Active user, time delay and channel joint estimation method for low-precision receiver

Abstract

The invention belongs to the technical field of information and communication, and relates to an active user, time delay and channel joint estimation method for a low-precision receiver. The problem that when a receiver adopts a one-bit analog-to-digital converter, transmission delay and channel state estimation precision are too low due to high quantization errors is solved. Performing correlation peak detection on the received signal and the known leader sequence, and preliminarily estimating active users and transmission time delay; the nonlinear receiving model is equivalent to a linear model based on the Bussgang decomposition principle; an expectation maximization algorithm framework is adopted, related parameters are jointly estimated through inner and outer layer iteration, an inner layer constructs a factor graph model based on a Markov chain to update an equivalent channel and a user state, and an outer layer adopts a local search algorithm to update a transmission delay estimation value. The method is used for a 1-bit low-precision receiver in an asynchronous large-scale machine type communication system, and realizes joint acquisition of active user detection, transmission delay estimation and channel state information.

Description

Technical Field

[0001] This invention belongs to the field of information and communication technology, and relates to a joint estimation method for active users, delay and channel for low-precision receivers. Background Technology

[0002] Massive Machine-Type Communication (mMTC) has become a key technology for various emerging intelligent services, with applications spanning industrial automation, public safety, the Internet of Things, healthcare, transportation, and remote manufacturing. Unlike traditional human-type communication, mMTC possesses three unique service characteristics: (i) uplink transmission primarily uses very short data packets; (ii) it employs unlicensed data transmission, where active devices send frames without prior scheduling, thus improving transmission efficiency; and (iii) the probability of each machine being active at the same time is very low, exhibiting sparsity. In the paper L. Liu and W. Yu, “Massive connectivity with massive mimo—part i: Device activity detection and channel estimation,” IEEE Trans. Signal Process., vol. 66, no.11, pp. 2933–2946, 2018, the authors proposed an active device detection and channel estimation algorithm based on an approximate message passing algorithm for transceiver synchronization scenarios. In the paper W. Zhu, M. Tao, X. Yuan, and Y. Guan, “Deep-learned approximate message passing for asynchronous massive connectivity,” IEEE Trans. Wireless Commun., vol. 20, no. 8, pp. 5434–5448, 2021, the authors proposed a joint active device detection, time delay, and channel estimation method based on a deep learning approximate message passing algorithm for asynchronous transceiver scenarios.

[0003] In mMTC, receivers typically employ massive MIMO (Multi-Input Multiple-Output) technology, which can improve spectral efficiency without requiring additional spectrum resources. However, massive MIMO increases receiver complexity; with a significant increase in the number of antennas, both hardware costs and operating power consumption rise substantially. In the receiver's RF link, the power consumption generated by the analog-to-digital converter (ADC) dominates the total power consumption. In RH Walden's paper, "Analog-to-digital converter survey and analysis," IEEE J. Sel. AreasCommun., vol. 17, no. 4, pp. 539–550, Apr. 1999, the authors demonstrate that sampling rate and quantization bit depth are two key factors affecting ADC power consumption, with the quantization bit depth having a more significant impact. Therefore, using a coarse-quantized (1-4 bit) ADC can significantly reduce receiver power consumption and is more suitable for massive MIMO systems. This patent considers a more extreme case of a 1-bit ADC, where the in-phase and quadrature components of the received sample are quantized as positive and negative one, respectively. This scheme shows significant appeal in large-scale MIMO systems. Specifically, each RF link contains only a simple comparator, eliminating the need for automatic gain control mechanisms. This advantage can significantly reduce the hardware cost and power consumption of the base station. Summary of the Invention

[0004] The purpose of this invention is to solve the technical problem that, in asynchronous massive machine-type communication (mMTC) scenarios, when the receiver uses a 1-bit low-precision analog-to-digital converter (ADC), it is difficult to simultaneously and accurately complete the joint estimation of active user detection, transmission delay estimation, and channel state due to transmission delay uncertainty and severe quantization distortion.

[0005] To achieve the above objectives, the present invention employs the following technical means:

[0006] This invention provides a joint estimation method for active users, delay, and channel for a low-precision receiver employing a 1-bit analog-to-digital converter. The method includes:

[0007] The quantized signal received by the receiving antenna is compared with the known preamble sequence in the known preamble sequence pool to detect the correlation peak, and the number of active users and their transmission delay within the observation window are initially estimated.

[0008] Based on the Bussgang decomposition principle, the original nonlinear signal receiving model is equivalent to a linear signal receiving model containing an equivalent channel matrix and equivalent noise.

[0009] Employing an expectation-maximization algorithm framework, the active user state, transmission delay, and channel state are jointly estimated through an inner and outer two-layer iterative process, wherein:

[0010] During the inner iteration process, a factor graph model is constructed based on the Markov chain. Through message passing between variable nodes and factor nodes, the estimated value of the equivalent channel matrix and the user activity status are updated.

[0011] During the outer iteration process, the estimated transmission delay is updated based on the channel estimation results obtained from the inner iteration using a local search algorithm.

[0012] When the preset iteration termination condition is met, the final active user detection result, transmission delay estimate, and channel state information are output.

[0013] In the above scheme, the 1-bit analog-to-digital converter (ADC) quantizes the in-phase component and quadrature component of the received signal into +1 or -1, respectively.

[0014] In the above scheme, the step of initially estimating the number of active users and their transmission delay within the observation window includes: at the receiving end, converting the received signal matrix... Correlation peak detection is performed with all leader sequences in the leader sequence pool to preliminarily estimate the number of users included in the observation window. and its delay And construct the leader sequence matrix ;

[0015] in, express A matrix composed of quantized signals received by each receiving antenna. The number of receiving antennas, The width of the observation window, This is the oversampling factor. This indicates that the k-th user has a delay. The preamble sequence signal at that location.

[0016] In the above scheme, the step of converting the nonlinear signal model into a linear signal model based on the Bussgang decomposition principle includes:

[0017] The quantized received signal matrix is ​​decomposed using Bussgang decomposition. Equivalent to linear form:

[0018]

[0019] in, The Bussgang decomposition coefficient matrix, , for The autocorrelation matrix, Take the diagonal matrix; This is the equivalent noise matrix; For pulse shaping filters and matched filter convolution The constructed Toplitz matrix has dimensions of ; This is the transmission delay vector; for The Toplitz matrix formed by this structure; This is the noise matrix; For delay The preceding sequence matrix; This is the equivalent channel matrix.

[0020] In the above scheme, the step of employing the expectation-maximization algorithm framework to jointly estimate the active user state, transmission delay, and channel state through inner and outer iterative processes includes the following inner iterative process:

[0021] Constructing a factor graph model based on Markov chains, defining... , ;

[0022] Based on the sum-product algorithm, in the... During each iteration, from the factor node Passed to variable node The message approximates a Gaussian distribution. ;

[0023] From variable node Passed down to factor nodes The message approximates a Gaussian distribution. ;

[0024] From variable node Propagated upwards to factor nodes The message approximates a Gaussian distribution. ;

[0025] From factor nodes Passed to variable node The message approximates a Gaussian distribution. ;

[0026] Calculate residuals and inverse residual variance ;

[0027] From factor nodes Passed to variable node The message approximates a Gaussian distribution. ;

[0028] renew In the posterior mean during each iteration and variance ;

[0029] Based on a set threshold Determine user activity level: If the matrix The power of the k-th row is greater than the set threshold. If the k-th user is active, then the k-th user is active.

[0030] Determine the termination condition for the inner iteration: If Less than the set inner iteration threshold or If the maximum number of inner iterations is exceeded, the inner iteration ends; otherwise, = +1, continue inner layer iteration.

[0031] In the above scheme, the outer iteration process includes:

[0032] Based on the channel estimation results obtained from the inner iteration, update the delay. ,in, Let it be a small positive real number to reduce the search complexity, such that

[0033]

[0034] maximum;

[0035] in, for The autocorrelation matrix, and They are respectively and The List, and They are respectively and The autocorrelation matrix;

[0036] Determine the termination condition of the outer iteration: If If the maximum number of outer iterations is exceeded, the outer iteration ends; otherwise, , Return to execute the inner iteration process.

[0037] In the above scheme, Gaussian distribution Their mean and variance are as follows:

[0038]

[0039] ,

[0040] in, for The To the All elements in the row;

[0041] Gaussian distribution Their mean and variance are respectively

[0042]

[0043] ,

[0044] in, and From the factor nodes respectively Passed up to the variable node The mean and variance are calculated as follows:

[0045]

[0046] ,

[0047] in, , Take the block diagonal matrix.

[0048] Gaussian distribution The calculation formula is:

[0049] ;

[0050] Gaussian distribution Their mean and variance are as follows:

[0051]

[0052] ,

[0053] in, for The element in the k-th column; for The element in the m-th row.

[0054] In the above scheme, the residual and inverse residual variance Calculated separately as follows:

[0055]

[0056] ,

[0057] in, and They are respectively and The (m) ) elements; and They are respectively and The (m) ) elements;

[0058] ,

[0059] in, Channel matrix elements The large-scale fading average, and:

[0060] .

[0061] Because the present invention employs the above-mentioned technical means, it has the following beneficial effects:

[0062] 1. This invention solves the serious nonlinear distortion problem introduced by 1-bit (ADC) quantization by transforming the original nonlinear signal receiving model into a linear signal receiving model containing an equivalent channel matrix and equivalent noise based on the Bussgang decomposition principle and by using oversampling and other techniques. It achieves the effect of transforming the complex nonlinear estimation problem into a relatively simple linear estimation problem, laying a theoretical foundation for subsequent parameter estimation.

[0063] 2. This invention solves the problem of high complexity in ordinary algorithms by constructing a factor graph model based on Markov chains in the inner iteration of the expectation-maximization (EM) algorithm framework and using the technique of updating the estimated value of the equivalent channel matrix and the user activity status by message passing between variable nodes and factor nodes. It achieves the effect of maintaining high channel estimation accuracy and active user detection accuracy under a low algorithm complexity environment.

[0064] 3. This invention solves the problem of performance degradation caused by transmission delay uncertainty in asynchronous communication scenarios by using a local search algorithm to update the estimated value of transmission delay based on the channel estimation result obtained from the inner iteration in the outer iteration of the expectation maximization (EM) algorithm framework. It achieves the effect of accurately capturing user transmission delay under low-precision reception conditions and overcomes the defect of delay estimation failure in traditional methods under 1-bit quantization.

[0065] 4. This invention achieves a significant synergistic effect by organically combining Bussgang decomposition linearization, an inner message passing algorithm based on Markov chains, and an outer delay calibration algorithm based on local search. Bussgang decomposition provides a linearized basic model for the message passing algorithm, the accurate channel estimation obtained from the inner iteration provides a reliable basis for the outer delay calibration, and the accurate delay estimation, in turn, optimizes the accuracy of the inner channel estimation. This mutually reinforcing mechanism of inner and outer iterations effectively solves the complex problem of joint estimation of active user detection, delay estimation, and channel state in asynchronous large-scale machine-type communication under extreme conditions of 1-bit quantization. It achieves the green communication goal of reducing receiver hardware complexity and power consumption without sacrificing system estimation performance. Attached Figure Description

[0066] Figure 1 System model.

[0067] Figure 2 Example of a factor plot, where, , .

[0068] Figure 3 Simulation results of channel normalized mean square error (NMSE) under different signal-to-noise ratio (SNR) conditions.

[0069] Figure 4 False detection probability under different signal-to-noise ratio (SNR) conditions Simulation results diagram. Detailed Implementation

[0070] The embodiments of the present invention will be described in detail below. Although the present invention will be described and illustrated in conjunction with some specific embodiments, it should be noted that the present invention is not limited to these embodiments. On the contrary, any modifications or equivalent substitutions made to the present invention should be covered within the scope of the claims of the present invention.

[0071] Furthermore, to better illustrate the present invention, numerous specific details are set forth in the following detailed embodiments. Those skilled in the art will understand that the present invention can be practiced without these specific details.

[0072] This invention proposes a joint estimation method for active users, delay, and channel for low-precision receivers, aiming to achieve joint detection and estimation of active users, transmission delay, and channel for low-precision receivers in asynchronous transmission scenarios at the transceiver end.

[0073] System model such as Figure 1 As shown, assume the receiver has One receiving antenna, simultaneously serving multiple single-antenna users, among which For pulse shaping filters, For matched filters, This is the oversampling factor. This is a 1-bit quantization. During uplink access, each active user randomly selects a preamble sequence from the preamble sequence pool and sends it to the receiver. Therefore, the 1st... The quantized signal received by each receiving antenna is:

[0074]

[0075] in, For active users To the Channel gain of each receiving antenna; for and Convolution; For the first Preamble sequence signals sent by an active user; For the first Transmission latency from one active user to the receiving end; This is environmental noise.

[0076] Considering the sliding listening window technique, then After sampling within any observation window, the following is obtained:

[0077]

[0078]

[0079] in, The estimated total number of users within the observation window; This is a user activity indicator coefficient, which is 1 when users are active and 0 otherwise. The width of the observation window; ; ; ; Let be the duration of a symbol. The matrix form of the above equation can be written as:

[0080]

[0081] in, ; ; ; The Toplitz matrix has the following structure:

[0082] ;

[0083] matrix It is a by The structure of the Topulitz matrix is ​​similar to... same.

[0084] Consider all The quantized signal obtained from the receiving antennas is:

[0085]

[0086] in, ; ; This is the equivalent channel matrix, where non-zero rows represent active users; As can be seen from Bussgang's decomposition, This can be further equivalent to the following linear form:

[0087]

[0088] in, , for The autocorrelation matrix, Take the diagonal matrix; For the equivalent noise matrix, For pulse shaping filters and matched filter convolution The constructed Toplitz matrix has dimensions of , For transmission delay vector, for The constructed Toplitz matrix, For the noise matrix, Let be the matrix of the preceding sequence with a time delay of τ. This is the equivalent channel matrix.

[0089] The joint detection and estimation method for active users, transmission delay, and channel used in this invention includes the following steps:

[0090] S1. At the receiving end, Correlation peak detection is performed with all leader sequences in the leader sequence pool to preliminarily estimate the number of users included in the observation window. and its delay ,in Indicates the first Initial transmission delay estimate for each active user.

[0091] S2, Definition , ,in for The To the Row all elements and get the following Figure 2The example factor plot shown below, where, , According to the sum-product algorithm, in the th... During each iteration, from the factor node Passed to variable node The message can be approximated by a Gaussian distribution. Their mean and variance are respectively

[0092]

[0093] ,

[0094] in, for The To the All elements of the row are passed to the variable node. , To account for the transmission delay of the preamble sequence matrix, Indicates the first The transmission delay vector estimated in the next outer iteration Indicates the first The equivalent channel matrix estimated in the next outer iteration has non-zero rows corresponding to active users. represents the conjugate transpose of the matrix, and n represents the row index in the matrix;

[0095] S3. From the variable node in the Markov chain Passed down to factor nodes The message approximates a Gaussian distribution. Their mean and variance are respectively

[0096]

[0097] ,

[0098] in, , for The To the All elements in the row, for The autocorrelation matrix, and From the factor nodes respectively Passed down to the variable node The mean and variance are calculated as follows:

[0099]

[0100] ,

[0101] in, ; It is a block diagonal matrix. The dimension is A matrix of all zeros express The zero matrix of .

[0102] S4. Similarly, from the variable node Propagated upwards to factor nodes The message approximates a Gaussian distribution. Their mean and variance are respectively

[0103]

[0104] ,

[0105] in, and From the factor nodes respectively Passed up to the variable node The mean and variance are calculated as follows:

[0106]

[0107] ,

[0108] in, ,in Represented as The identity matrix.

[0109] S5, combining S2 and S3, from the factor node Passed to variable node The message approximates a Gaussian distribution. The calculation is as follows:

[0110] .

[0111] S6, Residual and inverse residual variance Calculated separately as

[0112]

[0113] ,

[0114] in, and They are respectively and The (m) ) elements; and They are respectively and The (m) ) elements.

[0115] S7, From Factor Nodes Passed to variable node The message can be approximated by a Gaussian distribution. Their mean and variance are respectively

[0116]

[0117] ,

[0118] in, for The element in the k-th column; for The element in the m-th row.

[0119] S8 In the The posterior mean and variance of each iteration are respectively

[0120]

[0121] ,

[0122] in, To reduce the sparsity of active users, Channel matrix elements The average large-scale fading, and

[0123] .

[0124] S9, if the matrix The The power of the row is greater than the set threshold. Then the first One user is currently active, meaning:

[0125]

[0126] like If the value is less than the set threshold, the process ends; otherwise, proceed to step S3.

[0127] S10, Update Latency , among which, the element exist Search within the range, Let it be a small positive real number to reduce the search complexity, such that

[0128]

[0129] Maximum. If If the maximum number of iterations is exceeded, the process ends; otherwise, proceed to S3.

[0130] in, and They are respectively and The List, and They are respectively and The autocorrelation matrix, This indicates extracting the real part of a complex number. The trace operation represents the sum of the diagonal elements of a matrix. The equivalent noise covariance matrix The inverse matrix.

[0131] Example 1

[0132] This invention provides a joint estimation method for active users, delay, and channel for a low-precision receiver employing a 1-bit analog-to-digital converter (ADC). The method includes:

[0133] The quantized signal received by the receiving antenna is compared with the known preamble sequence in the preamble sequence pool to detect correlation peaks, and the number of active users and their transmission delay within the observation window are initially estimated; (corresponding to S1)

[0134] Based on the Bussgang decomposition principle, the original nonlinear signal receiving model is equivalent to a linear signal receiving model containing an equivalent channel matrix and equivalent noise; (corresponding to S1 and S2)

[0135] Employing an Expectation-Maximization (EM) algorithm framework, active user states, transmission delays, and channel states are jointly estimated through an inner and outer two-layer iterative process, wherein:

[0136] During the inner iteration process, a factor graph model is constructed based on a Markov chain. Through message passing between variable nodes and factor nodes, the estimated value of the equivalent channel matrix and the user activity status are updated; (inner iterations correspond to S2-S9)

[0137] During the outer iteration process, based on the channel estimation results obtained from the inner iteration, a local search algorithm is used to update the estimated transmission delay; (outer iteration corresponds to S10)

[0138] When the preset iteration termination condition is met, the final active user detection result, transmission delay estimate, and channel state information are output.

[0139] In the above scheme, the 1-bit analog-to-digital converter (ADC) quantizes the in-phase component and quadrature component of the received signal into +1 or -1, respectively.

[0140] In the above scheme, the step of initially estimating the number of active users and their transmission delay within the observation window includes: at the receiving end, converting the received signal matrix... Correlation peak detection is performed with all leader sequences in the leader sequence pool to preliminarily estimate the number of users included in the observation window. and its delay And construct the leader sequence matrix ;

[0141] in, express A matrix composed of quantized signals received by each receiving antenna. The number of receiving antennas, The width of the observation window, This is the oversampling factor. This indicates that the k-th user has a delay. The preamble sequence signal at that location.

[0142] In the above scheme, the step of converting the nonlinear signal model into a linear signal model based on the Bussgang decomposition principle includes:

[0143] The quantized received signal matrix is ​​decomposed using Bussgang decomposition. Equivalent to linear form:

[0144]

[0145] in, The Bussgang decomposition coefficient matrix, , for The autocorrelation matrix, Take the diagonal matrix; For the equivalent noise matrix, For pulse shaping filters and matched filter convolution The constructed Toplitz matrix has dimensions of , For transmission delay vector, for The constructed Toplitz matrix, For the noise matrix, Let be the matrix of the preceding sequence with a time delay of τ. This is the equivalent channel matrix.

[0146] In the above scheme, the step of employing the Expectation-Maximization (EM) algorithm framework to jointly estimate the active user state, transmission delay, and channel state through an inner and outer iterative process includes the following inner iterative process:

[0147] Constructing a factor graph model based on Markov chains, defining... , ;

[0148] Based on the sum-product algorithm, in the... During each iteration, from the factor node Passed to variable node The message approximates a Gaussian distribution. ,

[0149] From variable node Passed down to factor nodes The message approximates a Gaussian distribution. ;

[0150] From variable node Propagated upwards to factor nodes The message approximates a Gaussian distribution. ;

[0151] From factor nodes Passed to variable node The message approximates a Gaussian distribution. ;

[0152] Calculate residuals and inverse residual variance ;

[0153] From factor nodes Passed to variable node The message approximates a Gaussian distribution. ;

[0154] renew In the posterior mean during each iteration and variance ;

[0155] Based on a set threshold Determine user activity level: If the matrix The power of the k-th row is greater than the set threshold. If the k-th user is active, then the k-th user is active.

[0156] Determine the termination condition for the inner iteration: If Less than the set inner iteration threshold or If the maximum number of inner iterations is exceeded, the inner iteration ends; otherwise, = +1, continue inner layer iteration.

[0157] In the above scheme, the outer iteration process includes:

[0158] Based on the channel estimation results obtained from the inner iteration, update the delay. ,in, Let it be a small positive real number to reduce the search complexity, such that

[0159]

[0160] maximum;

[0161] in, for The autocorrelation matrix, and They are respectively and The List, and They are respectively and The autocorrelation matrix;

[0162] Determine the termination condition of the outer iteration: If If the maximum number of outer iterations is exceeded, the outer iteration ends; otherwise, , Return to execute the inner iteration process.

[0163] In the above scheme, Gaussian distribution Their mean and variance are as follows:

[0164]

[0165] ,

[0166] in, for The To the All elements in the row;

[0167] Gaussian distribution Their mean and variance are respectively

[0168]

[0169] ,

[0170] in, and From the factor nodes respectively Passed up to the variable node The mean and variance are calculated as follows:

[0171]

[0172] ,

[0173] in, , Take the block diagonal matrix.

[0174] Gaussian distribution The calculation formula is:

[0175] ;

[0176] Gaussian distribution Their mean and variance are as follows:

[0177]

[0178] ,

[0179] in, for The element in the k-th column; for The element in the m-th row.

[0180] In the above scheme, the residual and inverse residual variance The calculation formulas are as follows:

[0181]

[0182] ,

[0183] in, and They are respectively and The (m) ) elements; and They are respectively and The (m) ) elements;

[0184] ,

[0185] in, Channel matrix elements The large-scale fading average, and:

[0186] .

[0187] Example 2

[0188] In this specific embodiment, using the ZC sequence as the leading sequence, the nth element in the sequence is... , ,in, The length of each leader sequence. Specific parameter settings are shown in the table below:

[0189] Table 1 Main Simulation Parameters

[0190] According to the sum-product algorithm, in the... During each iteration, from the factor node The message can be approximated by a Gaussian distribution. Their mean and variance are as follows:

[0191]

[0192] ,

[0193] in, for The To the Row all elements.

[0194] S3. From the variable node in the Markov chain Passed down to factor nodes The message approximates a Gaussian distribution. Their mean and variance are respectively

[0195]

[0196] ,

[0197] in, , for The To the All elements in the row, and From the factor nodes respectively Passed down to the variable node The mean and variance are calculated as follows:

[0198]

[0199] ,

[0200] in, ; It is a block diagonal matrix.

[0201] S4. Similarly, from the variable node Propagated upwards to factor nodes The message approximates a Gaussian distribution. Their mean and variance are respectively

[0202]

[0203] ,

[0204] in, and From the factor nodes respectively Passed up to the variable node The mean and variance are calculated as follows:

[0205]

[0206] ,

[0207] in, .

[0208] S5, combining S2 and S3, from the factor node Passed to variable node The message approximates a Gaussian distribution. , calculated as

[0209] .

[0210] S6, Residual and inverse residual variance Calculated separately as

[0211]

[0212] ,

[0213] in, and They are respectively and The (m) ) elements; and They are respectively and The (m) ) elements.

[0214] S7, From Factor Nodes Passed to variable node The message can be approximated by a Gaussian distribution. Their mean and variance are respectively

[0215]

[0216] ,

[0217] in, for The element in the k-th column; for The element in the m-th row.

[0218] S8 In the The posterior mean and variance for each iteration are as follows:

[0219]

[0220] ,

[0221] in, Channel matrix elements The large-scale fading average, and:

[0222] .

[0223] S9, if the matrix The The power of the row is greater than the set threshold. Then the first Each user is active, that is...

[0224]

[0225] like If the value is less than the set threshold, the process ends; otherwise, proceed to step S3.

[0226] S10, Update Latency , among which, the element exist Search within the range, Let it be a small positive real number to reduce the search complexity, such that

[0227]

[0228] Maximum. Among them, for The autocorrelation matrix, and They are respectively and The List, and They are respectively and The autocorrelation matrix. If If the maximum number of iterations is exceeded, the process ends; otherwise, proceed to S3.

[0229] Figure 3 and Figure 4 This paper presents a performance comparison between the proposed algorithm and a traditional random access algorithm in existing systems (3GPP, “Evolved universal terrestrial radio access (E-UTRA); physical channels and modulation,” 3rd Generation Partnership Project (3GPP), Technical Specification (TS) 36.211, Dec. 2016). In the traditional random access algorithm, after estimating active users, a linear minimum variance algorithm is used to further estimate channel information. It can be seen that the proposed algorithm has superior channel estimation performance while maintaining a low false detection probability. Furthermore, it improves the oversampling rate. It helps improve the overall performance of the system.

Claims

1. A joint estimation method for active users, delay, and channel for low-precision receivers, characterized in that, The low-precision receiver employs a 1-bit analog-to-digital converter, and the method includes: The quantized signal received by the receiving antenna is correlated with the known preamble sequence in the preamble sequence pool to detect the correlation peak and make a preliminary estimate of the number of active users and their transmission delay within the observation window. Based on the Bussgang decomposition principle, the original nonlinear signal receiving model is equivalent to a linear signal receiving model containing an equivalent channel matrix and equivalent noise. Employing an expectation-maximization algorithm framework, the active user state, transmission delay, and channel state are jointly estimated through an inner and outer two-layer iterative process, wherein: During the inner iteration process, a factor graph model is constructed based on the Markov chain. Through message passing between variable nodes and factor nodes, the estimated value of the equivalent channel matrix and the user activity status are updated. During the outer iteration process, the estimated transmission delay is updated based on the channel estimation results obtained from the inner iteration using a local search algorithm. When the preset iteration termination condition is met, the final active user detection result, transmission delay estimate, and channel state information are output.

2. The method according to claim 1, characterized in that, The 1-bit analog-to-digital converter quantizes the in-phase and quadrature components of the received signal into +1 or -1, respectively.

3. The method according to claim 1, characterized in that, The steps for initially estimating the number of active users and their transmission delay within the observation window include: at the receiving end, converting the received signal matrix... Correlation peak detection is performed with all leader sequences in the leader sequence pool to preliminarily estimate the number of users included in the observation window. and its delay And construct the leader sequence matrix ; in, express A matrix composed of quantized signals received by each receiving antenna. The number of receiving antennas, The width of the observation window, This is the oversampling factor. This indicates that the k-th user has a delay. The preamble sequence signal at that location.

4. The method according to claim 1, characterized in that, The steps for converting a nonlinear signal model into a linear signal model based on the Bussgang decomposition principle include: The quantized received signal matrix is ​​decomposed using Bussgang decomposition. Equivalent to linear form: in, The Bussgang decomposition coefficient matrix, , for The autocorrelation matrix, Take the diagonal matrix; This is the equivalent noise matrix; For pulse shaping filters and matched filter convolution The constructed Toplitz matrix has dimensions of ; This is the transmission delay vector; for The constructed Toplitz matrix, This is the noise matrix; For delay The preceding sequence matrix; This is the equivalent channel matrix.

5. The method according to claim 1, characterized in that, The step of jointly estimating active user state, transmission delay, and channel state through an inner and outer iterative process using the expectation-maximization algorithm framework includes the following inner iterative process: Constructing a factor graph model based on Markov chains, defining... , ; Based on the sum-product algorithm, in the... During each iteration, from the factor node Passed to variable node The message approximates a Gaussian distribution. , From variable node Passed down to factor nodes The message approximates a Gaussian distribution. ; From variable node Propagated upwards to factor nodes The message approximates a Gaussian distribution. ; From factor nodes Passed to variable node The message approximates a Gaussian distribution. ; Calculate residuals and inverse residual variance ; From factor nodes Passed to variable node The message approximates a Gaussian distribution. ; renew In the posterior mean during each iteration and variance ; Based on a set threshold Determine user activity level: If the matrix The power of the k-th row is greater than the set threshold. If the k-th user is active, then the k-th user is active. Determine the termination condition for the inner iteration: If Less than the set inner iteration threshold or If the maximum number of inner iterations is exceeded, the inner iteration ends; otherwise, = +1, continue inner layer iteration.

6. The method according to claim 5, characterized in that, Executing the outer iteration process includes: Based on the channel estimation results obtained from the inner iteration, update the delay. ,in, Let it be a small positive real number to reduce the search complexity, such that maximum; in, for The autocorrelation matrix, and They are respectively and The List, and They are respectively and The autocorrelation matrix; Determine the termination condition of the outer iteration: If If the maximum number of outer iterations is exceeded, the outer iteration ends; otherwise, , Return to execute the inner iteration process.

7. The method according to claim 5, characterized in that, Gaussian distribution Their mean and variance are as follows: ， in, for The To the All elements in the row; Gaussian distribution Their mean and variance are respectively ， in, and From the factor nodes respectively Passed up to the variable node The mean and variance are calculated as follows: ， in, , Take the block diagonal matrix. Gaussian distribution The calculation formula is: ； Gaussian distribution Their mean and variance are as follows: ， in, for The element in the k-th column; for The element in the m-th row.

8. The method according to claim 7, characterized in that, residual and inverse residual variance Calculated separately as follows: ， in, and They are respectively and The (m) ) elements; and They are respectively and The (m) ) elements.

9. The method according to claim 8, characterized in that, ， in, Channel matrix elements The large-scale fading average, and: 。

Images