Iterative signal detection method for uplink NOMA systems based on OTFS-OFDM modulation
By employing a symbolic posterior probability-driven iterative detection method in the uplink NOMA system to process the mean and covariance matrix of residual interference, the problem of ineffective handling of residual interference in existing technologies is solved, and detection performance is improved.
Patent Information
- Application Number
- CN202611394274.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-09
- Publication Date
- 2026-11-17
AI Technical Summary
In uplink NOMA systems where high-speed and low-speed mobile users coexist, existing detection methods typically assume that the detected user signals can be accurately reconstructed and completely eliminated, resulting in residual interference in reality not being effectively handled and affecting detection performance.
An iterative detection method based on symbolic posterior probability is adopted. By calculating the mean and covariance matrix of residual interference, the equivalent noise statistical parameters are updated, and signal reconstruction and interference cancellation of the soft decision result are performed. Iterative detection of low-speed and high-speed users is performed alternately.
It improves the accuracy of multi-user interference statistical modeling, reduces signal reconstruction deviation and residual interference accumulation caused by erroneous hard decisions, and enhances the system's detection performance.
Smart Images

Figure CN123123093A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of information and communication, specifically to the construction of an uplink non-orthogonal multiple access system for the coexistence of high-speed and low-speed mobile users and the design of an iterative detection method for the base station and receiver. Background Technology
[0002] Next-generation wireless communication systems are designed for dynamic service scenarios such as high-speed railways, vehicle-to-everything (V2X) communication, low-altitude platforms, and integrated air-space-ground communication. They need to maintain link stability and reliability while ensuring high spectrum utilization, even with both high-speed and low-speed mobile users. Traditional Orthogonal Frequency Division Multiplexing (OFDM) technology can transform frequency-selective multipath channels into parallel flat fading channels on each subcarrier, offering advantages such as low equalization complexity and high spectral efficiency in low-speed mobile scenarios. However, when the relative movement speed between transceivers is high, significant Doppler spread can disrupt the orthogonality between subcarriers, causing inter-carrier interference and requiring more complex receiver equalization methods. Furthermore, rapid movement causes rapid time-varying channel variations, increasing channel estimation overhead and errors, ultimately degrading system performance. The literature [HADANI R, RAKIB S, TSATSANIS M, et al. Orthogonal time frequency space modulation[C] / / 2017 IEEE Wireless Communications and Networking Conference. San Francisco: IEEE, 2017: 1-6.] proposes Orthogonal Time Frequency Space (OTFS) modulation, which modulates and demodulates signals in the Delay-Doppler (DD) domain and uses a two-dimensional transform to achieve the conversion between the DD domain and the time-frequency domain. The literature [RAVITEJA P, PHAN KT, HONG Y, et al. Interference cancellation and iterative detection for orthogonal time frequency space modulation[J]. IEEE Transactions on Wireless Communications, 2018, 17(10): 6501-6515.] further studies interference cancellation and iterative detection in OTFS systems. OTFS modulation enables fast time-varying channels to appear as approximately stable sparse channels in the DD domain, and is one of the solutions for reliable transmission under high Doppler spread channel conditions. OFDM has been applied in 4G and 5G mobile networks and has a mature link structure and infrastructure.The paper [WU Y, ZHANG Z. Co-existence analysis of OTFS and OFDM waveforms for multi-mobility scenarios[C] / / 2022IEEE 95th Vehicular Technology Conference. Helsinki: IEEE, 2022: 1-5.] studies the coexistence problem of OTFS and OFDM waveforms in multi-mobility scenarios. Therefore, introducing an OTFS transmission mechanism for high-speed mobile users into existing OFDM systems is a practically significant evolutionary path for scenarios where high-speed and low-speed mobile users coexist.
[0003] In multi-user networks, multiple access technology is a crucial factor affecting system performance. Orthogonal multiple access (OMA) avoids inter-user interference by allocating mutually orthogonal time, frequency, or code domain resources to different users, resulting in relatively low signal detection complexity at the receiver. However, the limited availability of orthogonal resources restricts the number of users that can access the network simultaneously. Non-orthogonal multiple access (NOMA) allows signals from multiple users to be superimposed non-orthogonally in the power or code domain, enabling multiple users to share the same transmission resources. This allows for more simultaneous access and more flexible resource allocation among users. However, non-orthogonal superposition inevitably introduces multi-user interference, typically requiring complex detection methods such as Successive Interference Cancellation (SIC). The literature [DING Z, LEI X, KARAGIANNIDIS GK, et al. A survey on non-orthogonal multiple access for 5G networks: research challenges and future trends[J]. IEEE Journal on Selected Areas in Communications, 2017, 35(10): 2181-2195.] reviews the basic principles, research problems and development trends of NOMA technology. With the development of OTFS technology research, multiple access technology in OTFS systems has also attracted attention. When there are only high-speed mobile users in the system, near-orthogonal multiple access can be achieved by allocating different DD domain resources to different users and using OTFS modulation. The literature [KHAMMAMMETTI V, MOHAMMED S K. OTFS-based multiple-access in high Doppler and delay spread wireless channels[J]. IEEE Wireless Communications Letters, 2019, 8(2): 528-531.] studies the OTFS multiple access method in high Doppler spread and delay spread channels. Similar to NOMA in the time or frequency domain, multi-user NOMA can also be implemented in the DD domain, allowing multiple high-speed mobile users to use the same DD domain resources for transmission.The literature [CHATTERJEE A, RANGAMGARI V, TIWARI S, et al. Nonorthogonal multiple access with orthogonal time-frequency space signal transmission[J]. IEEE Systems Journal, 2021, 15(1): 383-394.] studies a nonorthogonal multiple access scheme using OTFS signal transmission. In practical applications, high-speed mobile users and low-speed mobile users are more likely to coexist in the same communication system. Since the OTFS modulation system can be implemented by adding inverse symplectic finite Fourier transform (ISFFT) and symplectic finite Fourier transform (SFFT) to the existing OFDM modulation system, it is a reasonable solution to the coexistence scenario of high-speed mobile users using OTFS modulation and low-speed mobile users continuing to use OFDM modulation. Based on the different resource reuse relationships between users, multiple access schemes for the coexistence of high-speed mobile users and low-speed mobile users can be divided into orthogonal multiple access schemes and non-orthogonal multiple access schemes. By utilizing the correspondence between DD domain resources and time-frequency domain resources, non-overlapping time-frequency domain resources or DD domain resources can be allocated to high-speed mobile users and low-speed mobile users, achieving orthogonal multiplexing of transmission resources between users and reducing interference between users. The literature [WU Y, ZHANG Z. Co-existence analysis of OTFS and OFDM waveforms for multi-mobility scenarios[C] / / 2022 IEEE 95th Vehicular Technology Conference. Helsinki: IEEE, 2022: 1-5.] addresses the coexistence scenario of high-speed OTFS users and low-speed OFDM users by using time-domain resource isolation or frequency-domain resource isolation to achieve orthogonal coexistence of the two types of waveforms. Non-orthogonal multiple access schemes allow high-speed mobile users and low-speed mobile users to share some or all of the transmission resources.In the scheme disclosed in [GEY, DENG Q, CHING PC, et al. OTFS signaling for uplink NOMA of heterogeneous mobility users[J]. IEEE Transactions on Communications, 2021, 69(5): 3147-3161.], users with different mobility speeds all use OTFS modulation and occupy all DD domain transmission resources, utilizing the differences in mobility characteristics between users to improve NOMA transmission performance. In the scheme disclosed in [DING Z, SCHOBER R, FAN P, et al. OTFS-NOMA: an efficient approach for exploiting heterogenous user mobility profiles[J]. IEEE Transactions on Communications, 2019, 67(11): 7950-7965.], high-speed mobile users use OTFS modulation, low-speed mobile users use OFDM modulation, the two types of users share the same time-frequency resources, non-orthogonal superposition transmission is performed in the power domain, and a SIC detection strategy is used at the receiver.
[0004] Signal detection is a crucial component of receiver design in wireless communication systems. Modulation and demodulation in OTFS modulation systems occur in the DD domain. Due to Doppler spread and time delay spread, inter-symbol interference exists between received symbols, making the equalization and detection processes complex. Currently, commonly studied detection algorithms in OTFS systems include Message Passing (MP) detection, Minimum Mean Square Error (MMSE) detection, and Maximum Ratio Combining (MRC) detection. The literature [RAVITEJA P, PHAN KT, JIN Q, et al. Low-complexity iterative detection for orthogonal time frequency spacemodulation[C] / / 2018 IEEE Wireless Communications and Networking Conference.Barcelona: IEEE, 2018: 1-6.] proposes a low-complexity iterative detection method for OTFS; the literature [SURABHIG D, CHOCKALINGAM A. Low complexity linear equalization for OTFS modulation[J]. IEEE Communications Letters, 2020, 24(2): 330-334.] studies a low-complexity linear equalization method for OTFS modulation; the literature [THAJ T, VITERBO E. Low complexity iterative rake detector for orthogonal time frequency space modulation[C] / / 2020 IEEE Wireless Communications and Networking Conference. Seoul: IEEE, 2020: 1-6.] proposes a low-complexity iterative rake detection method for OTFS. In multi-user systems, inter-user interference also exists, making the signal detection process more complex. In OTFS-NOMA systems, a SIC process is typically added to the OTFS equalization detection algorithm to reduce multi-user interference through user-by-user detection and interference cancellation.The literature [CHATTERJEE A, RANGAMGARI V, TIWARIS, et al. Nonorthogonal multiple access with orthogonal time-frequency spacesignal transmission[J]. IEEE Systems Journal, 2021, 15(1): 383-394.] uses an MMSE-SIC receiver for multi-user detection. In the DD domain, it uses the linear MMSE criterion to estimate user symbols and combines it with SIC to gradually eliminate interference from detected users. To further improve the interference cancellation effect, the literature [MCWADE S, FARHANG A, FLANAGAN M F. Low-complexity reliability-based equalization and detection for OTFS-NOMA[J]. IEEE Transactions on Communications, 2023, 71(11): 6779-6792.] combines iterative equalization, reliability region decision and interference cancellation. In each iteration, it identifies symbol estimates with higher reliability and uses the corresponding symbols for interference reconstruction and cancellation to improve the detection quality of subsequent iterations. For NOMA systems where high-speed and low-speed mobile users coexist, the literature [DING Z, SCHOBER R, FAN P, et al. OTFS-NOMA: an efficient approach for exploiting heterogenous user mobility profiles[J].IEEE Transactions on Communications, 2019, 67(11): 7950-7965.] studies NOMA systems where high-speed OTFS users and low-speed OFDM users coexist, and introduces methods such as frequency domain linear equalization and decision feedback equalization to suppress inter-symbol interference in the DD domain. During the receiver detection process, the signals of low-speed OFDM users are mainly detected in the time-frequency domain, while the signals of high-speed OTFS users are mainly equalized and detected in the DD domain. The interference cancellation process adopts the SIC strategy and includes the ideal interference cancellation assumption, that is, it is assumed that the signals of the detected users can be accurately reconstructed and effectively eliminated. However, in the actual NOMA signal detection process, due to factors such as noise, channel fading, and symbol decision errors, SIC is difficult to achieve completely ideal interference cancellation.The literature [KARA F, KAYA H. BER performances of downlink and uplink NOMA in the presence of SIC errors over fading channels[J]. IET Communications, 2018, 12(15): 1834-1844.] studies the bit error rate performance of uplink and uplink NOMA systems in the presence of SIC errors, and analyzes the problem of incomplete interference cancellation caused by erroneous decisions, leading to a decrease in bit error rate performance, based on the traditional SIC detection process. This study reveals the impact of SIC errors and residual interference on NOMA systems from the perspective of detection performance. To further improve interference cancellation in scenarios where high-speed and low-speed mobile users coexist, the literature [ZHANG H, NIU K, XU J, et al. Iterative SIC-based multiuser detection for uplinkheterogeneous NOMA system[C] / / 2022 IEEE Globecom Workshops. Rio de Janeiro:IEEE, 2022: 94-99.] proposes an iterative SIC multiuser detection method based on frequency domain equalization and MP detectors for uplink heterogeneous NOMA systems where low-speed OFDM users and high-speed OTFS users coexist. This method performs multiple rounds of iterative processing between low-speed OFDM user detection and high-speed OTFS user detection, gradually improving subsequent detection performance by utilizing the detection results of the previous round for interference reconstruction and cancellation. However, in actual iterative detection, there is still decision uncertainty for the detected symbols. If each iteration still relies on hard decision results for interference reconstruction, incorrect decisions will lead to incomplete interference cancellation in the current round and may cause residual interference accumulation and error propagation in subsequent iterations. Summary of the Invention
[0005] The purpose of this invention is to provide a signal detection method in an uplink NOMA system comprising one low-speed mobile user and one high-speed mobile user. In the system, the low-speed mobile user uses OFDM modulation, and the high-speed mobile user uses OTFS modulation, occupying the same time-frequency resources. The receiver's signal detection also employs a SiC detection strategy, and the potential for detection errors leading to incomplete interference cancellation is considered.
[0006] To achieve the above objectives, this invention employs the following technical solution: First, the signal is estimated based on the symbol posterior probability obtained during the detection process, and then this estimate is used for interference cancellation. Furthermore, to improve the interference cancellation effect and system detection performance, an iterative interference cancellation strategy for high- and low-speed user signals is introduced during the detection process. To quantify the impact of non-ideal interference cancellation on symbol detection, this paper derives the mean and covariance matrix of residual interference, and calculates the symbol posterior probability under the condition of residual interference, feeding it back into the subsequent iterative detection process.
[0007] The specific steps are as follows:
[0008] (1) Signal transmission and reception: Low-speed mobile users use Orthogonal Frequency Division Multiplexing (OFDM) to transmit information symbols in the time-frequency domain, while high-speed mobile users use Orthogonal Time Frequency Space (OTFS) modulation to transmit information symbols in the time-delay Doppler domain. The two users non-orthogonally multiplex the same time-frequency resources to form a Non-Orthogonal Multiple Access (NOMA) system. The base station performs Discrete Fourier Transform on the received superimposed signals from the two users to convert them to the time-frequency domain.
[0009] (2) First round detection of low-speed mobile user signal: Treat the high-speed mobile user signal as noise, detect the low-speed mobile user signal in the time and frequency domain, obtain the posterior probability of the low-speed mobile user symbol, obtain the estimate of the low-speed mobile user signal based on the posterior probability, and cancel the low-speed user signal in the received signal.
[0010] (3) High-speed mobile user detection: The time-frequency domain signal after cancellation of low-speed mobile user signal is transformed to the time-delay Doppler domain, and high-speed mobile user is detected. The posterior probability of high-speed mobile user symbol is obtained. The posterior probability is used to estimate the high-speed mobile user signal, and the high-speed mobile signal in the received signal is cancelled.
[0011] (4) Iterative detection of low-speed mobile users: The time-delayed Doppler domain signal after canceling the high-speed mobile user signal is transformed to the time-frequency domain, the low-speed mobile user signal is detected, and the posterior probability of the low-speed mobile user symbol, signal estimation and signal cancellation are updated.
[0012] (5) Iterative detection, stopping and output: Repeat steps (3) and (4). After each iteration detection of low-speed and high-speed mobile users, calculate the average Kullback-Leibler (KL) divergence between the symbol posterior probability distributions obtained from two adjacent iteration detections. When the KL divergence is less than the preset threshold or the number of iterations reaches the preset maximum value, stop the iteration. Make a decision on the symbols of low-speed and high-speed mobile users based on the posterior probability obtained in the last detection and output the decision result.
[0013] Furthermore, in the first round of detection of low-speed mobile users and detection of high-speed mobile users described in steps (2) and (3), the posterior mean of the symbol is calculated based on the posterior probability of the user symbol, and the posterior mean is used as the estimated value of the user symbol for signal reconstruction and interference cancellation; the corresponding posterior variance is calculated based on the posterior probability of the symbol, and the mean and covariance of the residual interference after signal cancellation are determined based on the posterior variance, and the residual interference and additive noise are used together as the equivalent noise in another user detection process.
[0014] Furthermore, in the iterative detection process of low-speed mobile users described in step (4), the time-delay Doppler domain observation signal after cancellation of high-speed mobile user signals is transformed to the time-frequency domain. After reconstructing based on the mean of the residual interference after high-speed mobile user signal cancellation Update the equivalent noise statistics parameters in low-speed moving user detection. and And recalculate the posterior probability, symbol estimate, and residual interference statistics of the low-speed mobile user symbols.
[0015] Furthermore, in step (5), during the iterative detection, stopping, and output process, high-speed mobile user detection and low-speed mobile user iterative detection are performed alternately, and the average Kullback-Leibler divergence between the symbol posterior probability distributions obtained in two adjacent iterations is calculated. When the average Kullback-Leibler divergence is less than a preset threshold, or when the number of iterations reaches a preset maximum value, the iteration is stopped, and the decision result of the two user symbols is output according to the symbol posterior probability obtained in the last iteration.
[0016] Compared with existing related studies, the present invention has the following beneficial technical effects: (1) Existing NOMA detection methods for the coexistence of high-speed OTFS users and low-speed OFDM users usually adopt the SIC strategy. In the receiver detection scheme disclosed in the literature [DING Z, SCHOBER R, FAN P, et al. OTFS-NOMA: an efficient approach for exploiting heterogenous usermobility profiles[J]. IEEE Transactions on Communications, 2019, 67(11):7950-7965.], low-speed OFDM users are mainly detected in the time-frequency domain, and high-speed OTFS users are mainly detected in the time-delay-Doppler domain, and subsequent user detection is based on the ideal interference cancellation condition that the detected user signal can be accurately reconstructed and effectively eliminated. The present invention does not adopt the ideal interference cancellation assumption, but treats the residual interference that still exists after the cancellation of the detected user signal as a random variable, calculates the mean vector and covariance matrix of the residual interference, and incorporates it into the equivalent noise statistical parameters in the detection process of another user, thus more accurately describing the received signal after the actual signal cancellation. (2) The literature [ZHANG H, NIU K, XU J, et al. IterativeSIC-based multiuser detection for uplink heterogeneous NOMA system[C] / / 2022IEEE Globecom Workshops. Rio de Janeiro: IEEE, 2022: 94-99.] proposes an iterative SIC multiuser detection method for uplink heterogeneous NOMA systems where low-speed OFDM users and high-speed OTFS users coexist. This method involves multiple iterations between the two types of user detection, and the symbol decision results from the previous round are used to reconstruct and cancel user interference. This type of interference reconstruction method based on hard decision results fails to reflect the uncertainty of symbol decision. When a hard decision is incorrect, a large deviation may occur between the reconstructed signal and the actual transmitted signal, causing error propagation in subsequent iterations. This invention obtains the posterior mean of the symbol based on the symbol posterior probability, and uses the posterior mean instead of the hard decision result for signal reconstruction and interference cancellation. This reduces the signal reconstruction deviation caused by erroneous hard decisions, and lowers the accumulation of residual interference and error propagation. (3) This invention not only uses the symbol posterior mean to reconstruct the user signal, but also calculates the symbol estimation variance based on the symbol posterior probability, and calculates the covariance matrix of the residual interference after cancellation of the user signal based on the estimated variance.Therefore, this invention can simultaneously utilize the estimated value information and reliability information in the soft detection results, so that the equivalent noise mean and variance in subsequent user detection are updated with the results of each iteration, thereby improving the accuracy of multi-user interference statistical modeling. Attached Figure Description
[0017] Figure 1 This is a communication system model of the present invention;
[0018] Figures 2 to 4 This is a curve showing the variation of BER (Path Loss Per Requirement) of low-speed mobile users under QPSK modulation with different large-scale path losses as a function of the mobile user's transmit power.
[0019] Figures 5 to 7 This is a curve showing the variation of BER (Path Loss Per Requirement) of low-speed mobile users under BPSK modulation with different large-scale path losses as a function of mobile user transmit power.
[0020] Figure 8 It is the average number of stopping iterations under QPSK modulation; Detailed Implementation
[0021] This paper studies the signal detection algorithm of a base station in an uplink NOMA system simultaneously serving one low-speed mobile user and one high-speed mobile user. The information bits of the two users are mapped by constellation to obtain modulation symbols. The two users share the same time-frequency resources, and the signals transmitted by the two users form a non-orthogonal superposition signal at the base station. Assuming the system bandwidth is B, it is divided into... Sub-channels, sub-channel bandwidth One OFDM symbol period is The signal sampling period is The duration of one OFDM symbol period is called a time slot. A frame is composed of several time slots, and a frame contains Each time-frequency resource unit (TFLU) is used. Symbol modulation for low-speed mobile users is performed in the time-frequency domain; symbols are mapped to TFLUs, transformed back to the time domain, and then transmitted through the channel. Symbol modulation for high-speed mobile users is performed in the DD domain, mapped to TFLUs... Delay index and On a DD domain grid composed of Doppler indices. The DD domain time delay resolution is... Doppler resolution is The DD domain symbols of high-speed mobile users are transformed into the time-frequency domain by ISFFT, and then transformed back into the time domain before transmission.
[0022] Let the modulation symbol on the m-th subcarrier in the n-th time slot of a low-speed mobile user be denoted as . ,in For slot indexing, For subcarrier index, This is the modulation symbol set for low-speed mobile users, where the subscript L indicates a low-speed mobile user. The table shows the time-frequency domain. The average power of the modulation symbol is normalized, i.e. The symbols on the M subcarriers of the nth time slot are denoted as vectors. The time-frequency domain symbols are transformed to the time domain by the Inverse Discrete Fourier Transform (IDFT) according to the time slot. The OFDM time-domain symbol vector of the nth time slot is...
[0023]
[0024] Insert a length of [length] before each OFDM symbol After the CP (Concurrent Probe) is applied, the signal is transmitted through the channel. The power of the transmitted signal is denoted as . .
[0025] Let the symbol for mapping high-speed mobile users to the DD domain be: ,in For Doppler indexing, For delay index, This is the modulation symbol set for high-speed mobile users, where the subscript H indicates a high-speed mobile user. , ,in This indicates taking the average value over 'a'. The DD symbols within a frame are represented using... dimensional matrix To represent. The DD domain symbol is obtained by ISFFT to obtain the time-frequency domain symbol as follows.
[0026]
[0027] Among them, the time-frequency domain symbol matrix The nth column That is, the first The frequency domain symbol vector of each time slot. Perform an IDFT on the frequency domain symbols of each time slot to obtain the time domain symbols:
[0028]
[0029] Then, a CP is inserted before the time-frequency symbol in each time slot before being sent into the channel for transmission, with a transmission power of [missing value]. .
[0030] For low-speed mobile users, the Doppler spread of the channel is small and can be ignored; the channel impulse response can be expressed as:
[0031]
[0032] in: It is a unit impulse function. For low-speed mobile user channels, the number of paths is [number]. and Representing the channel number respectively The complex gain of each path and the delay in units of the sampling period are considered. Given that the sampling period is sufficiently small, this paper only considers delays that are integer multiples of the sampling period. High-speed mobile user channels exhibit significant Doppler frequency shift, and the channel impulse response is expressed as...
[0033]
[0034] in: This represents the number of paths in the high-speed mobile user channel. , , and They represent the first Complex gain, delay, and path The integer and fractional parts of the Doppler frequency shift in units.
[0035] Symbols transmitted by low-speed and high-speed mobile users reach the base station after being transmitted through the channel. The base station first removes the initial CP (Content Change) from the beginning of the received symbol sequence in each time slot, and then performs a Discrete Fourier Transform (DFT) to transform the symbols into the time-frequency domain. The time-frequency domain received symbol vector is...
[0036]
[0037] in: This represents the time-domain received symbol vector after removing the CP in the nth time slot; It is additive Gaussian noise. Let I be the noise variance. M Denotes an M-dimensional identity matrix; and The nth time slots of the low-speed mobile user channel and the high-speed mobile user channel are respectively. 3D equivalent time-frequency domain channel matrix
[16] Low-speed mobile users experience slower channel changes, therefore it is assumed that... They are identical within a single frame and form a diagonal matrix:
[0038]
[0039] here , This indicates that the elements of a form a diagonal matrix. for
[0040]
[0041] in: , , , express A zero matrix of dimension 1. express An identity matrix of dimension 1.
[0042] The detection of low-speed mobile user symbols is performed in the time-frequency domain, while the detection of high-speed mobile user symbols needs to be performed in the DD domain. A matrix is formed by assembling the time-frequency domain symbol vectors of N time slots. SFFT can be used to transform time-frequency domain symbols into DD domain symbols:
[0043]
[0044] The relationship between received and transmitted symbols in the DD field is as follows:
[0045]
[0046] in: Additive Gaussian noise over the DD domain; DD domain symbol matrix for high-speed mobile users This is the NM-dimensional vector obtained by column vectorization, where This indicates that matrix A is vectorized into column vectors; To construct a matrix from N time-frequency domain symbols within a frame for a low-speed moving user The NM-dimensional vector is obtained by transforming the symbolic matrix into a DD-domain symbolic matrix and then vectorizing it. and The equivalent DD domain channel matrices for low-speed mobile users and high-speed mobile users are respectively:
[0047]
[0048]
[0049] Generally, low-speed mobile users are closer to the base station than high-speed mobile users, resulting in better channel conditions. A significant power difference exists between the signals of the two users when they reach the base station. Therefore, in this detection scheme, the base station first detects the signal of the low-speed mobile user, performs interference reconstruction and cancellation, and then detects the signal of the high-speed mobile user. Reconstructing the interference signal based on the hard decision result is a common method, but errors in hard decision-making lead to error propagation problems. This paper chooses to reconstruct the interference based on soft decision information and adopts an iterative interference cancellation method to improve detection performance in the detection of both low-speed and high-speed mobile user signals. Specifically, the base station first detects the low-speed mobile user signal in the time-frequency domain, obtains an estimate of the received signal of the low-speed mobile user based on its soft decision result, and cancels the signal from the low-speed mobile user in the received signal. Then, the canceled signal is transformed into the DD domain for the detection of the high-speed mobile user signal and interference cancellation, and then the signal is transformed back to the time-frequency domain for the detection of the low-speed mobile user signal. This process of alternating between the time-frequency domain and the DD domain for the detection of low-speed and high-speed mobile user signals and interference cancellation continues until the convergence condition or the maximum number of iterations is reached.
[0050] The equivalent time-frequency domain channel for low-speed mobile users is a diagonal matrix, and each subcarrier can be detected independently. The DD domain symbols for high-speed mobile users, after transformation to the time-frequency domain, are a linear combination of all NM DD domain symbols. Regardless of the type of symbol set (i.e., constellation diagram) used, they can all be considered Gaussian distributed after transformation to the time-frequency domain. During initial detection, the signal from the high-speed mobile user in the received signal is combined with the additive Gaussian noise of the channel, serving as the equivalent noise in the detection. According to equation (6), the... The equivalent noise vector in the received symbol vector of each time slot is: The mean vector is
[0051]
[0052] in, This indicates taking the nth column of matrix A. The equivalent noise covariance matrix is...
[0053]
[0054] in: It is a diagonal matrix, and the vector composed of the variances of the equivalent noise on each subcarrier symbol is: , This represents taking the elements on the diagonal of matrix A. Assume that the symbol values for low-speed users are equally probable prior. After the base station receives the signal, in the nth time slot, the... The posterior probability of low-speed user symbols on each subcarrier is
[0055]
[0056] in: for Time-frequency domain received symbol on the m-th subcarrier for The equivalent noise variance on the m-th subcarrier for The mean of equivalent noise on the m-th subcarrier.
[0057] The mean of the symbols for low-speed mobile users is obtained based on the posterior probability and used as an estimate of the symbols transmitted by low-speed users.
[0058]
[0059] The estimates of low-speed mobile user symbols on each subcarrier in each time slot are grouped into a vector:
[0060]
[0061] Using this estimated vector to cancel out signals from low-speed mobile users in the received signal, a time-frequency domain observation vector is obtained for detecting symbols from high-speed mobile users, i.e.
[0062]
[0063] Even after cancellation, a portion of the signal transmitted by low-speed moving users will still remain in the signal. From equation (15), we can see that the mean residual interference vector in the symbol vector of the nth time slot is...
[0064]
[0065] The covariance matrix of the residual disturbance is
[0066]
[0067] Among them, because It is a diagonal matrix. ,therefore
[0068]
[0069] After detecting low-speed mobile user signals, the base station combines the time-frequency domain observation vectors of N time slots in one frame into an observation signal matrix for high-speed mobile user signal detection. The SFFT transformation of this matrix yields the DD domain observation signal matrix:
[0070]
[0071] To facilitate the subsequent derivation, the above transformation process is written in another form that performs operations on vectors. , Column vectorization yields vectors of length NM. and The transformation process of equation (23) is equivalent to
[0072]
[0073] in, This is an NM-dimensional vector formed by linking the residual interference vectors from each time slot after low-speed mobile user signal cancellation. This is a vector formed by concatenating the noise vectors from each time slot. The residual interference from low-speed mobile users, after being transformed into the DD domain, can be treated as a Gaussian distributed random variable. When transformed together with additive Gaussian noise into the DD domain, it serves as the equivalent noise in high-speed mobile user detection; that is, the equivalent noise vector is... . The mean vector is
[0074]
[0075] The covariance matrix of the equivalent noise vector is
[0076]
[0077] Among them, the block diagonal matrix Let be the residual interference conditional covariance matrix for all time slots.
[0078] Combining equation (20), only the diagonal elements of the covariance matrix are taken as the equivalent noise variance vector for high-speed mobile user detection:
[0079]
[0080] in, Represents an MN-dimensional column vector of all 1s. This represents the element-wise squared modulo of matrix A.
[0081] After determining the distribution of equivalent noise, the signal of high-speed mobile users can be detected. Detection algorithms that can be used include MMSE, MP, and MRC. The MP algorithm is used in the simulation presented in this paper.
[11] Here is a brief introduction. The MP algorithm is a factor graph-based algorithm. (Sending symbols) For variable nodes , receive symbol For observation nodes , For variable nodes With observation nodes The correlation coefficient between values. Definition Indicates the relationship with the first A set of variable nodes connected to each observation node. Indicates the relationship with the first A set of observation nodes connected to a variable node. The MP detection algorithm updates probability messages iteratively between variable nodes and observation nodes.
[0082] When sending symbolic priors with equal probability, the initial probability message passed from the variable node to the observation node is:
[0083]
[0084] The message that observation node c transmits to variable node i is the mean and variance of the interference when the signals from other variable nodes besides variable node i in the received symbols corresponding to observation node c are considered as Gaussian interference:
[0085]
[0086]
[0087] in, Indicates in set The set excluding variable node i. Variable node i is determined by the observed node. The mean and variance of the transmitted interference can be used to calculate the interference based on the observed nodes. Resulting variable nodes The probability of sending the corresponding symbol:
[0088]
[0089] Furthermore, in addition to the observation nodes Other observation nodes are passed to the variable nodes. Multiply the probabilities of the messages and normalize them to get
[0090]
[0091] This round of variable nodes First observe the node The probability of transmission is updated based on the probability of the previous round, and the update formula is as follows:
[0092]
[0093] in, It is the damping factor. The transmitted symbol probability is updated to...
[0094]
[0095] After the iteration termination condition is met, the posterior probability of the transmitted symbol is obtained based on the messages passed by each observation node in the last iteration:
[0096]
[0097] After obtaining the posterior probability of the transmitted symbol, the estimated value of the high-speed mobile user symbol can be calculated:
[0098]
[0099] The symbol estimation vector for high-speed mobile users in the DD domain is defined as follows:
[0100]
[0101] By using this estimated vector to cancel out the signal from the high-speed mobile user in the received signal, the DD-domain equivalent observation vector for estimating the signal from the low-speed mobile user is obtained:
[0102]
[0103] Define the residual interference vector of high-speed mobile users as Its mean is
[0104]
[0105] The covariance matrix is
[0106]
[0107] in, It is a diagonal matrix.
[0108]
[0109] At this point, the iterative processes of high-speed mobile user detection, soft reconstruction, and interference cancellation are complete. The next step is to continue with low-speed mobile user interference reconstruction and re-detection.
[0110] After cancelling the detection of high-speed user signals, the detection of low-speed mobile user signals is performed again. The DD domain symbol vector after cancelling the high-speed mobile user signals, given by equation (38), is now... Organized in column order matrix The time-frequency domain observation matrix for low-speed mobile user signal detection is obtained by ISFFT.
[0111]
[0112] Similar to the process of analyzing the residual interference and noise mean and covariance matrix during high-speed user signal detection, the equivalent vector form of the above transformation process is:
[0113]
[0114] remember The equivalent noise is the combination of residual interference and additive Gaussian noise, and its mean vector is...
[0115]
[0116] The covariance matrix is
[0117]
[0118] in,
[0119]
[0120] The diagonal elements of the covariance matrix are the equivalent noise variance vector:
[0121]
[0122] here This indicates finding the square of the modulus of an element in matrix A.
[0123] Will Substituting the expression into the above formula, we take the vector. From the To the The elements yield the time-frequency domain equivalent noise variance vector for the nth time slot:
[0124]
[0125] in, , .
[0126] After obtaining the mean and variance of the residual interference plus noise in the observed signal matrix, the method for the first round of detection of low-speed user signals can be used to detect low-speed user signals. It is only necessary to use the variance in formula (15) for calculating the probability of each symbol. Replace with That's it. Then, calculate the estimate of the low-speed user symbol from the received signal based on the probability obtained in this round of detection, and cancel the low-speed user signal from the received signal to obtain the observation signal for high-speed user signal detection, and then perform high-speed user signal detection.
[0127] The above-mentioned detection and cancellation of low-speed user signals, plus the detection and cancellation of high-speed user signals, constitutes an iterative detection process. After each iterative detection, the KL divergence of the posterior probability distribution of low-speed mobile users in two adjacent iterations is calculated to assess the degree of change in the posterior probability during the iteration process.
[0128]
[0129] The superscripts "(t)" and "(t−1)" represent iterations t-th and t−1-th, respectively. Similarly, the KL divergence between the posterior probability distributions of high-speed mobile users in two adjacent iterations is calculated:
[0130]
[0131] The average of the posterior probabilities of low-speed and high-speed mobile users is calculated using the KL divergence. This serves as the basis for determining whether the iterative process has converged. When Less than the preset threshold When the iteration reaches a certain value, it indicates that the change in the probability of lag during the iteration process has become very small, and the iteration can be terminated. Alternatively, it indicates that the number of iterations has reached its maximum value. When the iteration ends, the iteration is terminated. After the iteration ends, the symbols for low-speed mobile users and high-speed mobile users are determined based on the posterior probability obtained from the last iteration. The determination criteria are as follows:
[0132]
[0133]
[0134]
[0135] The invention will now be described in further detail with reference to the accompanying drawings. Unless otherwise specified, the simulation parameters are set as follows: number of subcarriers. Number of time slots Circular prefix length carrier frequency Subcarrier spacing Low-speed mobile users and high-speed mobile users use the same transmission power, which is set to [value missing]. The number of propagation paths for both low-speed mobile user channels and high-speed mobile user channels is [number missing]. The delay index is The movement speed of high-speed mobile users is set to [specify]. The maximum Doppler frequency shift is Where c represents the speed of light, take Corresponding to the maximum normalized Doppler index Doppler index of each path in the simulation exist Random values are selected from within the channel. The power of the additive white Gaussian noise in the channel is fixed. For both high-speed and low-speed mobile users, the small-scale fading coefficients of each path are modeled as zero-mean complex Gaussian random variables with a variance of 1 / 4. The large-scale path losses are denoted as follows: and The path loss for slow-moving users is The path loss for fixed high-speed mobile users is The maximum number of iterations for iterative interference elimination is t. max =10, the KL divergence threshold for iteration termination is = To evaluate the performance of the proposed algorithm, a serial interference cancellation scheme, a soft interference cancellation scheme, and a user-free interference cancellation scheme were selected for comparison. In the serial interference cancellation scheme, firstly, subcarrier-by-subcarrier posterior probability detection is performed on low-speed mobile users in the time-frequency domain, and hard decision symbols are obtained based on the detection results. Then, the low-speed mobile user signal is reconstructed using these hard decision symbols to eliminate interference. Next, an MP detector is used in the DD domain to detect high-speed mobile users, further reconstructing their signals using the hard decision results of the high-speed mobile users, and updating the detection input of the low-speed mobile users for iterative detection. This is denoted as the Iterative SIC scheme. Correspondingly, the non-iterative soft interference cancellation scheme is denoted as Soft-IC, and the iterative soft interference cancellation scheme proposed in this paper is denoted as Iterative Soft-IC. In the user-free interference cancellation scheme, each user is detected independently, denoted as the No-MUI scheme. The bit error rate (BER) is used as the performance evaluation index in the simulation to compare the detection performance of the above schemes, and low-speed and high-speed mobile users are denoted as LM and HM, respectively.
[0136] Figures 2 to 4 The curves showing the BER (Path Loss Per Requirement) as a function of transmit power for different low-speed mobile users under QPSK modulation are presented. It can be seen that the BER of both high-speed and low-speed mobile users generally decreases with increasing transmit power. This is because as transmit power increases, the received signal power of both users increases simultaneously, reducing the impact of noise on symbol detection and thus improving the reliability of low-speed user detection in the time-frequency domain and high-speed user detection in the time-delay Doppler domain. Under the same transmit power and path loss conditions, the BER of traditional SIC is usually higher than that of the Iterative Soft-IC proposed in this paper. Traditional SIC directly uses hard-decision symbols to reconstruct interference. Once an error occurs in the first stage of detection, the interference component corresponding to the erroneous symbol will be incorrectly eliminated from the received signal, further affecting the detection of another user and causing error propagation. Iterative Soft-IC, on the other hand, uses the posterior mean for interference reconstruction and uses the posterior variance to describe the uncertainty of the current symbol estimation. When the posterior probability of a symbol is not yet concentrated, its posterior variance is large. The algorithm will treat the corresponding reconstruction error as residual interference in subsequent detections, unlike hard-decision SIC which treats unreliable symbols as completely correct. Therefore, Iterative Soft-IC can reduce the performance loss caused by non-ideal interference cancellation.
[0137] Further comparison Figures 2 to 4It can be observed that as the path loss of low-speed mobile users decreases from 70 dB to 50 dB, the received power advantage of low-speed mobile users relative to high-speed mobile users gradually increases, and the performance of Iterative Soft-IC also improves accordingly. This is because the receiver in this paper first detects low-speed mobile users with better channel conditions. When the path loss of low-speed mobile users is small, their posterior probabilities in the time-frequency domain are more concentrated, the posterior mean is closer to the actual transmitted symbol, and the posterior variance is also smaller, thus making the soft reconstruction signal of low-speed users more accurate. After the residual interference of low-speed users is reduced, the detection conditions of high-speed OTFS users in the DD domain are improved, and the soft information obtained from high-speed user detection can be used for the next round of low-speed user detection, thereby forming a positive iterative gain.
[0138] To verify the applicability of the proposed algorithm to different modulation schemes, Figures 5 to 7 The BER performance under BPSK modulation is presented. The overall trends of the algorithms under both modulation schemes are basically consistent. With increasing transmission power, the BER of each detection algorithm gradually decreases. As the path loss of low-speed users decreases and the difference in path loss between the two users increases, the reliability of initial detection and soft interference reconstruction for low-speed users improves, and the residual interference in high-speed user detection decreases accordingly, thus improving the overall system detection performance. Compared with QPSK results, BPSK typically has a lower BER under the same transmission power. This is because BPSK contains only two constellation points, has a larger minimum Euclidean distance under the same average symbol energy, and has a lower probability of crossing the decision boundary after being disturbed by noise and residual interference. Therefore, under BPSK conditions, the posterior probabilities of the two types of users are usually more likely to converge, and iterative detection with soft information is more likely to obtain reliable results. It can also be seen that when the transmission power is high and the path loss of low-speed users is small, some curves of Iterative Soft-IC, Iterative SIC, or No-MUI may reach similar lower limits of BER statistics. This is mainly due to the strong noise resistance inherent in BPSK itself.
[0139] Figure 8It can be seen that the average number of iterations for Iterative Soft-IC is approximately 3-5, significantly less than the set maximum of 10, indicating that the stopping criterion based on KL divergence can avoid a large number of invalid iterations. The average number of iterations does not decrease strictly monotonically across the entire transmission power range. Near 0 dBm, signal quality is poor, noise and residual interference are strong, and the posterior probabilities in each round may quickly enter a stagnant state with minimal change. Therefore, the KL divergence meets the stopping condition after a few iterations. When the transmission power increases to around 5 dBm, the detector can extract more effective information from the received signal, but the posterior probabilities of the two users still require multiple rounds of alternating updates to stabilize, thus the average number of iterations increases somewhat. When the transmission power continues to increase to 10-25 dBm, the initial detection of low-speed users becomes more reliable, the posterior mean reconstruction error and posterior variance both decrease, and the detection of high-speed users can obtain stable soft information more quickly. Therefore, the average number of iterations generally decreases and gradually stabilizes at around 3.
Claims
1. A signal iterative detection method for an uplink NOMA system based on OTFS-OFDM modulation, characterized in that: Includes the following steps: (1) Signal transmission and reception: Low-speed mobile users use Orthogonal Frequency Division Multiplexing (OFDM) to transmit information symbols in the time-frequency domain, while high-speed mobile users use Orthogonal Time Frequency Space (OTFS) modulation to transmit information symbols in the time-delay Doppler domain. The two users non-orthogonally multiplex the same time-frequency resources to form a Non-Orthogonal Multiple Access (NOMA) system. The base station performs Discrete Fourier Transform on the received superimposed signals from the two users to convert them to the time-frequency domain. (2) First round detection of low-speed mobile user signal: Treat the high-speed mobile user signal as noise, detect the low-speed mobile user signal in the time and frequency domain, obtain the posterior probability of the low-speed mobile user symbol, obtain the estimate of the low-speed mobile user signal based on the posterior probability, and cancel the low-speed user signal in the received signal. (3) High-speed mobile user detection: The time-frequency domain signal after cancellation of low-speed mobile user signal is transformed to the time-delay Doppler domain, and high-speed mobile user is detected. The posterior probability of high-speed mobile user symbol is obtained. Based on the posterior probability, the high-speed mobile user signal is estimated, and the high-speed mobile signal in the received signal is cancelled. (4) Iterative detection of low-speed mobile users: The time-delayed Doppler domain signal after canceling the high-speed mobile user signal is transformed to the time-frequency domain, the low-speed mobile user signal is detected, the posterior probability of the low-speed mobile user symbol and the signal estimation are updated, and the signal is canceled. (5) Iterative detection, stopping and output: Repeat steps (3) and (4). After each iteration detection of low-speed and high-speed mobile users, calculate the average Kullback-Leibler (KL) divergence between the symbol posterior probability distributions obtained from two adjacent iteration detections. When the KL divergence is less than the preset threshold or the number of iterations reaches the preset maximum value, stop the iteration. Make a decision on the symbols of low-speed and high-speed mobile users based on the posterior probability obtained in the last detection and output the decision result.
2. The signal iterative detection method for an uplink NOMA system based on OTFS-OFDM modulation according to claim 1, characterized in that: In the signal transmission and reception process described in step (1), the modulation symbol on the m-th subcarrier in the n-th time slot of the low-speed mobile user is: , For the modulation symbol set of low-speed mobile users, the modulation symbols on all M subcarriers form the OFDM frequency domain symbol vector. After transforming to the time domain using the inverse discrete Fourier transform, the power is expressed as... Transmission; Modulation symbols for high-speed mobile users on delay index l and Doppler index k , For high-speed mobile user modulation symbol set; all N×M modulation symbols form a matrix. After undergoing inverse symplectic finite Fourier transform to the time-frequency domain, and then performing an inverse discrete Fourier transform back to the time domain, the power is expressed as... Transmit; Base station time-frequency domain receive vector is The time-delayed Doppler domain receive vector is ,in: and These are the time-frequency domain channel matrices for low-speed and high-speed mobile users, respectively. and These are the time-delay Doppler domain channel matrices for low-speed and high-speed mobile users, respectively. For high-speed mobile users, the time-frequency domain symbol vector, and These are the time-delay Doppler domain symbol vectors for low-speed and high-speed mobile users, respectively. and These are the channel noise vectors in the time-frequency domain and the time-delay Doppler domain, respectively.
3. The signal iterative detection method for an uplink NOMA system based on OTFS-OFDM modulation according to claim 1, characterized in that: In the first round of low-speed mobile user detection described in step (2), the signal from the high-speed mobile user is regarded as noise, and together with the additive noise, it is used as the equivalent noise in the low-speed mobile user detection. The mean vector of the equivalent noise is... Covariance matrix variance vector The posterior probability of the low-speed mobile user symbol is calculated based on the high-speed user's transmit power, high-speed user channel matrix, and channel noise. in , and For the received symbol in the time-frequency domain on the m-th subcarrier, the equivalent noise variance and the equivalent noise mean are given; the mean of the transmitted symbol is obtained based on the posterior probability to estimate the transmitted symbol: The estimates of low-speed mobile user symbols on each subcarrier in each time slot are grouped into a vector. The low-speed mobile user signal in the received signal is canceled out based on the symbol estimation vector to obtain the time-frequency domain observation vector for detecting high-speed mobile user symbols. ,in The residual low-speed mobile user signal; the mean vector and covariance matrix of the residual signal are respectively... , It is calculated from the symbol posterior probability and the low-speed user channel matrix.
4. The signal iterative detection method for an uplink NOMA system based on OTFS-OFDM modulation according to claim 1, characterized in that: In the high-speed mobile user detection described in step (3), the received time-frequency domain signal after the low-speed mobile user signal has been canceled is first subjected to a symplectic finite Fourier transform to obtain the time-delay Doppler domain signal matrix. The residual low-speed user signal is considered as interference, and the interference plus noise vector in the time-delay Doppler domain is the mean vector. Covariance matrix The mean vector of the time-frequency domain interference plus noise vector can be obtained. Covariance matrix The variance of the interference and channel noise is calculated. Based on the mean and covariance matrix of the interference plus noise in the time-frequency domain, a message passing algorithm or other detection algorithm is used to detect high-speed mobile user signals, and the symbol posterior probability distribution is obtained. The mean of the symbol is calculated based on the posterior probability and used as an estimate of the symbol for high-speed mobile users. ; A vector composed of the estimated values After canceling out the signal from the high-speed mobile user in the received signal, the time-delay Doppler domain equivalent observation vector for estimating the signal from the low-speed mobile user is obtained as follows: ,in, The residual high-speed mobile user signal, its mean Covariance Matrix It is calculated based on the symbol posterior probability and the high-speed user channel matrix.
5. The signal iterative detection method for an uplink NOMA system based on OTFS-OFDM modulation according to claim 1, characterized in that: In step (4), during the iterative detection of low-speed mobile users, the time-delayed Doppler domain signal after canceling the high-speed mobile user signal is first transformed to the time-frequency domain through a symplectic finite Fourier transform to obtain the symbol matrix. The time-frequency domain equivalent noise is formed by combining the residual high-speed user signal with additive Gaussian noise. mean vector Covariance Matrix Based on the mean value of the residual high-speed mobile user signal in the time-delay Doppler domain Covariance matrix The mean and variance of the equivalent noise in the low-speed user signal detection in step (2) are obtained by calculating the channel noise variance; the mean and variance of the equivalent noise are replaced with the mean and variance of the channel noise variance. The mean and variance of the low-speed user symbols are used to calculate the posterior probability and estimate of the low-speed user symbols, and to cancel the low-speed user symbols in the received signal.
6. The signal iterative detection method for an uplink NOMA system based on OTFS-OFDM modulation according to claim 1, characterized in that: In step (5), during the iterative detection, stopping, and output, steps (3) and (4) are repeated to perform iterative detection and interference signal cancellation for high-speed mobile users and low-speed mobile users. After each high-speed mobile user signal detection, the average KL divergence between the symbolic posterior probability distributions obtained from two adjacent iterations is calculated. When the KL divergence is less than a preset threshold, it indicates that the iteration has converged. Alternatively, when the number of iterations reaches a preset maximum value, the iteration is stopped. A decision is made based on the symbolic posterior probability obtained from the last iteration, and the decision criterion is as follows: , Finally, output the decision results for the two user symbols.
