A one-bit MIMO signal detection method of fast local search

Abstract

The application provides a one-bit MIMO signal detection method of fast local search, and belongs to the technical field of signal detection.The one-bit MIMO signal detection method of fast local search of the application compensates the loss of received signal information caused by using a one-bit quantizer through the diversity gain of a large-scale MIMO system, takes the maximum likelihood signal detection algorithm with the optimal theoretical performance as a theoretical basis, avoids the high calculation complexity required by the traditional maximum likelihood signal detection algorithm which needs to search the whole set of constellation points through the idea of local search, and further reduces the calculation complexity of calculating the likelihood through function fitting.The simulation result shows that, compared with the existing signal detection algorithm, the fast local search signal detection algorithm provided by the application can greatly reduce the calculation complexity of the algorithm.

Description

TECHNICAL FIELD

[0001] The present application relates to a signal detection method of a wireless communication system, in particular, a signal detection method based on fast local search for a large-scale Multiple-in Multiple-out (MIMO) scattering communication system using a one-bit Analog-to-Digital Converter (ADC). BACKGROUND

[0002] The tropospheric scattering communication technology is a wireless communication technology using the scattering effect of the unevenness of the troposphere on radio waves, and its technical basis is the scattering propagation theory of the troposphere. The tropospheric scattering transmission system designed using the scattering propagation theory of the troposphere can realize over-the-horizon transmission, and has moderate transmission capacity, transmission performance and reliability, and strong anti-nuclear explosion and anti-ionosphere disturbance ability. The tropospheric scattering communication technology has an irreplaceable role in various communication transmission technologies due to its unique transmission characteristics.

[0003] The scattering communication has a development history of more than 60 years since its inception, and the capacity problem of the scattering communication has become the focus of attention. With the gradual application of the fifth generation mobile communication technology, high-throughput satellite technology and millimeter wave large-capacity microwave communication technology in communication, the transmission rate of the next generation communication system will be improved as a whole. The scattering communication also needs to further improve the transmission rate to avoid becoming a bottleneck of the whole network communication transmission.

[0004] The large-scale MIMO technology is one of the important technologies for improving the system capacity of the current mobile communication. Compared with the traditional single antenna, the theoretical channel capacity of the large-scale MIMO communication system increases with the increase of the number of antennas at the receiving and transmitting ends without the need for additional transmission power or frequency band. Therefore, it can be predicted that if the large-scale MIMO technology is introduced into the scattering communication, it can also provide considerable communication rate for the scattering communication system, thereby realizing large-capacity scattering communication.

[0005] With the increase of the number of antennas, the hardware implementation of the large-scale MIMO system becomes very complex, especially in the ADC unit of each receiving antenna. When configuring high-rate and high-complexity ADC, the exponential growth of resource and energy consumption is the main bottleneck for deploying a large-scale MIMO system, so combining a low-bit (usually 1-3 bit unit) quantizer with a large-scale MIMO technology can reduce the cost and power consumption of the system. Although the traditional maximum likelihood one-bit quantization signal detection algorithm has the best theoretical performance, it has an exponentially high computational complexity, which is difficult to use in actual communication systems. SUMMARY

[0006] In order to solve the problem of high computational complexity of signal detection algorithm for one-bit quantization based large-scale MIMO scattering communication system, the application provides a one-bit MIMO signal detection method with fast local search.

[0007] In order to achieve the above object, the technical scheme adopted by the application is:

[0008] A one-bit MIMO signal detection method with fast local search comprises a sending end operation process and a receiving end operation process.

[0009] The sending end operation process comprises the following steps:

[0010] Step 1: The sending end modulates the bit stream to be sent into a symbol sequence Wherein N s is the number of data streams to be sent, is a complex number set;

[0011] Step 2: The symbol sequence x is pre-coded to obtain a pre-coded symbol sequence z=P x, wherein the pre-coding matrix is defined as

[0012]

[0013] Wherein, N t represents the number of sending antennas, 1 N represents a vector with a dimension of N and all elements being 1, represents the floor operation on x;

[0014] Step 3: The symbol sequence z is sent to the receiving end, and after passing through a channel , a received signal is obtained

[0015] y=Q1(Hz+n)

[0016] Wherein N r is the number of receiving antennas, n~CN(0,σ 2 I) represents additive white Gaussian noise, σ 2 is the noise power, and Q1(·) represents a one-bit quantization operator

[0017] Q1(x)=sgn(Re{x})+jsgn(Im{x})

[0018] Wherein Re{x} and Im{x} respectively represent the real part and the imaginary part of x, sgn(·) represents a sign function, and is defined as

[0019]

[0020] The receiving end operation process includes the following steps:

[0021] Step 4: Let Calculate Where represents the pseudo-inverse matrix of ;

[0022] Step 5: For x ZF , the nth term x ZF (n) of the constellation point closest to x ZF (n) is found:

[0023]

[0024] Where, represents the set of modulation constellation points;

[0025] Search for constellation points within a distance of c units from x ZF (n), and form a set:

[0026]

[0027] Where c = 2.5 is a constant;

[0028] Step 6: Define the set

[0029]

[0030] Define

[0031]

[0032] And

[0033] y R = sgn(G R x R + n R )

[0034] Where

[0035]

[0036] Here n(n) represents the nth term of the noise vector n;

[0037] Solve

[0038]

[0039] Where

[0040]

[0041] Step 7: restore real signal to complex signal

[0042] The present application has the following advantages:

[0043] 1. The present application uses the diversity gain of large-scale MIMO system to compensate for the loss of received signal information caused by using a one-bit quantizer, and provides redundant information by simultaneously transmitting the same signal on multiple antennas, thereby compensating for the performance loss of the signal detection algorithm.

[0044] 2. The present application uses the maximum likelihood signal detection algorithm with the best theoretical performance as the theoretical basis, and uses the idea of local search to avoid the high computational complexity required by the traditional maximum likelihood signal detection algorithm, which requires a full set of constellation points to be searched. The computational complexity of calculating the likelihood function is further reduced by function fitting, i.e. the computational complexity of the algorithm is reduced at the cost of a small amount of performance loss in two aspects of the traditional algorithm. BRIEF DESCRIPTION OF DRAWINGS

[0045] Figure 1 is a structure block diagram of a one-bit quantized large-scale MIMO scattering communication system in an embodiment of the present application;

[0046] Figure 2 is a flow chart of a one-bit quantized large-scale MIMO scattering signal detection algorithm based on fast local search proposed by the present application;

[0047] Figure 3 is a performance comparison chart of different signal detection algorithms (modulation method is QPSK);

[0048] Figure 4 is a performance comparison chart of different signal detection algorithms (modulation method is 8PSK);

[0049] Figure 5 is a comparison chart of average running time of different signal detection algorithms. DETAILED DESCRIPTION

[0050] The present application will be described in detail below with reference to the accompanying drawings.

[0051] A one-bit MIMO signal detection method based on fast local search, which first uses the estimated value obtained by the zero-forcing algorithm as the initial search point, then performs a traversal search in the vicinity of the point, uses a function fitting method to greatly reduce the computational complexity of the search, and uses the constellation point with the maximum likelihood as the estimated value.

[0052] The method considers a one-bit quantized massive MIMO scattering communication system structure as shown in Figure 1 The flow is as shown in Figure 2 The steps of the method are as follows:

[0053] The sending end operation process is as follows:

[0054] Step 1: The sending end modulates the sent bit stream into a symbol sequence Where N s is the number of data streams sent.

[0055] Step 2: The sending symbol sequence x is precoded to obtain a precoded sending symbol sequence z, and the formula is

[0056] z=Px

[0057] Where the precoding matrix is defined as

[0058]

[0059] Where N t represents the number of sending antennas, 1 N represents a vector with a dimension of N and all elements being 1, represents the floor operation on x.

[0060] Step 3: The precoded sending symbol sequence z is sent to the receiving end, and after passing through the channel The receiving signal

[0061] y=Q1(Hz+n)

[0062] Where n~CN(0,σ 2 I) represents an additive white Gaussian noise, N r is the number of receiving antennas, σ 2 is the noise power, and Q1(·) represents a one-bit quantization operator

[0063] Q1(x)=sgn(Re{x})+jsgn(Im{x})

[0064] Where Re{x} and Im{x} represent the real part and the imaginary part of x respectively, and sgn(·) represents a sign function, which is defined as

[0065]

[0066] The receiving end operation process is as follows:

[0067] Step 4: Let Calculate Where represents the pseudo-inverse matrix of .

[0068] Step 5: For x ZF the nth term of x ZF (n), find the constellation point closest to x ZF (n)

[0069]

[0070] where represents the set of modulation constellation points. Search for the constellation point closest to x ZF (n) within c units of distance, to form the set

[0071]

[0072] where c = 2.5 is a constant.

[0073] Step 6: Define the set

[0074]

[0075] Define

[0076]

[0077] and

[0078] y R = sgn(G R x R + n R )

[0079] where

[0080]

[0081] Here n(n) represents the nth term of the noise vector n. Solve for

[0082]

[0083] where

[0084]

[0085] y R (i) represents the ith term of the vector y R , G R (i,:) represents the ith row of the matrix G R ;

[0086] Step 7: Convert the real signal back to a complex signal

[0087] Principle explanation:

[0088] The basic idea of the present application is to firstly obtain an initial detection value by a zero-forcing signal detection algorithm with extremely low computational complexity, then perform local search in the vicinity of the value to find all possible constellation point combinations, and calculate the likelihood functions of all combinations, and finally take the constellation point combination with the largest likelihood as the final estimation value. Since the computational complexity of the traditional standard normal distribution cumulative distribution function is too high in calculating the likelihood, the present application uses the function fitting method to fit the standard normal distribution cumulative distribution function with a sigmoid function with very low computational complexity, thereby reducing the computational complexity of the algorithm.

[0089] The simulation results are shown in Figure 3 、 Figure 4 and Figure 5 , in which the number of antennas at the transmitting and receiving ends is 16, and the number of data streams is 2. The transmitting end transmits signals of 100 time frames, each time frame contains 100 time slots, and each time slot transmits 1 symbol. The channel state information of different time frames is independently generated, and the scattering channel is randomly generated according to the 7-path large-scale MIMO Rayleigh channel model. The simulation is performed for 100 times of Monte Carlo simulation. Figure 3 and Figure 4 The horizontal coordinate is the bit signal-to-noise ratio (expressed in dB), and the vertical coordinate is the bit error rate, in which Figure 3 the modulation mode is QPSK, and Figure 4 the modulation mode is 8PSK; Figure 5 The horizontal coordinate is different signal detection algorithms, and the vertical coordinate is the average running time, and the modulation mode is QPSK. Figure 3 and Figure 4 The simulation results show that the signal detection method of the present application has better detection performance than the existing zero-forcing and gradient descent methods, and the performance is close to that of the two-step gradient descent algorithm which approximately reaches the theoretical optimum. Figure 5 The results show that the computational complexity of the signal detection method proposed by the present application is much lower than that of the gradient descent and two-step gradient descent algorithms. According to the above results, it can be found that compared with the two-step gradient descent algorithm, the fast local search signal detection algorithm proposed by the present application can greatly reduce the computational complexity of the algorithm at the cost of a small performance loss; and compared with the traditional zero-forcing signal detection algorithm, the algorithm proposed by the present application greatly improves the performance of signal detection at the cost of a certain computational complexity.

[0090] In summary, the application compensates the loss of received signal information caused by using a one-bit quantizer through the diversity gain of a large-scale MIMO system, takes the maximum likelihood signal detection algorithm with the optimal theoretical performance as a theoretical basis, avoids the excessively high calculation complexity required by the traditional maximum likelihood signal detection algorithm which needs to perform traversal search on the whole set of constellation points by using the idea of local search, and further reduces the calculation complexity of calculating the likelihood by using the function fitting mode.

Claims

1. A one-bit MIMO signal detection method with fast local search, characterized in that, The method comprises a sending end operation process and a receiving end operation process; The sending end operation process comprises the following steps: Step 1: The transmitter modulates the transmitted bit stream into a sequence of symbols where N s is the number of transmitted data streams, is a complex set; Step 2: Precoding the symbol sequence x to get the precoded symbol sequence z = Px, where the precoding matrix P is defined as is defined as where N t denotes the number of transmit antennas, 1 N denotes a vector of dimension N with all elements equal to 1, denotes the floor operation on x; Step 3: The symbol sequence z is sent to the receiving end through a channel and the received signal is obtained y = Q1 (Hz + n) where N r is the number of receive antennas, n ~ CN(0, σ 2 I) denotes additive white Gaussian noise, σ 2 is the noise power, and Q1(·) denotes a one-bit quantization operator Q1 (x) = sgn (Re{x}) + jsgn (Im{x}) Where Re{x} and Im{x} represent the real part and the imaginary part of x respectively, and sgn(·) represents a sign function defined as The receiving end operation process comprises the following steps: Step 4: Let Compute where denotes the pseudo-inverse matrix; Step 5: For x ZF the nth term x ZF (n), find the nearest constellation point: x ZF (n) wherein, represents a set of modulation constellation points; Search distance x ZF (n) The constellation points not more than c unit distances apart, form a set: Where c = 2.5 is a constant; Step 6: defining a set Define And y R = sgn(G R x R + n R ) Where Here n(n) represents the nth item of the noise vector n; Solve by traversing the set X Where Step 7: Reduction of real signal to complex signal

Images