WO2023236932A1

WO2023236932A1 - Llr value quantification method and apparatus, electronic device, and storage medium

Info

Publication number: WO2023236932A1
Application number: PCT/CN2023/098503
Authority: WO
Inventors: 李薿; 王大飞
Original assignee: 锐捷网络股份有限公司
Priority date: 2022-06-06
Filing date: 2023-06-06
Publication date: 2023-12-14

Abstract

Disclosed in the present application are a log likelihood ratio (LLR) value quantification method and apparatus, an electronic device, and a storage medium. The method comprises: obtaining signal to interference plus noise ratio (SINR) linear values of a plurality of received symbols obtained by demodulation; calculating, on the basis of the SINR linear values of the plurality of received symbols, mutual information of the plurality of received symbols; and determining, according to an average value of the mutual information of the plurality of received symbols, a scaling factor for quantizing a plurality of LLR values, the plurality of LLR values comprising an LLR value of each received symbol obtained by performing calculation on the plurality of received symbols on the basis of an LLR algorithm.

Description

LLR value quantification method, device, electronic equipment and storage medium

Cross-references to related applications

This application requests the priority of the Chinese patent application submitted to the China Patent Office on June 6, 2022, with the application number 202210635279.5 and the invention title "A quantification method and device for LLR values", and the request is filed on September 2, 2022 The priority of the Chinese patent application submitted to the China Patent Office with application number 202211074819.3 and the invention title is "A soft bit quantization processing method, device and electronic equipment", the entire content of which is incorporated into this application by reference.

Technical field

The present application relates to the field of communication technology, and in particular, to a quantification method, device, electronic equipment and computer-readable storage medium for LLR values.

Background technique

In the fifth generation New Radio (English: 5th generation New Radio, abbreviated as: 5G NR) system, the transmitter encodes, scrambles and modulates the information bits to be transmitted to obtain modulation symbols, and then performs layer mapping and After the antenna is mapped, it is modulated into a time domain signal through Orthogonal Frequency Division Multiplexing (English: Orthogonal Frequency Division Multiplexing, referred to as OFDM) technology, and finally the time domain signal is sent to the receiving end. The receiving end will demodulate the received signal, but due to the influence of noise, the existing demodulation method is no longer a simple inverse mapping process, but a log likelihood ratio based on the posterior probability criterion (English: Log Likelihood Ratio (abbreviated as: LLR) to calculate soft bit information. After demodulation a decoding step is required. Since the input of the decoder is a fixed-point number sequence with a fixed saturated bit width, the floating-point LLR value obtained by demodulation needs to be quantized.

Contents of the invention

Each exemplary embodiment of the present application provides a quantification method, device, electronic device, and computer-readable storage medium for LLR values to improve the accuracy of quantization and thereby reduce the bit error rate of the decoding process.

The first aspect provides a quantification method for LLR values, including:

Obtain multiple signal to interference plus noise ratio SINR linear values corresponding to multiple received symbols obtained by demodulation;

Calculate a plurality of mutual information corresponding to the plurality of received symbols based on the plurality of SINR linear values corresponding to the plurality of received symbols; and

Determine a scaling factor for quantizing multiple LLR values based on the average of the multiple mutual information corresponding to the multiple received symbols, where the multiple LLR values include log-likelihood based on the multiple received symbols than each calculated by the algorithm The LLR value of the received symbol.

In this embodiment, the scaling factor is determined based on the average value of the mutual information of the received symbols obtained by demodulation, taking into account different fading scenarios. The impact of the fading channel is converted into a measurable system through the parameter of average mutual information, so as to adjust the quantization process of the LLR value, making the quantization result more accurate, so that the input of the decoder can better adapt to the channel environment, improve decoding performance and reduce bit error rate.

In some embodiments, determining a scaling factor for quantizing the plurality of LLR values based on an average value of mutual information of the plurality of received symbols includes:

The scaling factor is determined according to the average value of the multiple mutual information corresponding to the multiple received symbols, the modulation order from the transmitting end, and the coding rate from the transmitting end.

In this embodiment, the modulation order and coding rate are added to the calculation process of the scaling factor, so that the quantization process not only considers the impact of the fading channel, but also takes into account the impact of the receiving end processing. It can be compatible with different fading channels and multiple modulation and coding methods, improving the adaptability of the solution.

In some embodiments, the method further includes:

Determine the SINR linear gain parameters of the multiple received symbols according to the average value of the multiple mutual information corresponding to the multiple received symbols; and

Determining the scaling factor based on the average value of multiple mutual information corresponding to the multiple received symbols, the modulation order from the transmitting end, and the coding rate from the transmitting end specifically includes:

The scaling factor is determined based on SINR linear gain parameters of the plurality of received symbols, the modulation order and the coding rate.

In some embodiments, the SINR linear gain parameter includes an average of SINR linear values.

In some embodiments, the average value of the mutual information of the multiple received symbols is calculated based on the following formula:

Wherein, I _equal is the average value of the multiple mutual information, M is the number of the multiple received symbols, sinr(m) is the SINR linear value of the m-th received symbol, and F(sinr(m)) is the The mutual information of the mth received symbol, where 1≤m≤M.

In some embodiments, the average value of the SINR linear value is calculated by the following formula:
SINR _equal =F ^-1 (I _equal ),

Wherein, SINR _equal is the average of the SINR linear values of the multiple received symbols, I _equal is the average of the multiple mutual information, and F ^-1 is the inverse function of the F function.

In some embodiments, the average value and modulation order of multiple mutual information corresponding to the multiple received symbols are and coding rate, determine the scaling factor, including:

Obtain a selection function determined in advance through machine learning; and

The average value of the mutual information corresponding to the multiple received symbols, the modulation order and the coding rate are used as independent variables of the selection function, and the dependent variable calculated by the selection function is used as the scaling factor .

In some embodiments, the selection function is obtained by using randomly generated random SINR linear values, random modulation orders and random coding rates as training samples and training based on an unsupervised learning algorithm, wherein the random modulation order and the The random coding rate is a control variable of the unsupervised learning algorithm, and the random SINR linear value is an independent variable of the unsupervised learning algorithm.

In some embodiments, the number of the multiple received symbols is set to be less than the total number of all received symbols obtained by demodulation.

In some embodiments, the method further includes: when the total number of all received symbols received is greater than a preset threshold, setting the number of the multiple received symbols to be less than the number of all received symbols obtained by demodulation. The total amount.

In some embodiments, the multiple received symbols are a set of P received symbols separated by S received symbols among all the received symbols obtained by demodulation, S is an integer greater than or equal to 1, and P is greater than or equal to 1. An integer equal to 1.

In some embodiments, the plurality of SINR linear values have a preset linear relationship with the SINR of a corresponding received symbol among the plurality of received symbols.

In some embodiments, the method is suitable for decoders that need to compute fixed-point LLR values as input.

In some embodiments, the average value of the mutual information of the plurality of received symbols is calculated by one of an arithmetic average, a weighted average, or a geometric average.

In the second aspect, a quantization device for LLR values is provided, including:

An acquisition unit, used to acquire the signal to interference plus noise ratio SINR linear values of the multiple received symbols obtained by demodulation;

A processing unit configured to calculate mutual information of the multiple received symbols based on the SINR linear values of the multiple received symbols;

The processing unit is further configured to determine a scaling factor for quantizing a plurality of LLR values based on an average of mutual information of the multiple received symbols, where the multiple LLR values include: The LLR value of each received symbol calculated by the numerical likelihood ratio algorithm.

In some embodiments, the processing unit is specifically used to:

The scaling factor is determined based on the average value of the mutual information of the plurality of received symbols, the modulation order from the transmitting end, and the coding rate from the transmitting end.

In some embodiments, the processing unit is also used to:

According to the average value of the mutual information of the multiple received symbols, the SINR linear value of the multiple received symbols is determined. average value;

The processing unit, when determining the scaling factor based on the average value of the mutual information of the multiple received symbols, the modulation order and the coding rate, is specifically used to:

The scaling factor is determined based on an average of SINR linear values of the plurality of received symbols, the modulation order and the coding rate.

In some embodiments, the processing unit is specifically used to:

Obtain the selection function determined in advance through machine learning through the acquisition unit;

The average value of the mutual information of the multiple received symbols, the modulation order and the coding rate are used as independent variables of the selection function, and the calculated dependent variable is used as the scaling factor.

In a third aspect, an electronic device is provided, including a controller and a memory. The memory is used to store computer-executed instructions, and the controller executes the computer-executed instructions in the memory to utilize the hardware resources in the controller to perform the operational steps of any method that may be implemented in the first aspect.

In a fourth aspect, a computer-readable storage medium is provided. The computer-readable storage medium stores instructions, which when run on a computer, cause the computer to execute the methods of the above aspects.

In addition, the beneficial effects of the second to fourth aspects can be referred to the beneficial effects described in the first aspect, and will not be described again here.

This application also provides a soft bit quantization processing method, device, electronic equipment and storage medium to optimize the decoding performance of the decoder and improve the accuracy of the final decoding result.

In a fifth aspect, this application provides a soft bit quantization processing method, which method includes:

Obtain the probability density distribution of the log-likelihood ratio LLR of the received symbol under the target signal-to-noise ratio; wherein the probability density distribution includes the uncertainty probability corresponding to each floating point value in the LLR, and the probability density distribution is Probability density distribution under additive white Gaussian Noise (English: Additive White Gaussian Noise, abbreviation: AWGN) conditions;

Based on the probability density distribution, N mapping values of the LLR are determined; where N is the number of LLR values of the received symbols;

The target signal-to-noise ratio is used to perform quantization processing on the N mapping values to obtain a quantization processing result.

Through the above method, taking into account the changes in signal-to-noise ratio in different fading scenarios, and then mapping the LLR in K sub-intervals based on the probability density distribution under the target signal-to-noise ratio, obtaining N mapping values, and mapping them based on the target signal-to-noise ratio. The above N mapping values are quantized and finally the quantization result is obtained. Inputting the quantization result obtained in this way into the decoder helps to optimize the decoding performance of the decoder and improve the accuracy of the final decoding result.

In a possible implementation, the target signal-to-noise ratio is obtained based on the following formula:

Among them, SNR is the target signal-to-noise ratio, E _s is the average symbol energy at the transmitter, T _s is the symbol period, N ₀ is the noise energy of the AWGN, and B _n is the noise bandwidth.

In a possible implementation, the E _s and the N ₀ are calculated based on the bit error rate inversion under the AWGN condition:

If the modulation method is BPSK, the bit error rate is calculated based on the following formula:

If the modulation method is QPSK, the bit error rate is calculated based on the following formula:

If the modulation method is M-QAM, the bit error rate is calculated based on the following formula:

Among them, Q() is the Q function, and the calculation formula of the Q function is as follows:

Among them, erfc() is the Gaussian complement function, and the calculation formula of the Gaussian complement function is as follows:

Through the above method, we further consider the actual fading scenario, select the theoretically optimal target signal-to-noise ratio, and then obtain the LLR quantification processing result based on this target signal-to-noise ratio. This also helps to optimize the decoding performance of the decoder and improve the decoder. decoding accuracy.

In a possible implementation, determining the N mapping values of the LLR includes: obtaining K sub-intervals of continuous distribution; wherein the K sub-intervals are divided by the saturated bit width of the decoder; dividing the LLR is mapped in the K sub-intervals to obtain mapped N mapping values.

Through the above method, considering the different saturated bit widths of different decoders, the LLR is mapped into K sub-intervals divided based on the saturated bit width, and N mapping values are obtained. Further, the quantization scaling factor is used to map the N mapping values. Processing can be applied to different fading scenarios to optimize decoding performance under different fading and long periods of silence.

In a possible implementation, mapping the LLR in the K sub-intervals includes: for a single sub-interval in the K sub-intervals, using the following formula to calculate the mapping in the single sub-interval R The mapping value y _n of the LLR of _i = {a _i-1 , a _i }, i = 1, 2,...K-1, then

If l∈R _K =[a _K-1 ,∞], then y _n =sign(x _n )·y _K-1 ;

Among them, x _n is the n-th floating point value of the LLR, sign() is the sign function of x _n , a _i is the right endpoint of the single interval, a _i-1 is the left endpoint of the single interval, l is the absolute value of x _n , p() is the target SNR under AWGN conditions probability density distribution.

Through the above method, considering the different saturated bit widths of different decoders, the LLR is mapped into K sub-intervals divided based on the saturated bit width, and N mapping values are obtained. The quantization scaling factor is further used to perform these N mapping values. The processing can be applied to different fading scenarios to optimize decoding performance under different fading conditions.

In a possible implementation, using the target signal-to-noise ratio to perform quantization processing on the N mapping values to obtain a quantization processing result includes: obtaining the linear signal-to-noise ratio of the received symbol, and Target signal-to-noise ratio; calculate the ratio of the linear signal-to-noise ratio to the target signal-to-noise ratio, and use the ratio as a quantized scaling factor; use the quantized scaling factor to scale the N mapping values Processing, using the result of scaling processing as the result of quantization processing.

Through the above method, the problem of the existing technology to determine the quantized scaling factor based on artificial experience can be solved, and the quantized scaling factor can be determined adaptively, and the scaling factor obtained in this way can be adapted to different fading scenarios, and can be implemented in these scenarios. Optimize the decoding performance.

In a possible implementation, using the quantized scaling factor to perform scaling processing on the N mapping values respectively, and using the scaling processing results as the quantization processing results includes: for the N mapping values For a single mapping value in , perform the following processing operations: calculate the product between the quantized scaling factor and the single mapping value, round it, and use the rounding result as the result of the scaling process of the single mapping value; Repeat the above processing operation to obtain the scaling processing results corresponding to each of the N mapping values, and use the scaling processing results as the quantization processing results.

Through the above method, considering the changes in signal-to-noise ratio in different fading scenarios, the LLR performs different mapping and scaling processes, and then performs rounding processing, which can enable the decoder to better identify the size relationship between LLRs, thus improving decoding. performance.

In a sixth aspect, the present application provides a soft bit quantization processing device, which includes:

The acquisition module obtains the probability density distribution of the log-likelihood ratio LLR of the received symbol under the target signal-to-noise ratio; wherein the probability density distribution includes the uncertainty probability corresponding to each floating point value in the LLR, and the probability density The distribution is the probability density distribution under the condition of additive Gaussian white noise AWGN;

The determination module determines N mapping values of the LLR based on the probability density distribution; where N is the number of LLR values of the received symbols;

A processing module uses the target signal-to-noise ratio to perform quantization processing on the N mapping values to obtain a quantization processing result.

In a possible implementation, the N mapping values of the LLR are determined, and the determination module is specifically configured to: obtain K sub-intervals of continuous distribution; wherein the K sub-intervals are determined by the saturation bits of the decoder. Obtained by wide division; map the LLR in the K sub-intervals to obtain mapped N mapping values.

In a possible implementation, the LLR is mapped in the K sub-intervals, and the determination module is specifically configured to: for a single sub-interval in the K sub-intervals, use the following formula to calculate the mapping The mapping value y _n of the LLR in the single sub-interval R _i ={a _i-1 , a _i }, i=1,2,...K-1, then

If l∈R _K =[a _K-1 ,∞], then y _n =sign(x _n )·y _K-1 ;

Among them, x _n is the n-th floating point value of the LLR, sign() is the sign function of x _n , a _i is the right endpoint of the single interval, a _i-1 is the left endpoint of the single interval, l is the absolute value of x _n , and p() is the probability density distribution of the target SNR under AWGN conditions.

In a possible implementation, the target signal-to-noise ratio is used to perform quantization processing on the N mapping values to obtain a quantization processing result. The processing module is specifically used to: obtain the linearity of the received symbol. signal-to-noise ratio, and the target signal-to-noise ratio; calculate the ratio of the linear signal-to-noise ratio to the target signal-to-noise ratio, and use the ratio as a quantized scaling factor; use the quantized scaling factor to N mapping values are scaled, and the scaled result is used as the quantization result.

In a possible implementation, the quantized scaling factor is used to perform scaling processing on the N mapping values respectively, and the result of the scaling processing is used as the quantization processing result. The processing module is specifically used to: For the N mappings For a single mapping value in the value, perform the following processing operation: calculate the product between the quantized scaling factor and the single mapping value, round it, and use the rounding result as the result of the scaling process of the single mapping value. ; Repeat the above processing operation to obtain the scaling processing results corresponding to each of the N mapping values, and use the scaling processing results as the quantization processing results.

In a seventh aspect, this application provides an electronic device, which includes:

Memory, used to store computer programs;

The processor is configured to implement the above-mentioned soft bit quantization processing method steps when executing the computer program stored on the memory.

In an eighth aspect, the present application provides a computer-readable storage medium. A computer program is stored in the computer-readable storage medium. When the computer program is executed by a processor, the steps of the above-mentioned soft bit quantization processing method are implemented. .

For various aspects in the above-mentioned sixth to eighth aspects and the technical effects that may be achieved by each aspect, please refer to the above-mentioned description of the technical effects that can be achieved by the fifth aspect or various possible solutions in the fifth aspect, which will not be repeated here.

Description of the drawings

In order to more clearly explain the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only for the purpose of the present application. Some examples.

Figure 1 is a schematic architectural diagram of a communication system provided by an embodiment of the present application;

Figure 2 is a schematic diagram of an information transmission process provided by an embodiment of the present application;

Figure 3 is a schematic flow chart of a quantification method for LLR values provided by an embodiment of the present application;

Figure 4 is a schematic flow chart of another quantification method for LLR values provided by an embodiment of the present application;

Figure 5 is a schematic structural diagram of an LLR value quantization device provided by an embodiment of the present application;

Figure 6 is a schematic structural diagram of an electronic device provided by an embodiment of the present application;

Figure 7 is a schematic diagram of a signal processing flow at a transmitter provided by an embodiment of the present application;

Figure 8 is a flow chart of a soft bit quantization processing method provided by an embodiment of the present application;

Figure 9 is a probability density distribution of LLR under different signal-to-noise ratios of a BPSK provided by the embodiment of the present application;

Figure 10 is a schematic diagram of a soft bit quantization processing device provided by an embodiment of the present application;

FIG. 11 is a schematic diagram of the structure of an electronic device according to an embodiment of the present application.

Detailed ways

In order to make the purpose, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are Some embodiments of the technical solution, rather than all embodiments. Based on the embodiments recorded in the application documents, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the technical solution of this application.

The terms "first" and "second" in the description and claims of this application and the above-mentioned drawings are used to distinguish different objects, rather than describing a specific sequence.

In the description of this application, "plurality" is understood to mean "at least two".

Furthermore, the term "includes" and any variations thereof are intended to cover non-exclusive protection. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes Other steps or units inherent to such processes, methods, products or devices. "Multiple" in this application may mean at least two, for example, it may be two, three or more, which is not limited by the embodiment of this application.

In addition, the term "and/or" in this article is only an association relationship that describes related objects, indicating that there can be three relationships. For example, A and/or B can mean: A alone exists, and A and B exist simultaneously. There are three cases of B alone. In addition, the character "/" in this article, unless otherwise specified, generally indicates that the related objects are in an "or" relationship.

In order to facilitate understanding of the quantization scheme for receiving symbols proposed in this application, the technical terms involved in this application are first introduced:

(1) Mutual information: Mutual information is a useful information measure in information theory. It can be regarded as the amount of information contained in one random variable about another random variable, or it can be said that one random variable is known to another random variable due to And reduce uncertainty. Simply put, it is the degree to which the uncertainty of event Y is reduced under the conditions in which event X occurs, that is, the correlation between the two events.

In the field of communication, according to information theory, due to the uncertainty of the channel, the sender and the receiver are generally not in a definite relationship, but a statistically dependent relationship. For example, taking the transmission of a certain information bit as an example, the entropy of the sending end is H(X), and it is known that the conditional entropy of the receiving end's information relative to the sending end is H(X|Y), then it can be defined according to the following formula (1) The mutual information of this information bit:
I(X;Y)＝H(X)-H(X|Y) Formula (1)

Among them, I(X;Y) is the mutual information of the information bits, H(X) is the entropy of the sending end, and H(X|Y) is the conditional entropy of the receiving end's information relative to the sending end.

(2) Constellation mapping: Constellation mapping refers to mapping the bit sequence carrying digital information into a symbol sequence suitable for transmission. Constellation mapping includes two elements, namely constellation diagram and constellation point mapping method. Among them, the constellation diagram represents a set of all values of the constellation mapping output symbol, and each constellation point in the constellation diagram corresponds to a value of the output symbol. The constellation point mapping method represents a specific mapping relationship from input bits (or bit groups) to constellation points, or a specific mapping relationship from constellation points to bits (or bit groups). Each constellation point in the constellation diagram corresponds to one bit (or a bit group composed of multiple bits). Different constellation mapping methods correspond to different constellation diagrams.

(3) Modulation and demodulation: Modulation is achieved by changing the high-frequency carrier (that is, changing the amplitude, phase or frequency of the carrier) so that it changes with the amplitude of the baseband signal. The encoded information bits can be modulated into symbols through constellation mapping of different modulation methods, such as π/2-BPSK, BPSK, QPSK, 16QAM, 64QAM, 256QAM and other modulation methods. Different modulation methods can correspond to different modulation orders. number. Demodulation is the reverse process of modulation, that is, extracting the baseband signal from the carrier (turning it into a low-frequency signal or directly into a data stream) to facilitate subsequent processing.

(4) Transmitting symbols and receiving symbols: Transmitting symbols are obtained by modulating the coded bits obtained by encoding through different modulation methods at the transmitting end. The received symbols are obtained by demodulating the received signals through different demodulation methods at the receiving end.

(5) Layer mapping and antenna mapping: In order to achieve transmission diversity, the symbols obtained by constellation mapping need to be divided into different layers through layer mapping, and then the data is mapped to the antenna port through precoding.

(6) Signal to Interference plus Noise Ratio (English: Signal to Interference plus Noise Ratio, abbreviation: SINR): refers to the strength of the useful signal received by the receiving end and the strength of the received interference signal (noise and interference) ratio.

(7) LLR value: The receiving end demodulates the received carrier signal to obtain the received symbol, and then uses the log-likelihood ratio algorithm based on the posterior probability criterion to determine the received symbol, the SINR of the received symbol, and the modulation order from the transmitting end. Calculate one or more LLR values corresponding to each received symbol. For different modulation modes at the sender, the number of LLR values calculated by the receiver is also different. For example, if the transmitter uses 16QAM modulation and its modulation order is 4, the number of LLR values calculated by the receiver is four times the number of received symbols. Alternatively, LLR values may also be called soft bits.

As mentioned earlier, the quantization accuracy of floating-point LLR values has a significant impact on the final decoding result. There are currently two mainstream quantization methods. One method is to set a scaling factor for quantization based on experience. This method does not take into account the influence of the fading channel, and the accuracy of quantization is low. Another method is the polling test method, which determines the scaling factor based on the current average signal-to-noise ratio. This method cannot well solve the impact of fading channels, and the quantization efficiency is low. For different coding rates and modulation methods, Applicability is also lower.

In the following, in order to facilitate understanding of the solutions, the exemplary embodiments involved in the solutions of the first to fourth aspects of the present application are described. introduce. First, a brief introduction to the system architecture is given. Refer to Figure 1, which is an architecture diagram of a communication system provided by an embodiment of the present application. It should be understood that the embodiments of the present application are not limited to the system shown in FIG. 1 . In addition, the device in Figure 1 may be hardware, software divided by function, or a combination of the above two.

As shown in Figure 1, the system architecture provided by the embodiment of the present application includes terminals and network equipment. The embodiments of this application do not limit the number of terminals and network devices included in the system. The communication system may also include core network equipment, which is not shown in Figure 1 .

User terminal (English: User Equipment, abbreviation: UE), also known as terminal equipment, mobile station (English: Mobile Station, abbreviation: MS), mobile terminal (English: Mobile Terminal, abbreviation: MT), etc., is a kind of Devices that provide voice and/or data connectivity to users, such as handheld devices with wireless connectivity, vehicle-mounted devices, etc. At present, some examples of terminals are: mobile phones, tablets, laptops, PDAs, mobile Internet devices (English: Mobile Internet Device, abbreviated as: MID), wearable devices, virtual reality (English: Virtual Reality, Abbreviation: VR) equipment, augmented reality (English: Augmented Reality, abbreviation: AR) equipment, wireless terminals in industrial control (Industrial Control), wireless terminals in self-driving (self driving), remote medical surgery (remote medical surgery) Wireless terminals in smart grids, wireless terminals in transportation safety, wireless terminals in smart cities, wireless terminals in smart homes, etc.

The network equipment involved in the embodiments of this application may be a base station, an access network equipment, an access node (English: Access Node, abbreviation: AN), etc. A base station is a public mobile communication base station and is an interface device for mobile terminals to access the Internet. The network equipment can specifically be an evolutionary base station (English: Evolutional Node B, abbreviated as: eNB or eNodeB) in the Long Term Evolution (English: Long Term Evolution, abbreviation: LTE) system, or the fifth generation (5th generation, 5G) mobile communication The next generation base station (English: Next Generation NodeB, abbreviation: gNB) in the system is not limited by this application.

Exemplarily, the core network equipment not shown in Figure 1 may include an access and mobility management entity (English: Access and Mobility Management Function, referred to as: AMF), a session management function entity (English: session management function, referred to as: SMF) etc.

Each exemplary embodiment of the present application provides a quantification method, device, electronic device, and storage medium for LLR values. By using the average mutual information of received symbols, the impact of the fading channel is equated into a measurable system, thereby improving the accuracy of the quantized LLR value, thereby improving the decoding performance and reducing the bit error rate.

The information transmission process is briefly introduced with reference to the communication system shown in Figure 1. For example, the information transmission process can be seen in the flow chart shown in Figure 2. When transmitting information, the sending end (which can be a terminal in the system architecture shown in Figure 1 or a network device) can encode the information bits to obtain coded bits. Then, you can encode The bits are scrambled to obtain scrambled bits. Further, constellation mapping of different modulation modes can be used to modulate the scrambling bits to obtain modulation symbols. Finally, after layer mapping, precoding and antenna mapping are performed on the modulation symbols, they are modulated into time domain signals through OFDM and sent to the receiving end (if the sending end is a terminal, the receiving end is a network device; if the sending end is a network device, then The receiving end is the terminal). The receiving end will perform the reverse process of the process shown in Figure 2. When demodulating, if the receiving end directly uses inverse mapping to demodulate, the bit error rate may become higher due to the influence of noise. For example, you can see the following formula (2):

in, is the received symbol obtained by demodulation using inverse mapping, s is the symbol sent by the transmitter, and n is the noise.

It can be seen from the above formula (2) that noise has a greater impact on the demodulation accuracy, which will also make the bit error rate of subsequent decoding higher. Therefore, it is proposed in the related art to use LLR calculation to demodulate the received signal. The essence of the calculated LLR is a probability value. If the LLR value is positive and the greater the absolute value, the greater the probability that the corresponding bit is 0; if the LLR value is negative and the greater the absolute value, the greater the probability the corresponding bit is 1; the closer the LLR value is to 0, the greater the probability of the corresponding bit. The higher. After demodulation a decoding step is required. However, the input bit width of the current decoder is fixed, so before decoding, the demodulated LLR values (or can also be called soft bits) need to be quantized to meet the input requirements of the decoder. For example, assume that the LLR value obtained by demodulation is: {X _LLRs }={x ₀ ,x ₁ ,x ₂ ,…,x _N-1 },-∞≤x _i ≤+∞. Then, for a decoder with a saturated bit width of K, the quantization result is: {Q _LLRs }={q ₀ ,q ₁ ,q ₂ ,…,q _N-1 },-2 ^K ≤q _i ≤2 ^K -1 . The result of quantization calculation is used as the input of the decoder. Therefore, the quantization calculation has a significant impact on the decoding result of the decoder. There are two main current mainstream quantification methods:

One quantification method is to empirically determine the scaling factor used for quantification and perform quantification directly. For example, see formula (3) below:

Among them, Q _LLRs is the quantized LLR value, _xi is the i-th LLR value to be quantized, a is the scaling factor determined based on experience, and N is the number of LLR values obtained by demodulation.

This quantization method does not take into account the influence of channel fading, and the quantization results are not accurate.

Another quantization method is the polling test method, which determines the scaling factor for quantization based on the average signal-to-noise ratio of the current channel. For example, see formula (4)-formula (6) below:

a=2 ^ML formula (5)

Among them, Q _LLRs is the quantized LLR value, x _i is the i-th LLR value to be quantized, and a is the shrinkage determined based on experience. Amplification factor, N is the number of LLR values obtained by demodulation, M is the fixed value corresponding to the lowest bit error rate during actual testing, snr _i is the signal-to-noise ratio of the i-th received symbol.

When using decoders from different manufacturers, this quantitative method of polling test method needs to be tested in advance on all coding rates and modulation methods supported by the 5G NR communication protocol and various different configuration conditions to select the best M value. , the adaptability and efficiency are low, and it cannot well solve the impact of fading channels.

An embodiment of the present application proposes a quantification method for LLR values. First, the mutual information of each received symbol can be calculated based on the SINR linear value of each received symbol obtained by demodulation, and the impact of the fading of the system channel is expressed as the average of the mutual information of all or part of the received symbols. The scaling factor used for quantization calculations is determined based on the average value of the mutual information.

The technical solution proposed in the embodiment of this application enables the impact of the fading channel on the quantization of the LLR value to be measured by averaging the mutual information of the received symbols. Therefore, the quantization scheme of the present application better considers the influence of the fading channel, thereby improving the accuracy of the quantization results, thereby reducing the bit error rate of the subsequent decoding process.

It should be noted that the quantization scheme of LLR values proposed in this application is applicable to any decoder that needs to calculate a fixed-point LLR value as input, and is not limited to the low-density parity check (English: Low) of the physical uplink and downlink shared channels in the 5G NR system. -Density Parity-Codes (abbreviation: LDPC) decoder, Turbo decoding or Polar decoding, etc.

Next, the quantification method of the LLR value proposed in one embodiment of the present application will be described in conjunction with the system architecture shown in Figure 1 and the communication information transmission process shown in Figure 2 . Optionally, please refer to FIG. 3 , which is a flow chart of a quantification method for LLR values provided by an embodiment of the present application. Optionally, the receiving end that performs the method process can be a terminal in the system shown in Figure 1, or a network device, depending on the information flow direction. In this embodiment, the method flow specifically includes the following steps.

Step 301: The receiving end obtains the SINR linear values of multiple received symbols obtained by demodulation.

Optionally, after receiving the carrier signal carrying the data stream from the transmitting end, the receiving end can use it as the input of the demodulation device and demodulate it to obtain multiple received symbols.

Further, after demodulation, the SINR of each received symbol is obtained. Furthermore, after obtaining the SINR of each received symbol, a preset linear correspondence relationship can be used to determine the SINR linear value of each received symbol. For example, the preset linear correspondence relationship may be y=k*x. If the SINR of a certain received symbol is a, then the linear SINR value of the received symbol may be k*a.

Step 302: The receiving end calculates the mutual information of each received symbol based on the SINR linear value of each received symbol.

Among them, since mutual information is a parameter used to characterize the correlation between two variables, the mutual information of a certain received symbol can be used to characterize the received symbol and the corresponding symbol sent by the sending end (i.e., the sending symbols).

Optionally, the mutual information used to characterize the correlation between a received symbol and a corresponding transmitted symbol can be determined by the SINR linear value of the received symbol.

Specifically, the receiving end calculates the mutual information of each received symbol based on the SINR linear value corresponding to each received symbol. For this purpose, an alternative implementation is to compute the mutual information of all received symbols. In another implementation, when the number of received symbols is large (for example, greater than the number threshold), the mutual information of some of the received symbols can be calculated according to the sampling parameter settings. In one embodiment, the partial received symbols may be a set of P received symbols every S received symbols in all modulated received symbols, N is an integer greater than or equal to 1, and P is an integer greater than or equal to 1. For example, when according to the ratio of 1:2, that is, according to S=1, P=1, from all the received symbols obtained by modulation, one received symbol can be selected as a partial received symbol from every other received symbol, and the corresponding part can be calculated One mutual information for each received symbol; or when according to the ratio of 1:3, that is, S=2, P=1, one received symbol can be selected from every two received symbols from all the received symbols obtained by modulation As partial received symbols, and calculate a mutual information for each of the corresponding partial received symbols; or when according to the ratio of 2:3, that is, S = 1, P = 2, each received symbol can be obtained from the modulation. Select two consecutive received symbols every other received symbol as partial received symbols, and calculate a mutual information for each of the corresponding partial received symbols.

Step 303: The receiving end determines a scaling factor for quantizing multiple LLR values based on the average value of mutual information corresponding to the multiple received symbols.

The plurality of LLR values include the LLR value of each received symbol calculated based on a log-likelihood ratio algorithm for the plurality of received symbols. For example, the receiving end can use the log-likelihood ratio algorithm to calculate the LLR value corresponding to each received symbol based on the received symbol, the SINR of the received symbol, and the modulation order from the transmitting end.

Optionally, after calculating the mutual information corresponding to the multiple received symbols (the mutual information of all received symbols, or the mutual information of some of the received symbols obtained based on the sampling parameters), the receiving end can calculate the multiple received symbols. The average value of the mutual information corresponding to the symbol (the average value of the mutual information will be referred to as the mean mutual information (English Mean Mutual Information, abbreviation: MMI) in the following), and the average mutual information is used to calculate the scaling factor. Then, the multiple LLR values obtained by the above demodulation can be quantized according to the calculated scaling factor. Specifically, the scaling factor may be used to perform quantization processing on the plurality of LLR values one by one.

Based on the above solution, this application proposes to determine the scaling factor based on the average mutual information of the received symbols obtained by demodulation, taking into account different fading scenarios, and equating the impact of the fading channel to a measurable parameter through the average mutual information. The system uses this to adjust the quantization process of LLR values to make the quantization results more accurate, so that the input of the decoder can better adapt to the channel environment, improve decoding performance and reduce the bit error rate.

As a possible implementation method, when the receiving end determines the scaling factor for quantizing the LLR value, the receiving end can also use the average mutual information, modulation order and coding rate of multiple received symbols to jointly determine the scaling factor. Among them, the coding rate is The modulation order is used by the sending end when encoding the information bits to be sent. The modulation order is used by the sending end when modulating the coded bits obtained after encoding. The coding rate and modulation order used by the receiving end come from the sending end. For example, the sending end may carry the coding rate and modulation order when sending data to the receiving end.

Optionally, an appropriate selection function can be used to calculate the average mutual information, modulation order and coding rate to obtain the scaling factor.

In the exemplary embodiment of the present application, it is proposed that the modulation order and coding rate are added to the calculation process of the scaling factor, so that the quantization process not only considers the impact of the fading channel, but also takes into account the impact of the receiving end processing. It can be compatible with different fading channels and multiple modulation and coding methods, improving the adaptability of the solution.

In some embodiments, after the receiving end calculates the average mutual information of multiple received symbols based on the SINR linear values of the received symbols, the receiving end further determines the SINR linear gain parameters of the multiple received symbols (for example, the SINR linear value of average value), and the scaling factor is calculated using the SINR linear gain parameters (for example, the average value of the SINR linear values) of the multiple received symbols, the modulation order, and the coding rate.

In order to facilitate understanding, the following is introduced with specific examples. For example, the following formula (7) can first be used to calculate the average mutual information of all received symbols:

Among them, I _equal is the average mutual information, M is the number of received symbols, sinr(m) is the SINR linear value of the m-th received symbol, and F(sinr(m)) is the mutual information of the m-th received symbol. Based on a simple modification of the above formula, those in the field can also use the mutual information of some received symbols to calculate the average mutual information, which will not be described in detail here.

Further, the average mutual information can be used to calculate the SINR linear gain parameters of multiple received symbols (for example, the average of SINR linear values). The SINR linear gain parameters of multiple received symbols (for example, the average of SINR linear values) can be calculated using It is used to characterize the SINR linear gain of the system. Alternatively, the following formula (8) can be used to calculate the average of the SINR linear values of multiple received symbols:
SINR _equal =F ^-1 (I _equal ); Formula (8)

Among them, SINR _equal is the average of the SINR linear values of multiple received symbols, I _equal is the average mutual information, and F ^-1 is the inverse function of F in formula (7) above.

After obtaining the average of the SINR linear values of multiple received symbols, the scaling factor can be determined using the average of the SINR linear values, the modulation order and the coding rate. In a possible implementation manner, a selection function for calculating the scaling factor through the average value of the SINR linear value, the modulation order and the coding rate can be obtained. Optionally, the selection function may be determined in advance through machine learning. For example, the machine learning process can be: randomly generate multiple training samples (including SINR linear values, modulation orders and coding rates), and combine the modulation orders and coding rates based on an unsupervised learning algorithm (such as the k-mean algorithm). The rate is used as the control variable of the algorithm, and the SINR linear value is used as the independent variable of the algorithm to perform iterative learning. It is used to determine the appropriate selection function for calculating the scaling factor from the average of multiple mutual information (or the average of SINR linear values) of multiple received symbols, the modulation order and the coding rate. Based on the predetermined selection function, the average value of the SINR linear values of multiple received symbols, the modulation order and the coding rate are used as inputs of the selection function, and the output of the selection function is used as the scaling factor. Optionally, the following formula (9) can be used to determine the scaling factor:
a _sf =U(SINR _equal ,Q _m ,R); Formula (9)

Among them, a _sf is the scaling factor, U is the selection function used to calculate the scaling factor through the average value of the SINR linear value, the modulation order and the coding rate, SINR _equal is the average value of the SINR linear value of multiple received symbols, Q _m is the modulation order, and R is the coding rate.

In other embodiments, after the receiving end calculates the average mutual information based on the SINR linear value of the received symbol, it can directly determine the scaling factor based on the average mutual information, modulation order, and coding rate. For example, a predetermined selection function for calculating scaling factors based on average mutual information, modulation order, and coding rate can be obtained, and the average mutual information, modulation order, and coding rate can be used as inputs to the selection function, and the output of the selection function can be as a scaling factor. Optionally, the following formula (10) can be used to determine the scaling factor:
a _sf =V(I _equal ,Q _m ,R); Formula (10)

Among them, a _sf is the scaling factor, V is the selection function used to calculate the scaling factor through the average mutual information, modulation order and coding rate, SINR _equal is the average of the SINR linear values of multiple received symbols, Q _m is the modulation order number, R is the coding rate.

Furthermore, after the scaling factor is determined, the scaling factor can be used to quantize the LLR value.

Alternatively, the following formula (11)-formula (12) can be used to perform quantitative calculations:

N=M·Q _m ; Formula (12)

Among them, q _i is the i-th LLR value after quantization, a _sf is the scaling factor, _xi is the i-th LLR value, N is the number of LLR values, M is the number of received symbols, and Q _m is the modulation order.

In the following, in order to further understand the solution proposed in this application, specific embodiments will be introduced. Refer to Figure 4, which is a flow chart of another quantification method for LLR values proposed in this application, which specifically includes the following steps.

Step 401: The receiving end receives the carrier signal from the transmitting end.

Step 402: The receiving end demodulates the carrier signal to obtain multiple received symbols, and determines the LLR value corresponding to each received symbol.

Optionally, after demodulating multiple received symbols, for any one of the received symbols, the LLR value corresponding to the received symbol can be calculated based on the received symbol, the SINR linear value of the received symbol, and the modulation order from the transmitting end.

Step 403: The receiving end obtains SINR linear values of multiple received symbols.

Specifically, the receiving end can obtain the SINR of multiple received symbols, and use a preset linear correspondence relationship to determine the linear SINR values of the multiple received symbols.

Step 404: The receiving end calculates the mutual information of a certain received symbol based on the SINR linear value of the received symbol.

For the specific calculation process and functions used, please refer to the introduction in the above embodiment, and will not be described in detail here.

Step 405: The receiving end calculates the average mutual information of multiple received symbols.

Alternatively, the receiving end may use the arithmetic average, weighted average, or geometric average of the mutual information of multiple received symbols as the average mutual information, which is not limited in this application.

Step 406: The receiving end calculates the average of the SINR linear values of multiple received symbols based on the average mutual information.

For details, please refer to formula (8) in the above embodiment.

Step 407: The receiving end determines the scaling factor used for quantization based on the average of the SINR linear values of multiple received symbols, the modulation order, and the coding rate.

Specifically, the receiving end can use the average of the SINR linear values of multiple received symbols, the modulation order, and the coding rate as the input of the pre-learned selection function, and use the output of the selection function as the scaling factor.

Step 408: The receiving end uses the scaling factor to quantize the LLR value.

For the specific quantitative calculation process, please refer to formula (11)-formula (12) above.

Based on the same concept as the above method, see FIG. 5 , which is a quantization device 500 for LLR values provided in an embodiment of the present application. The device 500 can be used to perform various steps in the above method. To avoid repetition, they will not be described one by one here. The device 500 includes: an acquisition unit 501 and a processing unit 502.

The acquisition unit 501 is used to acquire the signal to interference plus noise ratio SINR linear value of multiple received symbols obtained by demodulation;

The processing unit 502 is configured to calculate the mutual information of the multiple received symbols based on the SINR linear values of the multiple received symbols;

The processing unit 502 is further configured to determine a scaling factor for quantizing multiple LLR values based on the average value of the mutual information of the multiple received symbols. The multiple LLR values include scaling the multiple received symbols based on The LLR value of each received symbol calculated by the log-likelihood ratio algorithm.

In some embodiments, the processing unit 502 is specifically used to:

In some embodiments, the processing unit 502 is also used to:

The processing unit 502, when determining the scaling factor based on the average value of the mutual information of the multiple received symbols, the modulation order and the coding rate, is specifically used to:

In some embodiments, the processing unit 502 is specifically used to:

The selection function determined in advance through machine learning is obtained through the acquisition unit 501;

FIG. 6 shows a schematic structural diagram of an electronic device 600 provided by an embodiment of the present application. The electronic device 600 in the embodiment of the present application can also include a communication interface 603. The communication interface 603 is, for example, a network port. The electronic device can transmit data through the communication interface 603. For example, the communication interface 603 can implement the receiving from the network interface introduced in the above embodiment. Steps for transmitting time domain signals.

In this embodiment of the present application, the memory 602 stores instructions that can be executed by at least one controller 601. At least one controller 601 can be used to perform various steps in the above method by executing the instructions stored in the memory 602. For example, the controller 601 can realize the functions of the acquisition unit 501 and part of the functions of the processing unit 502 in Figure 5 mentioned above.

Among them, the controller 601 is the control center of the electronic device. It can use various interfaces and lines to connect various parts of the entire electronic device by running or executing instructions stored in the memory 602 and calling data stored in the memory 602. Optionally, the controller 601 may include one or more processing units. The controller 601 may integrate an application controller and a modem controller. The application controller mainly processes operating systems and application programs, and the modem controller Mainly deals with wireless communications. It can be understood that the above modem controller may not be integrated into the controller 601. In some embodiments, the controller 601 and the memory 602 can be implemented on the same chip, and in some embodiments, they can also be implemented on separate chips.

The controller 601 can be a general controller, such as a central controller (English: Central Processing Unit, referred to as: CPU), a digital signal controller, an application specific integrated circuit, a field programmable gate array or other programmable logic devices, discrete gates or transistors Logic devices and discrete hardware components can implement or execute the methods, steps and logical block diagrams disclosed in the embodiments of this application. A universal controller can be a microcontroller or any conventional controller, etc. The steps executed by the data statistics platform disclosed in the embodiments of this application can be directly executed by the hardware controller, or can be executed by a combination of hardware and software modules in the controller.

As a non-volatile computer-readable storage medium, the memory 602 can be used to store non-volatile software programs, non-volatile computer executable programs and modules. Memory 602 may include at least one type of storage medium, such as For example, it can include flash memory, hard disk, multimedia card, card-type memory, random access memory (English: Random Access Memory, abbreviation: RAM), static random access memory (English: Static Random Access Memory, abbreviation: SRAM), programmable read-only memory Memory (English: Programmable Read Only Memory, abbreviation: PROM), read-only memory (English: Read Only Memory, abbreviation: ROM), electrically erasable programmable read-only memory (English: Electrically Erasable Programmable Read-Only Memory, abbreviation: : EEPROM), magnetic memory, magnetic disks, optical disks, etc. Memory 602 is, but is not limited to, any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. The memory 602 in the embodiment of the present application can also be a circuit or any other device capable of realizing a storage function, used to store program instructions and/or data.

By designing and programming the controller 601, for example, the code corresponding to the neural network model training method introduced in the previous embodiment can be solidified into the chip, so that the chip can execute the aforementioned neural network model training method during runtime. The steps and how to design and program the controller 601 are well-known techniques to those skilled in the art, and will not be described again here.

Below, in order to facilitate understanding of the solution, exemplary embodiments related to the solutions of the fifth to eighth aspects of the present application are introduced with reference to FIGS. 7 to 11 .

With the evolution of communication systems, in the fifth generation new radio system (English: 5th generation New Radio, abbreviation: 5G NR), binary phase shift keying (English: Binary Phase Shift Keying, abbreviation: BPSK), normal Cross-phase shift keying (English: Quadrature Phase Shift Keying, abbreviation: QPSK), quadrature amplitude modulation (English: Quadrature Amplitude Modulation, abbreviation: 16QAM), 64QAM, 256QAM and other modulation methods are used to perform corresponding modulation and demodulation. Generally speaking, modulation is relatively simple because the modulation process is a direct mapping process; while demodulation is relatively complex because the demodulation process is also affected by noise, so the demodulation process is a probabilistic and statistical process.

In the 5G NR system, the transmitter can obtain the modulation symbols corresponding to the information bits through constellation mapping of different modulation methods, and then process the modulation symbols into signals for transmission through a series of operations. The receiving end receives the signal transmitted by the transmitting end and obtains the received symbol recovered from the signal. Here, the received symbol is a modulation symbol affected by noise during the transmission process. Therefore, if the received symbol is directly demodulated, it will result in the final obtained information bits. The bit error rate is high, that is, the demodulation result is inaccurate. In order to remove the influence of noise, the following operations are generally performed: calculate the log likelihood ratio (English: Log Likelihood Ratio, abbreviation: LLR) of the received symbol, and then send the calculation result to the decoder for decoding (demodulation).

Specifically, each LLR value calculated above can be understood as a probability value: if the LLR value is a positive number and the greater the absolute value of the LLR value, the greater the probability that the information bit is 1; if the LLR value is a negative number, And the greater the absolute value of the LLR value, the greater the probability that the information bit is -1; if the absolute value of the LLR value is closer to 0, the uncertainty of the information bit is higher.

It is worth mentioning that the above LLR refers to the log likelihood ratio, which means that one binary information bit is 0 or The probability form of 1. Generally speaking, tens of thousands of information bits are transmitted, which is represented by N here. That is, N information bits correspond to N LLR values when processed at the receiving end. In addition, the so-called modulation mapping is to map one or several information bits into a constellation symbol. For example, one QPSK symbol is mapped by two bits, so two LLR values can be calculated from one QPSK received symbol.

Based on this, by calculating the received symbols using the log-likelihood ratio method, a set of floating-point number sequences composed of floating-point LLR values can be obtained, where the floating-point numbers are numbers with an unfixed decimal point, and the floating-point LLR values X _LLRs can Expressed as: {X _LLRs }={x ₀ ,x ₁ ,x ₂ ,…,x _N-1 },-∞≤x _i ≤+∞.

However, starting from the implementation of the decoder, the decoder cannot directly process floating point numbers. The decoder can only process a fixed-point number sequence with a fixed saturated bit width. The fixed saturated bit width is a specified value range. Points are numbers with a fixed decimal point. Therefore, in order to meet the implementation of the decoder, the above floating-point LLR values need to be quantized. Specifically, a set of fixed-point number sequences composed of fixed-point LLR values with a fixed saturated bit width are obtained through quantization processing, where, if It has N input values, and the saturated bit width is K, then the fixed-point LLR value Q _LLRs can be expressed as: {Q _LLRs }={q ₀ , q ₁ , q ₂ ,..., q _N-1 },-2 ^K ≤q _i ≤2 ^K -1.

Obviously, the results of the above quantization processing will have an impact on the accuracy of the decoding results of the decoder. Therefore, in order to optimize decoding performance and improve the accuracy of decoding results, the existing technology proposes the following quantization processing method: setting a quantization scaling factor a based on human experience for the floating point LLR value x _i , and then passing n=0,1,…,N-1 to obtain the fixed-point LLR value Q _LLRs , that is, the existing technology relies on manual experience. However, in practical applications, various fading scenarios will be faced, and when the above-mentioned existing technology is applied to different fading scenarios, there is still a problem of inaccurate quantization processing results, which in turn leads to inaccurate decoding results of the decoder.

Each exemplary embodiment of the fifth to eighth aspects of the present application relates to the technical field of communication systems, and specifically relates to a quantization processing method, device and electronic equipment for soft bits, which are used to solve the problem of inaccurate decoding results of the decoder caused by inaccurate quantization processing results. Inaccurate questions. The method includes obtaining the probability density distribution of the log-likelihood ratio LLR of the received symbol under the target signal-to-noise ratio. The probability density distribution contains the uncertainty probability corresponding to each floating point value in the LLR. The probability density distribution is in the additive Gaussian white Probability density distribution under noisy AWGN conditions, and then based on this probability density distribution, determine N mapping values of LLR, where N is the number of LLR values of the received symbol, and then use the target signal-to-noise ratio to quantify the N mapping values. , to obtain the quantitative processing results. Based on the above method, taking into account the changes in signal-to-noise ratio in different fading scenarios, the decoding performance of the decoder can be optimized and the accuracy of the final decoding result can be improved.

Each exemplary embodiment of the fifth to eighth aspects of the present application provides a possible application scenario, specifically including: a transmitting end and a receiving end.

As shown in Figure 7, it is a schematic diagram of the signal processing process at the transmitter. For the transmitter, the information bits are first encoded to obtain the encoded bits; then the encoded bits are scrambled to obtain the scrambling process. The resulting scrambling bits are then processed through constellation mapping using different modulation methods to obtain modulation symbols after constellation mapping. Then, the modulation symbols are sequentially subjected to layer mapping processing, precoding processing, antenna mapping processing and orthogonal frequency division multiplexing. Technology (English: Orthogonal Frequency Division Multiplexing, OFDM for short) modulation processing. After OFDM modulation processing, a time domain signal is obtained, and this time domain signal is transmitted as a transmission signal.

The receiving end receives the transmission signal transmitted from the transmitting end, and then performs signal processing on the received transmission signal. It is easy to understand that those skilled in the art can regard the signal processing process of the receiving end as the inverse processing process of the signal processing process of the transmitting end. In This will not be elaborated upon.

In other words, the transmitter can obtain the modulation symbols corresponding to the information bits through constellation mapping of different modulation methods, and then process the modulation symbols into signals through a series of operations and transmit them. The receiving end receives the signal transmitted by the transmitting end and obtains the received symbol recovered from the signal. Here, the received symbol is a modulation symbol affected by noise during the transmission process. If the received symbol is directly demodulated, it will result in the final information bits being The bit error rate is high, that is, the demodulation result is inaccurate. Therefore, in order to remove the influence of noise, the following operations are generally performed: calculate the LLR of the received symbol, and then send the calculation result to the decoder for decoding (demodulation).

Among them, each LLR value calculated above is understood as a probability value: if the LLR value is a positive number and the greater the absolute value of the LLR value, the greater the probability that the information bit is 1; if the LLR value is a negative number and the LLR value The greater the absolute value of , the greater the probability that the information bit is -1; if the absolute value of the LLR value is closer to 0, the uncertainty of the information bit is higher.

It is worth noting that the above-mentioned LLR refers to the log likelihood ratio, which represents the probability form of one binary information bit being 0 or 1. Generally speaking, tens of thousands of information bits are transmitted, which is represented by N here. That is, N information bits correspond to N LLR values when processed at the receiving end. In addition, the so-called modulation mapping is to map one or several information bits into a constellation symbol. For example, one QPSK symbol is mapped by two bits, so two LLR values can be calculated from one QPSK received symbol.

Based on this, the received symbols are calculated by log likelihood ratio. Under the conditions of AWGN, the floating point LLR value X _LLRs can be expressed as: {X _LLRs }={x ₀ ,x ₁ ,x ₂ ,…, x _n ,…,x _N-1 },-∞≤x _n ≤+∞, where x _n is the nth floating-point LLR value.

However, starting from the implementation of the decoder, the decoder cannot directly process floating point numbers. The decoder can only process a fixed-point number sequence with a fixed saturated bit width. The fixed saturated bit width is a specified value range. Points are numbers with a fixed decimal point. Therefore, in order to meet the implementation of the decoder, the above floating-point LLR values need to be quantized. Specifically, through quantization processing, a set of fixed-point number sequences composed of fixed-point LLR values with a fixed saturated bit width are obtained, where the fixed-point LLR The fixed-point number sequence Q _LLRs composed of values can be expressed as: {Q _LLRs }＝{q ₀ ,q ₁ ,q ₂ ,…,q _n ,…,q _N-1 },-2 ^K ≤q _n ≤2 ^K - 1, where K is the saturated bit width that the decoder can recognize, -2 ^K is the left end point of the range that the decoder can recognize, and 2 ^K -1 is the right end point of the range that the decoder can recognize. Worth saying It is obvious that there may be differences in the saturated bit widths recognized by different decoders, that is, there are differences in the saturated bit width K recognized by the decoders.

Obviously, the results of the above quantization processing will have an impact on the accuracy of the decoding results of the decoder. Therefore, an optimized quantization processing method helps optimize decoding performance and improve the accuracy of decoding results.

This application provides a soft bit quantization processing method, device and electronic equipment to solve the problem of inaccurate decoding results of the decoder caused by inaccurate quantization processing results.

It is worth noting that the technical solutions provided by the embodiments of this application can be applied to any communication system that needs to perform quantization processing on LLR and/or any decoder that needs to perform quantization processing on LLR. Of course, it is especially suitable for physical uplink in 5G NR. Shared channel (English: Physical Uplink Shared Channel, abbreviation: PUSCH)/Physical Downlink Shared Channel (English: Physical Downlink Shared Channel, abbreviation: PDSCH) channel low density parity check code (English: Low Density Parity Check Code, abbreviation: LDPC) decoder.

Furthermore, the technical features contained in the embodiments of the present application can be used in any combination. Those skilled in the art should understand that, starting from the actual application situation, the technical solutions obtained by reasonably combining the technical features in the embodiments of the present application can also solve the same technical problems. problem or achieve the same technical effect.

According to the method provided by the embodiment of the present application, the first step is to obtain each probability density distribution of the log-likelihood ratio LLR of the received symbol under the target signal-to-noise ratio, that is, taking into account the changes in the signal-to-noise ratio under different fading scenarios. Then based on each probability density distribution, the LLR is mapped to the respective mapping values of the K sub-intervals to obtain N mapping values, that is, N is an integer greater than K and less than 2 ^K , and then the target signal-to-noise ratio is used to map the N mapping values. The value is quantized, and finally the quantization result is obtained, that is, the quantization is performed based on the probability density distribution under the target signal-to-noise ratio to obtain the quantization result. The quantization result obtained in this way is input into the decoder, which helps to optimize the decoder. decoding performance, improving the accuracy of the final decoding result.

The methods provided by the embodiments of the present application will be further described in detail below with reference to the accompanying drawings.

Referring to Figure 8, an embodiment of the present application provides a soft bit quantization processing method, as follows:

Step 801: Obtain the probability density distribution of the log-likelihood ratio LLR of the received symbol under the target signal-to-noise ratio.

In the embodiment of this application, a way to obtain the best signal-to-noise ratio is proposed, as follows. For example, it is assumed that the average symbol energy at the transmitter is E _s , the AWGN noise energy is N ₀ , the symbol period is T _s , and the noise The bandwidth is B _n , the noise power is N, and the transmitted signal power is S, then the bit error rate under AWGN conditions is as follows:

For BPSK:

For QPSK:

Assuming that the minimum amplitude symbol energy of M-QAM modulation is E _min , then for M-QAM:

The optimal linear signal-to-noise ratio required for decoding under AWGN can be obtained by performing inversion calculations based on the BER required by the system:

N＝N ₀ *B _n ;

Among them, the Q function is:

The erfc (Gaussian error complement) function is:

To sum up, the target signal-to-noise ratio can be obtained based on the following formula:

In the embodiment of this application, the probability density distribution includes the uncertainty probability corresponding to each floating point value in the LLR.

Specifically, the log-likelihood ratio LLR of the received symbol can be a sequence composed of floating point LLR values: {X _LLRs }={x ₀ ,x ₁ ,x ₂ ,…,x _n ,…,x _{N -1} },-∞≤x _n ≤+∞, where x _n is the n-th floating-point LLR value. Taking the modulation method of BPSK as an example, x _n (expressed as x in the formula) under different signal-to-noise ratios The probability density distribution function of can be seen in the following formula:

in, is the variance, is the noise power.

Based on the above probability density distribution function, the probability density distribution of LLR based on BPSK under different signal-to-noise ratios can be obtained.

For example, see Figure 9, which is a schematic diagram of the probability density distribution of LLR of BPSK under different signal-to-noise ratios. The abscissa is the floating-point LLR value, and the ordinate is the uncertainty probability corresponding to the floating-point LLR value. In detail, taking the floating-point LLR value as 0 as an example, when SNR=-5dB, the uncertainty probability corresponding to the floating-point LLR value is 0.18; when SNR=0dB When SNR=+5dB, the uncertainty probability corresponding to the floating-point LLR value is 0.05; when SNR=+5dB, the uncertainty probability corresponding to the floating-point LLR value is 0.003. That is to say, when the floating-point LLR value is 0 and SNR=+5dB, the uncertainty probability corresponding to the floating-point LLR value is the smallest.

In summary, through this step, each probability density distribution of the log-likelihood ratio LLR of the received symbol under different signal-to-noise ratios is obtained.

Step 802: Based on the probability density distribution, determine N mapping values of the LLR.

In the embodiment of this application, each floating-point LLR value of LLR can be understood as a probability value, that is, if the floating-point LLR value is a positive number and the absolute value of the floating-point LLR value is larger, the probability that the information bit is 1 The larger the floating-point LLR value is, the larger the absolute value of the floating-point LLR value is, the greater the probability that the information bit is -1; if the absolute value of the floating-point LLR value is closer to 0, the probability of the information bit is -1. The higher the uncertainty. In other words, when the absolute value of the floating-point LLR value approaches 0, its uncertainty is greatest. In order to optimize decoding performance, it is necessary to reduce the uncertainty when the absolute value of the floating-point LLR value approaches 0.

Therefore, in the embodiment of the present application, the LLR can be mapped to the respective mapping values of the K sub-intervals based on the probability density distribution under the target signal-to-noise ratio obtained in step 801. The calculated N mapping values, the aforementioned K sub-intervals Divided by the saturated bitwidth of the decoder.

In detail, due to the limitation of the saturated bit width of the decoder, if the floating point LLR value {X _LLRs }={x ₀ ,x ₁ ,x ₂ ,…,x _n ,…,x _N-1 },-∞≤x _n ≤ +∞, where x _n is the nth bit of the floating-point LLR value, and the fixed-point sequence Q _LLRs composed of fixed-point LLR values is: {Q _LLRs }={q ₀ ,q ₁ ,q ₂ ,…, q _n ,…,q _N-1 },-2 ^K ≤q _n ≤2 ^K -1, where K is the saturated bit width that the decoder can recognize, -2 ^K is the left endpoint of the range that the decoder can recognize, 2 ^K -1 is the right end point of the decoder's identifiable range. The decoder can only recognize up to 2 ^K+1 situations for the value of q _n . Therefore, quantization mapping is performed under the best signal-to-noise ratio. Here, [0, 2 ^K ] is divided into K-1 sub-intervals R _i which can be expressed as {a _i-1 , a _i }, i=1,2,...,K-1.

In addition, the above K-1 sub-interval is a (positive) finite sub-interval. On this basis, a sub-interval [a _K-1 ,∞] can be added to obtain the final K sub-interval. The aforementioned K is determined by different manufacturers. Determined by the decoder fixed-point settings.

After K sub-intervals are clearly divided, for a single sub-interval in the K sub-intervals, the mapping value y _n of the LLR mapped in this single sub-interval R _i ={a _i-1 , a _i } can be calculated, i= 1,2,…K-1, see the following formula:

For the last sub-interval, if l∈R _K =[a _K-1 ,∞], then y _n =sign(x _n )·y _K-1 ;

Among them, x _n is the nth floating-point value of LLR, sign() is the sign function of x _n , a _i is the right endpoint of a single interval, a _i-1 is the left endpoint of a single interval, and l is the absolute value of x _n value, p() is the probability density distribution of the target SNR under AWGN conditions.

Through the above method, the mapping values of the LLR mapping in K sub-intervals can be obtained, and the final N mapping values can be obtained.

It is worth mentioning that although the number of LLR values finally obtained is N, generally N is much larger than K, and K sub-intervals correspond to K types of mapping values. After mapping the N LLR values separately, we will get There are N mapping values, and LLR has a distinction between positive and negative signs, so there are at most ^2K kinds of N mapping values (the mapping values of negative numbers and positive numbers are opposite).

Step 803: Use the target signal-to-noise ratio to perform quantization processing on the N mapping values to obtain a quantization processing result.

In the embodiment of this application, the linear signal-to-noise ratio of the received symbol and the target signal-to-noise ratio are obtained, and then the ratio of the linear signal-to-noise ratio and the target signal-to-noise ratio is calculated, and the ratio is used as the quantization scaling factor, and then the quantization scaling is used The N mapping values are scaled by a factor, and the scaling result is used as the quantization result.

Here, the calculation process of the quantized scaling factor can be seen in the following formula:

Among them, a _n is the quantization scaling factor of the received symbol, snr _n is the linear signal-to-noise ratio of the received symbol, and snr _opt is the target signal-to-noise ratio.

After the quantization scaling factor is calculated, the quantization scaling factor is used to quantize the LLR to obtain the quantization processing result.

In other words, the above processing can be understood as a scaling process, which includes scaling N mapping values using quantized scaling factors, and using the scaling results as quantization processing results.

Specifically, N mapping values can be obtained based on step 801. For a single mapping value among the N mapping values, the following processing operations are performed: calculate the product between the quantized scaling factor and the single mapping value, round it to an integer, and then The rounding result is used as the result of the scaling process of a single mapping value; the above processing operation is repeatedly executed to obtain the results of the scaling processing corresponding to each of the N mapping values, and the result of the scaling processing is used as the result of the quantization processing, that is, the result of the above processing operation. The calculation process can be seen in the following formula:

Among them, q _n is the scaling processing result of a single mapping value, that is, the fixed-point LLR value, a _n is the quantization scaling factor, y _n is a single mapping value, and b is a specified value.

It is worth noting that the above specified value can be determined according to actual application requirements, and the value is usually 0.5.

Through the above processing operations, each mapping value y _n can obtain the corresponding scaling processing result q _n , thus obtaining N mapping values {y ₁ , y ₂ ,..., y _n ,..., y _N } The corresponding scaling processing results of n=1,2,…,N are {q ₁ , q ₂ ,…, q _n ,…, q _N }, n=1, 2,…, N. Here, these are The results of N scaling processes {q ₁ , q ₂ ,…, q _n ,…, q _N }, n=1, 2,…, N are used as the quantization processing results.

Through the method of the embodiment of this application, taking into account the changes in signal-to-noise ratio in different fading scenarios, based on the probability density distribution of the LLR of the received symbols under different signal-to-noise ratios of a certain modulation method, the target signal-to-noise ratio is selected, and based on the selected target The signal-to-noise ratio is used to obtain the quantization scaling factor, and then the N LLR values of the received symbol are mapped based on the saturated bit width of the decoder, specifically mapped to K different continuous sub-intervals divided by the saturated bit width of the decoder, and we get N mapping values, and a quantization scaling factor is used to scale the N mapping values, and then the scaling processing results are rounded to obtain the final quantization processing result. The quantization processing results obtained in this way can not only adapt to decoders of different manufacturers or models, but also help the decoder better identify the size relationship between LLRs, thereby improving the decoding performance of the decoder.

Furthermore, the above method takes into account the actual fading scenario, that is, a method is proposed to obtain the target signal-to-noise ratio. Based on this target signal-to-noise ratio, the LLR quantification processing result is obtained, which also helps to optimize the decoding performance of the decoder and improve decoding. The decoding accuracy of the device.

Optionally, in the embodiment of the present application, for the selection of the target signal-to-noise ratio, simulation can also be used, that is, the decoding performance of the LLR quantization processing results under different signal-to-noise ratios is simulated and tested. Further, the decoding is selected. The signal-to-noise ratio corresponding to the performance that meets the preset performance requirements is used as the target signal-to-noise ratio. Here, the decoding performance can specifically be the decoding rate of the decoder, the preset performance requirements can be set based on actual simulation tests, and the preset performance requirements can also be set according to the requirements of the decoder.

In some possible implementations, the corresponding relationship between the debugging mode and the signal-to-noise ratio can also be set, and this corresponding relationship also includes a priority relationship. That is, taking any debugging mode as an example, it has three preset signal-to-noise ratios. Ratio: signal-to-noise ratio 1, signal-to-noise ratio 2, signal-to-noise ratio 3. The order of priority is also signal-to-noise ratio 1, signal-to-noise ratio 2, and signal-to-noise ratio 3. During the simulation and debugging process, priority will be given to the signal-to-noise ratio. If the signal-to-noise ratio 1 is not the target signal-to-noise ratio that meets the conditions, then simulate and debug the signal-to-noise ratio 2, and so on. If none of the preset correspondences meets the conditions, the signal-to-noise ratio with the best decoding performance will be selected as the target signal-to-noise ratio among the signal-to-noise ratios participating in the simulation debugging. Of course, the corresponding relationship here can also be the value range of the signal-to-noise ratio corresponding to the debugging mode.

Optionally, in this embodiment of the present application, the selection of the target signal-to-noise ratio can also be obtained through model training. The model here can be built based on a neural network. Its input parameters include debugging methods and signal-to-noise ratio sequences, and the output parameters are the signal-to-noise ratio corresponding to the best decoding performance. Specifically, the training parameters of the neural network are first obtained, and then based on this training parameter, for a certain debugging method, the target signal-to-noise ratio in the specified range is used as the input parameter. After iterative training of the model, the output result is obtained, and the output result is used as Target signal-to-noise ratio.

Through the method provided by the embodiments of this application, the changes in signal-to-noise ratio in different fading scenarios can be considered. According to the probability density distribution of the LLR under the target signal-to-noise ratio, the target signal-to-noise ratio is used to quantize the LLR. During the quantization process Different mapping, scaling and rounding of LLRs can enable the decoder to better identify the size relationship between LLRs. Thereby improving decoding performance.

Based on the same inventive concept, this application also provides a soft bit quantization processing device to implement quantization processing of LLR in different fading scenarios and solve the problem of inaccurate decoding results of the decoder caused by inaccurate quantization processing results. Effectively optimize the decoding performance of the decoder and improve the accuracy of the final decoding result. See Figure 10. The device includes:

The acquisition module 1001 obtains the probability density distribution of the log-likelihood ratio LLR of the received symbol under the target signal-to-noise ratio; wherein the probability density distribution includes the uncertainty probability corresponding to each floating point value in the LLR, and the probability The density distribution is the probability density distribution under the condition of additive Gaussian white noise AWGN;

The determination module 1002 determines N mapping values of the LLR based on the probability density distribution; where N is the number of LLR values of the received symbol;

The processing module 1003 uses the target signal-to-noise ratio to perform quantization processing on the N mapping values to obtain a quantization processing result.

In a possible implementation, the determination module 1002 of determining the N mapping values of the LLR mapping is specifically configured to: obtain K sub-intervals of continuous distribution; wherein the K sub-intervals are determined by the decoder. Obtained by dividing the saturated bit width; mapping the LLR in the K sub-intervals to obtain mapped N mapping values.

In a possible implementation, the LLR is mapped in the K sub-intervals, and the determination module 1002 is specifically configured to: for a single sub-interval in the K sub-intervals, use the following formula to calculate The mapping value _yn of the LLR mapped in the single sub-interval R _i ={a _i-1 , a _i }, i=1,2,...K-1, then

If l∈R _K =[a _K-1 ,∞], then y _n =sign(x _n )·y _K-1 ;

In a possible implementation, the target signal-to-noise ratio is used to perform quantization processing on the N mapping values to obtain a quantization processing result. The processing module 1003 is specifically used to: obtain the received symbol Linear signal-to-noise ratio, and the target signal-to-noise ratio; calculate the ratio of the linear signal-to-noise ratio to the target signal-to-noise ratio, and use the ratio as a quantized scaling factor; use the quantized scaling factor to The N mapping values are scaled, and the scaled result is used as the quantization result.

In a possible implementation, the quantized scaling factor is used to perform scaling processing on the N mapping values respectively, and the result of the scaling processing is used as the quantization processing result. The processing module 1003 is specifically used to : For a single mapping value among the N mapping values, perform the following processing operation: calculate the product between the quantization scaling factor and the single mapping value, round it, and use the rounding result as the single mapping The result of the scaling process of the value; repeat the above processing operation to obtain the result of the scaling process corresponding to each of the N mapping values, and use the result of the scaling process as the result of the quantization process.

Based on the above device, taking into account the changes in signal-to-noise ratio in different fading scenarios, the probability density distribution of LLR under the target signal-to-noise ratio is based on a certain modulation method, and the quantized scaling factor is obtained based on the target signal-to-noise ratio, and then based on the saturation of the decoder The bit width performs mapping processing on K different continuous sub-intervals of the LLR to obtain N mapping values, and uses the quantized scaling factor to scale these N mapping values, and then takes the scaling processing results downward. The whole operation is carried out to obtain the final quantitative processing result. The quantization processing results obtained in this way can not only adapt to decoders of different manufacturers or models, but also help the decoder better identify the size relationship between LLRs, thereby improving the decoding performance of the decoder.

Based on the same inventive concept, embodiments of the present application also provide an electronic device, which can realize the function of the aforementioned soft bit quantization processing device. Referring to Figure 11, the electronic device includes:

At least one processor 1101, and a memory 1102 connected to the at least one processor 1101. The specific connection medium between the processor 1101 and the memory 1102 is not limited in the embodiment of this application. In Figure 11, the connection between the processor 1101 and the memory 1102 is Take the example of connecting via bus 1100. The bus 1100 is represented by a thick line in FIG. 11 , and the connection methods between other components are only schematically illustrated and not limited thereto. Bus 1100 can be divided into address bus, data bus Bus, control bus, etc., are only represented by a thick line in Figure 11 for ease of presentation, but this does not mean that there is only one bus or one type of bus. Alternatively, the processor 1101 may also be called a controller, and there is no restriction on the name.

In this embodiment of the present application, the memory 1102 stores instructions that can be executed by at least one processor 1101. By executing the instructions stored in the memory 1102, at least one processor 1101 can perform the soft bit quantization processing method discussed above. The processor 1101 can implement the functions of each module in the device shown in Figure 10.

Among them, the processor 1101 is the control center of the device and can use various interfaces and lines to connect various parts of the entire control device. By running or executing instructions stored in the memory 1102 and calling data stored in the memory 1102, the processor 1101 can Various functions of the device and process data to provide overall monitoring of the device.

In a possible design, the processor 1101 may include one or more processing units, and the processor 1101 may integrate an application processor and a modem processor, where the application processor mainly processes the operating system, user interface and application programs. etc., the modem processor mainly handles wireless communications. It can be understood that the above modem processor may not be integrated into the processor 1101. In some embodiments, the processor 1101 and the memory 1102 can be implemented on the same chip, and in some embodiments, they can also be implemented on separate chips.

The processor 1101 may be a general-purpose processor, such as a central processing unit (CPU), a digital signal processor, an application specific integrated circuit, a field programmable gate array or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, or Implement or execute each method, step and logical block diagram disclosed in the embodiments of this application. A general-purpose processor may be a microprocessor or any conventional processor, etc. The steps of the quantization processing method for soft bits disclosed in the embodiments of the present application can be directly implemented by a hardware processor, or executed by a combination of hardware and software modules in the processor.

As a non-volatile computer-readable storage medium, the memory 1102 can be used to store non-volatile software programs, non-volatile computer executable programs and modules. The memory 1102 may include at least one type of storage medium, for example, may include flash memory, hard disk, multimedia card, card-type memory, random access memory (English: Random Access Memory, referred to as: RAM), static random access memory (English: Static Random Access Memory (abbreviation: SRAM), programmable read-only memory (English: Programmable Read Only Memory, abbreviation: PROM), read-only memory (English: Read Only Memory, abbreviation: ROM), electrically erasable programmable read-only memory (English: Electrically Erasable Programmable Read-Only Memory, referred to as: EEPROM), magnetic memory, magnetic disks, optical disks, etc. Memory 1102 is, but is not limited to, any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer. The memory 1102 in the embodiment of the present application can also be a circuit or any other device capable of realizing a storage function, used to store program instructions and/or data.

By designing and programming the processor 1101, the code corresponding to the quantization processing method of soft bits introduced in the previous embodiment can be solidified into the chip, so that the chip can execute the soft bits of the embodiment shown in Figure 8 during operation. of Steps in the quantitative processing method. How to design and program the processor 1101 is a technology well known to those skilled in the art, and will not be described again here.

Based on the same inventive concept, embodiments of the present application also provide a storage medium that stores computer instructions. When the computer instructions are run on a computer, they cause the computer to execute the soft bit quantization processing method discussed above.

In some possible implementations, various aspects of the soft bit quantization processing method provided by this application can also be implemented in the form of a program product, which includes program code. When the program product is run on a device, the program code is used to The control device is caused to perform the steps in the quantization processing method of soft bits according to various exemplary embodiments of the present application described above in this specification.

Those skilled in the art will understand that embodiments of the present application may be provided as methods, systems, or computer program products. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment that combines software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.

The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to the present application. It will be understood that each process and/or block in the flowchart illustrations and/or block diagrams, and combinations of processes and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a controller of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that the instructions executed by the controller of the computer or other programmable data processing device produce a use A device for realizing the functions specified in one process or multiple processes of the flowchart and/or one block or multiple blocks of the block diagram.

These computer program instructions may also be stored in a computer-readable memory that causes a computer or other programmable data processing apparatus to operate in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including the instruction means, the instructions The device implements the functions specified in a process or processes of the flowchart and/or a block or blocks of the block diagram.

These computer program instructions may also be loaded onto a computer or other programmable data processing device, causing a series of operating steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby executing on the computer or other programmable device. Instructions provide steps for implementing the functions specified in a process or processes of a flowchart diagram and/or a block or blocks of a block diagram.

Although exemplary embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once the basic inventive concepts are understood. Therefore, it is intended that the appended claims be construed to include the exemplary embodiments and all changes and modifications that fall within the scope of this application.

Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. scope. In this way, if these modifications and variations of the present application fall within the scope of the claims of the present application and equivalent technologies, the present application is also intended to include these modifications and variations.

Claims

A quantification method for log-likelihood ratio LLR values, including:

Obtain multiple signal to interference plus noise ratio SINR linear values corresponding to multiple received symbols obtained by demodulation;

Calculate a plurality of mutual information corresponding to the plurality of received symbols based on the plurality of SINR linear values corresponding to the plurality of received symbols;

Determine a scaling factor for quantizing multiple LLR values based on the average of the multiple mutual information corresponding to the multiple received symbols, where the multiple LLR values include log-likelihood based on the multiple received symbols The LLR value of each received symbol calculated by the ratio algorithm.
The method according to claim 1, wherein determining the scaling factor for quantizing the plurality of LLR values based on an average of the plurality of mutual information corresponding to the plurality of received symbols includes:

The scaling factor is determined based on an average of the plurality of mutual information of the plurality of received symbols, a modulation order from the transmitting end, and a coding rate from the transmitting end.
The method of claim 2, further comprising:

Determine SINR linear gain parameters of the multiple received symbols according to the average value of the multiple mutual information corresponding to the multiple received symbols; and

Determining the scaling factor based on the average value of the multiple mutual information corresponding to the multiple received symbols, the modulation order from the transmitting end, and the coding rate from the transmitting end, specifically includes: :

The scaling factor is determined based on the SINR linear gain parameters of the plurality of received symbols, the modulation order, and the coding rate.
The method of claim 3, wherein the SINR linear gain parameter includes an average of SINR linear values.
The method according to any one of claims 1 to 4, wherein the average value of the mutual information of the plurality of received symbols is calculated based on the following formula:

Wherein, I equal is the average value of the multiple mutual information, M is the number of the multiple received symbols, sinr(m) is the SINR linear value of the m-th received symbol, and F(sinr(m)) is the The mutual information of the mth received symbol, where 1≤m≤M.
The method according to claim 5, wherein the average value of the SINR linear value is calculated by the following formula:
SINR equal =F -1 (I equal ),

Wherein, SINR equal is the average of the SINR linear values of the multiple received symbols, I equal is the average of the multiple mutual information, and F -1 is the inverse function of the F function.
The method according to any one of claims 2 to 4, wherein the average value of the plurality of mutual information according to the plurality of received symbols, the modulation order from the transmitting end and the The coding rate from the sending end determines the scaling factor, including:

Obtain a selection function determined in advance through machine learning; and

The average value of the mutual information corresponding to the multiple received symbols, the modulation order and the coding rate are used as independent variables of the selection function, and the dependent variable calculated by the selection function is used as The scaling factor.
The method according to claim 7, wherein the selection function is obtained by using randomly generated random SINR linear values, random modulation orders and random coding rates as training samples and training based on an unsupervised learning algorithm, wherein the random The modulation order and the random coding rate are the control variables of the unsupervised learning algorithm, and the random SINR linear value is the independent variable of the unsupervised learning algorithm.
The method of claim 1, further comprising:

When the total number of all received symbols is greater than the preset threshold, the number of the multiple received symbols is set to be less than the total number of all demodulated received symbols.
The method according to claim 9, wherein the plurality of received symbols is a set of P received symbols separated by S received symbols among all the received symbols obtained by demodulation, S is an integer greater than or equal to 1, and P is an integer greater than or equal to 1.
The method according to any one of claims 1 to 10, wherein the plurality of SINR linear values have a preset linear relationship with the SINR of a corresponding received symbol among the plurality of received symbols.
A method according to any one of claims 1 to 11, wherein said method is used in a decoder that requires the calculation of fixed-point LLR values as input.
A quantification device for log likelihood ratio LLR values, including:

An acquisition unit, configured to acquire multiple signal and interference plus noise ratio SINR linear values corresponding to the multiple received symbols obtained by demodulation;

A processing unit configured to calculate a plurality of mutual information corresponding to the plurality of received symbols based on the plurality of SINR linear values corresponding to the plurality of received symbols;

The processing unit is further configured to determine a scaling factor for quantizing multiple LLR values based on an average of the multiple mutual information corresponding to the multiple received symbols, where the multiple LLR values include The LLR value of each received symbol calculated based on the log-likelihood ratio algorithm.
The apparatus according to claim 13, wherein the plurality of received symbols based on SINR linear values, calculating the multiple mutual information corresponding to the multiple received symbols, including:

The scaling factor is determined according to the average value of the mutual information corresponding to the multiple received symbols, the modulation order from the transmitting end, and the coding rate from the transmitting end.
The device according to claim 14, wherein the processing unit is also used for:

Determine the SINR linear gain parameters of the multiple received symbols according to the average value of the multiple mutual information corresponding to the multiple received symbols;

Determining the scaling factor based on the average value of the mutual information corresponding to the multiple received symbols, the modulation order from the transmitting end, and the coding rate from the transmitting end includes:

The scaling factor is determined based on the SINR linear gain parameters of the plurality of received symbols, the modulation order, and the coding rate.
The device according to claim 14 or 15, wherein the average value of the mutual information corresponding to the plurality of received symbols, the modulation order from the transmitting end and the modulation order from the The coding rate of the sending end determines the scaling factor, including:

Obtain the selection function determined in advance through machine learning through the acquisition unit;

The average value of the mutual information corresponding to the multiple received symbols, the modulation order and the coding rate are used as independent variables of the selection function, and the dependent variable calculated by the selection function is used as The scaling factor.
A quantification processing method for LLR values, including:

Obtain the probability density distribution of the log-likelihood ratio LLR of the received symbol under the target signal-to-noise ratio, where the probability density distribution includes the uncertainty probability corresponding to each floating point value in the LLR, and the probability density distribution is Probability density distribution under additive Gaussian white noise AWGN conditions;

Based on the probability density distribution, N mapping values of the LLR are determined, where N is the number of LLR values of the received symbols; and

The target signal-to-noise ratio is used to perform quantization processing on the N mapping values to obtain a quantization processing result.
A quantization processing device for LLR values, including:

The acquisition module is used to obtain the probability density distribution of the log-likelihood ratio LLR of the received symbol under the target signal-to-noise ratio, wherein the probability density distribution includes the uncertainty probability corresponding to each floating point value in the LLR, and the The probability density distribution is the probability density distribution under the condition of additive white Gaussian noise AWGN;

A determination module configured to determine N mapping values of the LLR based on the probability density distribution, where N is the number of LLR values of the received symbols; and

A processing module configured to use the target signal-to-noise ratio to perform quantization processing on the N mapping values to obtain a quantization processing result.
An electronic device including:

Memory, for storing computer programs or instructions; and

A controller, configured to execute computer programs or instructions in the memory, so that the method according to any one of claims 1 to 12 or the method according to claim 17 is executed.
A computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions. When called by a computer, the computer-executable instructions cause the computer to execute any one of claims 1 to 12. The method described in the item or the method described in claim 17.