EP1728371A1

EP1728371A1 - Sequence estimation in presence of an interfering channel

Info

Publication number: EP1728371A1
Application number: EP05731946A
Authority: EP
Inventors: Sabah Badri-Hoeher; Peter Adam Hoeher; Claudiu Krakowski; Wen Xu
Original assignee: BenQ Mobile GmbH and Co OHG
Current assignee: Qualcomm Inc
Priority date: 2004-03-25
Filing date: 2005-03-15
Publication date: 2006-12-06
Also published as: WO2005094024A1

Abstract

The invention relates to a method and a communication device (MS) for sequence estimation of received data (y[k]) in a device (MS) of a cellular communication network, said method comprising the steps of - determining a number (w) of states (Kd, Ki) to be used for 10 said sequence estimation (SO), - assigning a corresponding number of said received data to the states to be used, and - executing of said sequence estimation with assigned data and states (S5), as well as - determining of a combination of states (Kd, Ki) to be used for the received data, and - assigning the data in correspondence with the optimal combination to the states (S3, S4) before executing of the sequence estimation (S5).

Description

SEQUENCE ESTIMATION IN PRESENCE OF AN INTERFERING CHANNEL

Background of the Invention

Field of the Invention

The invention relates to a method for sequence estimation in a cellular communication network according to pre- characterizing part of claim 1. Further, the invention relates to a communication device for sequence estimation of received data in a cellular communication network according to pre-characterizing part of claim 9.

Description of the Related Art

Currently, single antenna co-channel interference cancellation (SAIC) is being a hot topic, especially for the Global System for Mobile communications/Enhanced Data (GSM/EDGE) downlink. In field trails, tremendous capacity gains have recently been demonstrated [1: 3G Americas, "SAIC and synchronized networks for increased GSM capacity, " www.3gamericas.org, Sept. 2003], particularly for synchro- nized networks in urban areas. As a consequence, the upcoming GSM/EDGE release will be tightened with respect to interference cancellation.

One of the most challenging tasks is the design of the inter- ference canceller, especially if due to cost, volume, power consumption, and design aspects only one receive antenna is available. Most interference cancellers fail if the number of receive antennas do not exceed the number of co-channels .

Algorithms for co-channel interference rejection can be classified in filter-based interference cancellation techniques and multi-user detection techniques [11: C. Tidestav, M. Sternad, and A. Ahlen, "Reuse within a cell - Interference rejection or multi-user detection," IEEE Trans. Commun., vol. 47, pp. 1511-1522, Oct. 1999] . In this paper, focus is on multi-user detection techniques, because they typically offer a superior performance, especially in synchronized Time Division Multiple Access (TDMA) networks [1] . The optimal receiver in the sense of maximum-likelihood sequence estimation is a so-called Joint Maximum-Likelihood Sequence Estimator (JMLSE) . However, the computational complexity of the JMLSE is prohibitive, since it grows exponentially with the number of co-channels, and with the effective memory length of the co-channels . Numerous papers have been published in order to reduce the complexity of the JMLSE [10: P.A. Ranta, A. Hotti- nen, and Z.-C. Honkasalo, "Co-channel interference cancella- tion receiver for TDMA mobile systems," in Proc. IEEE ICC '95, Seattle, pp. 17-21, June 1995], [2: J.-T. Chen, J.-W. Liang, H.-S. Tsai, and Y.-K. Chen, "Low-complexity joint MLSE receiver in the presence of CCI," IEEE Commun. Letters, vol. 2, pp. 125-127, May 1998], [9: C. Luschi and B. Mulgrew, "Non-parametric trellis equalization in the presence of non- Gaussian interference," IEEE Trans. Commun., vol. 51, .no. 2, pp. 229-239, Feb. 2003], [6: A. Hafeez, D. Hui, and H. Arslan, "Interference cancellation for EDGE via two-user joint demodulation," in Proc. IEEE Veh. Techn. Conf., Oct. 2003], [7: D. Hui and R. Ramesh, "Maximum likelihood sequence estimation in the presence of constant envelope interference," in Proc. IEEE Veh. Techn. Conf., Oct. 2003], [8: K. Kim and G.L. Stϋber, "Interference canceling receiver for range extended reception in TDMA cellular systems," in Proc. IEEE Veh. Techn. Conf., Oct. 2003].

Further, [4: A. Duel-Hallen and C. Heegard, "Delayed decision-feedback sequence estimation," IEEE Trans. Commun., vol. 37, no. 5, pp. 428-436, May 1989], [5: M.V. Eyuboglu and S.U. Qureshi, "Reduced-state sequence estimation with set partitioning and decision feedback," IEEE Trans. Commun., vol. 36, no. 1, pp. 13-20, Jan. 1988] for single user systems it is proposed to use a reduced-state trellis-based equalization. In [10], [2], [6], [7], [8] special cases of this general framework are considered.

Object of the Invention

An object of the invention is to improve a method of sequence estimation in a cellular communication network and to improve a communication device using sequence estimation of received data in a cellular communication network.

Summary of the Invention

This object is solved by a method for sequence estimation in a cellular communication network according to the features of claim 1 and by a communication device for sequence estimation of received data in a cellular communication network according to features of claim 9.

In a preferred embodiment there is provided a method for sequence estimation of received data, in a device of a cellular communication network, said method comprising the steps of determining a number of states to be used for said sequence estimation, assigning a corresponding number of said received data to the states to be used, and executing of sequence estimation with assigned data and states as well as determining of an optimized, especially optimal combination of states to be used for the received data, and assigning the data corresponding to the optimized combination to the states before executing of the sequence estimation.

According to the preferred embodiment there is provided a communication device for communicating with another device in a cellular communication network, said communication device comprising a receiver unit for receiving of data sent from said other device via a communication channel and receiving data sent from at least one further device via an interfering co-channel, and a processing unit for processing of said received data and executing a sequence estimation algorithm with reduced number of states. The processing unit is designed for determining an optimized combination of states to be assigned to said sequence estimation algorithm according to such a method.

Especially, there is preferred a method, wherein the optimized combination is determined as an optimal number of states assigned to said data received via a channel between communicating devices on the one hand and on the other hand as an optimal number of states assigned to data received in the receiving one of said devices via an interfering channel. I. e. the reduced number of states is divided up to states for communication channel data and to states for interfering co-channel data.

Especially, there is preferred a method, wherein the optimized combination is determined by determining the minimum squared Euclidean distance of possible combinations of states .

Especially, there is preferred a method, wherein the optimized combination is determined by determining the minimum squared Euclidean distance of all possible combinations of states .

Especially, there is preferred a method, wherein the optimized combination is determined as the combination with the highest minimum squared Euclidean distance of all combinations .

Especially, there is preferred a method, wherein the sequence estimation is executed using a trellis-based algorithm, said number of states being states of a corresponding trellis diagram. Especially, there is preferred a method, wherein said cellular network is executed under General System for Mobile communications and wherein said sequence estimation is a de- layed-decision feedback sequence estimation using a Viterbi algorithm.

Especially, there is preferred a method, wherein said data are received via at least one channel and via at least one interfering co-channel .

Advantageously, embodiments are subject matter of dependent claims .

According to a preferred embodiment two novel aspects are tackled. First, the principle of joint reduced-state trellis- based equalization, originally proposed for single-user systems, is generalized to the multi-user case. Secondly, the performance is optimized by means of adaptive state allocation. Especially, single antenna co-channel interference cancellation for cellular TDMA networks by means of joint delayed- decision feedback sequence estimation is proposed. The performance can be increased by a preferred adaptive state allo- cation technique.

As a particular application, the GSM system is considered. Therefore, emphasis is on Delayed Decision-Feedback Sequence Estimation (DDFSE) [4] . For non-binary modulation, a gener- alization to Reduced-State Sequence Estimation (RSSE) [5] by additionally applying set-partitioning of the symbol constellation is straightforward.

Brief description of the drawings

An embodiment will be described hereinafter in further detail with reference to drawings . Fig. 1 illustrates a cellular communication network and a mobile station receiving data from a base station of a first cell and receiving co-channel interfering signals from a second base station;

Fig. 2 illustrates components of the mobile station used for sequence estimation of received data and a flow-chart of a preferred method for preparing received data;

Fig. 3 illustrates a raw bit error rate versus Carrier-to- interference ratio C/I for a JDDFSE with 16 states for a TU0 Channel Model with perfect channel knowledge; and

Fig. 4 illustrates a raw bit error rate versus C/I for a JDDFSE with 16 states of a TU50 Channel Model using joint least-squares channel estimation.

Detailed description of the preferred embodiment

As can be seen from Fig. 1 a mobile station MS used as a user station communicates via a radio channel with a first base station BS1 of a cellular communication network GSM. The mo- bile station stays in a first radio cell cl around the first base station BS1. There are some objects 0 disturbing communication between the mobile station MS and the base station BS1. A first object 0 interrupts direct channel path si between the first base station BS1 and the mobile station MS . The second object 0 arranged beside the direct communication path serves as a reflector for radio waves. Therefore, a second communication path s2 transmits radio waves sent by the first base station BS1 and being reflected by the second object 0. Therefore, mobile station receives first data via a first communication path si and second data via a second communication path s2. Signal strength of data received via different communication paths si, s2 is different one from each other. Further, data received over second communication path s2 are received to a later time than data received via direct first communication path si.

The communication network provides further base stations BS2, BS3 each having a communication cell c2, c3, said communication cells cl - c3 being arranged in an overlapping manner. A mobile station MS moving from a first cell cl to a second cell c2 changes from a first base station BS1 to a second base station BS2 after reaching handover regions ho of the overlapping cells . However, before reaching a handover region ho the mobile station MS receives interfering signals ill, il2, i21 from the other base stations BS2, BS3. Especially, the interfering signals are signals sent from the other base stations BS2, BS3 on the same frequency channel to further mobile stations MS*, MS' in their communication cells c2, c3. Therefore, the mobile station receives co-channel interference disturbing the data received from their own first base station BS1. In addition, the mobile station MS is disturbed by thermal noise n caused by the components of a Front-End device. The thermal noise is in general additive white Gaussian noise.

Usually, the base stations BS1 - BS3 of a cellular communica- tion network are connected via a base station controller BSC to other components of the communication network. To avoid co-channel interference base station controller BSC assigns different channels, especially frequency channels to neighboring cells cl - c3. According to a preferred embodi- ment same frequency channels shall be used in neighboring cells . To reduce co-channel interference in the following it is proposed to use sequence estimation, in the case of GSM a joint delayed-decision feedback sequence estimation to improve reconstruction of data received by the mobile station MS. Data y [k] received by the mobile station MS can be written as : y[k] = ∑hM k-l]+∑∑gjj[khi^k-^l]+"ikl ≤ k ≤ K-l 1=0 /=ι?=o ₍₁₎

The first terms describe the signals si, s2, ... received via communication paths i. e. communication channels si, s2 from the first base station BS1. These signal components si, s2 shall be used by sequence estimation to reconstruct data originally sent by the first base station BSl. Second sums and terms correspond to the signal components received via interfering communication paths ill, il2, 121 from second and third base stations BS2, BS3. These interfering signal components ill, il2, 121 has to be^"* distinguished by the joint se- quence estimation procedure in presence of additive white

Gaussian noise values n[k]. All signal components si, s2 and interfering signals components ill, 112, i21 has been sent by corresponding base stations BSl, BS2, BS3 using same channel especially using same frequency f(ij ]_).

As can be seen from Fig. 2, mobile station MS comprises a plurality of components . A transmitter or at least a receiver unit TX/RX receives the signal y[k]. A processing unit C serves for operating of the mobile station MS and for running processes for digital signal processing. One component of the processor unit C is designed to execute a Viterbi algorithm (VA) . Shown VA uses K = 2⁴ = 16 states for data processing. Further, mobile station MS comprises a memory MEM for storing data to be processed, data being processed and procedures and programs for processing of the data. Components of the mobile station are connected by a data bus B.

Usually, in a case like that of Fig. 2 there exist v = 5 possible combinations . These are Kd = M^κ* states for the de- sired user, wherein κ_d is a design parameter which is optimized as trade of between complexity and performance, for re- construction of data received from the first base station BSl, and Ki = M^κ' states for the interfering signals ill, 112, i21 received from second and third base station BS2, BS3, wherein K. is a design parameter which is optimized as trade of between complexity and performance. However, Viterbi detector of the mobile station MS shall use only K = 16 states (procedure step SO) .

In a first step SI it is determined the minimum squared Euclidean distance d² _mi_n,_z(Kd, Ki) with respect to z = 1, ..., v possible state combinations (Kd, Ki) . Determination has been done for all combinations of possible states (Kd, Ki) with the total number of states w = Kd*Ki = 16.

In a second step S2 the highest of the minimum squared

Euclidean distances d² _mi_nrZOp = πtax{d² _min,z (Kd, Ki)} will be determined to get optimal combination of states (Kd, Ki) .

Thereafter, in a third step S3 there is done a setting of the optimal state relation for the possible states Kd and Ki in the Viterbi detector VA like the state relation of the. combination (Kd, Ki) maximizing the minimum squared Euclidean distance d min-

Thereafter, (S4) the w states for the VA decoder are set with the corresponding state combination before starting (S5) of the Viterbi and the joint delayed-decision feedback sequence estimations .

Therefore, preferred method and procedure determines the best combination (Kd, Ki) for signal components and interfering components received by the mobile station MS from their own first base station BSl or interfering base stations BS2, BS3.

In the following there is offered an introduction of the equivalent discrete-time channel model under consideration, a short description of the metric used in the conventional DDFSE is given, which ignores CCI . Based on the conventional DDFSE, a joint reduced-state sequence estimator (JDDFSE) is derived, which takes CCI into account . The performance of JDDFSE can be improved by a preferred adaptive state alloca- tion technique, which is presented next.

The equivalent discrete-time channel model considered is given as y[k]= ∑hM"[k-ι]+∑∑g_j,t[k]b.[k-ι]+n[kl 0 ≤ k≤ K-l , <1) 1=0 j=\ 1=0

where y[k] C is the &-th baud-rate output sample of the analog receive filter, L is the effective memory length of the discrete-time ISI channel model, [fc]e C are the channel co- efficients of the desired user ( R|h[tJ C are the channel coefficients of the j -th interferer, l ≤ j ≤ J , J is the number of interferer, a[k] and are the fc-th independent identically distributed data symbols of the desired user and the /-th interferer, respectively, both randomly drawn over an M -ary alphabet ), n[k]e C is the Jfc-th sample of a Gaussian noise process N₀ /R_ ) , k is the time index, and K is the number of -ary data symbols per burst. All random processes are assumed to be mutu- ally independent. The channel coefficients h[&]:= [A₀[t],...,A_£[/t]]^r and g[fc]:=g_Λβ[fc],...,g-__/)t[fc]f comprise pulse shaping, the respective physical channel, analog receive filtering, the sampling phase, and the sampling frequency. Without loss of generality, the effective memory length, L , is assumed to be the same for all co-channels. Especially, some coefficients can also be zero.

In case of square-root Νyquist receive filtering and baud- rate sampling, the Gaussian noise process is white. This case is assumed in the following. The equivalent discrete-time channel model is suitable both for synchronous as well as asynchronous TDMA networks, because time-varying channel coefficients are considered.

In the following there is described a multi-user detector design and optimization.

A first case A relates to Delayed Decision-Feedback Sequence Estimation (DDFSE) .

First, there is considered the case of no co-channel interference. The corresponding baud-rate equivalent discrete-time channel model can be written as

For additive white Gaussian noise, the DDFSE is defined as [4], [5] a = (3)-

where K is a design parameter (O≤K≤Z,) . The number of states is M^κ . The third term in (3) is taken into account by state-dependent decision feedback called "parallel decision feedback equalization" (PDFE) .

A second case B relates to Joint Delayed Decision-Feedback Sequence Estimation (JDDFSE) .

Now, there is considered the case of J = l dominant interferer for the purpose of derivation of the JDDFSE:

A generalization to multiple interferers is straightforward. For convenience, the index j is dropped. It is assumed that the channel coefficients h[fc]and g[k] are known to the mobile station, they can be estimated by means of channel estimator such as Joint Least-Squares Channel Estimation (JLS-CE) or Joint Least-Mean-Squares Channel Estimation (JLMS-CE) . For additive white Gaussian noise, the JDDFSE can be defined as

(5)

where κ._d (Kd) and K₍. (Ki) are design parameters, which can be chosen independently within the range 0<κ_rf,K_; ≤ L .

In the general case, one design parameter is needed for each of the J+ l co-channels. According to the exemplary embodiment the total number of states is M^κ,1+K' , where M^Kj corresponds to the number of states Kd of the desired user and M^κ' corresponds to the number of states Ki of the dominant interferer. The last term in (5) is taken into account by state-dependent decision feedback called "joint parallel decision feedback equalization, JPDFE". For κ_d =κ_f =0, Joint

Decision-Feedback Equalizer (JDFE) is obtained. The other extreme is κ_d = κ. = L , where JMLSE is obtained.

A second case C relates to the Adaptive State Allocation.

As pointed out in the previous subsection, the design parameters κ_d and κ₍. can be chosen independently. This provides many degrees of freedom. For example, if the total number of states is fixed to be M^Kj+K' =16, possible choices are

( _rf = 0,κ,. = 4) , (κ_d =l,κ. = 3) , (κ_d = 2,κ. = 2) , (κ_rf = 3,κ,. =l) , and

(κ_d =4,κ,. =θ) . Conventionally, the design parameters would be selected so that on average the error performance is low. An alternative approach according to the preferred embodiment is to optimize the design parameters for each burst giving a fixed number of states and given h[k] and g[k] . Let us again assume that the total number of states is sixteen. If the dominant interferer is weak with respect to the desired user, (κ_d =3,κ,. =l), or ( _rf=4,κ,. =θ) shall be selected. Vice versa, if the dominant interferer is strong with respect to the desired user, (κ_rf=0,κ_z. =4) or ( _d=l, . =3) is likely to be better. This concept is dubbed adaptive state allocation (ASA) . Given essentially the same computational complexity compared to a fixed allocation plus some additional complexity for the selection rule, a lower average error rate can be expected. As a suitable selection criterion, the minimum squared Euclidean distance d^ between possible paths in the joint reduced trellis is chosen here, but other criteria like e.g. the energy of the channels may be suitable as well.

In [12: H. Zamiri-Jafarian and S. Pasupathy, "Complexity re- duction of the MLSD/MLSDE receiver using the adaptive state allocation algorithm," IEEE Trans . Wireless Commun . , vol. 1, pp. 101-111, Jan. 2002], an adaptive state allocation scheme has been proposed for single-user receivers . The number of states is adapted to the short-term received power, whereas in present approach the computational complexity is constant. Of course, in present approach the overall number of states can be made adaptive as well.

The proposed adaptive state allocation technique can be de- scribed for J = l interferer as follows: In the noiseless case, the squared Euclidean distance between the transmitted sequence y and any other sequence y is given as (6)

where α[&] and β[&] are called difference symbol of desired user and interferer, respectively, at time index k . For systems with binary antipodal modulation the difference symbols will be elements of e.g. -2, 0, and 2 (α[&],β[A;]e {0,2,-2}) . In order to simplify the computations, in (6) it is assumed that the JPDFE-part of the metric (c.f., (5)) cancels out due to correct decisions. The computation of the minimum squared distance, d^ , can be done in the so-called joint difference trellis . For systems with binary antipodal modulation, the joint difference trellis has 3^K,,+1' states. Typically, the error events do not exceed three or four times the constraint length L + l . Besides for adaptive state allocation,

{κ_d,κ. ) =

the minimum squared distance can simultaneously serve as an indicator for the instantaneous bit error rate: ~

Maximizing the squared Euclidean distance is similar to maxi- mizing the energy of the channel coefficients taken into account .

For the numerical results presented in Fig. 3 and Fig. 4, a synchronous GSM network with J = l dominant interferer is as- sumed. Fig. 3 illustrates a raw Bit-Error-Rate (BER) vs. a Carrier-to-interference ratio (C/I) for a JDDFSE with 16 states using TUO channel model, perfect channel knowledge, a synchronous GSM network, and Es/N₀ = 20 dB. Fig. 4 illustrates a raw BER vs. C/I for a JDDFSE with 16 states using TU50 channel model, joint least-squares channel estimation, a synchronous GSM network, and Es/N₀ = 20 dB.

The number of the training sequence code (TSC) of the desired user is assumed to be uniformly distributed over the 8 possi- ble sequences specified for GSM. The same applies to the TSC of the dominant interferer, with the exception that it is assumed to be different from the TSC of the desired user. This scenario can be managed by the network operator. Since co- channel interference is particularly a problem in densely populated areas, the GSM 05.05 Typical Urban (TU) channel model has been assumed, which is characterized by an efficient channel memory length of L = 3. In order to provide reliable results, 50 thousand statistically independent bursts have been generated. A JDDFSE equalizer with 16 states is used, both with and without adaptive state allocation. As a benchmark, performance results for the. conventional receiver ignoring CCI (8 states) and for JMLSE (64 states) are included as well. In Fig. 3 perfect channel knowledge is assumed, whereas in Fig. 4 joint least-squares channel estima- tion is done [10] .

As indicated by these figures, with manageable complexity near-optimum performance can be obtained even without additional adaptive prefiltering [6] .

According to preferred embodiment a class of reduced-state trellis-based multi-user detectors, called joint reduced- state delayed decision feedback estimation (JDDFSE) , is described and explored. The aim of JDDFSE is to eliminate co- channel interference and intersymbol interference jointly. In contrast to the JMLSE, the complexity/performance trade-off of JDDFSE is adjustable. Especially in conjunction with adap- tive state allocation, a computational complexity can be achieved which is comparable to the complexity of a conventional GSM receiver neglecting co-channel interference, although the performance loss compared to JMLSE is small. The concept of JDDFSE can easily be generalized to joint reduced- state sequence estimation (JRSSE) . The algorithm can also easily be modified to deliver soft outputs .

Claims

Claims :

1. Method for sequence estimation of received data (y[k]) in a device (MS) of a cellular communication network, said method comprising the steps of

- determining a number (w) of states (Kd, Ki) to be used for said sequence estimation (SO) ,

- assigning a corresponding number of said received data to the states to be used, and - executing of sequence estimation with assigned data and states (S5) , c h a r a c t e r i z e d , by

- determining of an optimized combination, especially an optimal combination of states (Kd, Ki) to be used for the re- ceived data, and

- assigning the data corresponding to the optimized combination to the states (S3, S4) before executing of the sequence estimation (S5) .

2. Method according to claim 1, wherein the optimized combination is determined as an optimal number .(Kd) of states assigned to said data received via a channel (si, s2) between communicating devices (MS, BSl) on the one hand and on the other hand as an optimal number (Ki) of states assigned to data received in the receiving one of said devices (MS) via an interfering channel (ill, il2, i21) .

3. Method according to claim 1 or 2 wherein the optimized combination is determined by determining the minimum squared Euclidean distance (d² _min,z(Kd, Ki) ) of possible combinations of states (Kd, Ki) .

4. Method according to claim 1 or 2 wherein the optimized combination is determined by determining the minimum squared Euclidean distance (d² _min,z (Kd, Ki) ) of all possible combinations of states (Kd, Ki) .

5. Method according to claim 3 or 4, wherein the optimized combination is determined as the highest minimum squared Euclidean distance (d² _min(Kd, Ki) ) of said determined combinations .

6. Method according to anyone of the preceding claims, wherein the sequence estimation is executed using a trellis- based algorithm, said number (Kd, Ki; w) of states being states of a corresponding trellis diagram.

7. Method according to anyone of the preceding claims, wherein said cellular network is executed under general system for mobile communications (GSM) and wherein said sequence estimation is a delayed-decision feedback sequence estima- tion.

8. Method according to anyone of the preceding claims, wherein said data are received via at least one channel (si, s2) and via at least one interfering co-channel (ill, il2, i21) .

9. Communication device (MS) for communicating with another device (BSl) in a cellular communication network (GSM) , said communication device (MS) comprising - a receiver unit (TX/RX) for receiving of data (y[k]) sent from said other device (BSl) via a communication channel (si, s2) and receiving data sent from at least one further device (BS2, BS3) via an interfering co-channel (ill, il2, i21) , and

- a processing unit (C) for processing of said received data (y[k]) and executing a sequence estimation algorithm with reduced number (w) of states (Kd, Ki) , c h a r a c t e r i z e d , in that

- said processing unit (C) is designed for determining an optimized combination of states to be assigned to said sequence estimation algorithm according to a method of anyone of the preceding claims .