An OFDM oversampling cyclic prefix based environment sensing method
By utilizing oversampled cyclic prefix and random strategy in OFDM system and combining with orthogonal approximate message passing algorithm, the problem of unutilized cyclic prefix environmental information is solved, efficient environmental perception reconstruction is achieved, and perception accuracy is improved.
Patent Information
- Application Number
- CN202311027643.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-08-14
- Publication Date
- 2025-10-10
- Estimated Expiration
- 2043-08-14
AI Technical Summary
In existing OFDM systems, the environmental information in the cyclic prefix is not utilized and the measurement matrix has poor randomness, resulting in poor environmental perception performance.
By utilizing OFDM oversampled cyclic prefix, combined with random strategy and dimension-by-dimension orthogonal approximate message passing algorithm, a channel model is established and the positions of environmental scattering points are reconstructed. The randomness and reconstruction performance of the measurement matrix are improved through compressed sensing method.
Without changing the existing communication system, the accuracy and reconstruction performance of environmental perception are significantly improved, and the perception effect under the number of OFDM symbols is enhanced.
Smart Images
Figure CN117221060B_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of wireless communications, and in particular relates to an environment perception method based on OFDM oversampling cyclic prefix. Background Art
[0002] Environmental perception, which extracts environmental information from wireless signals, is a key technology in Integrated Sensing and Communication (ISAC). This includes not only high-precision positioning but also detailed environmental imaging, gesture and posture recognition, and enhanced human perception. These key technologies will greatly expand the service scenarios of communication networks. Therefore, expanding environmental perception capabilities within existing network architectures is urgent. This approach can both save hardware development resources and maximize compatibility with existing hardware devices.
[0003] The CP in an OFDM system contains rich environmental information, and the upsampling and downsampling process provides a way to observe it with a wider bandwidth during actual transmission in hardware devices. The cyclic prefix, serving as a guard interval between OFDM symbols, mitigates inter-symbol interference (ISI) caused by multipath and ensures the integrity of the signal's spectral structure. This also results in highly correlated data within the CP. The CP contains rich environmental information, but this information is removed before the FFT transform at the receiver and thus goes unused. Furthermore, in actual system transmission, the sampling rate of the analog-to-digital converter / digital-to-analog converter (ADC / DAC) is often greater than the Nyquist bandwidth of the signal, necessitating upsampling and downsampling to match this difference in transmission rate. This process has the potential to provide higher-bandwidth observations for environmental perception. At the receiver, the original oversampled signal can be obtained from the ADC hardware without downsampling or low-pass filtering. However, when the CP is used to construct a measurement matrix for environmental perception, data correlation can lead to reduced randomness in the matrix. Furthermore, the unique attenuation characteristics of the channel can also degrade the performance of the measurement matrix, making reconstruction and solution difficult. Summary of the Invention
[0004] To address the shortcomings of the aforementioned prior art, the present invention provides an environmental perception method based on OFDM oversampled cyclic prefixes. This method enables environmental perception at the receiving end based on the oversampled CP in the transmitted OFDM signal in a universal SISO transceiver mode. This method utilizes the naturally existing oversampled cyclic prefixes in existing OFDM communication systems for perception, achieving integrated perception and communication without making significant changes to the existing communication system. Considering the poor performance of the measurement matrix constructed using cyclic prefixes for environmental perception, two randomized strategies and a dimension-by-dimension orthogonal approximate message passing algorithm are proposed to efficiently solve the environmental perception reconstruction problem.
[0005] The purpose of the present invention is achieved through the following technical solutions:
[0006] An environment perception method based on OFDM oversampling cyclic prefix is proposed, which includes the following steps:
[0007] According to the positional relationship between the transceiver and the environmental scattering surface, a channel model is established based on OFDM and with the parameters determined;
[0008] Determining the location of an environmental scattering point corresponding to each channel in the channel model on an environmental scattering surface using an ellipse cluster;
[0009] Upsampling the demodulated OFDM signal to obtain a desired oversampled cyclic prefix sequence, wherein the OFDM signal includes location information of environmental scattering points;
[0010] Using the oversampled cyclic prefix sequence and based on the channel model, establishing an environment perception model for solving the location of the environment scattering point, wherein the environment perception model involves the location problem of the environment scattering point;
[0011] Two random strategies are employed to improve the randomness of a measurement matrix composed of an oversampled cyclic prefix sequence in the environmental perception model, thereby converting the environmental scattering point location problem into a compressed sensing reconstruction problem. One random strategy processes the orthogonal frequency division multiplexing signal to be transmitted using a random symbol modulation scheme to increase the independence of data between each dimension of the measurement matrix. The other random strategy randomly extracts different measurement matrices to form a new measurement matrix to reduce the dimensionality of the original measurement matrix.
[0012] The compressed sensing reconstruction problem is solved by using the orthogonal approximate message passing method for dimension-by-dimension processing to obtain the reconstruction result of the environmental scatterer.
[0013] Furthermore, establishing the channel model specifically includes the following steps:
[0014] Step S11: The locations of the transmitter and receiver are known, and the distance is d. LOS The transmitted OFDM signal has a bandwidth of B and a number of fast Fourier transform points of N. fft , the cyclic prefix length is N CP , time domain sampling point interval ΔT = 1 / B;
[0015] Step S12: The propagation channel consists of two parts: the direct path from the transmitter to the receiver, i.e. the distance is equal to d LOS , and the indirect path from the transmitter to the ambient scattering surface and then to the receiver. The amplitude and phase of each channel in the indirect path are:
[0016]
[0017] where is the normalized gain of the transmit-receive antenna, j represents the imaginary unit, λ and f c represent the wavelength and frequency of the signal, σ i represents the radar cross section of the ith scatterer; d i = d TS + d SR represents the distance from the transmitter to the ith scatterer on the environment scattering surface d TS and the distance from the scatterer to the receiver d RS ; Δt i = (i-1) / f adc , represents the time delay difference between the ith non-line-of-sight channel and the line-of-sight channel, t LOS represents the propagation time delay of the line-of-sight, f adc is the sampling rate of the device digital-to-analog converter; channel response vector where i = 2, …, N+1, N = UN CP , U represents the signal oversampling ratio, and h1 represents the channel response of the line-of-sight.
[0018] Further, the determination of the position of the environment scattering point corresponding to each channel in the channel model on the environment scattering surface using the ellipse cluster includes the following steps:
[0019] Step S21, x s represents the environment scattering point corresponding to each channel, 1 if there is a scatterer, and 0 if there is not, and the entire vector is represented as
[0020]
[0021] The length of the environment scattering point vector and the channel coefficient vector is equal, i.e. M = N+1;
[0022] Step S22, according to the determined x s and h, the ith environment scattering point is located on an ellipse with a focal length of the focal distance between the transmitter and the receiver, According to the size L l of the environment scattering surface, L w and the center position, the set Γ surface of the environment scattering points on the environment scattering surface can be determined, so that two spatial constraints of the ith environment scattering point can be obtained:
[0023]
[0024] where P i , P TX , P RX represent the spatial positions of the ith environment scattering point, the transmitter and the receiver, respectively.
[0025] Furthermore, upsampling the demodulated OFDM signal to obtain a desired oversampled cyclic prefix sequence specifically includes the following steps:
[0026] The receiving end's digital-to-analog converter does not undergo low-pass filtering, and the OFDM signal obtained is demodulated. The demodulated symbol is subjected to inverse fast Fourier transform and then oversampled to obtain the oversampled cyclic prefix sequence corresponding to the kth OFDM symbol sent. Where dimension N = UN CP ; The received OFDM symbol is directly oversampled to obtain the oversampled cyclic prefix corresponding to the received k-th OFDM symbol as follows:
[0027] Furthermore, establishing an environmental perception model for solving the location of environmental scattering points specifically includes the following steps:
[0028] Assuming that the maximum channel delay spread is exactly equal to the cyclic prefix duration, the cyclic prefix of the kth OFDM symbol will be affected by the data of equal length at the end of itself and the previous OFDM symbol, and its value is equal to the cyclic prefix of the k-1th OFDM symbol. Therefore, it can be equivalent to the cyclic prefix of the kth OFDM symbol being affected by the cyclic prefixes of the kth and k-1th OFDM symbols; the oversampled cyclic prefix satisfies the same influence, and the number of sampling points is U times the cyclic prefix of the OFDM symbol. The influence of the channel can be expressed as:
[0029]
[0030] in represents Gaussian additive white noise, r k is the received oversampled cyclic prefix sequence, It represents a cyclic prefix pair sequence consisting of the k-1th and kth oversampled cyclic prefixes, which is specifically expressed as Used to represent the shift of the bitwise multiplication of the channel coefficients and the scattering point vector:
[0031]
[0032] In order to facilitate the solution of the reconstruction target, the relationship is rewritten according to matrix multiplication as follows:
[0033] r k =Φ k h⊙x s +n
[0034] Where ⊙ represents the Hadamard product, that is, the bitwise multiplication of vectors, Φ k Indicated by The cyclic shift matrix composed of the data elements in is specifically expressed as follows:
[0035]
[0036] Assuming the channel is quasi-static, sym OFDM symbols, namely K sym -1 The total number of observations is N×K sym , the complete perception model composed of multiple oversampled cyclic prefixes is as follows:
[0037]
[0038] in K sym The complete observation vector consisting of oversampled cyclic prefixes, Indicated by K sym indivual The measurement matrix composed of k = 1, ..., K sym -1, K sym .
[0039] Furthermore, converting the environmental scattering point location problem into a compressed sensing reconstruction problem specifically includes the following steps:
[0040] According to the perception model of multiple oversampled cyclic prefix sequences, random modulation and random decimation strategies are used to improve the properties of the measurement matrix.
[0041] Random modulation: From the generation process of OFDM symbols, we know that the data in the cyclic prefix is related to the data in the frequency domain. Therefore, in order to ensure that the measurement matrix Φ composed of different oversampled cyclic prefixes is k To ensure the independence of data, each transmitted OFDM symbol has a different symbol mapping method to generate data. The specific strategy is to randomly select from four mapping methods: BPSK, QPSK, 16QAM, and 64QAM.
[0042] Random sampling: Since the measurement matrix in the oversampling cyclic prefix perception model is composed of The cyclic shift is generated, and each row has a high correlation. Therefore, the measurement matrix Φ formed from each oversampled cyclic prefix is k Randomly select a row from the matrix to form a new measurement matrix
[0043] Under the above two strategies, a new oversampling cyclic prefix perception model can be obtained, which is expressed as
[0044]
[0045] z=h⊙x s
[0046] in, represents the new observation vector, Represented by random Φ k Each row in the new measurement matrix is formed, Represents the new reconstruction vector, which is composed of the Hadamard product of the channel coefficient and the reconstructed environmental scattering point vector; by estimating z and then removing the corresponding elements in h dimension by dimension, the recovered x is obtained based on the posterior probability judgment s .
[0047] Furthermore, solving the compressed sensing reconstruction problem using the dimension-by-dimension orthogonal approximate message passing method specifically includes the following steps:
[0048] Step S61: first set the environmental perception prior probability, and let the reconstructed environmental scattering point vector x s If the Bernoulli distribution is satisfied and the parameter is P, then the sparsity of z is also P, and the elements satisfy Pr(z i =0) = P, Pr(z i =h i )=1-P, Pr represents the probability of an element taking a certain value.
[0049] Step S62: Initialize the parameters of the approximate message passing algorithm. The input parameters include the measurement matrix Observation vector Channel response coefficient h, Bernoulli distribution parameter P and noise variance σ at different locations 2 The standard orthogonal approximate message passing algorithm includes a linear estimator (LE) and a nonlinear estimator (NLE). The input parameters start from the LE iteration, and the initial mean The variance is where k∈1,…,N+1,
[0050] Step S63, each iteration calculates the estimated mean of the minimum mean square error (LMMSE) of the LE part and variance
[0051]
[0052]
[0053] Among them S t Therefore is a diagonal matrix with diagonal elements. "° / " indicates Hadamard division by dimension, estimating the variance for A vector of diagonal elements, represents the new observation vector in step 6;
[0054] Step S64, complete the orthogonal operation in LE, record Mean as the mean of the calculated vector, calculate the average variance
[0055]
[0056] For k = 1, 2, … N + 1, when The orthogonal variance
[0057] When The orthogonal variance is updated as:
[0058]
[0059] The mean after orthogonal is updated as
[0060] The estimated mean output by the entire LE The estimated variance
[0061] Step S65, for k = 1, 2, …, N + 1, calculate the posteriori estimation r of each dimension of the demodulator part in NLE according to the estimated mean and variance transmitted by LE t [k] and variance v t [k], first calculate the posteriori distribution parameters and posteriori probability:
[0062]
[0063]
[0064] Calculate the posteriori estimation mean r t [k] and variance v t [k]
[0065] r t [k] = p post h k
[0066] v t [k] = p post (1-p post )|h k | 2
[0067] Step S66, for k = 1, 2, …, N + 1, complete the orthogonal calculation of the NLE part, obtain the estimated mean and variance of the NLE part, first calculate the average variance:
[0068]
[0069] For k = 1, 2, … N + 1, when The orthogonal variance
[0070] when When , the orthogonal variance is calculated as
[0071]
[0072] The mean after orthogonalization is updated to
[0073] Estimated mean of the entire LE output Estimated variance As the input of the next iteration LE part. The tth estimated scattering point vector
[0074] Step S67, t=t+1, based on the NLE obtained in the previous iteration and Start a new round of LE estimation.
[0075] Step S68: Repeat steps S63 to S67 until convergence is achieved to obtain the perception results of the environmental scattering points.
[0076] The present invention achieves the following beneficial effects: In a typical SISO transceiver system scenario, it implements environmental perception at the receiving end based on an OFDM oversampled CP. The two proposed randomized optimization strategies significantly improve performance when using an oversampled CP as the measurement matrix, effectively enhancing reconstruction performance compared to existing algorithms and significantly improving perception accuracy under varying OFDM symbol numbers. BRIEF DESCRIPTION OF THE DRAWINGS
[0077] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0078] Figure 1 A schematic diagram of an OFDM oversampling CP environment perception process and scenario provided by an exemplary embodiment;
[0079] Figure 2 The compressed sensing reconstruction algorithm provided by an exemplary embodiment estimates the relationship between the mean square error (MSE) and the signal-to-noise ratio (SNR). Comparison algorithm 1 is the generalized approximate message passing algorithm (GAMP), comparison algorithm 2 is the sparse Bayesian learning (SBL), and comparison algorithm 3 is the traditional orthogonal approximate message passing algorithm (OAMP).
[0080] Figure 3A graph showing the relationship between the estimated mean square error (MSE) and the number of oversampled cyclic prefixes of a compressed sensing reconstruction algorithm provided by an exemplary embodiment. Comparison algorithm 1 is the generalized approximate message passing algorithm (GAMP), and comparison algorithm 2 is the sparse Bayesian learning (SBL). DETAILED DESCRIPTION
[0081] In order to better understand the technical solution of the present application, the embodiments of the present application are described in detail below with reference to the accompanying drawings.
[0082] It should be clear that the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.
[0083] Consider a general single-send single-receive scenario, such as Figure 1 As shown in Figure 1, the transmitter (TX) transmits an OFDM signal through the environment to the receiver. The receiver (RX) processes the signal and perceives scattering points on the scattering surfaces in the environment. Because a large number of objects in the environment reflect and scatter electromagnetic waves, the communication signal propagates through multipath channels to the receiver, and the received signal contains some information about the environmental scatterers. Our goal is to leverage the OFDM framework for both transmission and reception, using an oversampled cyclic prefix (CP) to process and separate the environmental information from the communication signal, thereby enabling the perception of scattering points on the target scattering surface in the environment.
[0084] like Figure 1 As shown, in one embodiment, an environment perception method based on OFDM oversampling cyclic prefix is provided, comprising the following steps:
[0085] Step S1, establishing a channel model based on OFDM and with parameters determined according to the positional relationship between the transceiver and the environmental scattering surface;
[0086] In one embodiment, establishing a channel model specifically includes the following steps:
[0087] Step S11: The locations of the transmitter and receiver are known, and the distance is d. LOS The transmitted OFDM signal has a bandwidth of B and a number of fast Fourier transform points of N. fft , the cyclic prefix length is N CP , time domain sampling point interval ΔT = 1 / B;
[0088] Step S12: The propagation channel consists of two parts: the line of sight (LOS) from the transmitter to the receiver, i.e. the distance equal to d LOS, and non-line-of-sight (NLOS) from the transmitter to the environmental scattering surface to the receiver, the amplitude and phase of each channel in the NLOS are:
[0089]
[0090] wherein is the normalized gain of the transmitting and receiving antennas, j represents the imaginary unit, λ and f c represent the wavelength and frequency of the signal, σ i represents the radar cross section of the ith scattering point; d i = d TS + d SR represents the sum of the distance d TS from the transmitter to the ith scattering point on the environmental scattering surface and the distance d RS from the scattering point to the receiver; Δt i = (i-1) / f adc , represents the time delay difference between the ith NLOS channel and the LOS channel, tL O S represents the propagation time delay of the LOS, f adc is the sampling rate of the device digital-to-analog converter; the channel response vector wherein i = 2, …, N+1, N = UN CP , wherein U represents the signal oversampling rate, and h1 represents the channel response of the LOS.
[0091] Step S2, determining the position of the environmental scattering point corresponding to each channel in the channel model on the environmental scattering surface by using the ellipse cluster;
[0092] In an embodiment, the step of determining the position of the environmental scattering point corresponding to each channel in the channel model on the environmental scattering surface by using the ellipse cluster specifically comprises the following steps:
[0093] Step S21, using x s to represent the environmental scattering point corresponding to each channel, wherein if the scattering point exists, the value is 1, and if the scattering point does not exist, the value is 0, and the entire vector is represented as
[0094]
[0095] The length of the environmental scattering point vector and the channel coefficient vector is equal, i.e., M = N+1;
[0096] Step S22, according to the determined x s and h, the ith environmental scattering point is located on an ellipse with the positions of the transmitter and the receiver as the foci, and the length of the ellipse is the focal distance; according to the size L l , L wand the center position, the set of ambient scattering points Γ on the ambient scattering surface can be determined surface , so we can get two spatial constraints for the i-th environmental scattering point:
[0097]
[0098] Among them, P i , P TX , P RX represent the spatial positions of the i-th environmental scattering point, the transmitter, and the receiver, respectively.
[0099] Step S3, upsampling the demodulated OFDM signal to obtain a required oversampled cyclic prefix sequence, wherein the OFDM signal contains the location information of the environmental scattering points;
[0100] In one embodiment, upsampling the demodulated OFDM signal to obtain the required oversampled cyclic prefix sequence specifically includes the following steps:
[0101] The analog-to-digital converter (ADC) at the receiving end does not undergo low-pass filtering. The OFDM signal obtained is demodulated, and the demodulated symbol is subjected to inverse fast Fourier transform and then oversampled to obtain the oversampled cyclic prefix sequence corresponding to the kth OFDM symbol transmitted. Where dimension N = UN CP ; The received OFDM symbol is directly oversampled to obtain the oversampled cyclic prefix corresponding to the received k-th OFDM symbol as follows:
[0102] Step S4, using the oversampled cyclic prefix sequence and based on the channel model, establishing an environment perception model for solving the location of the environment scattering point, wherein the environment perception model involves the location problem of the environment scattering point;
[0103] In one embodiment, establishing an environment perception model for solving the location of environmental scattering points specifically includes the following steps:
[0104] Assuming that the maximum channel delay spread is exactly equal to the cyclic prefix duration, the cyclic prefix of the kth OFDM symbol will be affected by the data of equal length at the end of itself and the previous OFDM symbol, and its value is equal to the cyclic prefix of the k-1th OFDM symbol. Therefore, it can be equivalent to the cyclic prefix of the kth OFDM symbol being affected by the cyclic prefixes of the kth and k-1th OFDM symbols; the oversampled cyclic prefix satisfies the same influence, and the number of sampling points is U times the cyclic prefix of the OFDM symbol. The influence of the channel can be expressed as:
[0105]
[0106] in represents Gaussian additive white noise, r k is the received oversampled cyclic prefix sequence, It represents a cyclic prefix pair sequence consisting of the k-1th and kth oversampled cyclic prefixes, which is specifically expressed as Used to represent the shift of the bitwise multiplication of the channel coefficients and the scattering point vector:
[0107]
[0108] In order to facilitate the solution of the reconstruction target, the relationship is rewritten according to matrix multiplication as follows:
[0109] r k =Φ k h⊙x s +n
[0110] Where ⊙ represents the Hadamard product, that is, the bitwise multiplication of vectors, Φ k Indicated by The cyclic shift matrix composed of the data elements in is specifically expressed as follows:
[0111]
[0112] Assuming the channel is quasi-static, sym OFDM symbols, namely K sym -1 The total number of observations is N×K sym , the complete perception model composed of multiple oversampled cyclic prefixes is as follows:
[0113]
[0114] in K sym The complete observation vector consisting of oversampled cyclic prefixes, Indicated by K sym indivual The measurement matrix composed of k = 1, ..., K sym -1, K sym .
[0115] Step S5: Two random strategies are used to improve the randomness of the measurement matrix composed of the oversampled cyclic prefix sequence in the environmental perception model, thereby converting the environmental scattering point location problem into a compressed sensing reconstruction problem. One random strategy is to process the orthogonal frequency division multiplexing signal to be transmitted using a random symbol modulation method to increase the independence of data between each dimension of the measurement matrix. The other random strategy is to randomly extract different measurement matrices to form a new measurement matrix to reduce the dimension of the original measurement matrix.
[0116] In one embodiment, converting the environmental scattering point location problem into a compressed sensing reconstruction problem specifically includes the following steps:
[0117] According to the perception model of multiple oversampled cyclic prefix sequences, random modulation and random decimation strategies are used to improve the properties of the measurement matrix.
[0118] Random modulation: From the generation process of OFDM symbols, we know that the data in the cyclic prefix is related to the data in the frequency domain. Therefore, in order to ensure that the measurement matrix Φ composed of different oversampled cyclic prefixes is k To ensure the independence of data, each transmitted OFDM symbol has a different symbol mapping method to generate data. The specific strategy is to randomly select from four mapping methods: BPSK, QPSK, 16QAM, and 64QAM.
[0119] Random sampling: Since the measurement matrix in the oversampling cyclic prefix perception model is composed of The cyclic shift is generated, and each row has a high correlation. Therefore, the measurement matrix Φ formed from each oversampled cyclic prefix is k Randomly select a row from the matrix to form a new measurement matrix
[0120] Under the above two strategies, a new oversampling cyclic prefix perception model can be obtained, which is expressed as:
[0121]
[0122] z=h⊙x s
[0123] in, represents the new observation vector, Represented by random Φ k Each row in the new measurement matrix is formed, Represents the new reconstruction vector, which is composed of the Hadamard product of the channel coefficient and the reconstructed environmental scattering point vector; by estimating z and then removing the corresponding elements in h dimension by dimension, the recovered x is obtained based on the posterior probability judgment s .
[0124] Step S6: using the dimension-by-dimension orthogonal approximate message passing method to solve the compressed sensing reconstruction problem and obtain the reconstruction result of the environmental scatterer.
[0125] In one embodiment, solving the compressed sensing reconstruction problem using the dimension-by-dimension orthogonal approximate message passing method specifically includes the following steps:
[0126] Step S61: first set the environmental perception prior probability, and let the reconstructed environmental scattering point vector x sIf the Bernoulli distribution is satisfied and the parameter is P, then the sparsity of z is also P, and the elements satisfy Pr(z i =0) = P, Pr(z i =h i )=1-P, Pr represents the probability of an element taking a certain value.
[0127] Step S62: Initialize the parameters of the approximate message passing algorithm. The input parameters include the measurement matrix Observation vector Channel response coefficient h, Bernoulli distribution parameter P and noise variance σ at different locations 2 The standard orthogonal approximate message passing algorithm includes a linear estimator (LE) and a nonlinear estimator (NLE). The input parameters start from the LE iteration, and the initial mean The variance is where k∈1,…,N+1,
[0128] Step S63, each iteration calculates the estimated mean of the minimum mean square error (LMMSE) of the LE part and variance
[0129]
[0130]
[0131] Among them S t Therefore is a diagonal matrix with diagonal elements. ". / " indicates the Hadamard division by dimension, estimating the variance for A vector of diagonal elements, represents the new observation vector in step 6;
[0132] Step S64, complete the orthogonal operation in LE, record Mean as the mean of the calculated vector, and calculate the average variance
[0133]
[0134] For k = 1, 2, ... N + 1, when When the orthogonal variance
[0135] when The orthogonal variance is updated as:
[0136]
[0137] The mean after orthogonalization is updated to
[0138] Estimated mean of the entire LE output Estimated variance
[0139] Step S65, for k=1, 2, ..., N+1, calculate the posterior estimate r of each dimension of the demodulator part in the NLE based on the estimated mean and variance transmitted by the LE t [k] and variance v t [k], first calculate the posterior distribution parameters and posterior probability:
[0140]
[0141]
[0142] Calculate the posterior estimated mean r t [k] and variance v t [k]
[0143] r t [k]=p post h k
[0144] v t [k]=p post (1-p post )|h k | 2
[0145] Step S66: For k=1, 2, ..., N+1, complete the orthogonal calculation of the NLE part to obtain the estimated mean and variance of the NLE part. First, calculate the mean variance:
[0146]
[0147] For k = 1, 2, ... N + 1, when When the orthogonal variance
[0148] when When , the orthogonal variance is calculated as
[0149]
[0150] The mean after orthogonalization is updated to
[0151] Estimated mean of the entire LE output Estimated variance As the input of the next iteration LE part. The tth estimated scattering point vector
[0152] Step S67, t=t+1, based on the NLE obtained in the previous iteration and Start a new round of LE estimation.
[0153] Step S68: Repeat steps S63 to S67 until convergence is achieved to obtain the perception results of the environmental scattering points.
[0154] Computer simulation shows that: Figure 2 and Figure 3 As shown in FIG, the simulation verifies the environmental perception imaging effect of the algorithm of the present invention. The algorithm of the present invention effectively realizes environmental perception based on OFDM oversampling CP. Figure 2 It shows that with the increase of signal-to-noise ratio, the performance of the E-OAMP reconstruction algorithm of the method of the present invention is gradually improved, and the mean square error is greatly improved compared with other traditional compressed sensing algorithms. Figure 3 It shows that as the number of observed OFDM symbols increases, the mean square error of the environment perception reconstruction method using the orthogonal approximate message passing algorithm (E-OAMP) with dimension-by-dimension processing used in the present invention gradually decreases, and its performance is better than that of traditional algorithms.
[0155] The above are only preferred embodiments of one or more embodiments of this specification and are not intended to limit one or more embodiments of this specification. Any modifications made within the spirit and principles of one or more embodiments of this specification are not intended to limit this specification.
Claims
1. An environment perception method based on OFDM oversampling cyclic prefix, characterized in that: The following steps are involved: According to the positional relationship between the transceiver and the environmental scattering surface, a channel model is established based on OFDM and with the parameters determined. Establishing the channel model specifically includes the following steps: Step S11: The locations of the transmitter and receiver are known, and the distance is d. LOS ; The OFDM signal sent has a bandwidth of B and a number of fast Fourier transform points of N fft , the cyclic prefix length is N CP , time domain sampling point interval ΔT = 1 / B; Step S12: The propagation channel consists of two parts: the direct path from the transmitter to the receiver, i.e. the distance is equal to d LOS , and the indirect path from the transmitter to the ambient scattering surface and then to the receiver. The amplitude and phase of each channel in the indirect path are: in is the normalized gain of the transmitting and receiving antennas, j represents the imaginary unit, λ and f c Represent the wavelength and frequency of the signal, σ i represents the radar cross section of the i-th scattering point; d i =d TS +d SR Indicates the distance d from the transmitter to the i-th scattering point on the ambient scattering surface TS The distance d from the scattering point to the receiver RS The sum of Δt i =(i-1) / f adc , represents the delay difference between the i-th non-direct path channel and the direct path channel, t LOS represents the propagation delay of the direct path, f adc is the sampling rate of the device digital-to-analog converter; the channel response vector in N=UN CP , use U to represent the signal oversampling ratio, h1 represents the channel response of the direct path; Determining the location of an environmental scattering point corresponding to each channel in the channel model on an environmental scattering surface using an ellipse cluster specifically includes the following steps: Step S21, use x s Represents the environmental scattering point corresponding to each channel. If there is a scattering point, the value is 1, and if not, it is 0. The entire vector is represented as The length of the environmental scattering point vector and the channel coefficient vector is equal, that is, M = N + 1; Step S22, according to the determined x s and h, which is determined by the i-th environmental scattering point located at the transmitter and receiver positions as the focus, On a long ellipse whose length is the focal length; according to the size L of the scattering surface of the ambient scattering l ,L w and the center position, the set of ambient scattering points Γ on the ambient scattering surface can be determined surface , so we can get two spatial constraints for the i-th environmental scattering point: Among them, P i , P TX , P RX denote the spatial positions of the i-th environmental scattering point, the transmitter, and the receiver respectively; Upsampling the demodulated OFDM signal to obtain a desired oversampled cyclic prefix sequence, wherein the OFDM signal includes location information of environmental scattering points; Using the oversampled cyclic prefix sequence and based on the channel model, establishing an environment perception model for solving the location of the environment scattering point, wherein the environment perception model involves the location problem of the environment scattering point; The establishment of the environment perception model for solving the location of the environmental scattering points specifically includes the following steps: Assuming that the maximum channel delay spread is exactly equal to the cyclic prefix duration, the cyclic prefix of the kth OFDM symbol will be affected by the data of equal length at the end of itself and the previous OFDM symbol, and its value is equal to the cyclic prefix of the k-1th OFDM symbol. Therefore, it can be equivalent to the cyclic prefix of the kth OFDM symbol being affected by the cyclic prefixes of the kth and k-1th OFDM symbols; the oversampled cyclic prefix satisfies the same influence, and the number of sampling points is U times the cyclic prefix of the OFDM symbol. The influence of the channel can be expressed as: in represents Gaussian additive white noise, r k is the received oversampled cyclic prefix sequence, It represents a cyclic prefix pair sequence consisting of the k-1th and kth oversampled cyclic prefixes, which is specifically expressed as Used to represent the shift of the bitwise multiplication of the channel coefficients and the scattering point vector: In order to facilitate the solution of the reconstruction target, the relationship is rewritten according to matrix multiplication as follows: r k =Φ k h⊙x s +n Where ⊙ represents the Hadamard product, that is, the bitwise multiplication of vectors, Φ k Indicated by The cyclic shift matrix composed of the data elements in is specifically expressed as follows: Assuming the channel is quasi-static, sym OFDM symbols, namely K sym -1 The total number of observations is N×K. sym , the complete perception model composed of multiple oversampled cyclic prefixes is as follows: in K sym The complete observation vector consisting of oversampled cyclic prefixes, Indicated by K sym indivual The measurement matrix composed of k=1,…,K sym -1,K sym ; Two random strategies are employed to enhance the randomness of a measurement matrix composed of an oversampled cyclic prefix sequence in the environmental perception model, thereby converting the environmental scattering point location problem into a compressed sensing reconstruction problem. One random strategy processes the orthogonal frequency division multiplexing signal to be transmitted using a random symbol modulation scheme to increase the independence of data within each dimension of the measurement matrix. The other random strategy randomly extracts different measurement matrices to form a new measurement matrix, thereby reducing the dimensionality of the original measurement matrix. Converting the environmental scattering point location problem into a compressed sensing reconstruction problem specifically includes the following steps: According to the perception model of multiple oversampled cyclic prefix sequences, random modulation and random decimation strategies are used to improve the properties of the measurement matrix. Random modulation: From the generation process of OFDM symbols, we know that the data in the cyclic prefix is related to the data in the frequency domain. Therefore, in order to ensure that the measurement matrix Φ composed of different oversampled cyclic prefixes is k To ensure the independence of data between two OFDM symbols, each OFDM symbol sent has a different symbol mapping method to generate data. The specific strategy is to randomly select from four mapping methods: BPSK, QPSK, 16QAM, and 64QAM. Random sampling: Since the measurement matrix in the oversampling cyclic prefix perception model is composed of Cyclic shift generation, each row has a high correlation; therefore, the measurement matrix Φ formed from each oversampled cyclic prefix k Randomly select a row from the matrix to form a new measurement matrix Under the above two strategies, a new oversampling cyclic prefix perception model can be obtained, which is expressed as z=h⊙x s in, represents the new observation vector, Represented by random Φ k Each row in the new measurement matrix is formed, Represents the new reconstruction vector, which is composed of the Hadamard product of the channel coefficient and the reconstructed environmental scattering point vector; by estimating z and then removing the corresponding elements in h dimension by dimension, the recovered x is obtained based on the posterior probability judgment s ; The compressed sensing reconstruction problem is solved by using the orthogonal approximate message passing method for dimension-by-dimension processing to obtain the reconstruction result of the environmental scatterer, which specifically includes the following steps: Step S61: first set the environmental perception prior probability, and let the reconstructed environmental scattering point vector x s If the Bernoulli distribution is satisfied and the parameter is P, then the sparsity of z is also P, and the elements satisfy Pr(z i =0) = P,Pr(z i =h i ) = 1-P, where Pr represents the probability of an element taking a certain value; Step S62: Initialize the parameters of the approximate message passing algorithm. The input parameters include the measurement matrix Observation vector Channel response coefficient h, Bernoulli distribution parameter P and noise variance σ at different locations 2 The standard orthogonal approximate message passing algorithm includes a linear estimator (LE) and a nonlinear estimator (NLE). The input parameters start from the LE iteration, and the initial mean The variance is where k∈1,…,N+1, Step S63, each iteration calculates the estimated mean of the minimum mean square error (LMMSE) of the LE part and variance Among them S t Therefore is a diagonal matrix with diagonal elements; "° / " indicates Hadamard division by dimension, estimating the variance for A vector of diagonal elements, represents the new observation vector in step 6; Step S64, complete the orthogonal operation in LE, record Mean as the mean of the calculated vector, and calculate the average variance For k=1,2,…N+1, when When the orthogonal variance when The orthogonal variance is updated as: The mean after orthogonalization is updated to Estimated mean of the entire LE output Estimated variance Step S65, for k = 1, 2, ..., N + 1, calculate the posterior estimate r of each dimension of the demodulator part in the NLE based on the estimated mean and variance transmitted by the LE t [k] and variance v t [k], first calculate the posterior distribution parameters and posterior probability: Calculate the posterior estimated mean r t [k] and variance v t [k] r t [k]=p post h k v t [k]=p post (1-p post )|h k | 2 Step S66: For k=1, 2, ..., N+1, complete the orthogonal calculation of the NLE part to obtain the estimated mean and variance of the NLE part. First, calculate the mean variance: For k=1,2,…N+1, when When the orthogonal variance when When , the orthogonal variance is calculated as The mean after orthogonalization is updated to Estimated mean of the entire LE output Estimated variance As the input of the next iteration LE part; the t-th estimated scattering point vector Step S67, t=t+1, based on the NLE obtained in the previous iteration and Start a new round of LE estimation; Step S68: Repeat steps S63 to S67 until convergence is achieved to obtain the perception results of the environmental scattering points.
2. The environment perception method based on OFDM oversampling cyclic prefix according to claim 1, characterized in that: The orthogonal frequency division multiplexing signal in the OFDM after demodulation is up-sampled to obtain the required oversampled cyclic prefix sequence, which specifically includes the following steps: The receiving end's digital-to-analog converter does not undergo low-pass filtering, and the OFDM signal obtained is demodulated. The demodulated symbol is subjected to inverse fast Fourier transform and then oversampled to obtain the oversampled cyclic prefix sequence corresponding to the kth OFDM symbol sent. Where dimension N = UN CP ; The received OFDM symbol is directly oversampled to obtain the oversampled cyclic prefix corresponding to the received k-th OFDM symbol as follows: