

Suppose selected 1 bit, before k fades to k+1 constantly from sampling, introduce described feedback (soft decision).Subsequently when decisioing making when this bit no longer occurs in the state vector b of channelling mode (k+1) at moment k+1.Owing to feedback has been improved estimation, because Kalman filter (not having feedback) does not consider to have launched binary signal.Kalman filter is for example said and is supposed that impliedly gaussian probability distributes, that is the emission data of Gaussian Profile.Because feedback is included in and belongs to $\underset{-}{\overset{^}{b}}\left(k\right)$ And in (Gauss) probability distribution of P (k), the i rank (i of abridged state vector b (k) when fading to the transformation of k+1 from k ^Th) element is binary.The probability distribution that obtains has square

The advantageous particularly part is that the improved square of present embodiment calculates by the equation that is similar to the Kalman filter equation.Consequently can adopt identical in fact algorithm to this calculating.For obtaining mean vector ${\underset{-}{\overset{^}{b}}}_{*}\left(k\right):$ ${\underset{-}{\overset{^}{b}}}_{.}\left(k\right)=\underset{-}{\overset{^}{b}}\left(k\right)+g.\left(k\right)e.\left(k\right)$ Used improved kalman gain vector: $g_{.}\left(k\right)=\frac{\underset{-}{P}\left(k\right){\underset{-}{u}}_i}{{\underset{-}{u}}_i^T\underset{-}{P}\left(k\right){\underset{-}{u}}_i}$ And used estimation error to count e _*(k).For α (k) $e_{.}\left(k\right)=\mathrm{\α}\left(k\right)-{\underset{-}{u}}_i^T\underset{-}{\overset{^}{b}}\left(k\right).$ Can derive subsequently: $\mathrm{\α}\left(k\right)=\tanh {\frac{{\underset{-}{u}}_i^T\underset{-}{\overset{^}{b}}\left(k\right)}{{\underset{-}{u}}_i^T\underset{-}{P}\left(k\right){\underset{-}{u}}_i}}^{-}.$ Vector u _iBe unit vector, wherein i rank element is different from zero, that is equals 1.Suppose that the i rank element of state vector b (k) is chosen and from vector, take out.

Another embodiment with relevant emission data " firmly " judgement obtains like this, promptly replaces tanh (tanh) with sign function (sign) in the equation of α (k).

Covariance matrix there is improved Kalman filter equation $\underset{-}{P.}\left(k\right)=\underset{-}{P}\left(k\right)-g.\left(k\right){\underset{-}{u}}_i^T\underset{-}{P}\left(k\right)\mathrm{\β}\left(k\right)$ Have following correction factor: $\mathrm{\β}\left(k\right)=1-\frac{1-{\mathrm{\α}}^2\left(\underset{-}{k}\right)}{{\underset{-}{u}}_i^T\underset{-}{P}\left(k\right){\underset{-}{u}}_i}$ Thereby square Occur, this is to consider that the i rank element of state vector b (k) is 1 bit.Before changing to sampling moment k+1, from state vector b (k), omit this bit simultaneously, this bit judges it is according to estimation vector or mean vector ${\underset{-}{\overset{^}{b}}}_{*}\left(k\right)$ Coherent element make.In addition, can level off to the bit error probability, because square ${\underset{-}{\overset{^}{b}}}_{*}\left(k\right)$ And P (k) gives probability distribution.Can obtain for the bit error probability:

This quality standard can advantageously be calculated in the decoder in circuit downstream, and is more reliable so that information sends.

If be omitted at a plurality of bits when k is converted to k+1, then can correspondingly calculate the filter equation of correction usually, can insert improved square simultaneously ${\underset{-}{\overset{^}{b}}}_{*}\left(k\right)$ And P _*(k) to replace $\underset{-}{\overset{^}{b}}\left(k\right)$ And P (k).If no bit is omitted, then the filter equation of cancellation correction is shown out simultaneously ${\underset{-}{\overset{^}{b}}}_{*}\left(k\right)=\underset{-}{\overset{^}{b}}\left(k\right)$ And P _*(k)=P (k).If save feedback also will be this situation.

From sampling constantly k be that transformation equation according to state space description takes place to the transformation of k+1.The Kalman that follows to mean vector expects that equation is: ${\underset{-}{\overset{^}{b}}}_{*}(k+1)=\underset{-}{A}\left(k\right){\underset{-}{\overset{^}{b}}}_{*}\left(k\right).$ Can obtain covariance matrix: ${\underset{-}{P}}_{*}(k+1)=\underset{-}{A}\left(k\right){\underset{-}{P}}_{*}\left(k\right){\underset{-}{A}}^T\left(k\right)+{\underset{-}{O}}_{\mathrm{b\Δ}}\left(k\right).$ Matrix ${\underset{-}{O}}_{\mathrm{b\Δ}}\left(k\right)$ Be diagonal matrix, it comprises the covariance at each bit of moment k+1 emission, that is 1 value (otherwise being 0) in 1 row, and relevant bits appears among the unknown vector b (k+1) in this row.Because it is equally distributed supposing everybody, that is be not with average, so at estimation vector or mean vector ${\underset{-}{\overset{^}{b}}}_{*}(k+1)$ In corresponding line in occur zero.Described and had the prediction steps and the operating process (cycle) that ends this and state exchange dependence among equations connection.In the present embodiment, in signal processor, not carrying out matrix multiplication during the prediction steps, only store the operation that those have been made explanations with reference to state transition equation.

Summed up described wave detector and assessed the cost to compare with the optimization wave detector sharp and descend, increased and there is the wrong rate of tangible bit.In addition, this wave detector generates the estimation of bit error probability.This can be used in the decoder to strengthen the reliability of emission.This wave detector also adopts the multipath transmission to obtain superiority.Concerning the sign indicating number sequence, must consider unrestricted with respect to the time cycle.Allow user's difference and variable data rate.The user also needn't be by synchronously.All these is a unexistent advantage in the suboptimization wave detector.

Claims (8)

1. wave detector that is used for the cdma communication system receiver, wherein the wave detector of She Zhiing is used for the description according to channel between transmitter or different transmitters and receiver, and the sign indicating number sequence that produces by sign indicating number sequencer, calculate the estimation sequence of value of the data sequence of emission, the joint probability distribution that it is characterized in that transmission of data sequences is by calculating adjoint matrix and from the estimation sequence of the data sequence of emission its derivation being calculated.

2. the described wave detector of claim 1 is characterized in that joint probability distribution only uses the incompatible calculating of restriction set of square.

3. the described wave detector of claim 2 is characterized in that only calculating the first and second rank squares when calculating joint probability distribution.

4. the described wave detector of claim 3 is characterized in that adopting Kalman filter as wave detector.

5. claim 1,2, one of 3 or 4 described wave detectors it is characterized in that making according to estimation about the judgement of data sequence value, and these judgements are fed in wave detector.

6. the described wave detector of claim 5 when it is characterized in that on duty being fed, is considered the quality of estimating at the value that has been fed.

7. the described wave detector of claim 1 is characterized in that introducing the sign indicating number sequence that is used for spread spectrum in channel is described.

8. the described wave detector of claim 7 is characterized in that in order to introduce the sign indicating number sequence in channel is described, to provide channel being described the symbol manipulation of carrying out.

Images