APSO-VAMP detection method and system based on particle swarm algorithm training

Abstract

The present invention provides an APSO-VAMP detection method and system based on particle swarm algorithm training. The method generates a delay Doppler domain input signal by randomly generating a user bit stream; converts it into a time-frequency domain signal of a transmitting end; converts it into a time-domain signal of the transmitting end; sends it to a receiving end through a wireless channel; converts the time-domain signal of the receiving end into the time-frequency domain signal of the receiving end; converts it into a delay Doppler domain signal of the receiving end; adopts an adaptive particle swarm-vectorized message passing algorithm APSO-VAMP iterative detection, takes the bit error rate of a set position as a fitness function, takes the damping factor of the vectorized message passing algorithm as a training parameter in each iteration, obtains the optimal damping factor parameter through adaptive particle swarm algorithm training; inputs it into a VAMP algorithm for detection; and the parameters trained by the APSO algorithm can reduce the bit error rate compared with the fixed parameters, and the detection performance is significantly improved.

Description

Technical Field

[0001] The invention relates to an APSO-VAMP detection method and system based on particle swarm algorithm training, belonging to the technical field of wireless transmission. Background Art

[0002] With the rapid development of wireless communication technology, the speed of mobile devices and the coverage of wireless networks are increasing. However, traditional modulation technologies, such as orthogonal frequency division multiplexing (OFDM), face severe challenges in high-speed mobile scenarios. Although OFDM technology performs well in static or low-mobility environments, in high-mobility scenarios, due to the combined effects of Doppler effect and multipath effect, channel characteristics will change rapidly, causing inter-carrier interference (ICI) and signal distortion, seriously affecting communication performance.

[0003] In order to solve the Doppler shift problem of traditional OFDM waveform in high-speed mobility scenarios, a new waveform called Orthogonal Time-Frequency-Space (OTFS) has been recently proposed and is becoming a hot topic of research. Orthogonal Time-Frequency-Space Modulation has attracted widespread attention due to its ability to achieve reliable communication in high-mobility applications. Orthogonal Time-Frequency-Space (OTFS) provides diversity in time and frequency because each symbol is distributed in the time and frequency domains through a two-dimensional inverse symplectic finite Fourier transform. Compared with OFDM, ORTHODUX can achieve significant performance improvement in high-mobility scenarios. In addition, when the number of channel paths is small, the effective channel in the delay-Doppler domain is sparse, which allows efficient channel estimation and data detection using message passing techniques. However, the bit error rate is high and the detection performance is low. Summary of the invention

[0004] The purpose of the present invention is to provide an APSO-VAMP detection method and system based on particle swarm algorithm training to solve the problems in the prior art that the bit error rate is high and the detection performance needs to be improved.

[0005] The technical solution of the present invention is:

[0006] An APSO-VAMP detection method based on particle swarm algorithm training includes the following steps:

[0007] S1, randomly generate a user bit stream, and modulate the bit stream through quadrature amplitude modulation (QAM) to generate an input signal in the delay-Doppler domain;

[0008] S2, converting the input signal in the delay-Doppler domain generated in step S1 into a time-frequency domain signal of the transmitting end through an inverse symmetric fast Fourier transform IFFT;

[0009] S3, converting the time-frequency domain signal of the transmitting end obtained in step S2 into the time domain signal of the transmitting end through Heisenberg transformation;

[0010] S4, sending the time domain signal of the transmitting end obtained in step S3 to the receiving end through the wireless channel;

[0011] S5, converting the time domain signal of the receiving end into a time-frequency domain signal of the receiving end through Wigner transformation;

[0012] S6, converting the time-frequency domain signal of the receiving end obtained in step S5 into a delay-Doppler domain signal of the receiving end by fast Fourier transform FFT;

[0013] S7. Adopting adaptive particle swarm-vectorized message passing algorithm APSO-VAMP iterative detection on the delay Doppler domain signal of the receiving end obtained in step S6. For the current training process, the bit error rate of the set position is used as the fitness function. In each iteration, the damping factor of the vectorized message passing algorithm, i.e., the VAMP algorithm, is used as the training parameter. The optimal damping factor parameter is obtained by training the adaptive particle swarm algorithm, i.e., the APSO algorithm. The obtained optimal damping factor parameter is input into the VAMP algorithm for detection.

[0014] Furthermore, in step S1, a user bit stream is randomly generated, specifically, a bit stream of size N*M*M_bits is randomly generated according to the number of frames, the number of subcarriers and the modulation method adopted, wherein N is the number of time slots, M is the number of carrier frequencies, and M_bits is the bit size corresponding to each symbol.

[0015] Furthermore, in step S5, the formula of Wigner transformation is as follows:

[0016] ,

[0017] Among them, Y (t, f) is the time-frequency domain signal of the receiving end, t is time, f is the frequency sampling point, g* rx is the shaping function of the receiving end, t' is the time sampling point, e is a natural constant, j is an imaginary unit, and r(t') is the time domain signal of the receiving end.

[0018] Furthermore, in step S6, the formula of fast Fourier transform FFT is as follows:

[0019] ,

[0020] in, is the delayed Doppler domain signal at the receiving end, q is the Doppler frequency sampling point, is the delayed sampling point, t is the time, f is the frequency sampling point, N and M are the number of time slots and carrier frequencies respectively, e is a natural constant, j is an imaginary unit, and Y(t,f) is the time-frequency domain signal at the receiving end.

[0021] Further, in step S7, the obtained optimal damping factor parameters are input into the VAMP algorithm for detection, specifically,

[0022] S71, initialize the current number k = 0, and initialize the mixed signal mean value at the 0th iteration The variance at iteration 0 is ;

[0023] S72, calculate the estimated signal mean value of the kth iteration :

[0024] ,

[0025] in, is the minimum mean square error criterion denoiser, is the mixed signal mean of the kth iteration, is the variance of the kth iteration;

[0026] S73, calculate the diagonal matrix of the kth iteration , the prior variance of the kth iteration is , the prior mean of the kth iteration And the diagonal matrix of the kth iteration is :

[0027] ,

[0028] ,

[0029] ,

[0030] ,

[0031] Where Diag is a diagonal function, is the derivative of the minimum mean square error criterion denoiser, σ is the noise spectrum, H is the channel state information, and * is the conjugate transpose;

[0032] S74, calculate the estimated signal mean value 2 of the kth iteration , the diagonal element vector of the diagonal matrix of the kth iteration , the variance of the k+1th iteration is The mixed signal mean of the k+1th iteration is :

[0033] ,

[0034] ,

[0035] ,

[0036] ,

[0037] Where H is the channel state information, * is the conjugate transpose, is the delayed Doppler domain signal at the receiving end, q is the Doppler frequency sampling point, is the delayed sampling point, Diag is the diagonal function;

[0038] S75, using the optimal damping factor parameter mes to obtain the variance of the k+1th iteration The mixed signal mean of the k+1th iteration is :

[0039] ,

[0040] ;

[0041] S76, set the current number k=k+1, and output the estimated signal mean value of the kth iteration When the current number k ≤ the maximum number of loops K, return to step S72; when the current number k > the maximum number of loops K, complete the detection and end the loop.

[0042] An APSO-VAMP detection system based on particle swarm algorithm training using any of the above methods comprises a transmitting end and a receiving end,

[0043] Transmitter: randomly generate user bit streams, and modulate the bit streams through quadrature amplitude modulation (QAM) to generate input signals in the delay-Doppler domain; convert the generated input signals in the delay-Doppler domain through inverse fast Fourier transform (IFFT) to the time-frequency domain signals of the transmitter; convert the obtained time-frequency domain signals of the transmitter through Heisenberg transform to the time-domain signals of the transmitter; and send the obtained time-domain signals of the transmitter to the receiver through the wireless channel;

[0044] Receiving end: The time domain signal of the receiving end is converted into the time-frequency domain signal of the receiving end through Wigner transform; the obtained time-frequency domain signal of the receiving end is converted into the delay-Doppler domain signal of the receiving end through fast Fourier transform FFT; the obtained delay-Doppler domain signal of the receiving end is iteratively detected using adaptive particle swarm-vectorized message passing algorithm APSO-VAMP. For the current training process, the bit error rate of the set position is used as the fitness function. In each iteration, the damping factor of the vectorized message passing algorithm, i.e., the VAMP algorithm, is used as the training parameter. The optimal damping factor parameter is obtained through training with the adaptive particle swarm algorithm, i.e., the APSO algorithm, and the obtained optimal damping factor parameter is input into the VAMP algorithm for detection.

[0045] The beneficial effects of the present invention are as follows: the APSO-VAMP detection method and system based on particle swarm algorithm training calculates the bit error rate as the fitness function, and regards the damping factor of each layer iteration as an independent parameter. Finally, the parameters trained by the APSO-VAMP algorithm have significantly improved detection performance compared with fixed parameters. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 is a flow chart of an APSO-VAMP detection method based on particle swarm algorithm training according to an embodiment of the present invention;

[0047] Figure 2 is a schematic diagram illustrating how to obtain an optimal damping factor through training of an adaptive particle swarm algorithm in an embodiment;

[0048] Figure 3 It is a schematic diagram of comparing the signal-to-noise ratio-bit error rate of the APSO-VAMP detection method based on particle swarm algorithm training in the embodiment and the existing methods including the orthogonal time-frequency-space-minimum mean square error algorithm OTFS-MMSE and the orthogonal time-frequency-space-approximate message passing algorithm OTFS-AMP;

[0049] Figure 4 1 is a schematic diagram of comparing the signal-to-noise ratio-bit error rate of the APSO-VAMP detection method based on particle swarm algorithm training and the orthogonal time-frequency-space-vectorized message passing algorithm OTFS-VAMP under different fixed damping factors MES values ​​of 0.5 and 0.8;

[0050] Figure 5 It is a schematic diagram comparing the number of iterations-bit error rate of the APSO-VAMP detection method based on particle swarm algorithm training in the embodiment, the orthogonal time-frequency-space-approximate message passing algorithm OTFS-AMP, and the orthogonal time-frequency-space-vectorized message passing algorithm OTFS-VAMP. DETAILED DESCRIPTION

[0051] The preferred embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0052] The embodiment provides an APSO-VAMP detection method based on particle swarm algorithm training, such as Figure 1 , including the following steps,

[0053] S1. Randomly generate a user bit stream and modulate the bit stream through quadrature amplitude modulation (QAM) to generate an input signal in the delay-Doppler domain.

[0054] In step S1, a user bit stream is randomly generated. Specifically, a bit stream of size N*M*M_bits is randomly generated according to the number of frames, the number of subcarriers and the modulation method adopted, where N is the number of time slots, M is the number of carrier frequencies, and M_bits is the bit size corresponding to each symbol.

[0055] S2, converting the delay-Doppler domain input signal generated in step S1 into a time-frequency domain signal of a transmitting end through an inverse symmetric fast Fourier transform IFFT.

[0056] In step S2, the formula of inverse symmetric fast Fourier transform IFFT is as follows:

[0057]

[0058] in, is the input signal in the delay Doppler domain, n is the time sampling point, m is the frequency sampling point, q is the Doppler frequency sampling point, is the delayed sampling point, N and M are the number of time slots and carrier frequencies respectively, X i [n,m] is the time-frequency domain signal of the transmitter, e is a natural constant, and j is an imaginary unit.

[0059] S3. Convert the time-frequency domain signal of the transmitting end obtained in step S2 into the time domain signal of the transmitting end through Heisenberg transformation.

[0060] In step S3, the formula of Heisenberg transformation is as follows:

[0061]

[0062] Among them, s(t) is the time domain signal of the transmitter, X i [n,m] is the time-frequency domain signal at the transmitter, n is the time sampling point, m is the frequency sampling point, N and M are the number of time slots and carrier frequencies respectively. g tx is the shaping function of the transmitter, t is the time, T is the symbol period, e is the natural constant, j is the imaginary unit, and △f is the subcarrier spacing.

[0063] S4. The time domain signal of the transmitting end obtained in step S3 is sent from the transmitting end to the receiving end through a wireless channel.

[0064] S5. The time domain signal of the receiving end is converted into a time-frequency domain signal of the receiving end through Wigner transformation.

[0065] In step S5, the formula of Wigner transformation is as follows:

[0066]

[0067] Among them, Y (t, f) is the time-frequency domain signal of the receiving end, t is time, f is the frequency sampling point, g*rx is the shaping function of the receiving end, t` is the time sampling point, e is a natural constant, j is an imaginary unit, and r (t') is the time domain signal of the receiving end.

[0068] S6. Convert the time-frequency domain signal of the receiving end obtained in step S5 into a delay-Doppler domain signal of the receiving end through Fast Fourier Transform (FFT).

[0069] In step S6, the formula of fast Fourier transform FFT is as follows:

[0070] ,

[0071] in, is the delayed Doppler domain signal at the receiving end, q is the Doppler frequency sampling point, l is the delayed sampling point, t is the time, f is the frequency sampling point, N and M are the number of time slots and carrier frequencies respectively, e is a natural constant, j is an imaginary unit, and Y(t,f) is the time-frequency domain signal at the receiving end.

[0072] S7, perform APSO-VAMP iterative detection on the delay Doppler domain signal of the receiving end obtained in step S6. For the current training process, the bit error rate of the set position is used as the fitness function, and the damping factor of the vectorized message passing algorithm is used as the training parameter in each iteration. The optimal damping factor parameter is obtained by training the adaptive particle swarm algorithm, i.e., the APSO algorithm, such as Figure 2 The obtained optimal damping factor parameters are input into the VAMP algorithm for detection.

[0073] S71, initialize the current number k = 0, and initialize the mixed signal mean value at the 0th iteration The variance at iteration 0 is ;

[0074] S72, calculate the estimated signal mean value of the kth iteration :

[0075] ,

[0076] in, is the minimum mean square error criterion denoiser, is the mixed signal mean of the kth iteration, is the variance of the kth iteration;

[0077] S73, calculate the diagonal matrix of the kth iteration , the prior variance of the kth iteration is , the prior mean of the kth iteration And the diagonal matrix of the kth iteration is :

[0078] ,

[0079] ,

[0080] ,

[0081] ,

[0082] Among them, Diag is the diagonal function, is the derivative of the minimum mean square error criterion denoiser, σ is the noise spectrum, H is the channel state information, and * is the conjugate transpose;

[0083] S74, calculate the estimated signal mean value 2 of the kth iteration , the diagonal element vector of the diagonal matrix of the kth iteration , the variance of the k+1th iteration is The mixed signal mean of the k+1th iteration is :

[0084] ,

[0085] ,

[0086] ,

[0087] ,

[0088] Where H is the channel state information, * is the conjugate transpose, is the delayed Doppler domain signal at the receiving end, q is the Doppler frequency sampling point, is the delayed sampling point, Diag is the diagonal function;

[0089] S75, using the optimal damping factor parameter mes to obtain the variance of the k+1th iteration The mixed signal mean of the k+1th iteration is :

[0090] ,

[0091] ;

[0092] S76, set the current number k=k+1, and output the estimated signal mean value of the kth iteration When the current number k ≤ the maximum number of loops K, return to step S72; when the current number k > the maximum number of loops K, complete the detection and end the loop.

[0093] The embodiment also provides an APSO-VAMP detection system based on particle swarm algorithm training using any of the above methods, including a transmitting end and a receiving end.

[0094] Transmitter: randomly generate user bit streams, and modulate the bit streams through quadrature amplitude modulation (QAM) to generate input signals in the delay-Doppler domain; convert the generated input signals in the delay-Doppler domain through inverse fast Fourier transform (IFFT) to the time-frequency domain signals of the transmitter; convert the obtained time-frequency domain signals of the transmitter through Heisenberg transform to the time-domain signals of the transmitter; and send the obtained time-domain signals of the transmitter to the receiver through the wireless channel;

[0095] Receiving end: The time domain signal of the receiving end is converted into the time-frequency domain signal of the receiving end through Wigner transform; the obtained time-frequency domain signal of the receiving end is converted into the delay-Doppler domain signal of the receiving end through fast Fourier transform FFT; the obtained delay-Doppler domain signal of the receiving end is iteratively detected using adaptive particle swarm-vectorized message passing algorithm APSO-VAMP. For the current training process, the bit error rate of the set position is used as the fitness function. In each iteration, the damping factor of the vectorized message passing algorithm, i.e., the VAMP algorithm, is used as the training parameter. The optimal damping factor parameter is obtained through training with the adaptive particle swarm algorithm, i.e., the APSO algorithm, and the obtained optimal damping factor parameter is input into the VAMP algorithm for detection.

[0096] The APSO-VAMP detection method and system based on particle swarm algorithm training generates random bits and modulates the bit stream through quadrature amplitude modulation (QAM) to generate an input signal in the delay Doppler domain; then the signal is subjected to an inverse symmetric fast Fourier transform (ISFFT) to map the input signal in the delay Doppler domain to the time-frequency domain and then subjected to a Heisenberg transform to convert the time-frequency domain signal of the transmitter into the time-domain signal of the transmitter; the time-domain signal of the transmitter is sent to the receiver through a wireless channel; the received signal is subjected to a Wigner transform to obtain a time-frequency domain signal, and the user's time-frequency domain signal is taken out for SFFT transform to obtain the corresponding delay Doppler domain signal; the obtained signal is subjected to APSO-VAMP iterative detection: the optimal damping factor parameter is obtained through optimization by an adaptive particle swarm algorithm, and the optimal damping factor parameter is put into the VAMP detection algorithm for detection to complete the symbol detection process of the system. This method calculates the bit error rate as the fitness function, and regards the damping factor of each layer iteration as an independent parameter, taking into account both performance and complexity. Finally, the parameters trained by the APSO-VAMP algorithm have significantly improved the detection performance compared with the fixed parameter system. The present invention can transmit multi-user signals without interference, and does not require additional system overhead.

[0097] The APSO-VAMP detection method and system based on particle swarm algorithm training of the embodiment are experimentally verified as follows:

[0098] S1. Randomly generate 8*64*16 bit streams according to the number of frames, the number of subcarriers and the modulation method used, modulate the bit streams through QAM, and generate input signals in the delay Doppler domain. Xo=[-0.9487+0.9487i,-0.9487+0.3162i,0.9487-0.9487i,-0.9487-0.3162i

[0099] -03162+0.9487i,-0.3162+0.3162i,-0.3162-0.9487i,-0.3162-0.3162i

[0100] 0.9487+0.9487i,0.9487+0.3162i,0.9487-0.9487i,0.9487-0.3162i

[0101] 0.3162+0.9487i,0.3162+0.3162i,0.3162-0.9487i,0.3162-0.3162i]

[0102] Among them, Xo is the input signal in the delay-Doppler domain after modulation, and i is the imaginary unit.

[0103] S2. Convert the delay-Doppler domain input signal generated in step S1 into a time-frequency domain signal through IFFT transformation.

[0104] S3, converting the time-frequency domain signal obtained in step S2 into a time domain signal through Heisenberg transformation.

[0105] S4. The time domain signal obtained in step S3 is sent from the transmitting end to the receiving end through a wireless channel.

[0106] In step S4, the complex signal Receive_signal after transmission through the wireless channel is: Receive_signal=[1.3200+0.5425i,-0.5883+5.2097i,0.1775+5.8187i,3.1403-1.8521i -0.7043+3.5683,0.6457+1.4394i,3.2810+0.7321i,0.2655+2.0543i]

[0107] S5. The time domain signal of the receiving end is converted into a time-frequency domain signal of the receiving end through Wigner transformation.

[0108] S6. Performing FFT transformation on the time-frequency domain signal of the receiving end obtained in step S5 to convert the time-frequency domain signal of the receiving end into a delay-Doppler domain signal of the receiving end.

[0109] S7, performing APSO-VAMP iterative detection on the delay Doppler domain signal of the receiving end obtained in step S6. For the current training process, the bit error rate of the set position is used as the fitness function, and the damping factor of the VAMP algorithm is a trainable parameter in each iteration. The optimal damping factor parameter is obtained through APSO algorithm training.

[0110] In step S7, the damping factor is used as an input parameter in each iteration, and the value is between 0 and 1. The bit error rate when the signal-to-noise ratio is 21db is used as the fitness function. The initial population size is 20, and the inertia weight, self-learning factor and group learning factor are dynamically adjusted. The initial minimum inertia weight and maximum inertia weight are 0.4 and 0.9 respectively, the initial self-learning factor and group learning factor are 0.3 and 0.4 respectively, and the maximum number of iterations is 10 times. The adaptive particle swarm algorithm is used to train the individual optimal values ​​of ten particles: pbest=[0.2724, 0.5050, 0.2686, 0.6766, 0.8561, 0.8848, 0.5201, 0.5260, 0.7287, 0.2984, 0.8815, 0.2888, 0.9026, 0.8992, 0.3712, 0.5298, 0.9019, 0.2426, 0.8015, 0.8003], and the group optimal value gbest=0.9019 is the optimal damping factor parameter.

[0111] S8. Input the optimal damping factor parameter obtained in step S7 into the VAMP algorithm for detection.

[0112] Figure 3 It is a schematic diagram comparing the signal-to-noise ratio-bit error rate of the APSO-VAMP detection method based on particle swarm algorithm training in the embodiment and the existing methods including orthogonal time-frequency-space-minimum mean square error algorithm OTFS-MMSE and orthogonal time-frequency-space-approximate message passing algorithm OTFS-AMP. Figure 3 In the figure, APSO-VAMP is the bit error rate curve when the optimal damping factor parameter is 0.9019. At the bit error rate of 10 to the power of -2, it can be seen that the APSO-VAMP of the embodiment has a performance gain of 2db relative to the orthogonal time-frequency-space-approximate message passing algorithm OTFS-AMP, and a performance gain of 3dm relative to the orthogonal time-frequency-space-minimum mean square error algorithm OTFS-MMSE. The embodiment method has better detection performance. For practical applications, channels with low signal-to-noise ratios have no practical use value, and the actual point of action starts from a signal-to-noise ratio of 15db.

[0113] Figure 4 3 is a schematic diagram of signal-to-noise ratio-bit error rate comparison of the APSO-VAMP detection method based on particle swarm algorithm training and the orthogonal time-frequency-space-vectorized message passing algorithm OTFS-VAMP under different fixed damping factors MES values ​​of 0.5 and 0.8 in the embodiment. Figure 4In the figure, APSO-VAMP is the bit error rate curve when the optimal damping factor parameter is 0.9019. When the signal-to-noise ratio is 21, the bit error rate detected by APSO-VAMP in the embodiment has reached the -5th power level, and the bit error rate of the VAMP algorithm with fixed damping factors of 0.5 and 0.8 is less than the -4th power level, which verifies that the embodiment method can effectively reduce the bit error rate. For practical applications, channels with low signal-to-noise ratios have no practical use value, and the actual action point starts from a signal-to-noise ratio of 15db.

[0114] Figure 5 It is a schematic diagram comparing the number of iterations-bit error rate of the APSO-VAMP detection method based on particle swarm algorithm training in the embodiment, the orthogonal time-frequency-space-approximate message passing algorithm OTFS-AMP, and the orthogonal time-frequency-space-vectorized message passing algorithm OTFS-VAMP. Figure 5 In the figure, APSO-VAMP is the bit error rate curve when the optimal damping factor parameter is 0.9019. After 10 iterations, the detection bit error rate of APSO-VAMP in the embodiment is lower than 10 to the power of -4, which is better than the orthogonal time-frequency-space-vectorized message passing algorithm OTFS-VAMP with a fixed damping factor parameter, and is much better than the orthogonal time-frequency-space-approximate message passing algorithm OTFS-AMP. It verifies that the embodiment method can effectively reduce the bit error rate.

[0115] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An APSO-VAMP detection method based on particle swarm algorithm training, characterized in that: The following steps are included: S1, randomly generate a user bit stream, and modulate the bit stream through quadrature amplitude modulation (QAM) to generate an input signal in the delay-Doppler domain; S2, converting the input signal in the delay-Doppler domain generated in step S1 into a time-frequency domain signal of the transmitting end through an inverse symmetric fast Fourier transform IFFT; S3, converting the time-frequency domain signal of the transmitting end obtained in step S2 into the time domain signal of the transmitting end through Heisenberg transformation; S4, sending the time domain signal of the transmitting end obtained in step S3 to the receiving end through the wireless channel; S5, converting the time domain signal of the receiving end into a time-frequency domain signal of the receiving end through Wigner transformation; S6, converting the time-frequency domain signal of the receiving end obtained in step S5 into a delay-Doppler domain signal of the receiving end by fast Fourier transform FFT; S7, adopting adaptive particle swarm algorithm-vectorized message passing algorithm APSO-VAMP iterative detection on the delay Doppler domain signal of the receiving end obtained in step S6, for the current training process, taking the bit error rate of the set position as the fitness function, taking the damping factor of the vectorized message passing algorithm, i.e., the VAMP algorithm, as the training parameter in each iteration, obtaining the optimal damping factor parameter through adaptive particle swarm algorithm, i.e., the APSO algorithm training, and inputting the obtained optimal damping factor parameter into the VAMP algorithm for detection; In step S7, the obtained optimal damping factor parameters are input into the VAMP algorithm for detection, specifically, S71, initialize the current number k = 0, and initialize the mixed signal mean value at the 0th iteration The variance at iteration 0 is ; S72, calculate the estimated signal mean value of the kth iteration : ， in, is the minimum mean square error criterion denoiser, is the mixed signal mean of the kth iteration, is the variance of the kth iteration; S73, calculate the diagonal matrix of the kth iteration , the prior variance of the kth iteration is , the prior mean of the kth iteration And the diagonal matrix of the kth iteration is : ， ， ， ， Where Diag is a diagonal function, is the derivative of the minimum mean square error criterion denoiser, σ is the noise spectrum, H is the channel state information, and * is the conjugate transpose; S74, calculate the estimated signal mean value 2 of the kth iteration , the diagonal element vector of the diagonal matrix of the kth iteration , the initial variance of the k+1th iteration is The initial mixed signal mean of the k+1th iteration is : ， ， ， ， Where H is the channel state information, * is the conjugate transpose, is the delayed Doppler domain signal at the receiving end, q is the Doppler frequency sampling point, is the delayed sampling point, Diag is the diagonal function; S75, using the optimal damping factor parameter mes to obtain the variance of the k+1th iteration The mixed signal mean of the k+1th iteration is : ， ； S76, set the current number k=k+1, and output the estimated signal mean value of the kth iteration When the current number k ≤ the maximum number of loops K, return to step S72; when the current number k > the maximum number of loops K, complete the detection and end the loop.

2. The APSO-VAMP detection method based on particle swarm algorithm training as claimed in claim 1, characterized in that: In step S1, a user bit stream is randomly generated. Specifically, a bit stream of size N*M*M_bits is randomly generated according to the number of frames, the number of subcarriers and the modulation method adopted, where N is the number of time slots, M is the number of carrier frequencies, and M_bits is the bit size corresponding to each symbol.

3. The APSO-VAMP detection method based on particle swarm algorithm training as claimed in claim 1, characterized in that: In step S5, the formula of Wigner transformation is as follows: ， Among them, Y (t, f) is the time-frequency domain signal of the receiving end, t is time, f is the frequency sampling point, g* rx is the shaping function of the receiving end, t' is the time sampling point, e is a natural constant, j is an imaginary unit, and r(t') is the time domain signal of the receiving end.

4. The APSO-VAMP detection method based on particle swarm algorithm training according to any one of claims 1 to 3, characterized in that: In step S6, the formula of fast Fourier transform FFT is as follows: ， in, is the delayed Doppler domain signal at the receiving end, q is the Doppler frequency sampling point, l is the delayed sampling point, t is the time, f is the frequency sampling point, N and M are the number of time slots and carrier frequencies respectively, e is a natural constant, j is an imaginary unit, and Y(t,f) is the time-frequency domain signal at the receiving end.

5. An APSO-VAMP detection system based on particle swarm algorithm training using the method described in any one of claims 1 to 4, characterized in that: Including the sending end and the receiving end, Transmitter: randomly generate user bit streams, and use quadrature amplitude modulation (QAM) to modulate the bit streams to generate input signals in the delay-Doppler domain; convert the generated input signals in the delay-Doppler domain into time-frequency domain signals at the transmitter through inverse symmetric fast Fourier transform (IFFT); The obtained time-frequency domain signal of the transmitting end is converted into the time domain signal of the transmitting end through Heisenberg transformation; the obtained time domain signal of the transmitting end is sent from the transmitting end to the receiving end through a wireless channel; Receiving end: The time domain signal of the receiving end is converted into the time-frequency domain signal of the receiving end through Wigner transform; the obtained time-frequency domain signal of the receiving end is converted into the delay-Doppler domain signal of the receiving end through fast Fourier transform FFT; the obtained delay-Doppler domain signal of the receiving end is iteratively detected using adaptive particle swarm-vectorized message passing algorithm APSO-VAMP. For the current training process, the bit error rate of the set position is used as the fitness function. In each iteration, the damping factor of the vectorized message passing algorithm, i.e., the VAMP algorithm, is used as the training parameter. The optimal damping factor parameter is obtained through training with the adaptive particle swarm algorithm, i.e., the APSO algorithm, and the obtained optimal damping factor parameter is input into the VAMP algorithm for detection.

Images