An asynchronous high-dimensional phase region orthogonal and differential large-scale MIMO communication system

Abstract

The application provides an asynchronous high-dimensional phase domain orthogonal and differential large-scale MIMO communication system, and belongs to the technical field of wireless communication and high-dimensional signal processing. The system comprises the following steps: adopting a time division duplex architecture, relying on the quasi-orthogonal characteristics of a high-dimensional random two-phase sequence to realize multi-user differentiation, and completing demodulation through energy domain non-coherent combining and differential energy decision. The system does not need channel matrix inversion and explicit channel state estimation throughout the process, and can stably work under the compound extreme working conditions of underdetermined antenna ratio, low signal-to-noise ratio, large sampling frequency offset and Rayleigh deep fading. The system is additionally provided with a hysteresis type phase domain adaptive regulation logic, a zero pilot overhead channel quality sensing function, a dynamic adjustment function of sequence dimension and access scale, a function of maximizing the access capacity and transmission rate under the premise of guaranteeing the target bit error rate, and a function of adapting to scenarios such as massive Internet of Things and satellite communication.

Description

Technical Field

[0001] This invention relates to the field of cross-disciplinary technology of wireless communication and high-dimensional signal processing, and particularly to an asynchronous high-dimensional phase-domain orthogonal and differential large-scale MIMO communication system. Background Technology

[0002] Multiple-input multiple-output (MIMO) technology is one of the core technologies for improving the capacity and reliability of modern wireless communication systems. The channel capacity of traditional MIMO systems is limited by the rank of the channel matrix, which is determined by the minimum number of transmitters and receivers. This means that to achieve concurrent transmission for K users, the receiver needs to be equipped with at least K antennas. In scenarios such as massive IoT access and satellite communication, the number of concurrent users is often much greater than the number of receiver antennas, forming an underdetermined MIMO channel. In this case, traditional spatial multi-user detection algorithms such as zero-forcing (ZF) and minimum mean square error (MMSE) will fail due to the singularity of the channel matrix. Orthogonal resource partitioning schemes such as time division, frequency division, and code division will split the limited spectrum resources into multiple sub-resources, significantly reducing the access capacity of a single resource block and failing to support the concurrent access needs of hundreds of thousands or even millions of users.

[0003] In addition, existing underdetermined MIMO schemes suffer from several performance shortcomings under extreme conditions: First, noise amplification is severe under low signal-to-noise ratio (SNR). In scenarios such as deep space communication, underwater communication, and covert communication, the channel SNR often drops to 0 dB or even lower, at which point the signal power is comparable to the noise power. Traditional multi-user detection relies on channel matrix inversion to separate the signal components of different users. Matrix inversion amplifies the base noise exponentially, ultimately leading to a demodulation error rate approaching the level of random guessing, resulting in a complete communication link interruption. Second, synchronization distortion is difficult to resolve under large sampling frequency offsets. The crystal oscillator accuracy of low-cost terminals, deep space probes, and other equipment is limited, often resulting in sampling frequency offsets on the order of several percent at both the transmitting and receiving ends, causing timing stretching or compression of the received signal. In existing synchronization schemes, linear interpolation smooths out the high-frequency transition edges of the binary phase-shift keying (BPS) sequence, destroying the orthogonality of the sequence itself and continuously introducing inter-symbol interference; FFT frequency domain resampling is affected by the non-periodic boundaries of the signal, generating Gibbs ringing interference, adding additional noise, and further degrading demodulation performance under low SNR. In short, existing solutions cannot maintain the orthogonality of the spreading sequence itself while compensating for large frequency offsets. Third, Rayleigh deep fading and broadband detection have technical blind spots. Rayleigh flat deep fading in wireless channels causes the received signal gain of a single antenna to approach zero, creating a decision blind spot; traditional coherent detection requires real-time estimation of complete channel state information, incurring huge computational overhead, and cannot automatically filter out receiving branches with poor signal quality. On the other hand, traditional square-law energy detection has an inherent defect of "phase blindness," meaning that signals with opposite signs become identical after squaring, losing the phase-carrying bit information. This limits it to narrowband binary transmission and cannot be extended to broadband high-speed modulation methods such as quadrature phase shift keying.

[0004] In summary, all existing MIMO multiple access schemes rely on the spatial dimension to distinguish users, heavily depend on channel matrix inversion, high-precision clock synchronization, and real-time channel estimation. They cannot simultaneously accommodate extreme conditions such as large-scale underdetermined antenna configuration, 0dB extreme signal-to-noise ratio, large-scale sampling frequency offset, Rayleigh deep fading, and broadband high-speed transmission, making it difficult to meet the underlying technical requirements of next-generation wireless communication systems. Summary of the Invention

[0005] This invention provides an asynchronous high-dimensional phase-domain orthogonal and differential massive MIMO communication system, which breaks away from the traditional spatial domain user differentiation design approach. It achieves multi-user differentiation by relying on the random orthogonal characteristics of high-dimensional sequences. The entire process does not require channel matrix inversion or explicit channel state information estimation. Under the combined extreme conditions of underdetermined antenna configuration, low signal-to-noise ratio, large sampling frequency offset, and Rayleigh deep fading, it achieves stable concurrent demodulation of massive users. Traditional MIMO systems rely heavily on high-precision clock synchronization and phase-locked loop circuits. When there is a percentage sampling frequency offset at the transmitting and receiving ends, the synchronization link is directly lost. However, this solution relies on the energy domain differential detection mechanism of high-dimensional sequences, combined with a cubic spline interpolation asynchronous alignment algorithm, to reduce the clock synchronization requirement to only coarse-grained time slot alignment. It achieves true asynchronous clockless communication capability from the physical layer mechanism.

[0006] This invention provides an asynchronous high-dimensional phase-domain orthogonal and differential massive MIMO communication system, comprising several transmitting devices and at least one receiving device configured with multiple receiving antennas. The transmitting devices include a high-dimensional random orthogonal sequence generation module and a dual-path bit modulation module for performing high-dimensional phase-domain random orthogonal modulation processing, wherein: The high-dimensional random orthogonal sequence generation module is used to assign two sets of independent random binary phase shift keying sequences of length N to each user to be accessed. The sequence length N is greater than the maximum number of concurrent users in the system. The random binary phase shift keying sequences form an N-dimensional linear Hilbert phase space, and multi-user differentiation is achieved by relying on the quasi-orthogonality of random sequences. The dual-path bit modulation module is used to control two physical transmission channels using a time-division multiplexing architecture. Each symbol period is divided into a reference time slot and a modulation time slot of equal length. The first physical transmission channel corresponds to the reference time slot and transmits a first set of random binary phase shift keying sequences as a reference sequence. The second physical transmission channel corresponds to the modulation time slot and selects to transmit either the first set of random binary phase shift keying sequences or the second set of independent random binary phase shift keying sequences according to the bits to be transmitted. The receiving device includes a timing alignment module, an implicit weight calculation module, a spatial subset merging module, and a differential detection and demodulation module, used to perform multi-level serial demodulation processing, wherein: The timing alignment module is used to acquire multiple mixed noisy received signals, and to perform timing resampling on two sets of random binary phase shift keying sequences stored locally using a high-order polynomial interpolation algorithm to compensate for the time axis distortion caused by sampling frequency offset, thereby achieving asynchronous clockless alignment. The implicit weight calculation module is used to calculate the incoherent channel weight of each receiving antenna corresponding to the target user using the received signal corresponding to the first physical transmission channel. The spatial diversity merging module is used to perform multi-antenna weighted merging on the received signal corresponding to the second physical transmission channel according to the non-coherent channel weight, and to aggregate the spatial diversity gain. The differential detection and demodulation module is used to calculate the L2 norm squared correlation energy of the weighted and merged second physical transmission channel signal and the two sets of local random two-phase shift keying sequences, respectively. The phase domain linear correlation result is converted into the energy domain decision space through L2 norm squared nonlinear mapping, and the original transmission bits are recovered by differential decision based on the numerical magnitude of the two sets of energy.

[0007] Preferably, the higher-order polynomial interpolation algorithm is a cubic spline interpolation algorithm; The cubic spline interpolation algorithm preserves the high-frequency transition edges of random binary phase shift keying sequences during time-series resampling.

[0008] Preferably, the formula for calculating the non-coherent channel weights is: ,in, The incoherent channel weights for user k corresponding to the m-th receiving antenna are: The signal received by the first physical transmission channel is collected by the m-th receiving antenna. The first set of local random binary phase shift keying sequences corresponding to user k; The incoherent channel weights corresponding to receiving antennas experiencing Rayleigh deep fading automatically approach zero.

[0009] Preferably, the multiple access interference generated by the remaining concurrent users contributes an equal amount of common-mode background noise to the two sets of L2 norm squared correlation energy, and the physical cancellation of the global multiple access interference is achieved by comparing the difference between the two sets of energy. The demodulation processing of the receiving device does not employ the L1 norm absolute value detection scheme.

[0010] Preferably, the transmitting device further includes a broadband extension module for performing broadband multi-code multiplexing extension processing, wherein: The broadband extension module is used to allocate four sets of pairwise orthogonal random binary phase shift keying sequences to each code channel of a single user. The two physical transmission channels select the corresponding in-phase or out-of-phase sequences for transmission according to the orthogonal phase shift keying double-bit symbols. The differential detection and demodulation module of the receiving device is also used to perform L2 norm square law differential detection on the four sets of sequences respectively, to eliminate the phase blindness defect of square law detection and realize broadband high-speed orthogonal phase shift keying transmission.

[0011] Preferably, the length N of the random binary phase shift keying sequence is ≥4096, the number of concurrent users K is ≥200, and the number of receiving antennas M is ≤20, so as to achieve asynchronous stable demodulation under an underdetermined antenna ratio of K / M ≥10; In narrowband operating mode, the phase domain processing gain is no less than 36dB.

[0012] The present invention provides a transmitting device for a large-scale MIMO communication system with asynchronous high-dimensional phase domain orthogonal and differential modulation, wherein the high-dimensional random orthogonal sequence generation module and the dual-path bit modulation module are configured to perform the high-dimensional phase domain random orthogonal modulation processing.

[0013] The present invention provides a receiving device for a large-scale MIMO communication system with asynchronous high-dimensional phase domain orthogonality and difference, wherein the timing alignment module, implicit weight calculation module, spatial subset merging module and differential detection demodulation module are configured to perform the multi-level serial demodulation processing.

[0014] Compared with the prior art, the beneficial effects of this application are as follows: 1. Breaking through capacity limits and supporting massive access. Relying on the random quasi-orthogonal characteristics of a high-dimensional space of thousands of dimensions, it can demodulate tens of times the number of concurrent users with only a few receiving antennas. It is not constrained by the capacity formula of traditional MIMO and can support the concurrent access needs of millions of IoT terminals.

[0015] 2. It avoids noise amplification problems from a mechanism perspective and is suitable for extremely low signal-to-noise ratios. There are no matrix inversion operations throughout the process, which avoids the noise amplification problem under low signal-to-noise ratios at the source; the processing gain provided by the high-dimensional sequence, combined with the diversity gain of multiple antennas, can improve the physical channel signal-to-noise ratio of 0dB to an equivalent decision signal-to-noise ratio of over 20dB, and the demodulation performance far exceeds that of traditional solutions.

[0016] 3. Compatible with large sampling frequency offsets and low orthogonality loss. Timing alignment is achieved using cubic spline interpolation, effectively preserving the sequence's transition characteristics while compensating for large sampling frequency offsets. It does not introduce significant inter-symbol interference or additional noise, making it compatible with low-cost crystal oscillators in terminal equipment.

[0017] 4. Autonomous resistance to deep fading with low computational complexity. By implicitly calculating channel weights through the baseline path, deep fading branches are automatically downweighted and masked, eliminating the need for real-time channel estimation and complex equalization calculations. Energy domain incoherent combining is used to avoid phase cancellation, resulting in baseband computational overhead that is far lower than traditional coherent detection architectures, making it easy to deploy on lightweight terminals and embedded devices.

[0018] 5. Dual-mode compatibility, applicable to a wide range of scenarios. Narrowband mode supports massive concurrent user access, adapting to scenarios such as IoT and satellite multiple access; broadband mode solves the phase blindness problem through four-sequence extension, enabling high-speed data transmission and balancing the needs of both low-power access and high-speed communication.

[0019] 6. Asynchronous operation without clock synchronization. This system requires no strict clock synchronization between the transmitting and receiving ends, no phase-locked loop, and no high-precision synchronous clock circuit. The receiving device only needs to perform timing resampling of the local sequence using cubic spline interpolation to compensate for ultra-large sampling frequency offsets of up to 5%, and autonomously fit the distorted time axis. Cubic spline interpolation completely preserves the high-frequency transition edges of the BPSK sequence, with no high-frequency attenuation, no Gibbs ringing distortion, and lossless orthogonality of the high-dimensional sequence throughout. This mechanism fundamentally eliminates the dependence on synchronous clocks, making it suitable for extreme asynchronous scenarios with limited crystal oscillator accuracy and huge clock drift, such as deep space communication, low-cost IoT terminals, and underwater acoustic communication. The transmitting and receiving ends can operate in completely independent clock domains.

[0020] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0021] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0022] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a structural diagram of an asynchronous high-dimensional phase-domain orthogonal and differential large-scale MIMO communication system according to an embodiment of the present invention. Detailed Implementation

[0023] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0024] This invention provides an asynchronous high-dimensional phase-domain orthogonal and differential large-scale MIMO communication system, such as... Figure 1 As shown, the system includes several transmitting devices and at least one receiving device equipped with multiple receiving antennas. The transmitting devices are deployed on the user terminal side to perform high-dimensional phase domain signal modulation and transmission. The receiving devices are deployed on the base station or receiving center side to perform demodulation and data recovery of multiple mixed signals. This scheme uses a time-division multiplexing architecture to achieve the transmission and separation of dual signals. Each symbol transmission cycle is divided into continuous reference time slots and modulation time slots, corresponding to the transmission windows of two physical transmission channels, respectively. The receiving device can directly separate the two signal components through the time slot boundaries without additional demodulation and separation operations.

[0025] This scheme is based on the slow fading channel assumption. Within a single symbol period (including the reference time slot and the modulation time slot), the amplitude and phase of the wireless channel remain approximately constant. Therefore, the channel state weights calculated from the reference time slot can be directly used for the combination of modulation time slot signals within the same symbol period.

[0026] In this embodiment, the structure and implementation principle of the transmitting device are as follows: The transmitting device includes a high-dimensional random orthogonal sequence generation module and a dual-path bit modulation module, which are used to perform high-dimensional phase domain random orthogonal modulation processing. The modules are connected and cooperated in the order of signal generation to transmission.

[0027] The core function of the high-dimensional random orthogonal sequence generation module is to generate a unique orthogonal signal basis for each access user.

[0028] The high-dimensional phase domain of this scheme refers to an N-dimensional Hilbert vector space constructed from discrete binary sequences of length N. User identifiers and bit information are carried on the spatial direction and phase state of the sequence vectors. Different users are distinguished by mutually quasi-orthogonal high-dimensional sequence vectors, which differs from the traditional fixed orthogonal resource partitioning method in the time domain, frequency domain, and code domain. When the sequence dimension N is much larger than the total number of concurrent users K, the randomly generated sequence vectors naturally possess extremely low cross-correlation characteristics, which can support the parallel differentiation of massive numbers of users.

[0029] The high-dimensional phase domain of this scheme is not a simple linear Hilbert space, but rather a two-stage construction method employing a linear sequence space and a nonlinear energy mapping: The first stage is a linear construction, using N-dimensional linear vector spaces constructed from N-length BPSK random sequences, leveraging the quasi-orthogonality of the random sequences to achieve multi-user differentiation; the second stage is a nonlinear construction, where the receiver maps the linear correlation results to the energy domain through L2 norm squared operations, forming a nonlinear decision space that converts phase coherence information into energy amplitude information, thereby achieving incoherent detection. This nonlinear construction makes the system naturally immune to channel phase fluctuations, enabling reliable demodulation without channel phase estimation, a fundamental characteristic that distinguishes it from traditional coherent MIMO.

[0030] Random binary phase shift keying sequences, also known as BPSK sequences, have each sampling point taking only two values: +1 or -1, corresponding to phase 0 and phase 1, respectively. The two-phase modulation; independent means that the two sets of sequences are generated through independent random processes, and the cross-correlation peak of the two is much lower than the autocorrelation peak, which has quasi-orthogonal characteristics.

[0031] Specifically, a linear feedback shift register can be used to generate a pseudo-random sequence, assigning a unique initial seed to each user to ensure that the sequences of different users are independent of each other. Alternatively, a true random number generator can be used to generate a true random sequence, further improving the orthogonality and anti-interception performance of the sequence. After the sequence is generated, it is stored in local memory for the modulation module to call.

[0032] For example, when the maximum number of concurrent users in the system is 200, the sequence length N can be 4096, that is, a 4096-dimensional vector space is constructed; Each user's two sets of sequences are generated through independent Bernoulli random processes. The normalized cross-correlation of any two different user sets is a zero-mean random variable with a standard deviation of approximately [missing value]. It naturally possesses quasi-orthogonal properties.

[0033] The dual-path bit modulation module is used to control two parallel physical transmission channels to complete the bit-to-sequence mapping. The two physical transmission channels are implemented using time-division multiplexing, that is, the reference time slot and the modulation time slot are divided sequentially within each symbol period. The two time slots are of equal length and each has N sampling points. The single-antenna terminal can continuously transmit the two signals by alternating time slots, and the receiving device can separate the two signal components by synchronizing the time slots.

[0034] The first physical transmission channel (corresponding to the reference time slot) constantly transmits the first set of random binary phase shift keying sequences as a benchmark reference for the receiving device to calculate the channel weight, and its output does not change with the bits to be transmitted.

[0035] The second physical transmission channel (corresponding to the modulation time slot) switches the output sequence according to the binary bits to be transmitted: when the bit to be transmitted is 1, the first set of random binary phase shift keying sequences is transmitted; when the bit to be transmitted is 0, the second set of independent random binary phase shift keying sequences is transmitted.

[0036] Specifically, in the baseband digital domain, a data gating unit is used to select the corresponding output sequence data based on the input bit value and the time slot counting signal. The output serial sequence is then transmitted through the terminal antenna after digital-to-analog conversion, up-conversion, and power amplification.

[0037] For example, if a user's bit sequence to be transmitted is 1, 0, 1, which takes three symbol periods, the first set of sequences is always output during the first half of each symbol period (reference time slot); during the second half of each symbol period (modulation time slot), the first set, the second set, and the first set of sequences are output sequentially, and the bit information is carried by switching the sequences within the modulation time slot.

[0038] In this embodiment, the structure and implementation principle of the receiving device are as follows: The receiving device includes a timing alignment module, an implicit weight calculation module, a spatial diversity merging module, and a differential detection demodulation module, used to perform multi-stage serial demodulation processing. Each module is connected sequentially according to the signal flow, with the output of the previous module serving as the input of the next, forming a complete demodulation pipeline. The receiving device first completes time slot synchronization, separates the reference time slot signal and the modulation time slot signal from the received signal, and then performs subsequent processing sequentially.

[0039] The timing alignment module is used to compensate for the time axis distortion caused by sampling frequency offset. It uses a high-order polynomial interpolation algorithm to perform timing resampling on two sets of random binary phase shift keying sequences stored locally, so that the time axis of the local sequence matches the time axis of the received distorted signal.

[0040] Sampling frequency offset refers to the deviation between the clock frequency of the transmitting end and the receiving device, which causes the sampling points of the received signal to be stretched or compressed relative to the transmitting end. For example, a 5% positive sampling frequency offset means that the sampling rate of the receiving device is 5% higher than that of the transmitting end. A transmission sequence of length N will correspond to N×1.05 sampling points in the receiving device. Without compensation, this will lead to related peak shift and energy loss.

[0041] Higher-order polynomial interpolation uses polynomial functions of degree three or higher to fit the sampling points of the original sequence. The fitted continuous curve is then resampled to obtain a new sequence that matches the distortion time axis.

[0042] Preferably, the high-order polynomial interpolation algorithm employs cubic spline interpolation. Cubic spline interpolation divides the entire sequence into multiple adjacent small intervals, fitting a curve using a cubic polynomial within each interval, ensuring the continuity of the first and second derivatives at the connection points of adjacent intervals. Compared to linear interpolation and FFT resampling, cubic spline interpolation effectively preserves the transition edge characteristics of random binary phase shift keying sequences, significantly reduces the attenuation of high-frequency components, and avoids Gibbs ringing distortion caused by FFT resampling, thus maximizing the maintenance of the inherent orthogonality of high-dimensional sequences.

[0043] Specifically, the process involves: first, coarse time slot synchronization is achieved using the sliding correlation method; then, sampling frequency offset estimation is performed based on the offset of the correlation peaks of adjacent symbols to obtain the current sampling frequency offset value and determine the resampling factor; next, three spline interpolation operations are performed on each set of random sequences stored locally to generate a resampled sequence of the corresponding length; the resampled sequence is used for subsequent cross-correlation operations.

[0044] If the system has a 5% positive sampling frequency offset, the local 4096-point reference sequence can be resampled to 4301 points (4096×1.05) through cubic spline interpolation, and then a sliding correlation can be performed with the received signal to eliminate the timing misalignment caused by the frequency offset and ensure the energy accuracy of the correlation peak.

[0045] The implicit weight calculation module is used to calculate the non-coherent channel weight of each receiving antenna corresponding to the target user using the received signal (i.e., the reference time slot signal) corresponding to the first physical transmission channel.

[0046] Incoherent channel weights refer to the aggregation coefficients obtained using only the signal energy information without extracting channel phase information. They are used to characterize the reception quality of the target user signal on a single receiving antenna. The higher the weight value, the better the signal quality of that antenna, and the higher its weight ratio during aggregation. Since the reference time slot always transmits a fixed reference sequence, the magnitude of the cross-correlation energy between the received signal and the local reference sequence directly reflects the gain of the corresponding channel for that antenna. The lower the channel gain, the smaller the correlation energy, and the lower the corresponding weight.

[0047] The formula for calculating the weight of the incoherent channel is: ,in, The incoherent channel weights for user k corresponding to the m-th receiving antenna; This refers to the nth sampling point of the reference time slot received signal acquired by the mth receiving antenna; This represents the nth sampling point of the first set of local random binary phase-shift keying sequences corresponding to user k. The cross-correlation operation of the corresponding sequences is summed, and the correlation energy value is obtained by taking the modulus and squaring. This weighting method does not require explicit channel state information estimation or channel matrix inversion. When a receiving antenna experiences Rayleigh deep fading, the channel gain approaches zero, and the corresponding correlation energy also approaches zero. The weights... It will automatically approach zero, which is equivalent to automatically shielding the receiving branch with degraded signal, without the need for additional fading detection and branch switching logic.

[0048] Specifically, for each receiving antenna, the received signal in the reference time slot is cross-correlated with the target user's local reference sequence. The square of the magnitude of the correlation peak is taken as the weight of the user corresponding to that antenna. The calculation is completed for each user and each antenna to obtain the corresponding weight matrix. For example, if the system has a total of 20 receiving antennas, and the 5th antenna encounters Rayleigh deep fading with a channel gain of only 1% of that of a normal antenna, then the calculated weight of that antenna is only one ten-thousandth of that of a normal antenna. It will hardly contribute in the subsequent combining process, thus automatically avoiding the decision error caused by deep fading.

[0049] The spatial diversity merging module is used to perform multi-antenna weighted merging on the received signal (i.e., the modulation time slot signal) corresponding to the second physical transmission channel according to the calculated noncoherent channel weight, to converge the spatial diversity gain and suppress the decision error caused by deep fading of a single antenna.

[0050] Spatial diversity combining refers to combining signals received by multiple antennas according to certain rules to improve the overall signal-to-noise ratio of the received signal. This scheme adopts the energy domain noncoherent maximum ratio combining (MRC) mechanism, which first eliminates the random phase influence of each antenna channel through squaring operation, and then combines them according to channel quality weighting to avoid mutual cancellation of branch signals with opposite phases, thereby maximizing the output signal-to-noise ratio after combining.

[0051] Specifically, the process involves: first, cross-correlation of the received signal from the modulation time slot of each antenna with two sets of local random sequences to obtain two sets of complex correlation values; then, taking the square of the modulus of the correlation value for each antenna to obtain two sets of correlation energies for a single antenna branch; next, multiplying the single-branch energy of each antenna by the weight of the incoherent channel corresponding to that antenna; finally, summing the weighted energies of all antennas to obtain two sets of combined total correlation energies, which are directly used for subsequent bit decisions. For example, in a network of 20 receiving antennas, 18 are in normal channel conditions and 2 are in deep fading. After weighted combining, the signal from the deep fading antennas hardly affects the total energy result, and the overall signal-to-noise ratio is improved by about 13dB compared to a single antenna, effectively resisting the effects of channel fading.

[0052] The differential detection and demodulation module is used to recover the original transmitted bits based on the two sets of total correlation energies output by spatial aggregation and differential decision by the magnitude of the two sets of energies.

[0053] L2 norm squared correlation energy is defined as the energy value of a signal obtained by cross-correlating the received signal with the local sequence and taking the square of the magnitude of the result. The L2 norm, or the squared norm, corresponds to energy calculation. In this scheme, the total correlation energy after merging is essentially a weighted sum of the L2 norm squared correlation energies of each antenna branch.

[0054] Decision rule: The total energy corresponding to the first set of sequences after merging is denoted as... The total energy corresponding to the second set of sequences is denoted as ;like If the current bit is 1, then the current transmitted bit is determined to be 1; otherwise, it is determined to be 0.

[0055] The formula for calculating total energy is: ; ;in, This represents the nth sampling point of the signal received in the modulation time slot of the m-th antenna. This is the second set of local random sequences corresponding to user k.

[0056] The principle of multiple access interference: The signals of other concurrent users in the system are random interference to the target user. Since both local sequences are independent and identically distributed random sequences, the cross-correlation results between the multiple access interference and the two local sequences are both zero-mean random variables, statistically having approximately equal variances and means. Therefore, the statistical distributions of the two sets of correlated energies have common mode characteristics. When making decisions based on energy differences, no systematic decision bias is introduced, avoiding decision threshold bias caused by multiple access interference. Stable demodulation performance can be maintained without additional multi-user iterative interference cancellation algorithms. It should be noted that this mechanism does not reduce the total power of multiple access interference, but rather eliminates decision bias by utilizing its statistical symmetry. In scenarios where the number of users is much lower than the sequence dimension, the overall variance of the interference is within an acceptable range.

[0057] This scheme does not use the L1 norm absolute value detection scheme because the L1 norm absolute value operation will lose the phase coherence information in the sequence correlation process, resulting in a significant reduction in the processing gain brought by high-dimensional sequences, a serious deterioration in demodulation performance, and an inability to achieve reliable communication in low signal-to-noise ratio and underdetermined scenarios.

[0058] In this embodiment, the magnitudes of the two sets of total correlation energies are directly compared using a numerical comparator, and the corresponding binary bits are output. For example, when the transmit bit is 1, the first set of sequences is transmitted in the modulation time slot, so the total energy after merging with the first set of sequences is... It will be significantly higher than The interference signals from the remaining 199 users contribute roughly the same amount to the statistical values ​​of the two energies. When comparing their magnitudes, no systematic decision bias is generated, and bit 1 can be correctly determined in the end.

[0059] As a further extension, the transmitting device also includes a broadband extension module, and the differential detection and demodulation module of the receiving device supports quaternary differential decision, jointly realizing the broadband multi-code multiplexing extension function. This extension scheme adopts the basic architecture of time-division dual-path, and each symbol period is still divided into a reference time slot and a modulation time slot: the reference time slot continuously transmits a fixed reference sequence for channel weight calculation, which is completely consistent with the narrowband mode; the modulation time slot carries high-order modulation information, allocating four pairs of independent random binary phase shift keying sequences to each code channel of a single user, corresponding to the four symbol states of QPSK (00, 01, 10, 11). In the modulation time slot, according to the currently transmitted two-bit symbol, one of the four sequences is selected for transmission. The receiving device also follows the energy domain merging process of single-antenna correlation → squaring to obtain the energy of a single branch → weighted accumulation to obtain the total energy, calculating the total correlation energy for each of the four sequences, comparing the magnitude of the four energies, and taking the symbol corresponding to the sequence with the largest energy as the decision result to recover the two-bit information.

[0060] The core reason for the phase blindness problem in traditional square-law detection is that in traditional schemes, bit information is carried by the phase sign of the same carrier (i.e., the positive and negative signs of the same sequence are reversed). After squaring, the energy of A and -A are completely equal, making it impossible to distinguish the phase state. However, in the four-sequence extended architecture of this scheme, two bits of information are carried by four independent and different sequences themselves, rather than the phase reversal of the same sequence. Each sequence corresponds to a unique symbol state. The squaring operation only changes the dimension of energy and does not change the correlation energy difference between different sequences and the received signal. Therefore, the phase ambiguity problem of traditional square-law detection is not present in the mechanism. This breaks through the limitation that incoherent detection can only support binary modulation and realizes broadband high-speed orthogonal phase-shift keying transmission.

[0061] Code channel orthogonal implementation: Multiple code reuse is achieved by assigning each code channel an independent family of random sequences, relying on the quasi-orthogonal characteristics of high-dimensional sequences to suppress inter-code channel interference; the number of code channels can be flexibly configured according to the transmission rate requirements. The higher the number of code channels, the higher the total transmission rate, but the accumulation of inter-code channel interference will also lead to a decrease in demodulation performance. In engineering, a trade-off must be made between rate and reliability.

[0062] The mapping rule is as follows: the four sets of sequences are denoted as S00, S01, S10, and S11, respectively, corresponding to the two-bit symbols 00, 01, 10, and 11. Within the modulation time slot, the corresponding sequence is selected for output based on the input two-bit data.

[0063] Specifically, a single user is allocated multiple mutually orthogonal code channels, each carrying an independent QPSK symbol. The transmitting end converts the bit stream from serial to parallel into two-bit QPSK symbols, and then maps them to the corresponding sequences. The receiving device calculates the correlation energy of four sets of sequences for each code channel, takes the symbol corresponding to the set with the highest energy as the decision result, and finally converts the bits of all code channels from parallel to serial and merges them into a complete data stream. For example, a single user is allocated 20 orthogonal code channels, and the sequence length of each code channel is N=256. The four symbol states of QPSK correspond to four independent sequences. The receiving device recovers the two bits through quaternary energy comparison, which doubles the transmission rate compared to the single code channel in binary mode.

[0064] As a preferred system parameter configuration, the length N of the random binary phase shift keying sequence is ≥4096, the number of concurrent users K is ≥200, and the number of receiving antennas M is ≤20, achieving asynchronous stable demodulation with an underdetermined antenna ratio of K / M ≥10. The phase domain processing gain described in this scheme is the ratio of the peak autocorrelation energy to the mean cross-correlation energy of the high-dimensional sequence, essentially representing the signal-to-noise ratio improvement factor brought about by high-dimensional spread spectrum, determined by the sequence length N, and expressed in decibels. In narrowband operating mode, when N=4096, the phase domain processing gain is approximately 36.1dB, which is not less than 36dB.

[0065] This system achieves multi-user differentiation through the random orthogonal characteristics of the high-dimensional phase domain, effectively breaking through the spatial degree-of-freedom capacity limitation of traditional MIMO systems. It does not require channel matrix inversion or explicit channel state information estimation throughout the process. It can stably demodulate under extreme conditions of a 10:1 extremely underdetermined antenna ratio, 0dB limiting signal-to-noise ratio, 5% large sampling frequency offset, and Rayleigh deep fading. It also supports dual modes of narrowband massive access and broadband high-speed transmission, combining the advantages of high access capacity, strong robustness, and low computational complexity.

[0066] The specific implementation process and performance of the present invention will be further described in detail below, using two typical scenario examples: Example 1: Narrowband Extremely Understressed Massive Access Scenario: This embodiment is designed for 6G ultra-massive machine-type communication scenarios, verifying the system's concurrent demodulation performance under extremely unstable configuration ratios.

[0067] System parameter configuration: High-dimensional sequence dimension: Sequence length N=4096, sequence value is +1 / -1, generated by independent Bernoulli random processes; Time slot structure: Each symbol period is equally divided into a reference time slot and a modulation time slot, and each time slot has a length of 4096 sampling points; System scale: concurrent single-antenna users K=200, base station receiving antennas M=20, understaffing ratio is 10:1; Channel environment: 20×200 independent and identically distributed Rayleigh flat fading channel, with channel gain generated independently for each antenna and each user, and Doppler frequency offset set to 10Hz to meet the slow fading condition; superimposed 0dB additive white Gaussian noise (AWGN), i.e., the average signal power is equal to the average noise power; Asynchronous condition: The receiving device has a global sampling frequency offset of 5%.

[0068] The sending end execution process is as follows: (1) The high-dimensional random orthogonal sequence generation module generates two independent 4096-point random binary phase shift keying sequences for 200 users respectively. Each user corresponds to a unique sequence combination, and the sequences between users are quasi-orthogonal to each other; (2) The dual-path bit modulation module controls the signal output according to the time slot: within the reference time slot of each symbol period, all users output their own first set of sequences at a constant value; within the modulation time slot of each symbol period, the output is switched according to the current bit to be transmitted: when the bit is 1, the first set of sequences is output, and when the bit is 0, the second set of sequences is output; (3) The transmission signals of 200 users are superimposed through the wireless channel and arrive at the receiving device.

[0069] The receiving device execution process is as follows: (1) Time slot synchronization and timing alignment: The receiving device first completes coarse time slot synchronization through sliding correlation, and separates the reference time slot signal and the modulation time slot signal from the received signal of each antenna; then, based on the correlation peak offset of adjacent symbols, it completes the 5% sampling frequency offset estimation, and performs cubic spline interpolation resampling on all user sequences stored locally. The resampling factor is 1.05 to match the timing stretching caused by the sampling frequency offset; the resampled sequence retains the main transition edge characteristics, and the orthogonality loss can be ignored; (2) Implicit weight calculation: For each user, the non-coherent channel weight is calculated antenna by antenna: the reference time slot received signal of the m-th antenna is cross-correlated with the first set of local sequences of user k, and the square of the magnitude of the correlation peak is taken as the weight. ; The weight of the antenna that experiences deep fading automatically approaches 0; (3) Spatial grouping and merging: For each user, the received signal of the modulation time slot of each antenna is cross-correlated with the two sets of local sequences, the modulus squared is taken to obtain the correlation energy of a single antenna, and then multiplied by the weight of the corresponding antenna and accumulated to obtain the total correlation energy of the two sets after merging. and (4) Differential detection demodulation: directly compare the magnitudes of the total correlation energy of the two groups. The decision is bit 1, otherwise the decision is bit 0.

[0070] Actual performance: This embodiment is based on Based on simulation sample statistics, the original physical layer bit error rate of the system under the aforementioned extreme conditions is approximately 0.04%. All errors are sporadic errors caused by isolated single-frame deep fading, with no continuous error clusters. After using the 5G NR standard 1 / 2 rate LDPC error correction code, the system bit error rate is lower than... It can meet the high reliability transmission requirements of scenarios such as deep space and tactical communication.

[0071] Example 2: 40MHz Broadband Single-User High-Speed ​​Transmission Scenario: This example is for a broadband high-speed transmission scenario to verify the system's broadband expansion performance.

[0072] System parameter configuration: Phase domain dimension: Single code channel sequence length N=256, which can provide a processing gain of approximately 24.1dB to suppress multipath inter-symbol interference; Time slot structure: The symbol period of each code channel is equally divided into a reference time slot and a modulation time slot, and each time slot has a length of 256 sampling points; System configuration: single user, 20 receiving antennas, channel bandwidth 40MHz, single user allocated 20 orthogonal code channels; Modulation method: Each code channel adopts four-sequence quadrature phase shift keying modulation, and adopts the time-division architecture of reference time slot + modulation time slot; the four independent sequences correspond to the four symbol states of QPSK respectively, and the mapping rule is: 00 corresponds to S00, 01 corresponds to S01, 10 corresponds to S10, and 11 corresponds to S11; the corresponding sequence is selected for transmission according to the symbol within the modulation time slot; Channel conditions: 0dBAWGN with multipath Rayleigh fading, 5 paths, maximum delay spread is [value missing]. .

[0073] The transmitting end execution process is as follows: (1) The broadband extension module allocates four sets of two independent 256-point random binary phase shift keying sequences to each code channel. The sequence groups of different code channels are independent of each other, and the interference between code channels is suppressed by relying on the high-dimensional quasi-orthogonal characteristics; (2) The bit stream to be transmitted is serialized and parallelized into two-bit QPSK symbols, with one symbol corresponding to each code channel; a fixed reference sequence is sent in the reference time slot of each symbol period, and the corresponding sequence is selected for output according to the symbol state in the modulation time slot; (3) The signals of 20 code channels are superimposed in parallel and transmitted, and arrive at the receiving device after passing through the channel.

[0074] The receiving device execution process is as follows: (1) Timing alignment: Code channel synchronization and time slot synchronization are completed by sliding correlation, and sampling frequency offset and time delay are compensated by cubic spline interpolation; (2) Implicit weight calculation: Based on the reference time slot signal of each code channel, the non-coherent channel weight of each antenna is calculated; (3) Spatial aggregation and merging: For the received signal of the modulation time slot of each code channel, cross-correlation is performed with four sets of local sequences and the modulus square is taken to obtain the single antenna energy, which is then multiplied by the corresponding antenna weight and accumulated to obtain the total correlation energy of the four sets; (4) Differential detection and demodulation: For each code channel, the magnitude of the total correlation energy of the four sets is compared, and the symbol corresponding to the maximum energy value is taken as the decision result to restore the two-bit information; Finally, the bits of all code channels are converted from parallel to serial to restore the complete transmission data stream.

[0075] Throughput calculation and measured performance: Throughput calculation process: Single code channel symbol rate = channel bandwidth / single time slot sequence length = 40MHz / 256 = 156.25ksps; each symbol carries 2 bits, and a total of 20 code channels run in parallel, so the original physical layer rate = 156.25ksps × 2bit × 20 = 6.25Mbps; after using 3 / 4 code rate LDPC error correction code, the effective throughput = 6.25Mbps × 3 / 4 = 4.6875Mbps, approximately 4.69Mbps.

[0076] Actual performance: Raw bit error rate is lower than 0dB at a signal-to-noise ratio. The bit error rate curve exhibits a standard waterfall shape, enabling stable broadband communication.

[0077] Both the transmitting and receiving devices in this solution can be implemented using digital signal processors (DSPs), field-programmable gate arrays (FPGAs), or application-specific integrated circuits (ASICs). The core operations only include basic linear operations such as sequence correlation, summation of squares, interpolation, and weighted accumulation. The hardware implementation is simple and easy to mass-produce.

[0078] As a further optimization, the receiving device also includes a phase-domain adaptive modulation module for executing a hysteresis closed-loop parameter modulation process. The entire modulation process is executed serially in the order of blind channel sensing → parameter decision → parameter synchronization → closed-loop verification, specifically including: Based on the incoherent channel weights and differential detection energy results output for each symbol period, the equivalent decision signal-to-noise ratio (SNR) and soft estimation bit error rate (BER) of the current system are statistically obtained. The core function of this step is to acquire real-time channel quality data without consuming additional transmission resources. Here, blind channel sensing refers to evaluating channel quality entirely based on the existing demodulation results of the receiving device, without requiring the transmitter to insert dedicated pilots or training sequences, thus avoiding additional spectral overhead. This step outputs two core metrics: the equivalent decision SNR and the soft estimation BER.

[0079] The equivalent decision signal-to-noise ratio (SNR) refers to the equivalent SNR at the decision input after spatial diversity and phase domain processing, and directly determines the bit error rate level of demodulation.

[0080] Calculation formula: ,in,: The sum of the incoherent channel weights of all receiving antennas for the target user corresponds to the total energy of the combined signal correlation. The noise power at a single sampling point can be obtained through variance statistics of the sidelobe region in the correlation calculation results; N is the currently used sequence length; M is the total number of receiving antennas. In this formula, the numerator is the total correlated energy of the combined signal, and the denominator is the mean of the total correlated energy of the combined noise. Both are energy units, and the ratio is a dimensionless value. After taking the logarithm, the unit is decibel (dB), which conforms to the physical definition of signal-to-noise ratio. After cross-correlation of the noise with a random sequence of length N, the mean square energy of the output noise is... The independent noise energy of multiple antennas can be directly accumulated, and the calculation logic perfectly matches the operational characteristics of incoherent detection in the energy domain. This formula is derived for low-load scenarios where thermal noise is limited. When the number of concurrent users in the system is large and multiple access interference is dominant, the estimated value of the equivalent signal-to-noise ratio will be too optimistic. In this case, the soft estimation bit error rate is used as the main basis for adjustment to ensure control accuracy.

[0081] Specifically, the process involves: directly reading the antenna weight values ​​output by the implicit weight calculation module and summing them to obtain the total signal energy; statistically analyzing the noise variance through sampling in the non-peak region of the relevant results; substituting these values ​​into the equivalent decision signal-to-noise ratio formula to perform logarithmic calculations and outputting the equivalent signal-to-noise ratio value. For example, if the current sequence length N=4096, the receiving antenna M=20, the noise power at a single sampling point is 1mW, and the weight sum of the 20 antennas is 20×40962mW, substituting these values ​​into the formula yields an equivalent decision signal-to-noise ratio of approximately 36.1dB, which is consistent with the theoretical value of the phase domain processing gain.

[0082] Soft-estimated bit error rate refers to the bit error rate estimate obtained by relying solely on the relative relationship between two sets of related energies without needing to know the transmitted bits, and is used for closed-loop verification.

[0083] Calculation formula: ,in, , These are the two sets of total correlation energies output by differential detection; It is the right-tail function of the standard normal distribution, used to calculate the bit error probability under Gaussian noise. That is, under the premise of the same total received energy, the larger the difference between the two sets of energy, the higher the confidence of the decision and the lower the estimated bit error rate. Under the premise of the same energy difference, the higher the total energy, the stronger the signal component, and the higher the confidence of the decision. The normalized decision margin reflects the bit error probability, which is consistent with the statistical characteristics of energy detection under Gaussian noise.

[0084] Specifically, the process involves: reading two sets of energy values ​​output by the differential detection module, calculating the normalized decision margin, and obtaining the bit error probability corresponding to the Q function using a lookup table; performing a moving average on the estimated values ​​for multiple consecutive symbols to obtain a smoothed soft-estimated bit error rate, such as the bit error rate under a certain symbol. =100, =20, the total energy is 120, and the normalized decision margin is approximately 2.58, corresponding to a Q function value of approximately 0.005, meaning the estimated bit error rate of this symbol is approximately 0.5%.

[0085] Using the decision signal-to-noise ratio (SNR) threshold corresponding to the preset target bit error rate as a benchmark, and combining the real-time equivalent decision SNR, the optimal sequence length is calculated using the optimal solution formula in the phase domain dimension. In narrowband mode, the maximum number of concurrent users is derived based on the optimal sequence length; in wideband mode, the optimal number of code channels is derived by combining quasi-orthogonal load constraints. After the sequence length is adjusted, the corresponding time slot length and symbol rate are synchronized. Parameter switching adopts a hysteresis step mechanism, updating parameters only when the channel quality continuously exceeds the threshold range. This step uses the target bit error rate as a constraint to calculate the optimal system parameters under the current channel, achieving a dynamic trade-off between reliability and transmission efficiency. After the sequence length is adjusted, the number of sampling points in the corresponding reference time slot and modulation time slot changes synchronously, and the symbol period and symbol rate are adjusted in conjunction with the sequence length, maintaining the system chip rate and channel bandwidth matching throughout the process.

[0086] Decision signal-to-noise ratio threshold This refers to the minimum equivalent signal-to-noise ratio (SNR) required at the input terminal to achieve a preset target bit error rate. It is determined through system simulation or actual measurement and serves as the benchmark for parameter decision-making. If the target original bit error rate requirement is... The corresponding calibrated decision signal-to-noise ratio threshold is approximately 9 dB; if the target bit error rate is... The corresponding threshold is approximately 14 dB.

[0087] The optimal solution formula for phase domain dimension is used to calculate the optimal sequence length that meets the target bit error rate requirement. It minimizes the sequence dimension while ensuring reliability, freeing up resources to increase capacity. The calculation formula is as follows: ,in, The length of the sequence currently in use is the calculation basis for dimension adjustment; The engineering margin factor, typically 3dB, is used to reserve a certain performance margin to combat instantaneous channel fluctuations. `round` performs rounding, with the final output being an integer power of 2. This is intended to adapt to the parallel computing architecture and addressing design of digital hardware, reducing hardware implementation difficulty. It follows the basic rule that the channel input signal-to-noise ratio + phase domain processing gain ≥ target threshold, where the phase domain processing gain and sequence length satisfy... The correspondence. Current equivalent signal-to-noise ratio. It is the reference length The measured values ​​can be used to inversely deduce the sequence length required to meet the target threshold; the lower the channel signal-to-noise ratio (SNR), the higher the required phase domain processing gain, and the longer the corresponding optimal sequence length; the higher the channel SNR, the shorter the required sequence length. Specifically: read the current sequence length. The theoretically optimal sequence length is calculated by substituting the real-time equivalent signal-to-noise ratio, a preset target threshold, and engineering margin into the formula, and then aligned to the nearest power of 2 for output. For example, the current baseline sequence length... =4096, the equivalent decision signal-to-noise ratio was measured. =24 dB, target threshold =9 dB, engineering margin =3dB; Substituting this into the calculation, the theoretical length is approximately 258. After aligning to powers of 2, the optimal sequence length is 256.

[0088] The formula for calculating the maximum concurrent users in narrowband mode is as follows: ,in, It is a quasi-orthogonal load factor, with a typical value range of 0.05 to 0.1, used to ensure that the cross-correlation interference between users is within an acceptable range; This is for floor function operations. A shorter sequence length allows for a larger number of concurrent users, dynamically maximizing access capacity, such as the optimal sequence length. =4096, Loading factor =0.05, then the maximum number of concurrent users is 204.8, which is rounded down to 204 users.

[0089] The formula for deriving the optimal number of code channels in wideband mode is as follows: ,in, This is the code channel orthogonality load factor, typically ranging from 0.08 to 0.15, used to constrain cross-correlation interference between code channels within an acceptable range. Multiplexing relies on the quasi-orthogonality of sequences to distinguish code channels, and it involves parallel transmission at the same frequency. The more code channels there are, the higher the total transmission rate. Therefore, under quasi-orthogonal interference constraints, maximizing the number of code channels will achieve the highest possible rate.

[0090] Total rate calculation: The symbol rate per code channel is Each symbol carries bits, total throughput is ;in, The system chip rate is determined by the channel bandwidth. Modulation order, in binary mode =2, in quaternary QPSK mode =4, such as =256, chip rate =40 Mcps, modulation order =4, load factor =0.1, then the optimal number of code channels is 25; the corresponding total physical layer rate is approximately 25×(40M / 256)×2≈7.81 Mbps.

[0091] The hysteresis stepping mechanism refers to setting two threshold ranges for parameter switching, one high and one low. Parameters are only updated when the channel quality continuously exceeds the boundary of the threshold range, avoiding frequent parameter jumps caused by fast fading. Specifically, an upward and downward threshold are set, with a hysteresis range of approximately 3dB between them. The sequence length is shortened only when the equivalent signal-to-noise ratio (SNR) is higher than the upward threshold for L consecutive symbol periods; the sequence length is increased only when the equivalent SNR is lower than the downward threshold for L consecutive symbol periods. For example, the upward threshold is set to 0dB and the downward threshold is set to -3dB. When the SNR briefly fluctuates to 1dB, no switching is triggered; parameter adjustment is only performed when the SNR is higher than 0dB for 10 consecutive symbols, avoiding parameter jitter caused by instantaneous fading.

[0092] The updated sequence configuration parameters are sent to all transmitting devices via the downlink broadcast link. The transmitting and receiving ends synchronously switch the generation configuration of the high-dimensional random orthogonal sequence based on the unified superframe time slot boundary. When the sequence length is switched, the time slot length and symbol rate take effect synchronously. This step is used to achieve parameter configuration alignment at both ends of the transmitting and receiving ends to ensure the correctness of demodulation.

[0093] The downlink broadcast link refers to the downlink channel through which the base station sends broadcast signaling to all terminals. It is used to distribute system configuration parameters without sending them individually to each user, resulting in low signaling overhead.

[0094] Unified time slot boundaries refer to the agreement between the transmitting and receiving ends to synchronously switch parameters at the start of a specific symbol period, avoiding demodulation errors caused by asynchronous parameter switching. When the sequence length is switched, the time slot length and symbol rate change synchronously, and the changes take effect synchronously at both the transmitting and receiving ends according to the agreed superframe boundaries.

[0095] Specifically: the receiving device encapsulates the updated sequence length, number of code channels, and other configuration information into broadcast signaling and broadcasts it through the downlink channel; after receiving the signaling, all transmitting devices synchronously update the configuration parameters of the sequence generation module at the agreed start time of the next superframe to complete the parameter switching. For example, if the system agrees that every 1024 symbol periods constitute a superframe, the parameter switching is only performed at the start time of the superframe; the receiving device sends out new parameters within the current superframe, and all terminals synchronously switch to the new sequence length and time slot configuration at the start of the next superframe.

[0096] After the parameters are updated, the soft-estimated bit error rate is continuously calculated. If the target bit error rate requirement is not met, the sequence dimension is fine-tuned step by step to form a continuous closed-loop control. This step is used to verify the actual effect of the parameter adjustment, form a closed-loop feedback, and ensure that the system performance always meets the requirements.

[0097] Step-by-step fine-tuning refers to adjusting the sequence length in small increments by a fixed step size if the bit error rate still does not meet the requirements after parameter adjustment, gradually converging to the optimal value and avoiding performance fluctuations caused by large adjustments.

[0098] Specifically: After parameter updates, the soft-estimated bit error rate (BER) within the sliding window is continuously calculated; if the BER of multiple consecutive windows is higher than the target value, the sequence length is increased by a step size of 2; if the BER is much lower than the target value, the sequence length is decreased by a step size until performance and efficiency are balanced. For example, if the estimated BER after parameter updates is... Exceeding the target If the sequence length is increased from 32 to 64, the performance is checked again until the bit error rate meets the requirements.

[0099] The beneficial effects of the above technical solution are: compared with a system with fixed parameters, it can increase the access capacity or transmission rate several times when the channel conditions are good, and automatically improve the anti-interference capability when the channel deteriorates, thus greatly improving the overall performance and scenario adaptability of the system under dynamic channels.

[0100] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. An asynchronous high-dimensional phase-domain orthogonal and differential massive MIMO communication system, comprising a plurality of transmitting devices and at least one receiving device configured with multiple receiving antennas, characterized in that, The transmitting device includes a high-dimensional random orthogonal sequence generation module and a dual-path bit modulation module, used to perform high-dimensional phase-domain random orthogonal modulation processing, wherein: The high-dimensional random orthogonal sequence generation module is used to assign two sets of independent random binary phase shift keying sequences of length N to each user to be accessed. The sequence length N is greater than the maximum number of concurrent users in the system. The random binary phase shift keying sequences form an N-dimensional linear Hilbert phase space, and multi-user differentiation is achieved by relying on the quasi-orthogonality of random sequences. The dual-path bit modulation module is used to control two physical transmission channels using a time-division multiplexing architecture. Each symbol period is divided into a reference time slot and a modulation time slot of equal length. The first physical transmission channel corresponds to the reference time slot and transmits a first set of random binary phase shift keying sequences as a reference sequence. The second physical transmission channel corresponds to the modulation time slot and selects to transmit either the first set of random binary phase shift keying sequences or the second set of independent random binary phase shift keying sequences according to the bits to be transmitted. The receiving device includes a timing alignment module, an implicit weight calculation module, a spatial subset merging module, and a differential detection and demodulation module, used to perform multi-level serial demodulation processing, wherein: The timing alignment module is used to acquire multiple mixed noisy received signals, and to perform timing resampling on two sets of random binary phase shift keying sequences stored locally using a high-order polynomial interpolation algorithm to compensate for the time axis distortion caused by sampling frequency offset, thereby achieving asynchronous clockless alignment. The implicit weight calculation module is used to calculate the incoherent channel weight of each receiving antenna corresponding to the target user using the received signal corresponding to the first physical transmission channel. The spatial diversity merging module is used to perform multi-antenna weighted merging on the received signal corresponding to the second physical transmission channel according to the non-coherent channel weight, and to aggregate the spatial diversity gain. The differential detection and demodulation module is used to calculate the L2 norm squared correlation energy of the weighted and merged second physical transmission channel signal and the two sets of local random two-phase shift keying sequences, respectively. The phase domain linear correlation result is converted into the energy domain decision space through L2 norm squared nonlinear mapping, and the original transmission bits are recovered by differential decision based on the numerical magnitude of the two sets of energy.

2. The asynchronous high-dimensional phase-domain orthogonal and differential large-scale MIMO communication system according to claim 1, characterized in that, The higher-order polynomial interpolation algorithm is a cubic spline interpolation algorithm; The cubic spline interpolation algorithm preserves the high-frequency transition edges of random binary phase shift keying sequences during time-series resampling.

3. The asynchronous high-dimensional phase-domain orthogonal and differential large-scale MIMO communication system according to claim 1, characterized in that, The formula for calculating the non-coherent channel weight is as follows: ,in, The incoherent channel weights for user k corresponding to the m-th receiving antenna are: The signal received by the first physical transmission channel is collected by the m-th receiving antenna. The first set of local random binary phase shift keying sequences corresponding to user k; The incoherent channel weights corresponding to receiving antennas experiencing Rayleigh deep fading automatically approach zero.

4. The asynchronous high-dimensional phase-domain orthogonal and differential large-scale MIMO communication system according to claim 1, characterized in that, The multiple access interference generated by other concurrent users contributes an equal amount of common-mode background noise to the two sets of L2 norm squared correlation energy. The physical cancellation of the global multiple access interference is achieved by comparing the difference between the two sets of energy. The demodulation processing of the receiving device does not employ the L1 norm absolute value detection scheme.

5. The asynchronous high-dimensional phase-domain orthogonal and differential large-scale MIMO communication system according to claim 1, characterized in that, The transmitting device further includes a broadband extension module for performing broadband multi-code multiplexing extension processing, wherein: The broadband extension module is used to allocate four sets of pairwise orthogonal random binary phase shift keying sequences to each code channel of a single user. The two physical transmission channels select the corresponding in-phase or out-of-phase sequences for transmission according to the orthogonal phase shift keying double-bit symbols. The differential detection and demodulation module of the receiving device is also used to perform L2 norm square law differential detection on the four sets of sequences respectively, to eliminate the phase blindness defect of square law detection and realize broadband high-speed orthogonal phase shift keying transmission.

6. The asynchronous high-dimensional phase-domain orthogonal and differential large-scale MIMO communication system according to claim 1, characterized in that, The length N of the random binary phase shift keying sequence is ≥4096, the number of concurrent users K is ≥200, the number of receiving antennas M is ≤20, and asynchronous stable demodulation is achieved under an underdetermined antenna ratio of K / M ≥10. In narrowband operating mode, the phase domain processing gain is no less than 36dB.

7. A transmitting apparatus for use in the asynchronous high-dimensional phase-domain orthogonal and differential large-scale MIMO communication system as described in claim 1, characterized in that, The high-dimensional random orthogonal sequence generation module and the dual-path bit modulation module are configured to perform the asynchronous high-dimensional phase-domain random orthogonal modulation processing described in claim 1.

8. A receiving device for a large-scale MIMO communication system based on asynchronous high-dimensional phase-domain orthogonality and difference as described in claim 1, characterized in that, The timing alignment module, implicit weight calculation module, spatial subset merging module, and differential detection demodulation module are configured to perform the multi-level serial demodulation process described in claim 1.

Images