CN106211235A - Parcel data transmission method in a kind of wireless network and system - Google Patents

Abstract

The invention provides the parcel data transmission method in a kind of wireless network and system.Method carries out length extension at transmitting terminal to parcel data, is then re-fed into channel and is transmitted；At receiving terminal, use normalized BOMP algorithm to recover parcel data.The present invention makes a number of transmitting terminal can share same running time-frequency resource to carry out the transmission of parcel data, allow active transmission terminal number can be more than the antenna number of receiving terminal, and receiving terminal is not required to know the definite information enlivening transmitting terminal, this just eliminates transfer resource in traditional method and distributes the heavy signaling consumption brought；Meanwhile, by using normalized BOMP algorithm, improve signal recovery performance.

Description

Method and system for transmitting packet data in wireless network

Technical Field

The invention belongs to the field of wireless communication, and particularly relates to a method and a system for transmitting packet data in a wireless network.

Background

With the rapid popularization of intelligent terminals such as smart phones, tablets, and the like, the number of users of the packet service is rapidly increasing. In response, packet data generated in communication networks is also growing rapidly. Typical packet services include instant messaging services such as Tencent QQ, social networking platforms such as Facebook, and the like. It can be said that a typical characteristic of service data generated by an intelligent terminal is that the data is a packet. The small packet service has greatly changed the way users (especially young user groups) live, learn, work, etc.

Packet services, as the name implies, mean that the data packets generated by the service are very short; basically, the length of the small data packet is shorter than the minimum specification of the communication device, i.e. shorter than the time-frequency resource block length. However, even if the packet length is short, overhead such as a preamble sequence included in the existing one-frame data cannot be small, and thus the utilization rate of the packet data frame is low; on the other hand, the huge user group of packet services and the irregularity of service requests cause frequent disconnection and reconnection between the mobile terminal and the network (e.g. the base station), which in turn causes heavy access signaling overhead. Overall, packet traffic imposes a heavy overhead burden on the network.

In the face of the new features of packet services, the traditional Medium Access Control (MAC) protocol is not satisfactory due to its low efficiency. In the above example, the conventional MAC protocol allocates orthogonal resource blocks to each access user to avoid mutual interference, i.e. orthogonal access. The orthogonal access can obtain a very good effect when the number of users needing service is small; however, in the face of a very large number of users, the amount of resources required for orthogonal access cannot be satisfied, and the cost is that resource scheduling causes a large service delay, thereby reducing quality of service (QoS). It is noted that the communication field covered by the 5G communication system is increasing, including cellular communication, machine to machine communication (M2M, machine to machine), and so on, and it can be said that the 5G communication embodies the concept of internet of things (IoT). In 5G communications, bursty traffic will account for a significant proportion, and another typical characteristic of these services is the short packet length. 5G communications need to meet requirements such as high spectral efficiency, high access and low latency services.

The problem that the existing orthogonal access technology cannot meet the future communication requirement is currently focused by the academic and industrial fields. Hot multiple access candidates in the 5G uplink are classified into an Orthogonal Multiple Access (OMA) mode and a non-orthogonal multiple access (NOMA) mode. The orthogonal access method mainly refers to Orthogonal Frequency Division Multiple Access (OFDMA) (non-orthogonal single carrier frequency division multiple access (SC-FDMA) is generally used for the uplink, because OFDMA brings a relatively high peak-to-average power ratio). Since the non-orthogonal multiple access method can improve the system throughput and accommodate more users, it is widely concerned and considered to be a promising access technology for 5G communication. Non-orthogonal access methods that have been proposed at present include NOMA [8] in the power domain, sparse code division multiple access (SCMA), multi-user shared access (MUSA), and the like. To summarize, these non-orthogonal techniques are not yet mature, and they all have their own problems that need to be overcome at present.

It is worth mentioning that 5G does not support only one access method. Since 5G has a concept of preliminary internet of things (IoT), communication scenarios covered by the IoT are many, some scenarios may satisfy the access amount requirement by an orthogonal multiple access method, and some scenarios may require a non-orthogonal multiple access method to accommodate more communication users.

In order to more effectively meet the transmission requirement of small packet data, it is very important to reduce the transmission overhead. One way to reduce the transmission overhead is to reduce the resource allocation signaling before sending small data packets. The other approach is that the receiving end automatically detects the user behavior, and jointly completes connection detection and symbol demodulation. The second approach is to complete the detection task belonging to the high layer according to the original protocol at the physical layer. To achieve the objective of the second approach, an effective method is to use a Compressed Sensing (CS) based multi-user detection technique.

Since the publication of a foundational article on compressed sensing, a great enthusiasm has arisen in the past few years for compressed sensing research. From a communication perspective, the compressed sensing model is highly consistent with the multi-user access model. Furthermore, the compressed sensing model represents a multi-user access model that is typically non-orthogonal. By applying the compressed sensing technique to communication, connection detection and symbol demodulation can be jointly performed at the physical layer, and the communication system is allowed to be an underdetermined linear system. Connection detection is performed by a high layer in the previous communication protocol, but the multi-user detection method based on compressed sensing can realize connection detection at a physical layer, which is a very interesting transition; it is not uncommon to implement symbol demodulation at the physical layer, such as a CDMA system is a typical example.

In fact, there has been considerable work on applying compressed sensing techniques to multi-user access detection. These efforts are essentially due to the fact that: sparsity is a natural property of the access channel, i.e. users that have data to transmit at the same time account for only a small fraction of the total number of users. These research works show that better recovery performance than single-user detection can be obtained with sparse recovery algorithms such as Orthogonal Matching Pursuit (OMP). Meanwhile, research results show that the multiple access technology based on compressed sensing can avoid the identity information overhead of users, and compared with some common access technologies, the compressed sensing-based multiple access technology can reduce the decoding delay (namely, the time span from the time when a user has a data sending request to the time when data is successfully decoded at a receiving end). Both the advantages come from the compressed sensing recovery method that the users do not need to know which information is sent by the users, and as long as the sending behaviors of the users are sparse, the decoding end can complete signal recovery meeting certain precision; meanwhile, as the resource allocation is not needed, the user can send data, and the decoding time delay is greatly reduced. Meanwhile, the non-orthogonal multi-user access method based on the compressed sensing does not need signaling to participate in the distribution and coordination process of transmission resources, and if the appointed access signaling refers to the expense of coordinating resource distribution when the user accesses, the method can realize the multi-address transmission of zero signaling.

Disclosure of Invention

Technical problem to be solved

The invention aims to provide a method and a system for transmitting packet data in a wireless network, which solve the problem of heavy signaling overhead caused by the transmission of a large amount of packet data to the wireless communication network.

(II) technical scheme

The invention provides a packet data transmission method in a wireless network, wherein the wireless network comprises a receiving end and a plurality of sending ends, and the receiving end is in communication connection with the sending ends, and the method comprises the following steps:

s1, at the sending end, expanding the length of the packet data to be sent;

s2, sending the small packet data with the expanded length from the sending end to the receiving end;

and S2, at the receiving end, carrying out sparse recovery on the small packet data after the length expansion.

Further, in step S1, for the packet data sent by the nth sending end, a corresponding precoding matrix P is used_nFor the symbol vector s in the small packet data_nPerforming length expansion to obtain length expanded packet data, signal x thereof_nExpressed as:

wherein,t represents that each time-frequency resource block can bearThe number of the carried message symbols, d, represents the length of the packet data after the length expansion.

Further, a precoding matrix P_nAnd the sending end is allocated to the sending end when the sending end is accessed to the receiving end through a wireless network.

Further, in step S2, the length-expanded packet data is sent to the receiving end via a channel, and the signal Y of the packet data acquired by the receiving end is represented as:

where ρ is₀Is the signal-to-noise ratio of the channel uplink,and the channel state vector corresponding to the nth sending end. Z is the noise matrix.

Further, in step S3, before recovering the packet data, the signal Y of the packet data is also vectorized to obtain a vectorized signal Y:

wherein B ═ B₁，B₂，…，B_N]，z represents a noise vector.

Further, in step S3, recovering the length-expanded packet data by using a normalized BOMP algorithm, specifically including:

s31, obtaining a vectorization signal y and corresponding parameters B and rho₀K and all channel response vectors, wherein K represents the iteration number of the BOMP algorithm;

s32, initializing iteration index k as 1, sender side set Λ₀Phi, residual signal r₀＝y；

S33, calculating r₀And the index in B belongs to the user set j ∈ {1, …, N } \ Λ_k-1Normalized coefficient c between sub-blocks of_j，k：

Represents a sub-matrix B_jConjugate transpose of (r)_k-1Representing the residual signal after the (k-1) th iteration, h_jRepresenting the channel response vector from the user j to the receiving end;

s34, finding out the sub-block lambda satisfying the condition_k：

S35, updating the sending end set:

Λ_k＝Λ_k-1∪{λ_k}；

and S36, updating the signal by adopting a least square method:

wherein s is₀Representing a symbol vector of all elements conforming to the constellation mapping,is represented by j ∈Λ_kSub-matrix B of_jA matrix of formations;

s37, updating the residual signal and the iteration index:

k＝k+1；

s38, ending and outputting result approaching signalI.e. a recovered signal is obtained.

The invention also provides a packet data transmission system in a wireless network, comprising:

the expansion module is used for carrying out length expansion on the packet data to be sent at the sending end;

the sending module is used for sending the small packet data with the expanded length from the sending end to the receiving end;

and the recovery module is used for performing sparse recovery on the small packet data after the length expansion at the receiving end.

Furthermore, aiming at the packet data sent by the nth sending end, the expansion module adopts a corresponding precoding matrix P_nFor the symbol vector s in the small packet data_nPerforming length expansion to obtain length expanded packet data, signal x thereof_nExpressed as:

wherein,t represents the number of the message symbols which can be carried by each time frequency resource blockAnd d represents the length of the packet data after the length expansion.

Further, a precoding matrix P_nAnd the sending end is distributed to the sending end when accessing to the receiving end through the wireless network.

Further, the sending module sends the packet data with the extended length to a receiving end through a channel, and a signal Y of the packet data obtained by the receiving end is represented as:

where ρ is₀Is the signal-to-noise ratio of the channel uplink,and the channel state vector corresponding to the nth sending end. Z is the noise matrix.

Further, before recovering the packet data, the recovery module performs vectorization on a signal Y of the packet data to obtain a vectorized signal Y:

wherein B ═ B₁，B₂，…，B_N]，z represents a noise vector.

Further, the recovery module recovers the packet data after the length expansion by using a normalized BOMP algorithm, which specifically includes:

s31, obtaining a vectorization signal y and corresponding parameters B and rho₀K and all channel response vectors, wherein K represents the iteration number of the BOMP algorithm;

s32, initializing iteration index k as 1, sender side set Λ₀Phi, residual signal r₀＝y；

S33, calculating r₀And the index in B belongs to the user set j ∈ {1, …, N } \ Λ_k-1Normalized coefficient c between sub-blocks of_j，k：

Representing the conjugate transpose of the sub-matrix Bj, r_k-1Representing a residual signal after the k-1 iteration, hj representing a channel response vector from a user j to a receiving end;

s34, finding out the sub-block lambda satisfying the condition_k：

S35, updating the sending end set:

Λ_k＝Λ_k-1∪{λ_k}；

and S36, updating the signal by adopting a least square method:

wherein s is₀Representing a symbol vector of all elements conforming to the constellation mapping,is represented by j ∈Λ_kSub-matrix B of_jA matrix of formations;

s37, updating the residual signal and the iteration index:

k＝k+1；

s38, ending and outputting result approaching signalI.e. a recovered signal is obtained.

(III) advantageous effects

In addition, the invention provides a system model based on structured Compressed Sensing (CS), according to the model, the receiving end can carry out sparse recovery, including carrying out user detection and symbol demodulation on a physical layer, and the method enables a certain number of packet service users to simultaneously share the same resource to carry out transmission of small data packets under a given multi-antenna system; furthermore, the receiving end does not need to know which users and how many users have transmitted the packet data. Therefore, the invention can avoid the huge access signaling overhead generated by resource scheduling in the service packet service of the traditional orthogonal access method, and the invention allows the number of the users to be transmitted simultaneously to be more than the number of the antennas at the receiving end. On the other hand, the invention proposes a normalized BOMP (normalized BOMP, NBOMP) algorithm for improving the signal recovery performance.

Drawings

Fig. 1 is a flowchart of a method for transmitting packet data in a wireless network according to the present invention.

Fig. 2 is a schematic diagram of a precoding process at a transmitting end according to the present invention.

FIG. 3 is a schematic block diagram of a BOMP algorithm in the prior art.

FIG. 4 is a functional block diagram of the normalized BOMP algorithm provided by the present invention.

Fig. 5A and 5B are examples of performance simulations of packet data transmission according to the present invention.

FIG. 6 is a flow chart of constructing a precoding matrix according to the present invention.

Fig. 7 is a diagram illustrating performance simulation of different precoding matrices for packet data transmission according to the present invention.

Detailed Description

The invention provides a method and a system for transmitting packet data in a wireless network. The precoding matrix constructed by the invention has the following advantages: a sufficient number of precoding matrices can be generated for a plurality of user connections; obtaining a better signal recovery accuracy than a randomly generated precoding matrix; the complexity of packet data transmission is reduced; so that the precoded symbols obtain statistically good PAPR characteristics.

Fig. 1 is a flowchart of a packet data transmission method in a wireless network provided by the present invention, as shown in fig. 1, the method includes:

s0, constructing a precoding matrix;

s1, at the transmitting end, using the constructed pre-coding matrix to expand the length of the packet data to be transmitted;

s2, sending the small packet data with the expanded length from the sending end to the receiving end;

s3, at the receiving end, the packet data after length expansion is recovered.

The invention firstly provides a scene of multiple access transmission of a large number of small data packets: assume that a wireless network has a Base Station (BS) as a receiving end, and the number of base station antennas is M. When a single antenna user (transmitting end) successfully accesses the network, the connection is established with the BS. Assume that there are N user connections (total connections) in the wireless network at time t, where N is_aThe individual connections have data transfer requirements (active connections). It should be noted that: (1) n is a radical of_aUnknown to the BS; (2) sparsity in transmission behaviour (access channel), i.e. N_aN (3) with no special convention_a≤M。

A complete transmission of a message comprises the transmission of a series of symbols. It is assumed that each time-frequency resource block can carry T message symbols (one time-frequency resource block has T cells), and all symbols are within the channel coherence time. It is assumed that the channel state information of all users is known at the BS.

Assume the number of valid information bits per packet is b_n，iNumber of bits of CRC field b_n，c，b_n＝b_n，i+b_n，cThe length of the data packet after channel coding, i.e. symbol mapping, is d. Assume the original complex-field data symbol vector produced by the nth connections_n，k(1. ltoreq. k. ltoreq. d) is s_nThe kth symbol of (1); the symbol mapping is implemented by precoding, and then the precoding matrix of the nth connection is assumed to beP_nColumn vector P corresponding to the kth column of (1)_n，kIs s_n，kThe precoding vector of (2). Connection n

Corresponding to the transmitted signal asAs shown in fig. 2. The precoding matrix is assigned and given when the user registers to access the network, and is effective until the user leaves the coverage area of the wireless network, here, it is assumed that the base station knows the precoding matrix corresponding to all connections, and the user only knows the precoding matrix of the user.

The signal received by the BS may be expressed as:

where ρ is₀Is the signal-to-noise ratio of the uplink,is the channel state vector corresponding to connection n. Z is the noise matrix.

Vectorizing the signals at the receiving end to finally obtain the following receiving signal models:

where Y ═ ec (Y) is the vector representation of the matrix Y,B＝[B₁，B₂，…，B_N]，where B is called a dictionary in sparse detection and consists of a sub-matrix of dictionaries for all users. Z ═ υ ec (Z) is a noise vector.

Considering that each connection transmission behavior in the packet service has obvious time domain sparsity, i.e.Is very highAnd if the probability is a zero vector, s has a block sparsity characteristic, and meanwhile, considering that B is a fat matrix, the formula (2) represents a typical compressed sensing model with block sparsity. It should be noted that formally, the model given by equation (2) can be regarded as a CDMA multiple access transmission, which is characterized in that:

1. for the kth symbol s connected to n_n，kBy precoding vectors P_n，kThe kth vector of (a) is extended into T units of the time-frequency resource block;

2. the measurement matrix B determines the multiple access transmission capability of the system, and is formed by a channel matrix h_n(N is more than or equal to 1 and less than or equal to N) and precoding matrix P_nAnd (N is more than or equal to 1 and less than or equal to N) is generated. The process realizes the cooperation of a space domain, a time domain and a frequency domain;

3. the multiple access approach does not require that the precoding vectors for all symbols be orthogonal vectors. In fact, since we are concerned with the scenario MT < Nd, equation (2) represents an underdetermined system, and the orthogonality between different connections is not satisfied.

The introduction of the precoding extension length method is also due to the fact that: the small data packet length d is shorter than the specification T of the network equipment, thereby bringing about a serious waste problem of precious resources. By precoding extension, a given transmission resource can be fully utilized. Compared with the existing multi-user detection model based on sparse recovery, the multi-access transmission model established based on the idea of pre-coding, expanding the whole small data packet and then transmitting has the following characteristics:

1. the combination of space domain, time domain and frequency domain multiple access transmission is realized, and the dimensionality of multiple access space is expanded;

2. different symbols of the same connection use different precoding vectors, so that the interference between symbols has different patterns. This operation averages the interference between symbols, which is advantageous for improving the accuracy of symbol demodulation.

Due to the established block sparse model, the existing block sparse recovery algorithm can be directly used for signal recovery. When each parameter (including the total connection number, the active connection number, the receiving end antenna number, the noise level and the like) meets a certain condition, the correct detection of the active connection can be realized. According to the foregoing description, we propose a multiple access technique that enables zero-signaling multiple access transmission. Zero signalling means that the connection only requires signalling participation at set-up and termination and no more signalling participation in the transmission during the lifetime of the connection. This completely avoids the heavy signaling problem caused by the transmission of small data packets according to the traditional access protocol.

The length expansion of the packet data is performed by the present invention, and a sparse model of block compressed sensing is obtained, and the recovery problem of the packet data in the present invention is described below.

There are many algorithms that are well established among CS algorithms. The Orthogonal Matching Pursuit (OMP) algorithm is fast in operation speed, and is convenient for practical application and widely applied. The BOMP algorithm is generated by applying the OMP algorithm to the block compression perception model, and the BOMP algorithm is consistent with the principle of the OMP algorithm. Specifically, in each iteration, it will select one of the remaining dictionary sub-arrays with the largest correlation with the residual signal (residual signal) (similar to the codeword correlation calculation in CDMA) to expand the basis functions, and then update the data with the selected basis functions by the Least Squares (LS) algorithm until the iteration is finished. The omp (bomp) algorithm embodies the idea of interference cancellation, but it is clearly different from interference cancellation in that the data of each of the previously selected possible non-zero positions is updated in each subsequent iteration. Fig. 5 is a schematic block diagram of a BOMP algorithm in the prior art, as shown in fig. 5, the calculation process is as follows:

1: input matrix B, signal vector y, parameter ρ₀All channel response vectors, and the number of iterations K of the BOMP algorithm, are, in general,

2, initializing, namely, setting the iteration index k to be 1, and collecting Λ users₀Phi, residual signal r₀＝y

3: iteration step: when K is less than or equal to K

1): calculating r₀And the index in B belongs to the user set j ∈ {1, …, N } \ Λ_k-1The correlation coefficient between the sub-blocks of (a):

2): finding sub-blocks that satisfy the condition:

3): updating the user set:

Λ_k＝Λ_k-1∪{λ_k}

4): update the signal with a Least Squares (LS) algorithm:

5): update residual signal and iteration index:

k＝k+1

4: and finishing and outputting the result: approximation signal of block-spark signal

Through calculation analysis of the correlation coefficient in BOMP operation, the result shows that when j represents active connection, the correlation coefficient c_j，kExpected value of andthe second moment of (a) is related to the first moment; when j is not an active connection, the correlation coefficient c_j，kExpected value of andis related to the first moment of (a). Thus, the channel gains of different connections cause an increase in uncertainty in the magnitude of the correlation coefficient. To eliminate this effect, the invention usesFor correlation coefficient c_j，kNormalization is carried out to obtain normalized correlation coefficientThen BOMP is required to use the normalized correlation coefficient for connection detection. Experiments show that the connection detection performance can be greatly improved after the correlation coefficient is normalized, so that the signal demodulation performance is also improved. The invention refers to such a recovery algorithm as normalized BOMP (normalized BOMP, NBOMP) algorithm, and its functional block diagram is shown in fig. 6, and the specific calculation process is as follows:

1: inputting: matrix B, signal vector y, parameter ρ₀All channel response vectors, and the number of iterations K of the BOMP algorithm, are, in general,

2, initializing, namely, setting the iteration index k to be 1, and collecting Λ users₀Phi, residual signal r₀＝y

3: iteration step: when t is less than or equal to K

1): calculating r₀And B middle cordThe reference belongs to the user set j ∈ {1, …, N } \ Λ_k-1Normalized coefficient between subblocks:

2): finding sub-blocks that satisfy the condition:

3): updating the user set:

Λ_k＝Λ_k-1∪{λ_k}

4): update the signal with a Least Squares (LS) algorithm:

5): update residual signal and iteration index:

k＝k+1

4: and finishing and outputting the result: approximation signal of block-spark signal

Fig. 5A and 5B are exemplary embodiments of performance of packet data transmission according to the present invention, where fig. 5A is a User Detection Success Rate (UDSR) under different number of transmission users, and fig. 5B is a Symbol Error Rate (SER) for packet data recovery under different number of transmission usersrate) performance, wherein the abscissa is E_s/N₀(dB)，E_sAverage symbol energy, N, for transmission by the transmitting end₀For noise power spectral density, the other parameters are (M, N, d, T, K) — (8, 80, 100, 500, 30), it should be noted that the precoding matrix used in the simulation of fig. 5A and 5B is randomly generated, as shown in fig. 5A and 5B, the NBOMP algorithm is much better than the BOMP algorithm in terms of UDSR or SER performance index.

It should be noted that the precoding matrix plays a crucial role in packet data transmission. Firstly, the precoding matrix solves the problem that the length of a small data packet is smaller than the granularity of system transmission resources, so that the system resources are fully utilized; secondly, the invention establishes a packet data transmission model with sparse blocks according to the pre-coding matrix, so that the recovery of packet data becomes possible even if the number of users is more than that of base station antennas; thirdly, compared with a compressed sensing model with no structure available, the compressed sensing model with a block structure enables better data recovery performance, and meanwhile enables the data recovery (namely the proposal of the ICBOMP algorithm) to be substituted by error correction and detection codes commonly used in communication. Besides, the precoding matrix also determines the interference characteristics among users, which directly affects the user detection and symbol demodulation; meanwhile, the precoding matrix also directly determines the complexity of the implementation of the invention, including the computational complexity (the precoding computational complexity of the transmitting end and the sparse recovery complexity of the receiving end), the precoding distribution complexity, the system storage complexity and the like. The precoding matrix adopted by the invention can be obtained by randomly generating a matrix and then performing energy normalization, and the precoding matrix is enough to indicate the effectiveness of the small data packet transmission method provided by the invention. On the other hand, the recovery performance brought by the randomly generated precoding matrix is not optimal in fact, or a plurality of precoding matrices can bring better sparse recovery accuracy; meanwhile, since the randomly generated precoding matrix is generally a dense matrix, the precoding complexity of the transmitting end and the signal recovery complexity of the receiving end in the transmission method are relatively large.

Therefore, in order to achieve a better effect of packet data transmission, the present invention further provides a precoding matrix, which is used for packet data transmission when a user accesses a wireless network in a non-orthogonal manner, and a construction method of the precoding matrix is shown in fig. 6:

step 1, a Hadamard matrix with smaller dimension is inserted into 0 to expand the dimension of the matrix, and the column orthogonality of the expanded matrix is kept. The method mainly aims to construct a matrix with orthogonal columns and sparse non-zero elements and consistent non-zero element amplitude, and the quantity of constructed matrices is large. These matrices are constructed from Hadamard matrices. The specific process is as follows:

constructing a matrix with sparse elements by using a Hadamard matrix, and assuming that a precoding matrix of T × T is to be constructed, the existing Hadamard matrix O of T × T has the element O in the ith row and the jth column_i，jTaking T-2 and T-4 as an example, the following element-sparse matrix Q is a matrix of orthogonal columns satisfying column sparsity β -50%, and the non-zero element amplitudes are all 1.

In fact, when the dimension parameters T and T are larger, a more column sparse satisfactory matrix can be constructed.

And 2, generating a plurality of precoding matrix clusters by the matrix with sparse elements. The step carries out random row-column exchange on the above Q matrix, and simultaneously, the element symbols of each row and each column can be randomly changed. I.e. the matrix after transformation of QWherein R is₁And R₂Each row and column of (a) has one and only one non-zero element (1 or-1). Random selectionAnd normalizing the energy of each column to generate an optional precoding matrix, with different R₁And R₂A very large number of precoding matrix clusters can be generated.

And 3, calculating the correlation among different new matrixes, and selecting N matrixes with smaller correlation as precoding matrixes. The precoding matrix constructed in step 2 can be many, and the correlation (inter-block correlation) between the precoding matrices is not large as long as the generation of the precoding matrix is sufficiently random. Since enough precoding matrixes can be selected, N matrixes with smaller correlation are selected as the precoding matrixes for the data transmission of the small packets. The inter-block correlation metric for each matrix isWhereinIs composed ofThe k-th singular value of (a). The algorithm of the step is relatively simple, namely the correlation measurement omega between all matrixes in the generated matrix cluster is measured_i，jOrdering from big to small, some ω that is the largest_i，jAnd discarding the corresponding matrixes until the remaining N matrixes are used as the precoding matrixes in the non-orthogonal access method. Column orthogonality is beneficial for signal recovery, and if the correlation of the channel is not considered, the small inter-block correlation indicates that the interference between users is small, which is beneficial for user detection and symbol demodulation. Meanwhile, the precoding matrix with sparse non-zero elements and consistent non-zero element amplitude can reduce the precoding complexity of the transmitting end, because of the precoding matrixThe precoding matrix does not require the sending end to carry out multiplication operation; in addition, when the correlation check user is carried out, a lot of multiplication operations can be avoided. Likewise, storage complexity may also be reduced. Finally, it can be statistically demonstrated that the symbols after such precoding can obtain good PAPR characteristics. If the number of the non-zero elements of each row of the precoding matrix is consistent, the PAPRs of all precoded symbols accord with the same statistical rule.

And 4, verifying the accurate performance of the precoding matrix selected in the step 3. And if the performance of the precoding matrix is not ideal enough, returning to the step 2 to reconstruct.

Fig. 7 is a performance simulation diagram of different precoding matrices for packet data transmission in the present invention, where fer (frame error rate) is a frame error rate, and parameters are selected as follows: QPSK symbol modulation, CRC check of 24bits, convolutional code of 1/2, and the other parameters are (M, N, d, T, K) ═ 8, 1280, 200, 1040, 30. In the simulation result notation, the precoding matrix corresponding to the "random generation" is generated by normalizing the random matrix, "H-sparsity 51/52" represents the precoding matrix constructed by the 20 × 20 Hadamard matrix according to the method of the present invention, and "H-sparsity 19/20" represents the precoding matrix constructed by the 52 × 52 Hadamard matrix according to the method of the present invention. As can be seen from fig. 4, the precoding matrix constructed by the method of the present invention may bring a gain of 1dB in terms of FER performance recovered from the packet relative to the randomly generated precoding matrix.

The precoding matrix constructed according to the above method may bring the following advantages:

1. a very large number of precoding matrices can be generated to serve a large number of users;

2. better signal recovery accuracy than a randomly generated precoding matrix can be obtained;

3. the complexity of realizing the method can be reduced, including the complexity of signal recovery, the complexity of precoding calculation, the complexity of precoding matrix allocation and storage, and the like;

4. the precoded signal can obtain a statistically good peak to average power ratio (papr) characteristic.

The invention can be applied to the scene that a large amount of packet data are sent, for example, a mobile phone sends a small data packet in mobile communication, and the base station receiving is the most typical application. And for example, data in meter reading (meter reading) is uploaded to a data collection node, machine to machine (machine to machine) communication, and the like. For the transmission of long packet data, the contention-free method proposed in the present invention can be used for reference, in addition to the conventional access method. Specifically, the long packet data may be divided into several small packets, and the small packets each have a complete communication signal structure. Thus, the invention can be uniformly used to transmit the data including the original small data packet or the small data packet which is divided into small data packets.

The above-mentioned embodiments are intended to illustrate the objects, technical solutions and advantages of the present invention in further detail, and it should be understood that the above-mentioned embodiments are only exemplary embodiments of the present invention, and are not intended to limit the present invention, and any modifications, equivalents, improvements and the like made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims (12)

1. A method for transmitting packet data in a wireless network, wherein the wireless network comprises a receiving end and a plurality of transmitting ends, and the receiving end is in communication connection with the transmitting ends, the method comprises the following steps:

s1, at the sending end, performing length expansion on the packet data to be sent;

s2, sending the small packet data with the expanded length from the sending end to the receiving end;

and S2, at the receiving end, performing sparse recovery on the small packet data after the length expansion.

2. The method for transmitting packet data in a wireless network according to claim 1, wherein in step S1, for the packet data sent by the nth sending end, a corresponding precoding matrix P is used_nFor the symbol vector s in the small packet data_nPerforming length expansion to obtain length expanded packet data, signal x thereof_nExpressed as:

wherein,t represents the number of message symbols which can be carried by each time-frequency resource block, and d represents the length of the packet data after the length expansion.

3. Method for packet data transmission in a wireless network according to claim 2, characterized in that the precoding matrix P_nAnd the sending end is allocated to the sending end when the sending end is accessed to the receiving end through a wireless network.

4. The method according to claim 2, wherein in step S2, the length-expanded packet data is sent to the receiving end via a channel, and a signal Y of the packet data acquired by the receiving end is represented as:

$Y=\sqrt{{\mathrm{\&rho;}}_0}{\&Sigma;}_{n=1}^N{h}_n{x}_n^T+Z,$

where ρ is₀Is the signal-to-noise ratio of the channel uplink,and the N is a complex field channel state vector corresponding to the nth sending end, and the N represents the total number of the sending ends. Z is the noise matrix.

5. The method for transmitting packet data in a wireless network according to claim 4, wherein in step S3, before recovering the packet data, the signal Y of the packet data is also vectorized to obtain a vectorized signal Y:

$y=\sqrt{{\mathrm{\&rho;}}_0}Bs+z,$

wherein B ═ B₁，B₂，…，B_N]，z represents a noise vector.

6. The method for transmitting packet data in a wireless network according to claim 5, wherein in step S3, the recovering the packet data after length expansion by using a normalized BOMP algorithm specifically includes:

s31, obtaining a vectorization signal y and corresponding parameters B and rho₀K and all channel response vectors, wherein K represents the iteration number of the BOMP algorithm;

s32, initializing iteration index k as 1, sender side set Λ₀Phi, residual signal r₀＝y；

S33, calculating r₀And B middle cordThe reference belongs to the user set j ∈ {1, …, N } \ Λ_k-1Normalized coefficient c between sub-blocks of_j，k：

$c_{j,k}=\left|\right|B_j^H{r}_{k-1}|{|}_2^2/|\left|h_j\right|{|}_2^2,$

Represents a sub-matrix B_jConjugate transpose of (r)_k-1Representing the residual signal after the (k-1) th iteration, h_jRepresenting the channel response vector from the user j to the receiving end;

s34, finding out the sub-block lambda satisfying the condition_k：

${\mathrm{\&lambda;}}_k=\arg \underset{j\&Element;\{1,...,N\}\backslash {\mathrm{\&Lambda;}}_{k-1}}{max}{C}_{j,k};$

S35, updating the sending end set:

Λ_k＝Λ_k-1∪{λ_k}；

and S36, updating the signal by adopting a least square method:

${\stackrel{\&OverBar;}{s}}_k={\mathrm{argmax}}_{s_0}\left|\right|y-\sqrt{{\mathrm{\&rho;}}_0}{B}_{{\mathrm{\&Lambda;}}_k}{s}_0|{|}_2,$

wherein s is₀Representing a symbol vector of all elements conforming to the constellation mapping,is represented by j ∈Λ_kSub-matrix B of_jA matrix of formations;

s37, updating the residual signal and the iteration index:

$r_k=y-\sqrt{{\mathrm{\&rho;}}_0}{B}_{{\mathrm{\&Lambda;}}_k}{\stackrel{\&OverBar;}{s}}_k$

k＝k+1；

s38, ending and outputting result approaching signalI.e. a recovered signal is obtained.

7. A packet data transmission system in a wireless network, the wireless network comprises a receiving end and a plurality of sending ends, the receiving end is connected with the sending ends in a communication way, and the system is characterized by comprising:

the expansion module is used for expanding the length of the packet data to be sent at the sending end;

the sending module is used for sending the small packet data with the expanded length from the sending end to the receiving end;

and the recovery module is used for performing sparse recovery on the small packet data after the length expansion at the receiving end.

8. The system of claim 7, wherein the spreading module uses a precoding matrix P corresponding to the packet data transmitted from the n-th transmitting end_nFor the symbol vector s in the small packet data_nPerforming length expansion to obtain length expanded packet data, signal x thereof_nExpressed as:

wherein,t represents the number of message symbols which can be carried by each time-frequency resource block, d represents the length of the packet data after the length expansion, and N represents the total number of the sending ends.

9. Root of herbaceous plantSystem for small packet data transmission in a wireless network according to claim 8, characterized in that said precoding matrix P_nAnd the sending end is allocated to the sending end when the sending end is accessed to the receiving end through a wireless network.

10. The system according to claim 8, wherein the sending module sends the length-extended packet data to the receiving end via a channel, and a signal Y of the packet data obtained by the receiving end is represented as:

$Y=\sqrt{{\mathrm{\&rho;}}_0}{\&Sigma;}_{n=1}^N{h}_n{x}_n^T+Z,$

where ρ is₀Is the signal-to-noise ratio of the channel uplink,and the channel state vector corresponding to the nth sending end. Z is the noise matrix.

11. The system of claim 10, wherein the recovery module further performs vectorization on the signal Y of the packet data before recovering the packet data to obtain a vectorized signal Y:

$y=\sqrt{{\mathrm{\&rho;}}_0}Bs+z,$

wherein B ═ B₁，B₂，…，B_N]，z represents a noise vector.

12. The system for packet data transmission in a wireless network according to claim 11, wherein the recovery module recovers the packet data after length expansion by using a normalized BOMP algorithm, specifically comprising:

s31, obtaining a vectorization signal y and corresponding parameters B and rho₀K and all channel response vectors, wherein K represents the iteration number of the BOMP algorithm;

s32, initializing iteration index k as 1, sender side set Λ₀Phi, residual signal r₀＝y；

S33, calculating r₀And the index in B belongs to the user set j ∈ {1, …, N } \ Λ_k-1Normalized coefficient c between sub-blocks of_j，k：

$c_{j,k}=\left|\right|B_j^H{r}_{k-1}|{|}_2^2/|\left|h_j\right|{|}_2^2,$

Represents a sub-matrix B_jConjugate transpose of (r)_k-1Representing the residual signal after the (k-1) th iteration, h_jRepresenting the channel response vector from the user j to the receiving end;

s34, finding out the sub-block lambda satisfying the condition_k：

${\mathrm{\&lambda;}}_k=\arg \underset{j\&Element;\{1,...,N\}\backslash {\mathrm{\&Lambda;}}_{k-1}}{max}{c}_{j,k};$

S35, updating the sending end set:

Λ_k＝Λ_k-1∪{λ_k}；

and S36, updating the signal by adopting a least square method:

${\stackrel{\&OverBar;}{s}}_k={\mathrm{argmax}}_{s_0}\left|\right|y-\sqrt{{\mathrm{\&rho;}}_0}{B}_{{\mathrm{\&Lambda;}}_k}{s}_0|{|}_2,$

wherein s is₀Representing a symbol vector of all elements conforming to the constellation mapping,is represented by j ∈Λ_kSub-matrix B of_jA matrix of formations;

s37, updating the residual signal and the iteration index:

$r_k=y-\sqrt{{\mathrm{\&rho;}}_0}{B}_{{\mathrm{\&Lambda;}}_k}{\stackrel{\&OverBar;}{s}}_k$

k＝k+1；

s38, ending and outputting result approaching signalI.e. a recovered signal is obtained.