WO2018108286A1

WO2018108286A1 - Resource spread multiple access detection

Info

Publication number: WO2018108286A1
Application number: PCT/EP2016/081279
Authority: WO
Inventors: Sergei SEMANOV; Alberto Diego JIMENEZ FELTSTROM; Chaitanya TUMULA; Junshi Chen; Neng Wang; Zuleita HO; Thanos DIMITRIOU
Original assignee: Huawei Technologies Co., Ltd.
Priority date: 2016-12-15
Filing date: 2016-12-15
Publication date: 2018-06-21

Abstract

A device is disclosed, the device comprising a receiver and a processor, wherein the receiver is configured to receive a resource spread multiple access (RSMA) signal comprising transmissions by K transmitters over N resource elements, and 5 the processor is configured to choose an NxK sparse matrix as an indicator matrix for the received RSMA signal, wherein each row corresponds to a resource element and each column corresponds to a transmitter and wherein the indicator matrix is representable by a bipartite factor graph comprising K variable nodes and N function nodes, use a Message Passing Algorithm (MPA) on the bipartite factor 10 graph to detect the symbols transmitted by each transmitter, and output the detected symbols transmitted by each transmitter.

Description

RESOURCE SPREAD MULTIPLE ACCESS DETECTION

TECHNICAL FIELD

[0001 ] The present application relates to the field of wireless

communications, and more particularly to a base station and a receiver therein receiving signals from multiple transmitters.

BACKGROUND

[0002] Telecommunication technologies have evolved from voice centric first generation to data centric all internet protocol (IP) fourth generation. During this evolution both the number of connected devices and the volume of data routed over the telecommunication networks have grown exponentially, but a key resource - radio spectrum has been limited. This has necessitated multiple schemes of sharing the radio spectrum resources with multiple users. The so called multiple access technologies like Orthogonal Frequency Division Multiple Access

(OFDMA), Code Division Multiple Access (CDMA) and Time Division Multiple Access (TDM A) aspire to address this issue by sharing the available radio spectrum resources with multiple users efficiently. These technologies, though adequate till now, may not be sufficient for future telecom networks for example the planned fifth generation networks, as the number of devices, the number of users and the amount of traffic is predicted to increase manifold in near future. New access technologies like Non Orthogonal Multiple Access (NOMA) are intended to address this issue. Resource Spread Multiple Access (RSMA) is one such NOMA technology wherein all users being transmit at on the same frequency. RSMA may use long pseudorandom spreading codes for a spreading signal over a channel to provide multiple access.

SUMMARY

[0003] This summary is provided to introduce a selection of concepts in a simplified form that are further described below in the detailed description. This summary is not intended to identify key features or essential features of the claimed subject matter, nor is it intended to be used to limit the scope of the claimed subject matter. [0004] It is an object of the invention to enable an improved device with affordable complexity capable of receiving Resource Spread Multiple Access (RSMA) signals from multiple transmitters. The foregoing and other objects are achieved by the features of the independent claims. Further implementation forms are apparent from the dependent claims, the description and the figures.

[0005] According to a first aspect a device is provided, the device

comprising: a receiver and a processor, wherein the receiver is configured to receive a resource spread multiple access (RSMA) signal comprising transmissions by K transmitters over N resource elements; and the processor is configured to: choose an NxK sparse matrix as an indicator matrix for the received RSMA signal, wherein each row corresponds to a resource element and each column corresponds to a transmitter and wherein the indicator matrix is representable by a bipartite factor graph comprising K variable nodes and N function nodes; use a Message Passing Algorithm (MP A) on the bipartite factor graph to detect the symbols transmitted by each transmitter; and output the detected symbols transmitted by each transmitter. Choosing an arbitrary sparse matrix with desirable properties enables usage of a message passing algorithm to detect an RSMA signal combined with projection based interference cancellation. This enables RSMA signal detection with an affordable complexity at high spectral efficiency.

[0006] In a first possible implementation of the device according to the first aspect, to perform the MPA the processor is configured to iteratively: traverse the matrix row by row and for each row, cancel the interference caused by transmitters corresponding to zeros in the row using projection based interference cancellation; traverse the matrix row by row and for each transmitter corresponding to a one in the current row, calculate a message from function node to variable node comprising a log likelihood ratio (LLR); traverse the matrix column by column and for each one in the column, calculate a message from variable node to function node comprising an LLR; calculate a weighted sum LLR for each transmitter based on the LLRs calculated; and stop the iteration when LLR values converge. Since in RSMA all transmitters collide, cancelling interference according to the pattern of zeros in the chosen indicator matrix makes the resultant, interference cancelled RSMA signal amenable to application of MPA wherein LLRs may be calculated with affordable complexity.

[0007] In a second possible implementation of the device according to the first implementation of the first aspect, while traversing the matrix row by row, calculating a message from function node to variable node for a transmitter corresponding to a one in the current row comprises calculation of an LLR comprising a numerator and a denominator; wherein the likelihood in the numerator of the LLR corresponds to the probability of the event that the symbol transmitted by the current transmitter takes a particular value from an alphabet, given received signal and symbols transmitted by remaining interfering

transmitters; wherein the interfering transmitters are indicated by an interference cancellation pattern represented by positions of ones in the current row of indicator matrix other than the one for which the LLR is being calculated; and wherein the likelihood in the denominator of the LLR corresponds to the probability of the event that the symbol transmitted by the current transmitter takes some fixed value from the alphabet given the same input as for the likelihood in the numerator.

[0008] In a third possible implementation of the device according to the second implementation of the first aspect, while calculating the LLR for the transmitter corresponding to a one in the current row, the processor is further configured to: calculate LLRs for the other rows containing one in the current column of indicator matrix corresponding to the current transmitter using the interference cancellation pattern corresponding to the current row of indicator matrix ; perform a combining of the computed LLRs of the current column and assign it to the LLR value of the transmitter corresponding to a one in the current row. Using the same interference cancellation pattern for multiple rows improves performance of decoding process.

[0009] In a fourth possible implementation of the device according to the first aspect as such or according to any of the preceding implementations of the first aspect, calculating an LLR comprising message from variable node to function node comprises combining LLRs calculated as messages from function nodes to variable nodes corresponding to ones in the current column of indicator matrix; wherein the message from function node to variable node corresponding to the current row of indicator matrix is not combined except in the last iteration; and wherein in the last iteration all LLRs indicated by the ones in the current column of indicator matrix are combined.

[001 0] In a fifth possible implementation of the device according to the first aspect as such or according to any of the preceding implementations of the first aspect, the processor is configured to stop iterating when LLRs comprised in messages from a variable node to a function node converge; or when a maximum number of iterations is reached.

[001 1 ] In a sixth possible implementation of the device according to the first aspect as such or according to any of the preceding implementations of the first aspect, the processor is configured to choose the indicator matrix such that the minimum cycle length of the corresponding bipartite factor graph is greater than 4 or greater than 6. The indicator matrix is chosen with properties that facilitate application of MPA and consequently decoding the RSMA signal becomes less complex.

[001 2] In a seventh possible implementation of the device according to the first aspect as such or according to any of the preceding implementations of the first aspect, the processor is configured to choose the indicator matrix such that the indicator matrix does not contain stopping sets. The indicator matrix is chosen with properties that facilitate application of MPA and consequently decoding the RSMA signal becomes less complex.

[001 3] In an eighth possible implementation of the device according to the first aspect as such or according to any of the preceding implementations of the first aspect, the processor is configured to: choose an indicator matrix according to the measured received power of transmitters in such a way that in at least one row of the chosen indicator matrix, zeroes correspond to transmitters with received power lower than a threshold. Having zeroes correspond to transmitters with low signal power in a row makes detection of RSMA signal more robust, improving the quality of the chosen sparse matrix as the indicator matrix. [001 4] In a ninth possible implementation of the device according to the first aspect as such or according to any of the preceding implementations of the first aspect, the projection based interference cancellation comprises: projecting a vector corresponding to the received RSMA signal onto a vector space which is orthogonal to an interference vector subspace; wherein the interference vector subspace comprises a vector space spanned by the columns corresponding to interference in a matrix representing a generalized communication channel, comprising multiplication of matrix of channel coefficients by matrix of

transmitters' signatures, over which the RSMA signal is received; and wherein the matrix representing the generalized communication channel is split into columns representing the transmitters corresponding to ones in the row of the indicator matrix and columns representing the transmitters corresponding to zeroes in the row of indicator matrix, wherein the signals transmitted by transmitters

corresponding to zeroes in the row comprise the interference to be cancelled.

Projecting the received RSMA signal to a vector subspace orthogonal to the signal subspace is used to cancel the interference from users corresponding to zeros in the current row of MPA traversal. This enables easier detection the symbols

transmitted by transmitters corresponding to ones in the current row.

[001 5] In a tenth possible implementation of the device according to the first aspect as such or according to any of the preceding implementations of the first aspect, the projection based interference cancellation includes a noise correlation matrix in the projection. Including a noise correlation matrix in the projection prevents excessive noise enhancement.

[001 6] In a eleventh possible implementation of the device according to the first aspect as such or according to any of the preceding implementations of the first aspect, the processor is further configured to use a second iterative process to detect and decode the received RSMA signal and to change the indicator matrix for every subsequent iteration of the second iterative process after the first iteration. Changing the indicator matrix for a second iterative process allows interference from different transmitters to be cancelled in different iterations of the second iterative process. It may improve channel diversity for different iterations. [001 7] In a twelfth possible implementation of the device according to the first aspect as such or according to any of the preceding implementations of the first aspect, the processor is further configured to change the indicator matrix for every subsequent iteration of the second iterative process by permuting the columns of an initially chosen indicator matrix. Changing the indicator matrix by permuting the columns provides a low complexity method of choosing new indicator matrices with desired qualities.

According to a second aspect a method is provided, the method comprising:

choosing an NxK sparse matrix as an indicator matrix for a received RSMA signal, wherein each row corresponds to a resource element and each column corresponds to a transmitter and wherein the indicator matrix is representable by a bipartite factor graph comprising K variable nodes and N function nodes; using a Message Passing Algorithm (MP A) on the bipartite factor graph to detect the symbols transmitted by each transmitter; and outputting the detected symbols transmitted by each transmitter. Choosing an arbitrary sparse matrix with desirable properties enables usage of an MPA to detect RSMA signal combined with projection based interference cancellation. This enables detection RSMA signal with an affordable complexity at high spectral efficiency enabling use cases like enhanced mobile broadband.

[001 8] In a first possible implementation of the method according to the second aspect, applying an MPA comprises: traversing the matrix row by row and for each row cancel the interference caused by transmitters corresponding to zeros in the row using projection based interference cancellation; traversing the matrix row by row and for each transmitter corresponding to a one in the current row calculating a log likelihood ratio (LLR); traversing the matrix column by column and for each one in the column calculating a LLR; and calculating a weighted sum LLR for each transmitter based on the LLRs calculated iterating the method till values of LLR converge. Since in RSMA all transmitters collide, cancelling interference according to the pattern of zeros in the chosen indicator matrix makes the resultant, interference cancelled RSMA signal amenable to application of MPA wherein LLRs may be calculated with affordable complexity. [001 9] In a second possible implementation of the method according to the second aspect, while traversing the matrix row by row, calculating an LLR for a transmitter corresponding to a one in the current row comprises using an

interference cancellation pattern; and calculating an LLR for the transmitter corresponding to a one in the current row comprises calculating an LLR is for each one in the current column using the same interference cancellation pattern. Using the same interference cancellation pattern for multiple rows may improve performance of the method.

[0020] In a third possible implementation of the method according to the second aspect or the first and second implementations of the second aspect, computer program comprising a program code is configured to perform the method when executed.

[0021 ] Many of the attendant features will be more readily appreciated as they become better understood by reference to the following detailed description considered in connection with the accompanying drawings.

DESCRIPTION OF THE DRAWINGS

[0022] The present description will be better understood from the following detailed description read in light of the accompanying drawings, wherein:

[0023] FIG. 1 illustrates a schematic representation of a device configured to receive and decode RSMA signal from multiple transmitters according to an embodiment;

[0024] FIG. 2 illustrates a schematic representation of an exemplary sparse matrix chosen as an indicator matrix according to an embodiment;

[0025] FIG. 3 illustrates a schematic representation of a bipartite factor graph corresponding to the indicator matrix of FIG. 2 according to an embodiment;

[0026] FIG. 4 illustrates a flowchart showing a method of receiving and decoding an RSMA signal according to an embodiment of the invention.

[0027] Like references are used to designate like parts in the accompanying drawings. DETAILED DESCRIPTION

[0028] The detailed description provided below in connection with the appended drawings is intended as a description of the embodiments and is not intended to represent the only forms in which the embodiment may be constructed or utilized. However, the same or equivalent functions and structures may be accomplished by different embodiments.

[0029] In a Resource Spread Multiple Access system all transmitters transmit over all resource elements and at high spectral efficiency, detection may be computationally intensive.

[0030] FIG. 1 illustrates device 100 comprising a processor 101, configured to receive a communication from a receiver 102 and output a decoded

communication to an output 103, according to an embodiment. The receiver 102 may comprise one or more antennas and associated circuitry. In an embodiment the receiver 102 may be integral to the processor 101. In another embodiment, the processor may comprise discrete components like a decoder, a detector, etc. The processor 101 may be an application specific integrated circuit, a general purpose microprocessor configured by instructions for example firmware or may comprise a mix of application specific and general purpose circuitry.

[0031 ] Referring to FIG. 1, the receiver 102 is configured to receive a Resource Spread Multiple Access (RSMA) signal comprising transmissions by multiple transmitters. The receiver 102 communicates the received RSMA signal to the processor 101 and the processor 101 processes the signal to output detected symbols to output 103. The output 103 may be another module comprising circuitry and/or instructions to process or forward the detected symbols. The processor 101 may comprise multiple components, each configured to perform a specific task including, but not limited to, detection, demodulation, decoding, and modulation. The components comprising the processor 101 may be discrete or integrated. According to an embodiment, the processor 101 may be integrated with the receiver 102.

[0032] Referring to FIG.1 , the processor is configured to detect and decode symbols in the received RSMA signal comprising transmissions by multiple transmitters. The processor 101 takes the RSMA signal and chooses a sparse matrix to use as an indicator matrix for decoding the signal. The sparse matrix is chosen so that the number of rows is equal to the number of resource elements in the signal and number of columns is equal to the number of transmitters

transmitting over the resource elements. Further for a candidate sparse matrix to be chosen as the indicator matrix, the processor ensures that it has certain desirable properties. For example the corresponding bipartite factor graph does not have stopping sets and cycle length in the graph is higher than 4. Further, if the processor 101 has prior knowledge of the received signal power of the transmitters, it chooses the sparse matrix such that at least one row exists in the matrix where the zeroes correspond to transmitters with low signal power. The processor 101 then applies a Message Passing Algorithm (MP A) on the indicator matrix to decode the received RSMA signal. The indicator matrix chosen by processor 101 is

representable as a bipartite factor graph comprising function nodes corresponding to resource elements in the RSMA signal and variable nodes corresponding to the transmitters transmitting on the resource elements.

[0033] Referring to FIG. 1, the MPA implemented by the processor 101 is an iterative in nature and in each iteration the processor 101 is configured to suppress interference from transmitters designated by zeroes in the current row by using projection based Interference Cancellation (IC). In an embodiment, the projection based IC may also take noise into account, for example, by including a noise covariance matrix in the projection. Further the processor 101, calculates a message from a function node to corresponding variable node which comprises a Log Likelihood Ratio (LLR) for each one in the current row. Subsequently, the processor 101 traverses the indicator matrix column by column and calculates a message from a variable node to a function node which comprises an LLR. The processor 101, for each one in the matrix, then combines the LLRs corresponding to it.

[0034] In an embodiment, the processor 101 may calculate a message from function node to variable node for a transmitter corresponding to a one in the current row as an LLR having a numerator and a denominator. The numerator of the LLR corresponds to the probability of the event that the symbol transmitted by the current transmitter takes a particular value from an alphabet, given received signal and symbols transmitted by remaining interfering transmitters indicated by an interference cancellation pattern represented by positions of ones in the current row of indicator matrix other than the one for which the LLR is being calculated. The denominator of the LLR corresponds to the probability of the event that the symbol transmitted by the current transmitter takes some fixed value from the alphabet given the same input as for the likelihood in the numerator.

[0035] Referring to Fig. 1, according to an embodiment, the processor 101 while calculating the LLR for a transmitter represented by a one in the current row may further calculate LLRs for the transmitter in other rows where a one exists in the row in the current column using the same interference cancellation pattern as in the current row. The LLRs of the current column may then be combined by the processor 101 and assigned to the current message from function node to variable node.

[0036] According to yet another embodiment, the processor 101 calculates an LLR comprising message from variable node to function node by combining LLRs calculated as messages from function nodes to variable nodes corresponding to ones in the current column of indicator matrix. However the processor 101 does not combine the message from a function node to a variable node corresponding to the current row of indicator matrix except in the last iteration and in the last iteration the processor 101 combines all LLRs indicated by the ones in the current column of indicator matrix.

[0037] FIG. 2 illustrates an exemplary sparse matrix 200 which may be chosen as an indicator matrix by processor 101 to detect incoming RSMA signal forwarded by receiver 102. FIG. 3 illustrates the bipartite factor graph 300 corresponding to the indicator matrix of FIG. 2.

[0038] Referring to FIG. 2 the indicator matrix chosen by processor 101 as represented by the bipartite factor graph 300 of FIG. 3 is selected such that the minimum cycle length in the graph 300 is more than 4 and it contains no stopping sets. Each row of the indicator matrix 200 corresponds to a resource element (RE) in the received RSMA signal and each column corresponds to a transmitter.

[0039] Since in RSMA all transmitters collide over each resource element, the real indicator matrix of an RSMA signal may not correspond to a sparse matrix such as 200. The sparse matrix 200 is chosen arbitrarily with the constraints discussed above to aid in detection of the signal. Interference by transmitters corresponding to zeroes in the indicator matrix is removed by the processor 101 using projection based IC. The dimensions of the sparse matrix chosen as indicator matrix 200 are NxK where N is the number of REs and K is the number of transmitters. The indicator matrix 200 is by way of illustration, in practice the size of a sparse matrix chosen as an indicator matrix for an RSMA signal depends on the number of REs and the number of transmitters. The indicator matrix may be chosen by using, for example, combinatorial designs. The position of ones in the nth row of the indicator matrix F_NxK denotes the set of transmitters that contribute their data at the nth symbol, while its kth column represents the set of symbols over which transmitter spreads its data. The maximum number of ones in each column d_u indicates the maximum number of nonzero spread symbols, which can be located for each transmitter among the N possible time-frequency resources.

Further each spread symbol will collide in the channel with d_c (maximum number of ones in row) symbols from other transmitters.

[0040] To process the incoming RSMA signal so that it corresponds to the indicator matrix 200, the processor 101 performs successive interference

cancellation for each iteration of the message passing algorithm. The signal may be modelled as y = Gx + n where y is the received signal, x is the vector comprising transmitted symbols of all transmitters and n is the noise vector. Matrix G is the generalized channel matrix including both channel coefficients and transmitter signatures. Denoting the part of vector x corresponding to signal of interest by x_T and the other part of vector x corresponding to interference by x_Q. Matrix G is split into two parts corresponding to signal of interest and interference and can be represented as G = [T, Q], where T and Q correspond to signal of interest and interference. This allows the signal to be represented as y = Gx + n = Tx_T + Qx_Q + n.

[0041 ] After projecting the received signal y onto null space of Q the processor 101 calculates:

= Tx_r + Qx_e +n- Q(Q^HQ) ¹Q^HTx_r - Q(Q^HQ) ¹Q^HQx₈

Q(Q^HQ) ¹Q^Hn = P_Q ¹Tx_r + P_Qn

[0042] Denoting vector P^ as , vector P^n as ϊϊ and matrix P^T as G .

Then the above equation may be represented as:

y = Gx_r ϊϊ

[0043] This projection based interference cancellation is carried out for each row as, while applying an MPA, the processor 101 traverses the matrix before calculating messages from a function node corresponding the row being traversed.

[0044] According to an embodiment, to avoid excessive noise enhancement, and to take noise amplification into account, the projection is modified in accordance with a Minimum Mean Square Estimation (MMSE) solution as

P_Q ¹y = (i -Q(Q^/iQ + ^₀ ²i)-¹Q^/i )y

where σ₀ ² represents the noise variance.

[0045] Referring to Fig. 3, the bipartite factor graph 300 represents the indicator matrix of FIG. 2. Processor 101 may utilize a bipartite graphical representation of a chosen indicator matrix to apply a Message Passing Algorithm (MPA). In this graph 300, the upper (variable) nodes {¾ },k=l,...,K are connected to the K transmitted symbols denoted by xi, x₂....Xk; the lower (function) nodes {c_n }, n=l,...,N represent N REs carrying the encoded information and connected to the observations at these REs {y_n },n=l,...,N; the edges between variable and function nodes indicate which REs are occupied by a transmitter. There is an edge between pair (ci, Uj) if and only if the matrix element Fy is nonzero. Further each node c_n is connected to d_c nodes and each node Uk is connected to d_u nodes. An edge e_{n k} is the edge that connects a function node c_n to a variable node u_k. ^c_n^u_k and ^c_n→u_k are the messages sent along edge e_{n k} between variable node u_k and function node c_n respectively. The message ^c_n^u_k gives an updated inference of x_k based on the observation taken at symbols y_m, m 6 _fc\ ,where ^_k is the set of d_u function nodes connected to the variable node u_k defined by the k^th column of index matrix function node c_n being excluded from the set, and is calculated by the processor 101 according to the equation;

where · 6 Λ is the corresponding element of the constellation alphabet Λ.

Constellation alphabet Λ may be a set of all modulation symbols possible in the RSMA signal.

[0046] At the function node c_n, the inference of x_k is updated and is calculated by the processor 101 according to the equation:

c_n→u_k 'S

' ' j = l, ... , \A\ - l.

[0047] According to an embodiment, the complete message being sent from the function node c_n to the variable node u_k onto the edge e_{n k} is calculated by processor 101 according to the formula:

where

denotes message from variable node u_z to function node c_n

corresponding to vector and function max^* is defined as max^* (a, b) =

log(e^a + e^b) = max( , b) + log(l + e^_'^a_ft') . After the message coming to variable nodes u_k, k = Ι, .,. , Κ have converged or the number of maximum iteration has been met, all messages received from all connected edges are used to calculate the final estimated inference for symbol x_k. Before the processor 101 computes the above equations, it replaces y with y = y and matrix G by a skewed matrix G = P_QG to be able to cancel interference caused by transmitters corresponding to zeroes in a row.

[0048] According to an embodiment, the computation above may be completed for all components of vector y corresponding to non-zero elements of column k in indicator matrix F except the component with index n. The processor 101 implements this by replacing y with vector y^, where vector y^ comprises vector y with zeros at the same positions where zeros are located in column k in indicator matrix F, and one more zero is located in the component with index n and vector by matrix G^ . The processor then adds the corresponding LLRs according to the formula:

j = l, ... , \A\ - l,

[0049] According to an embodiment, the complexity needed to decode the incoming RSMA signal may be kept affordable and/or substantially manageable at high spectral efficiency. This may enable enhanced mobile broadband connectivity using RSMA.

[0050] According to an alternative embodiment, the device 100 may implement a second iterative process of detecting and decoding, for example, turbo equalization processing on the received RSMA signal wherein from second iteration onwards of the turbo equalization process, the processor 101 chooses a new indicator matrix 200. New indicator matrices may be obtained from the initial indicator matrix 200 for example by permuting the columns of indicator matrix 200 for every iteration of the turbo equalization process.

[0051 ] FIG. 4 illustrates a flow chart of a method of decoding RSMA signal comprising transmissions by multiple transmitters. The method may be, for example carried out at a base station. The method comprises steps 410, 411, 412 and 418.

[0052] Referring to FIG. 4, Step 410 includes receiving an RSMA signal.

The received RSMA signal may comprise transmissions by multiple transmitters on shared resource elements.

[0053] Step 411 includes choosing a sparse matrix as an indicator matrix

300 to be used to detect the received RSMA signal. The sparse matrix chosen is such that the number of rows is equal to the number of resource elements in the RSMA signal and the number of columns is equal to the number of transmitters transmitting on the RSMA signal. The sparse matrix is chosen such that the corresponding bipartite factor graph has a minimum cycle length of 4 or more and does not contain stopping sets.

[0054] Step 412 includes applying a Message Passing Algorithm on the bipartite factor graph 300 corresponding to the chosen indicator matrix 200. The message passing algorithm may include cancelling interference by transmitters corresponding to zeroes of the current row using projection based interference cancellation, before calculating messages at the function nodes.

[0055] Step 416 includes outputting the detected symbols as detected using the MPA.

[0056] According to an embodiment, Step 412 may further include 413, 414,

415, 416, and 417.

[0057] Step 413 may include traversing the indicator matrix 300 row by row and for each row cancelling interference by transmitters corresponding to zeroes in the current row. The interference may be cancelled using projection based interference cancellation as disclosed in embodiments of FIG. 2. [0058] Step 414 may include traversing the indicator matrix row by row and for each transmitter corresponding to a one in the current row calculating a message from the corresponding function node to variable node on the bipartite factor graph 300. This message may comprise a LLR and may be calculated as disclosed in embodiments of FIG. 2 and 3.

[0059] Step 415 may include traversing the indicator matrix column by column and for each transmitter corresponding to a one in the current column, calculating a message from a corresponding variable node to the function node. The message may comprise an LLR and may be calculated as described in embodiments of FIG. 2 and 3.

[0060] Step 416 may include calculating a weighted sum of the LLRs calculated in step 415.

[0061 ] Step 417 may include stopping the iterations of MPA. The iteration may be stopped when all the symbols are detected or when some other stopping conditions are met.

[0062] It should be noted that any indicator matrix, the values of elements therein, the bipartite factor graph corresponding to the indicator matrix are by way of example and by no means shall they be construed as a limitation. They are provided as an example to better illustrate the disclosed embodiments and not as the only possible implementations of the present embodiments.

[0063] The functionality described herein can be performed, at least in part, by one or more hardware logic components. According to an embodiment, the device 100 comprises a processor 101 configured by program code, when executed, to execute the embodiments of the operations and functionality described. Alternatively, or in addition, the functionality described herein can be performed, at least in part, by one or more hardware logic components. For example, and without limitation, illustrative types of hardware logic components that can be used include Field-programmable Gate Arrays (FPGAs), Program- specific Integrated Circuits (ASICs), Program- specific Standard Products (ASSPs), System-on-a-chip systems (SOCs), Complex Programmable Logic Devices (CPLDs), Graphics Processing Units (GPUs). [0064] Any range or device value given herein may be extended or altered without losing the effect sought. Also any embodiment may be combined with another embodiment unless explicitly disallowed.

[0065] Although the subject matter has been described in language specific to structural features and/or acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Rather, the specific features and acts described above are disclosed as examples of implementing the claims and other equivalent features and acts are intended to be within the scope of the claims.

[0066] It will be understood that the benefits and advantages described above may relate to one embodiment or may relate to several embodiments. The embodiments are not limited to those that solve any or all of the stated problems or those that have any or all of the stated benefits and advantages. It will further be understood that reference to 'an' item may refer to one or more of those items.

[0067] The steps of the methods described herein may be carried out in any suitable order, or simultaneously where appropriate. Additionally, individual blocks may be deleted from any of the methods without departing from the spirit and scope of the subject matter described herein. Aspects of any of the

embodiments described above may be combined with aspects of any of the other embodiments described to form further embodiments without losing the effect sought.

[0068] The term 'comprising' is used herein to mean including the method, blocks or elements identified, but that such blocks or elements do not comprise an exclusive list and a method or apparatus may contain additional blocks or elements.

[0069] It will be understood that the above description is given by way of example only and that various modifications may be made by those skilled in the art. The above specification, examples and data provide a complete description of the structure and use of exemplary embodiments. Although various embodiments have been described above with a certain degree of particularity, or with reference to one or more individual embodiments, those skilled in the art could make numerous alterations to the disclosed embodiments without departing from the spirit or scope of this specification.

Claims

1. A device (100) comprising a receiver (102) and a processor (101), wherein the receiver (102) is configured to receive a resource spread multiple access (RSMA) signal comprising transmissions by K transmitters over N resource elements; and the processor (101) is configured to:

choose an NxK sparse matrix (200) as an indicator matrix for the received RSMA signal, wherein each row corresponds to a resource element and each column corresponds to a transmitter and wherein the indicator matrix is representable by a bipartite factor graph (300) comprising K variable nodes and N function nodes;

use a Message Passing Algorithm (MPA) on the bipartite factor graph (300) to detect the symbols transmitted by each transmitter; and

output the detected symbols transmitted by each transmitter.

2. The device (100) of claim 1, wherein to perform the MPA, the processor (101) is configured to iteratively:

traverse the indicator matrix (200) row by row and for each row, cancel the interference caused by transmitters corresponding to zeros in the row using projection based interference cancellation;

traverse the indicator matrix (200) row by row and for each transmitter corresponding to a one in the current row, calculate a message from a function node to a variable node comprising a log likelihood ratio (LLR);

traverse the indicator matrix (200) column by column and for each one in the column, calculate a message from a variable node to a function node comprising an LLR;

calculate a weighted sum LLR for each transmitter based on the LLRs calculated; and

stop the iteration when LLR values converge.

3. The device (100) of claim 2, wherein while traversing the indicator matrix (200) row by row, calculating a message from function node to variable node for a transmitter corresponding to a one in the current row comprises

calculation of an LLR comprising a numerator and a denominator;

wherein the likelihood in the numerator of the LLR corresponds to the probability of the event that the symbol transmitted by the current transmitter takes a particular value from an alphabet, given received signal and symbols transmitted by remaining interfering transmitters;

wherein the interfering transmitters are indicated by an interference cancellation pattern represented by positions of ones in the current row of indicator matrix (200) other than the one for which the LLR is being calculated; and

wherein the likelihood in the denominator of the LLR corresponds to the probability of the event that the symbol transmitted by the current transmitter takes some fixed value from the alphabet given the same input as for the likelihood in the numerator.

4. The device (100) of claim 3, wherein while calculating the LLR for the

transmitter corresponding to a one in the current row, the processor (101) is further configured to:

calculate LLRs for the other rows containing one in the current column of indicator matrix (200) corresponding to the current transmitter using the interference cancellation pattern corresponding to the current row of indicator matrix (200);

perform a combining of the computed LLRs of the current column and assign it to the LLR value of the transmitter corresponding to a one in the current row.

5. The device (100) of any preceding claim, wherein calculating an LLR

comprising message from variable node to function node comprises combining LLRs calculated as messages from function nodes to variable nodes

corresponding to ones in the current column of indicator matrix (200); wherein the message from a function node to a variable node corresponding to the current row of indicator matrix is not combined except in the last iteration; and

wherein in the last iteration all LLRs indicated by the ones in the current column of indicator matrix are combined.

6. The device (100) of any preceding claim, wherein the processor is configured to stop iterating when:

LLRs comprised in messages from a variable node to a function node converge; or

when a maximum number of iterations is reached.

7. The device (100) of any preceding claim, wherein the processor (101) is

configured to choose the indicator matrix (200) such that the minimum cycle length of the corresponding bipartite factor graph (300) is greater than 4 or greater than 6.

8. The device (100) of any preceding claim, wherein the processor (101) is

configured to choose the indicator matrix (200) such that the indicator matrix (200) does not contain stopping sets.

9. The device (100) of any preceding claim, wherein the processor (101) is

configured to:

choose an indicator matrix (200) according to the measured received power of transmitters in such a way that in at least one row of the chosen indicator matrix, zeroes correspond to transmitters with received power lower than a threshold.

10. The device (100) of any preceding claim, wherein the projection based

interference cancellation comprises: projecting a vector corresponding to the received RSMA signal onto a vector space which is orthogonal to an interference vector subspace;

wherein the interference vector subspace comprises a vector space spanned by the columns corresponding to interference in a matrix representing a generalized communication channel, comprising multiplication of matrix of channel coefficients by matrix of transmitters' signatures, over which the RSMA signal is received; and

wherein the matrix representing the generalized communication channel is split into columns representing the transmitters corresponding to ones in the row of the indicator matrix (200) and columns representing the transmitters corresponding to zeroes in the row of indicator matrix (200), wherein the signals transmitted by transmitters corresponding to zeroes in the row comprise the interference to be cancelled.

11. The device (100) of any preceding claim, wherein the projection based

interference cancellation includes a noise correlation matrix in the projection.

12. The device of any preceding claim, wherein the processor (101) is further

configured to use a second iterative process to detect and decode the received RSMA signal and to change the indicator matrix (101) for every subsequent iteration of the second iterative process after the first iteration.

13. The device of any preceding claim wherein the processor (101) is further

configured to change the indicator matrix (200) for every subsequent iteration of the second iterative process by permuting the columns of an initially chosen indicator matrix (200).

14. A method comprising:

choosing an NxK sparse matrix as an indicator matrix (200) for a received RSMA signal, wherein each row corresponds to a resource element and each column corresponds to a transmitter and wherein the indicator matrix (200) is representable by a bipartite factor graph (300) comprising K variable nodes and N function nodes;

using a Message Passing Algorithm (MP A) on the bipartite factor graph (300) to detect the symbols transmitted by each transmitter; and

outputting the detected symbols transmitted by each transmitter.

15. The method of claim 14 wherein applying an MP A comprises:

traversing the indicator matrix (200) row by row and for each row cancel the interference caused by transmitters corresponding to zeros in the row using projection based interference cancellation;

traversing the indicator matrix (200) row by row and for each transmitter corresponding to a one in the current row calculating a log likelihood ratio (LLR);

traversing the indicator matrix (200) column by column and for each one in the column calculating a LLR; and

calculating a weighted sum LLR for each transmitter based on the LLRs calculated

iterating the method till values of LLR converge.

16. The method of claim 14 wherein while traversing the indicator matrix (200) row by row, calculating an LLR for a transmitter corresponding to a one in the current row comprises using an interference cancellation pattern; and calculating an LLR for the transmitter corresponding to a one in the current row comprises calculating an LLR is for each one in the current column using the same interference cancellation pattern.

17. A computer program comprising a program code configured to perform a method according to any of claims 14 to 16, when the computer program is executed on a computer.