CN111555787A - Iterative correction of artificial noise weights and low-bit feedback method for multi-transmit and single-receive systems - Google Patents

Abstract

The invention discloses an artificial noise weight iterative correction and low-bit feedback method for a multi-transmission and single-receive system. In the TDD communication mode, the N-antenna sender Alice obtains an erroneous channel estimation value in an access time slot.

An initial main beam weight w ₀ and an artificial noise weight q _0,i are generated. The receiving end Bob calculates the error data Er after receiving the signal containing artificial noise, and feeds back the result after quantization. After receiving the data, the Alice end updates the main beam weight w and the artificial noise weight q _i (i=1,...,N-1) and performs transmission and error feedback in the next time slot. After a certain number of iterations, the original mismatched weights will be corrected to the actual channel response. The present invention can realize the function of weight correction by using a small amount of return bits in each uplink time slot without changing the existing TDD communication mode and without using additional equipment, has the advantages of low complexity, and can meet the communication requirements. real-time requirements.

Description

Iterative correction of artificial noise weights and low-bit feedback method for multi-transmit and single-receive systems

technical field

The invention belongs to the field of wireless communication, and in particular relates to a method for iterative correction of artificial noise weights and low-bit feedback in a multi-transmission and single-receive system.

Background technique

With the development of wireless communication technology and the popularization of intelligent terminal equipment, people pay more and more attention to information security in the process of wireless communication. Due to its natural openness and borderless nature, wireless communication signals are easily intercepted and stolen, which is one of the root causes of information leakage. Therefore, how to ensure the validity and reliability of communication without resorting to the encryption of the upper-layer protocol has become the primary problem to solve the security risks. In view of this, research on the secure transmission of information at the physical layer has become the key to research, and the physical layer security transmission technology, which is a powerful complement to traditional encryption technology, has also received extensive attention.

For the traditional Multiple Send Single Receive (MISO) system, the widely accepted zero-space artificial noise scheme is an effective wireless physical layer security transmission method. The null space artificial noise scheme requires the sender to transmit the useful signal carrying confidential information in the direction of the main channel, and at the same time transmit the artificial noise signal uniformly in the null space of the main channel. By reducing the cost of power utilization at the transmitting end, a gap in transmission performance is artificially introduced between legal channels and illegal channels.

However, in an actual TDD system, since the result of channel estimation is based on channel reciprocity, the uplink end uses the received signal to estimate the uplink channel, and uses the estimated value of the uplink channel as the estimation result of the downlink channel. Due to inaccurate estimation, incomplete reciprocity of uplink and downlink channels, and inter-channel errors that exist objectively in the actual system, there will be a gap between the channel estimation results and the real channel response. For the main lobe signal, such an error is not critical, and the signal gain in the main lobe direction is still guaranteed. However, for the artificial noise scheme, using the null space weights generated by the channel estimation results with errors for transmission will introduce noise in the direction of legitimate users. Due to the characteristics of artificial noise nulling, the additionally introduced noise will have a significant impact on the original characteristics.

The traditional engineering scheme is based on channel calibration, and is dedicated to correcting the error between the multi-antenna transmitting and receiving channels to meet the channel reciprocity condition, so that the channel estimation result is close to the real channel response. However, traditional offline and online calibration, wired and wireless calibration, all require additional equipment and more tedious processing in terms of implementation, and the accuracy of calibration is limited by the accuracy of existing equipment, and it is difficult to achieve a high level. It is even more difficult to deal with the influence of estimation errors and slow channel changes other than inter-channel errors, and it is difficult to ensure that artificial noise does not affect the information transmission in the direction of legitimate users.

The above facts show that in the actual communication engineering, there is still a lack of a truly simple, effective and practical method for correcting the artificial noise weights of the MISO system.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a method for iterative and low-bit feedback of distributed beam weights at the transmitting end in view of the above problems in the prior art, which can realize the iterative method through single-parameter feedback without using additional equipment and communication time slots. The method realizes the correction of transmission weights, and provides a low-bit quantization feedback scheme to enhance the safe transmission performance of artificial noise. Purpose.

In order to achieve the above purpose, the method for iterative and low-bit feedback of the distributed beam weight at the transmitting end of the present invention includes the following steps:

在MISO视距模型下，认为信道慢变，信道估计值

为Nx1维列向量。根据已有信息生成主波束权值w和N-1组人工噪声权值q_i(i＝1，...，N-1)，在发射信号时同时发送人工噪声对有用信号进行保护。Step 1, Alice, the N-antenna sender, establishes a communication link with the single-antenna receiver Bob according to the access information, and obtains the initial channel estimation result by receiving the uplink pilot signal.

Under the MISO line-of-sight model, the channel is considered to change slowly, and the channel estimation value

is an Nx1 dimensional column vector. The main beam weight w and N-1 groups of artificial noise weights q _i (i=1, .

Step 2, after the single-antenna receiving end Bob contains the artificial noise signal at the receiving end, calculates the error between the received signal and the known signal according to the known pilot signal, and performs differential quantization on the error. Feed back the quantized result to Alice.

Step 3, the N-antenna sender Alice uses the received data to obtain the current error value, corrects the weight of the main beam according to the error value, and updates the main beam and the artificial noise beam according to the corrected result. Continue to send the next time slot signal using the updated weight, and go to step 2.

其中，s(t)代表承载用于训练的已知信号，y_i(t)代表人工噪声信号服从均值为零，能量为1的标准高斯分布，σ_i为第i个人工噪声y_i(t)的复幅度，h为信道真实响应，w为发射有用信号的主波束，在初始的时候为信道估计值

即

q_i为第i个人工噪声权值，取自于当前主波束权值的零空间

则其又可以表示为q_i＝U₀(：，i)，n(t)为环境噪声。在视距环境下，认为信道真实响应h在迭代收敛的过程中保持不变。The communication system on which the weight correction method of the present invention is based includes an N-antenna sender Alice and a single-antenna expected receiver Bob, and the two form a communication pair through access. In the TDD communication mode, the N-antenna sender Alice sends a signal containing artificial noise, and the signal received by the single-antenna receiver Bob can be expressed as

Among them, s(t) represents the known signal bearing for training, y _i (t) represents the artificial noise signal obeying the standard Gaussian distribution with mean zero and energy 1, σ _i is the ith artificial noise y _i (t ), h is the real response of the channel, w is the main beam transmitting the useful signal, and it is the channel estimation value at the initial time

which is

q _i is the ith artificial noise weight, which is taken from the null space of the current main beam weight

Then it can be expressed as qi =U ₀ (:, _i ), and n(t) is the environmental noise. In the line-of-sight environment, the true channel response h is considered to remain unchanged during the iterative convergence process.

生成主波束的初始值w₀，并相应生成零空间Ω₀来得到人工噪声的权值q_i，与传统人工噪声生成方案一致。The described step 1 is to use the channel estimation value in the access slot

The initial value w ₀ of the main beam is generated, and the null space Ω ₀ is generated accordingly to obtain the weight q _i of the artificial noise, which is consistent with the traditional artificial noise generation scheme.

The step 2 is that the receiving end Bob generates the quantized bit of the return error according to the known signal after receiving the signal, and the specific operation steps are as follows:

对于Bob而言有用信号s(t)完全已知，由此可以得到接收信号中的误差项e(t_k)＝x(t)-s(t)。根据接收信号的表达不难得到

这里Λ₀＝diag(σ₁，σ₂，..，σ_N-1)且Y(t)＝(y₁(t)，y₂(t)...，y_N-1(t))。Step 2.1, the single-antenna receiver Bob receives the pilot signal containing artificial noise from Alice, and the received signal is:

The useful signal s(t) is completely known to Bob, from which the error term e(t _k )=x(t)-s(t) in the received signal can be obtained. According to the expression of the received signal, it is not difficult to obtain

Here Λ ₀ =diag(σ ₁ ,σ ₂ ,...,σ _N-1 ) and Y(t)=(y ₁ (t),y ₂ (t)...,y _N-1 (t)) .

Assuming that there are known signals s(t _k ) and error terms e(t _k ) respectively in a certain time slot, calculate the cross-correlation value Er _k =E{s(t _k )e( t _k ) ^* } and normalize it by the energy of the useful signal s(t _k ).

Step 2.2: Perform Q-format quantization on the real and imaginary parts of the normalized complex number Er _k respectively, and perform L-bit truncation on the quantized bit sequence. L represents the effective bit length of the real part or the imaginary part when the sign bit is not included, and the specific value depends on the noise introduced and the environmental signal-to-noise ratio SNR. According to the rough calculation results of the quantization length and the equivalent quantization noise, when L>SNR/6, it can be considered that the introduced quantization noise is lower than the environment and has no effect on the subsequent convergence performance.

The error bit sequence after the truncated real part/imaginary part is E _k [l], and the bit sequence obtained by the difference with the bit sequence E _k-1 [l] at the previous moment is D _k [l]. Due to the convergent nature of the algorithm, the high-order digits of the differential bits will quickly become 0 during the iteration and will not change in the subsequent iterations. If the effective number of bits of the differential bit sequence is at the lower M bits at this time, then the M bits are truncated and transmitted. The selection rule of M is determined according to the error Er ₀ of the initial weight, and the maximum representation value d _max of the last M digits should be equal to Er ₀ . For the differential data that exceeds the expression range after M is the direct acquisition sequence, it can be returned at the same time. For sequences whose differential data exceeds the expression range of the last M bits, they are split according to the following rules, and are fed back in sequence in multiple time slots, and no additional calculation and quantization are performed during the multiple feedback process:

When the data value of the real part or imaginary part of the difference sequence returned in a certain time is greater than the maximum value d _max expressed by the last M bits, the Q with the larger value is divided into Q=n*d _max +q, where q is the value that can be M is the numerical value expressed after use. In the subsequent return process, the sequence of all 1s of M bits is returned n times, and the sequence corresponding to one transmission q occupies n+1 times of return, so that the processing of the differential value at the receiving end always maintains the addition operation. The other side of the real and imaginary parts is split in the same way, and 0 is added for transmission.

It is worth noting that because the all-one sequence has a special meaning in the number of times, when the error data just corresponds to the all-one sequence and does not exceed the expression range, the last bit of the corresponding differential sequence is modified to 0.

In step 2.3, the single-antenna receiver Bob sends back the generated real-imaginary difference bit sequence through the uplink time slot.

The step 3 is that the N-antenna sender Alice obtains the error data Er _k after receiving the differential bit sequence of the Bob end, and calculates the main beam weight and the artificial noise beam weight of the next time slot accordingly. The specific operation steps as follows:

Step 3.1, Alice, the sender of the N antenna, calculates the difference result _Δk according to the received difference bits under the premise that the L bits are differentially quantized and then the M bits are intercepted, and accumulates the error data Er, that is, Erk ₌ Er _k-1 + _Δk .

At this time, different operations are taken according to the difference of received differential bits: if the received bit sequence contains all 1s, go to step 3.3, that is, the weights are not updated; if the received bit sequence does not contain all 1s, Go to step 3.2, that is, correct the weights.

其中w_k为当前波束形成的权值，w_k+1为更新后的权值，μ为迭代步长，U_0，k为当前权值的零空间即当前的人工噪声权值空间，Λ₀＝diag(σ₁，σ₂，..，σ_N-1)为人工噪声的能量分配，Y(t)＝(y₁(t)，y₂(t)...，y_N-1(t))为各路人工噪声序列。在更新完主波束权值w_k+1后，同样生成响应的零空间U_0，k+1，更新各路人工噪声权值并转到步骤3.3。Step 3.2, after the receiving end obtains the error value Er _k at this time according to the feedback bits, it updates the original beamforming weight, and calculates

where w _k is the weight of the current beamforming, w _k+1 is the updated weight, μ is the iteration step, U _{0, k} is the null space of the current weight, that is, the current artificial noise weight space, Λ ₀ =diag(σ ₁ ,σ ₂ ,...,σ _N-1 ) is the energy allocation of artificial noise, Y(t)=(y ₁ (t), y ₂ (t)..., y _N-1 ( t)) is the artificial noise sequence of each channel. After updating the main beam weight w _k+1 , the corresponding null space U _0,k+1 is also generated, and the artificial noise weights of each channel are updated and go to step 3.3.

Step 3.3, according to the main beam weight w and the artificial noise weight U ₀ , transmit the signal of the next time slot. The communication of the next time slot is performed, and the operation goes to step 2 to iteratively revise the weights.

Compared with the prior art, the present invention has the following advantages: the computational complexity is low, and the real-time requirements of communication can be met in actual use; the time slot design is suitable for mainstream communication, and under the traditional TDD time slot communication structure, it does not occupy Additional time-frequency code domain resources; the design starts from the background of artificial noise, makes full use of multi-antenna airspace resources, gives full play to the role of multi-beam, and can cover pilot frequency slots with artificial noise to enhance safety performance. That is, the present invention can overcome the influence of channel mismatch with minimum resources and optimize the performance of artificial noise beams without changing the existing TDD communication framework.

Description of drawings

Fig. 1 is a schematic diagram of the iterative correction process flow of artificial noise weights under MISO time-division communication of the present invention

Fig. 2 is the flow chart of the differential quantization method of the present invention

Fig. 3 is the convergence curve of the SINR output of the receiving end Bob with the number of iterations under the given conditions of the present invention

Detailed ways

The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

As shown in FIG. 1 , the communication system adopted in the present invention includes an N-antenna sender Alice, a single-antenna expected receiver Bob, and the dual-transmission uplink and downlink communication is time-sharing according to the TDD communication method. The pilot sequence at the front end of the transmission slot is completely known.

其中，s(t)代表承载用于训练的已知信号，y_i(t)代表人工噪声信号服从均值为零，能量为1的标准高斯分布，σ_i为第i个人工噪声y_i(t)的复幅度，h为信道真实响应，w为发射有用信号的主波束，在初始的时候为信道估计值

q_i为第i个人工噪声权值，取自于当前主波束权值的零空间

则其又可以表示为q_i＝U₀(：，i)，n(t)为环境噪声。In the TDD communication mode, the N-antenna sender Alice sends a signal containing artificial noise, and the signal received by the single-antenna receiver Bob can be expressed as

Among them, s(t) represents the known signal bearing for training, y _i (t) represents the artificial noise signal obeying the standard Gaussian distribution with mean zero and energy 1, σ _i is the ith artificial noise y _i (t ), h is the real response of the channel, w is the main beam transmitting the useful signal, and it is the channel estimation value at the initial time

q _i is the ith artificial noise weight, which is taken from the null space of the current main beam weight

Then it can be expressed as qi =U ₀ (:, _i ), and n(t) is the environmental noise.

So far, the initial weight generation in step 1 is completed, which is consistent with the traditional method for generating artificial noise weights.

此时对于Bob对于有用信号s(t)完全已知，由此可以得到接收信号中的误差项e(t_k)＝x(t)-s(t)。之后具体计算误差值与差分量化的操作步骤如下：In the process of downlink communication, after receiving the signal, Bob at the single-antenna receiving end generates back-transmission error quantization bits according to the known signal. The single-antenna receiver Bob receives the pilot signal containing artificial noise from Alice, and the received signal is

At this time, Bob's useful signal s(t) is completely known, and thus the error term e(t _k )=x(t)-s(t) in the received signal can be obtained. After that, the specific operation steps for calculating the error value and differential quantization are as follows:

Step 2.1, assuming that there is a known signal s(t _k ) and an error term e(t _k ) in a certain communication, calculate the cross-correlation value Er _k =E{s(t _k ) of the known signal and the error term at this time e(t _k ) ^* }, normalized by the energy of the wanted signal s(t _k ).

Step 2.2: Perform Q-format quantization on the real and imaginary parts of the normalized complex number Er _k respectively, and perform L-bit truncation on the quantized bit sequence. L represents the effective bit length of the real part or the imaginary part when the sign bit is not included, and the specific value depends on the noise introduced and the environmental signal-to-noise ratio SNR. According to the rough calculation results of the quantization length and the equivalent quantization noise, when L>SNR/6, it can be considered that the introduced quantization noise is lower than the environment and has no effect on the subsequent convergence performance.

The error bit sequence after the truncated real part or imaginary part is E _k [l], and the bit sequence obtained by difference with the bit sequence E _k-1 [l] at the previous moment is D _k [l]. Due to the convergent nature of the algorithm, the high-order digits of the differential bits will quickly become 0 during the iteration and will not change in the subsequent iterations. If the effective number of bits of the differential bit sequence is at the lower M bits at this time, then the M bits are truncated and transmitted. The selection rule of M is determined according to the error Er ₀ of the initial weight, and the maximum representation value d _max of the last M digits should be equal to Er ₀ . For differential data that exceeds the range of expression after M is directly taken, it will be returned at the same time. Among them, when the error data just corresponds to the all-one sequence and does not exceed the expression range, the last bit of the corresponding differential sequence is modified to 0. For sequences whose differential data exceeds the expression range of the last M bits, they are split according to the following rules, and are fed back in sequence in multiple time slots, and no additional calculation and quantization are performed during the multiple feedback process:

When the data value of the real part or imaginary part of the difference sequence returned in a certain time is greater than the maximum value d _max expressed by the last M bits, the Q with the larger value is divided into Q=n*d _max +q, where q is the value that can be M is the numerical value expressed after use. In the subsequent return process, the sequence of all 1s of M bits is returned n times, and the sequence corresponding to one transmission q occupies n+1 times of return, so that the processing of the differential value at the receiving end always maintains the addition operation. The other side of the real and imaginary parts is split in the same way, and 0 is added for transmission.

Step 2.3, the single-antenna receiver Bob sends back the generated real part and imaginary part differential bit sequence through the uplink time slot.

In the process of uplink communication, after receiving the error difference bits fed back by Bob, the sender Alice of the N antenna accumulates the error data Er _k , and calculates the main beam weight and artificial noise beam weight of the next time slot accordingly. value. The specific operation steps for accumulating error data and updating weights are as follows:

Step 3.1, Alice, the sender of the N antenna, calculates the difference result _Δk according to the received difference bits under the premise that the L bits are differentially quantized and then the M bits are intercepted, and accumulates the error data Er, that is, Erk ₌ Er _k-1 + _Δk .

At this time, different operations are taken according to the difference of received differential bits: if the received bit sequence contains all 1s, go to step 3.3, that is, the weights are not updated; if the received bit sequence does not contain all 1s, Go to step 3.2, that is, correct the weights.

其中w_k为当前波束形成的权值，w_k+1为更新后的权值，μ为迭代步长，U_0，k为当前权值的零空间即当前的人工噪声权值空间，Λ₀＝diag(σ₁，σ₂，..，σ_N-1)为人工噪声的能量分配，Y(t)＝(y₁(t)，y₂(t)...，y_N-1(t))为各路人工噪声序列。在更新完主波束权值w_k+1后，同样生成响应的零空间U_0，k+1，更新各路人工噪声权值并转到步骤3.3。Step 3.2, after the receiving end obtains the error value Er _k at this time according to the feedback bits, it updates the original beamforming weight, and calculates

where w _k is the weight of the current beamforming, w _k+1 is the updated weight, μ is the iteration step, U _{0, k} is the null space of the current weight, that is, the current artificial noise weight space, Λ ₀ =diag(σ ₁ ,σ ₂ ,...,σ _N-1 ) is the energy allocation of artificial noise, Y(t)=(y ₁ (t), y ₂ (t)..., y _N-1 ( t)) is the artificial noise sequence of each channel. After updating the main beam weight w _k+1 , the corresponding null space U _0,k+1 is also generated, and the artificial noise weights of each channel are updated and go to step 3.3.

Step 3.3, according to the main beam weight w and the artificial noise weight U ₀ , transmit the signal of the next time slot.

与信道真实响应h之间存在10％的增益误差与5度相位误差，环境信噪比30dB条件下的优化性能。其中每帧导频长度为127bit，截位长度选取6bit，差分长度选取3bit，故每次反馈数据长度为8bit。原本的人工噪声性能只能在输出端信干噪比20dB左右。It is not difficult to see that the iterative process of this algorithm is consistent with the uplink and downlink transmission modes, no additional convergence conditions are required, and it has a tracking effect on the slowly changing channel in the process of constant correction. By simulating this method in the channel estimation

There is a gain error of 10% and a phase error of 5 degrees between the actual channel response h, and the optimal performance under the condition of an environmental signal-to-noise ratio of 30dB. The pilot frequency of each frame is 127 bits long, the truncation length is 6 bits, and the difference length is 3 bits, so the length of each feedback data is 8 bits. The original artificial noise performance is only about 20dB at the output signal-to-interference noise ratio.

It can be seen from Fig. 3 that under the above conditions, the method of the present invention is used to correct the artificial noise weights, so that the output signal-to-interference-noise ratio of the Bob end converges to the environmental signal-to-noise ratio after a limited number of iterations, that is, the transmitted artificial noise is for The communication performance of the legitimate receiving end is not affected, and the price paid is 8 return bits per time slot.

To sum up, the theoretical analysis of the technical solution and the simulation results can verify the effective effect of the method of the present invention in modifying the artificial noise weight and optimizing the transmission performance.

The above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention in any form. Those skilled in the art should understand that, without departing from the spirit and principles of the present invention, the present invention can also Several modifications or simple substitutions may be made, which also fall within the scope of the appended claims.

Claims (5)

1. An artificial noise weight iteration correction and low bit feedback method of a multi-transmitter-receiver system is characterized by comprising the following steps:

step 1, an N-antenna sender Alice establishes a communication link with a single-antenna receiver Bob according to access information and obtains an initial channel estimation result by receiving an uplink pilot signal

Under the MISO line-of-sight model, the channel is considered to be slowly changed, and the channel estimation value

Generating a main beam weight w and N-1 groups of artificial noise weights q for an Nx1 dimensional column vector according to the existing information_i(i 1.., N-1), and when transmitting a signal, simultaneously sending artificial noise to protect a useful signal;

step 2, after the single-antenna receiving end Bob contains artificial noise signals at the receiving end, calculating errors between the received signals and the known signals according to the known pilot signals, carrying out differential quantization on the errors, and feeding back quantized results to the Alice end;

and 3, using the received data to obtain a current error value by the Alice of the N-antenna sender, correcting the weight of the main beam according to the error value, updating the main beam and the artificial noise beam according to the corrected result, and using the updated weight to continuously send a signal of the next time slot.

2. The artificial noise weight iterative correction and low bit feedback method for multi-transmission single-reception system according to claim 1, wherein: the communication system comprises an N-antenna sender Alice and a single-antenna expected receiver Bob;

in TDD communication mode, the N-antenna sender Alice sends a signal containing artificial noise, and the signal received at the single-antenna receiver Bob can be represented as

Where s (t) represents the carrier of a known signal for training, y_i(t) represents the standard Gaussian distribution of the artificial noise signal with mean value of zero and energy of 1, sigma_iIs ith artificial noise y_i(t) complex amplitude, h is the true response of the channel, w is the main beam transmitting the desired signal, and initially the channel estimate

Namely, it is

q_iIs the ith artificial noise weight, which is taken from the null space of the current main beam weight

It may be represented again as q_i＝U₀And (i), n (t) is environmental noise, and in a line-of-sight environment, the real response h of the channel is considered to be kept unchanged in the iterative convergence process.

3. The artificial noise weight iterative correction and low bit feedback method for multi-transmission single-reception system according to claim 1, wherein: step 2 and step 3 are to correct the weight value by feeding back the error and feed back the error information by using a small number of bits.

4. The iterative artificial noise weight correction and low bit feedback method according to claim 3, wherein the method for correcting the weight by feeding back the error data comprises the following steps:

step 2.1, the single-antenna receiver Bob receives a pilot signal containing artificial noise from the Alice terminal, and the received signal is

The useful signal s (t) is completely known to Bob, from which the error term e (t) in the received signal can be derived_k) X (t) -s (t), which is not difficult to obtain from the expression of the received signal

Here Λ₀＝diag(σ₁，σ₂，..，σ_N-1) And y (t) ═ y₁(t)，y₂(t)...，y_N-1(t))；

Suppose there is a known signal s (t) under a certain time slot_k) And error term e (t)_k) And calculating the cross-correlation value Er of the known signal and the error term at the moment_k＝E{s(t_k)e(t_k)^*And applying a useful signal s (t) thereto_k) Normalizing the energy, and feeding the error data back to an Alice end;

step 3.2, the receiving end obtains the error value Er at the moment according to the feedback bit_kThen, the original beam forming weight value is updated and calculated

Wherein w_kAs a weight of the current beam forming, w_k+1For the updated weight, μ is the iteration step, U_0，kThe null space of the current weights, i.e. the current artificial noise weight space, Λ₀＝diag(σ₁，σ₂，..，σ_N-1) For the energy allocation of artifacts, y (t) ═ y₁(t)，y₂(t)...，y_N-1(t)) is the artificial noise sequence of each path, and the main beam weight w is updated after the updating_k+1Then, the same appliesNull space U for generating response_0，k+1。

5. The method for iterative modification of artificial noise weight and low bit feedback in a multi-transmission single-reception system according to claim 3, wherein the error information is transmitted through a small number of bits, comprising the steps of:

step 2.2, the normalized complex number Er_kRespectively carrying out Q-format quantization on the real part and the imaginary part, carrying out L-bit truncation on a bit sequence after quantization, wherein L represents the effective bit length which is obtained under the condition that the real part or the imaginary part does not comprise a sign bit, and the specific value of the effective bit length depends on introduced noise and an environment signal-to-noise ratio (SNR). according to the rough calculation result of the quantization length and equivalent quantization noise, when L is more than SNR/6, the introduced quantization noise is considered to be lower than the environment, and the convergence performance after the quantization is not influenced;

the error bit sequence after the real part/imaginary part truncation is E_k[l]Bit sequence E from the previous instant_k-1[l]And the bit sequence obtained by the difference is D_k[l]After interception, M bits are transmitted, and the selection rule of M is according to the error Er of the initial weight₀Determining the maximum representation d of the last M digits_maxTo be reacted with Er₀Equivalently, for the sequence of the differential data which is directly taken when the differential data exceeds the expression range, M is the expression range, the sequence of the differential data which exceeds the expression range of M bits is split according to the following rule, the feedback is carried out in sequence in a plurality of time slots, and no additional calculation and quantization are carried out in the process of multiple feedback:

when the real part or imaginary part data value of a certain returned difference sequence is larger than the maximum value d expressed by the M rear bits_maxDividing the larger Q into Q ═ n × d_max+ q, where q is a numerical value that can be expressed by the last M, and n times of returning M-bit full 1 sequence and a sequence corresponding to 1 time of q transmission occupy n +1 times of returning, so that the processing of the receiving end on the differential numerical value always keeps addition operation, and the other side of the real and imaginary parts is split in the same way to complement 0 transmission;

it is worth noting that because the all-1 sequence has a special meaning in the times, when the error data just corresponds to the all-1 sequence and does not exceed the expression range, the tail bit of the corresponding differential sequence is modified to be 0;

step 3.1, the sender Alice of the N antenna calculates a difference result delta according to the received difference bits under the premise that the sender Alice of the N antenna intercepts M bits after the known L bit difference is quantized_kAnd accumulating error data Er, i.e. Er_k＝Er_k-1+Δ_k；

At this time, different operations are taken according to different received differential bits: if the received bit sequence contains the all 1 sequence, turning to step 3.3, namely not updating the weight; if the received bit sequence does not contain all 1 bit, go to step 3.2, i.e. modify the weight.

Images