Linear quantization method for input LLR signal of decoder, computer program product, storage medium and electronic equipment

Abstract

The invention provides a linear quantization method for an input LLR signal of a decoder, and the method comprises the following steps: obtaining the conditional probability distribution of an output LLR signal of a demodulator based on a current communication system parameter; determining a saturation level for a linear quantizer and a bit width of the linear quantizer by using a maximum mutual information algorithm according to the conditional probability distribution; determining a quantization factor of the linear quantizer based on the determined bit width and saturation level; and performing linear quantization processing on the LLR signal by using the quantization factor to serve as an input LLR signal of the decoder. According to the invention, efficient linear quantization of the LLR signal output by the demodulator can be realized, and the bit width of the input LLR signal is reduced, so that hardware resources and power consumption of a subsequent decoder are saved.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to linear quantization methods for input LLR signals of decoders, computer program products, computer-readable storage media, and electronic devices. Background Technology

[0002] With the continuous development of modern communication technology, digital receivers typically improve the sensitivity of communication systems by increasing bandwidth and using error correction coding. To enhance performance, current mainstream communication systems generally use "soft decision" information as input to the decoder. Unlike demodulators / decoders using "hard decision" (where the decision received information bits have only two levels: "0" or "1"), soft decision demodulators output "soft" bits, consisting of multiple levels. In some implementations, such as digital signal processor (DSP) demodulators, the output can be floating-point type. Compared to hard decision, under the same channel conditions, these soft bits can provide more information about the data transmitted by the transmitter. Therefore, decoders based on these "soft decision" inputs offer better error correction capabilities compared to decoders with "hard decision" inputs.

[0003] In practical system design, "soft decision information" is usually provided to the decoder in the form of log-likelihood ratio (LLR). LLR is usually represented by a signed real number. A large positive number indicates that the received bit is more likely to be a "1" sent by the transmitter, a large negative number indicates that the received bit is more likely to be a "0" sent by the transmitter, and an LLR with an amplitude close to 0 indicates that the demodulator cannot provide much information about the bits sent by the transmitter, that is, it is difficult to determine the value of the original bit.

[0004] However, in practical applications (especially application-specific integrated circuits such as ASICs), the accuracy of LLR signals is often limited because the decoder's area and power consumption are strongly correlated with the input bit width and accuracy of the LLR. Whether it's a convolutional code (BCC), low-density parity-check code (LDPC), Turbo code, or Polar code, the higher the bit width and accuracy of the LLR input, the greater the hardware resource and power consumption. Therefore, in hardware implementation, the high-precision LLR signal output by the demodulator needs to be quantized before being input to the decoder, converting continuous (or discrete high-precision) "soft information" into discrete values ​​with a finite bit width. If the LLR signal is directly input to the decoder without quantization, it will lead to a significant increase in hardware area and power consumption, making it difficult to meet the requirements of actual productization.

[0005] Therefore, how to provide a design method for an LLR quantizer that can save hardware resources and power consumption of the back-end decoder while meeting decoding performance requirements remains a key problem that urgently needs to be solved in this field. Summary of the Invention

[0006] This application aims to solve at least one of the technical problems existing in the prior art or related art. To this end, this application provides a linear quantization method for the input LLR signal of a decoder, a computer program product, a computer-readable storage medium, and an electronic device.

[0007] According to a first aspect of this application, a linear quantization method for the input LLR signal of a decoder is provided, comprising the following steps:

[0008] Based on the current communication system parameters, obtain the conditional probability distribution of the demodulator's output LLR signal;

[0009] Based on the conditional probability distribution, the saturation level and bit width of the linear quantizer are determined using the maximum mutual information algorithm.

[0010] Based on the determined bit width and saturation level, determine the quantization factor of the linear quantizer; and

[0011] The LLR signal is linearly quantized using a quantization factor and used as the input LLR signal for the decoder.

[0012] In some embodiments, communication system parameters include at least one or more of modulation and coding schemes, channel type, and signal-to-noise ratio.

[0013] In some embodiments, the conditional probability distribution is obtained through offline system simulation.

[0014] In some embodiments, the maximum mutual information algorithm employs a linear programming method that maximizes the mutual information of a nonlinear quantizer.

[0015] In some embodiments, the nonlinear quantizer is a quantizer designed for binary input discrete memoryless channels that maximizes the mutual information between the quantizer input and the quantizer output.

[0016] In some embodiments, the method further includes: determining an upper bound for the maximum mutual information of the linear quantizer using a maximum mutual information algorithm based on a conditional probability distribution; and

[0017] The bit width of the linear quantizer is determined based on the upper bound of the maximum mutual information.

[0018] In some embodiments, determining the bit width for the linear quantizer includes:

[0019] Set a mutual information performance threshold, calculate the mutual information value of the linear quantizer for each bit width by traversing different bit widths, and select the smallest bit width that can meet the mutual information performance threshold as the bit width of the linear quantizer.

[0020] In some embodiments, the mutual information performance threshold is 90% to 100% of the maximum mutual information upper bound.

[0021] In some embodiments, an exhaustive search method is used to traverse different bit widths, with the bit width ranging from 3 to 9.

[0022] In some embodiments, the mutual information value of the linear quantizer is calculated based on the sum of partial mutual information of the output of the linear quantizer.

[0023] In some embodiments, the formula for calculating the quantization factor of a linear quantizer is:

[0024]

[0025] Where k is the quantization factor, m is the determined bit width, and T is the saturation level.

[0026] In some embodiments, linear quantization of the LLR signal using a quantization factor includes:

[0027] The LLR signal is multiplied by a quantization factor for scaling.

[0028] Round the scaled values ​​to the nearest integer; and

[0029] The rounded result is saturated and limited to generate an integer signal with a finite bit width.

[0030] According to a second aspect of this application, a computer program product is also provided, the computer program product including computer instructions that, when executed by a processor, implement any of the methods described above.

[0031] According to a third aspect of this application, a non-volatile computer-readable storage medium is also provided, on which computer instructions are stored, which, when executed by a processor, implement any of the methods described above.

[0032] According to a fourth aspect of this application, an electronic device is also provided, comprising: a memory, a processor, and computer instructions stored in the memory and executable on the processor, wherein the computer instructions, when executed by the processor, implement any of the methods described above.

[0033] In a linear quantization scheme for the input LLR signal of a decoder provided in this application embodiment, the conditional probability distribution of the demodulator output LLR signal is obtained, and combined with the quantizer mutual information maximization algorithm, the saturation level and bit width of the optimal quantizer (usually a nonlinear quantizer) can be obtained. Furthermore, these parameters can serve as the saturation level and output bit width of the linear quantizer, providing a basis for the subsequent optimization design of linear quantizer parameters (e.g., quantization factor). Therefore, on the one hand, this application embodiment avoids the high dependence on system simulation and the risk of parameters getting trapped in local optima in traditional methods, effectively improving the accuracy and practicality of the quantization process; on the other hand, this application embodiment can achieve efficient linear quantization of the demodulator output LLR signal, reducing the bit width of the input LLR signal, thereby saving hardware resources and power consumption of the subsequent decoder. Attached Figure Description

[0034] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. The drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 A structural block diagram of a communication system applicable according to an embodiment of this application is shown.

[0036] Figure 2 A flowchart illustrating a linear quantization method for the input LLR signal of a decoder according to a first embodiment of this application is shown.

[0037] Figure 3A and Figure 3B The diagrams show the conditional probability distributions of LLR signals under different signal-to-noise ratio environments.

[0038] Figure 4 This diagram shows the equivalent structural block diagram of the communication system to which the embodiments of this application apply.

[0039] Figure 5 The pseudocode flowchart shown in this application illustrates the calculation of the optimal bit width m for a linear quantizer.

[0040] Figure 6 A flowchart illustrating a method for generating a parameter lookup table for linear quantization of the input LLR signal of a decoder, according to a second embodiment of this application, is shown.

[0041] Figure 7 The flowchart illustrates a method for obtaining a conditional probability mass function as shown in this application.

[0042] Figure 8A flowchart illustrating a method for determining the quantization factor for a linear quantizer, as shown in this application, is presented.

[0043] Figure 9 A flowchart illustrating the method for designing a lookup table according to a second embodiment of this application is shown.

[0044] Figure 10 A flowchart illustrating a linear quantization method for the output LLR signal of a demodulator according to a third embodiment of this application is shown.

[0045] Figure 11 A flowchart illustrating the linear quantization processing method exemplified in this application is shown.

[0046] Figure 12 A schematic diagram of the structure of an LLR signal quantizer according to an embodiment of this application is shown.

[0047] Figure 13 A flowchart illustrating the online quantization method of the third embodiment of this application is shown.

[0048] Figure 14 A schematic diagram of the structure of an electronic device according to an embodiment of this application is shown. Detailed Implementation

[0049] The technical solution of this application will now be clearly and completely described with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0050] To enable those skilled in the art to better understand the present invention, firstly, in conjunction with Figure 1 The technical basis of the communication system and LLR quantizer applicable to the embodiments of this application is explained.

[0051] Figure 1 A structural block diagram of a communication system applicable according to an embodiment of this application is shown. The communication system 1 includes a transmitter side and a receiver side.

[0052] The transmitter side mainly includes modules such as FEC encoder 101, mapper 102, modulator 103, driver 104, and channel 105. Specifically, the information data to be transmitted is first input to FEC encoder 101 in the form of a binary bit stream. After forward error correction (FEC), the output coded bit stream is symbol-mapped by mapper 102 (such as QPSK, 16QAM, 64QAM, etc.), then converted by modulator 103 into a real or complex signal suitable for transmission through the physical channel, and finally amplified by driver 104 before being transmitted to the receiver through channel 105. The channel can be wireless (single antenna or multiple antennas) or wired, without being limited by the specific implementation method.

[0053] The receiver side mainly includes modules such as analog front-end 106, digital front-end 107, channel estimator 108, demodulator 110, inverse mapper 111, LLR quantizer 113, and FEC decoder 115. The received signal is first processed by analog front-end 106 and digital front-end 107, and then sent to demodulator 110. Demodulator 110, in conjunction with inverse mapper 111, outputs a soft-decision bit information stream, namely the log-likelihood ratio (LLR) signal y, for use by subsequent decoders. This signal is the object of quantization processing in this application.

[0054] Meanwhile, the output signal of the digital front-end 107 is also sent to the channel estimation module 108, where channel state information (CSI), delay spread, channel amplitude / phase, signal-to-noise ratio (SNR), etc. are calculated based on the received signal, providing a basis for subsequent quantization parameters.

[0055] Currently, many communication systems, such as Digital Subscriber Line (DSL), Modem, 5G, Digital Video Broadcasting (DVB), Wi-Fi, and Ultra-Wideband (UWB), widely employ error correction coding technology to improve the reliability and efficiency of information transmission. With technological advancements, decoder complexity continues to increase, and the bit width of its input signal has an increasingly significant impact on the decoder's hardware complexity and power consumption.

[0056] The core LLR quantizer 113 of this application is located between the demodulator 110 and the decoder 115. It is used to quantize the high dynamic range, high bit width real LLR signal y output by the demodulator into an integer z with a finite bit width. This measure aims to significantly reduce the hardware complexity and power consumption of the back-end decoder while ensuring decoding performance. The quantized LLR signal z is finally input to the decoder 115 to complete error correction decoding to recover the original information data.

[0057] In the design of LLR quantizers, they are currently mainly divided into linear quantizers and non-linear quantizers. Non-linear quantizers are better at preserving the mutual information between the quantizer input and output, but they require complex non-linear operations during implementation, which is not conducive to hardware implementation and consumes a lot of resources. Therefore, in engineering, linear quantizers, which have a simpler structure and are easier to implement in hardware, are preferred.

[0058] The aforementioned technical foundation constitutes the communication system for implementing the embodiments of this application. The specific embodiments of this application will be described in detail below based on this system.

[0059] [First Embodiment]

[0060] The first embodiment of this application proposes a linear quantization method for the input LLR signal of a decoder. This method details the entire process from parameter determination to signal quantization, aiming to achieve the given performance requirements for the input / output mutual information of the quantizer while maximizing the saving of hardware resources and power consumption of the back-end decoder.

[0061] Figure 2 A flowchart illustrating a linear quantization method for the input LLR signal of a decoder according to a first embodiment of this application is shown, including the following steps S201 to S204.

[0062] Step S201: Based on the current communication system parameters, obtain the conditional probability distribution of the output LLR signal of the demodulator.

[0063] In this step, the communication system parameters include at least one or more of the following: modulation and coding scheme (MCS), channel type, and signal-to-noise ratio (SNR).

[0064] Modulation-coding scheme (MCS) describes the combination of modulation scheme (such as QPSK, 64QAM, 4096QAM) and channel coding rate used for data transmission. Different MCS levels correspond to different transmission rates and robustness. MCS information can be obtained by the receiver through demodulation of the current frame preamble signal, or it can be known in advance by other means.

[0065] Channel types can include channels with delay spread characteristics, multipath channels, and other types of channels, such as additive white Gaussian noise (AWGN) channels, block fading channels (B), frequency selective fading channels (F), etc.

[0066] It is understandable that in practical communication systems (such as LTE, Wi-Fi, UWB, 5G), MCS is used to select modulation and coding methods, channel type is used to describe the channel environment, and SNR is used to measure signal quality.

[0067] As an example, this conditional probability distribution can be obtained through offline system simulation.

[0068] The specific operation is as follows: A real or approximate communication channel model and demodulator model can be used. For given communication system parameters, the conditional probability mass function (pmf) of the demodulator model's output LLR signal is obtained offline. This means obtaining the conditional probability distribution (Py|x) of the demodulator's output LLR signal under different bit inputs (x=0 or x=1). It can be understood that the conditional probability mass function P(y|X=x) describes the probability distribution of the LLR output y for each possible discrete value, given the transmitted bit x (which is 0 or 1).

[0069] For easier intuitive understanding, Figure 3A and Figure 3B The diagrams show the conditional probability distribution of LLR signals under different signal-to-noise ratio (SNR) environments, illustrating the probabilistic relationship between received signals and transmitted bits (typically "0" and "1") under different SNR conditions.

[0070] like Figure 3A As shown, under high SNR, the distribution curves of P(y|x=0) and P(y|x=1) have high separation, indicating that in a high signal-to-noise ratio environment, the noise interference is small, the received signal can clearly distinguish whether the transmitted signal is "0" or "1", and the signal decision accuracy is strong.

[0071] like Figure 3B As shown, under low SNR, the distributions of P(y|x=0) and P(y|x=1) have a large overlap. In a low signal-to-noise ratio environment, the noise interference is large, and the received signal is difficult to clearly distinguish between the transmitted "0" and "1", which significantly increases the probability of misjudgment.

[0072] It is worth noting that this example employs offline simulation rather than theoretical analysis. The main reason is that, even under the same channel conditions, the probability distribution of the output differs for different types of demodulators (e.g., zero-force demodulators, maximum likelihood detection demodulators, or rake receivers). Furthermore, due to the complex effects of multipath propagation, fading, interference, and noise introduced by the physical channel, it is generally difficult to obtain the conditional probability distribution of the equivalent channel LLR output analytically. Therefore, it is difficult to accurately describe the distribution law of LLR using formulas or theoretical analysis. In view of this, this embodiment uses simulation to obtain the conditional probability distribution Py|x.

[0073] Step S202: Based on the conditional probability distribution, use the maximum mutual information algorithm to determine the saturation level and bit width for the linear quantizer.

[0074] To further clarify the technical solutions of the embodiments of this application, see [link to relevant documentation]. Figure 4 , Figure 4 This diagram illustrates the equivalent structural block diagram of the communication system in an embodiment of this application. Specifically, it combines... Figure 1 The communication system can be equivalently modeled as a structure consisting of an equivalent channel 117 formed by the FEC encoder 101 to the demodulator 110 cascaded with an LLR quantizer 113. This equivalent model provides a criterion for maximizing the input / output mutual information of the LLR quantizer 113 in its parameter design. Specifically, mutual information... The mutual information value is used to measure the degree of information correlation between the transmitted bit x and the output z of the LLR quantizer 113. The higher the mutual information value, the more information about the original transmitted bit x is retained in the quantized output z. The decoder can recover the original data more effectively, thereby significantly improving the decoding performance and the overall system performance.

[0075] In this step, the maximum mutual information algorithm refers to minimizing information loss during the quantization process by maximizing the mutual information between the input and output of the quantizer. Based on this algorithm, the optimal values ​​of the following parameters can be obtained: (1) saturation level T; (2) bit width m.

[0076] The optimal saturation level T refers to the amplitude limiting of the input LLR signal when its amplitude exceeds T, confining it to within ±T. A properly set saturation level T helps suppress the impact of abnormally large LLR signals on subsequent decoding performance and hardware resources, improving system robustness. The optimal quantizer refers to the quantizer Q among all available options (including linear and nonlinear quantizers) that maximizes input / output mutual information. quantizer Its mathematical expression is as follows:

[0077]

[0078] Typically, the optimal quantizer obtained through the above optimization process is a nonlinear quantizer. The optimal bit width *m* is the input / output mutual information *I* that enables the optimal quantizer. m (X;Z) The bit width parameter that reaches its maximum value.

[0079] Therefore, in this embodiment, by obtaining the conditional probability distribution of the demodulator output LLR signal and combining it with the quantizer mutual information maximization algorithm, the saturation level T and bit width m of the optimal quantizer (usually a nonlinear quantizer) can be obtained. Furthermore, these parameters can also be used as the saturation level and output bit width of a linear quantizer, providing a basis for subsequent optimization design of linear quantizer parameters. Thus, while reducing hardware implementation complexity and power consumption, decoder performance and system reliability can be maximized. In addition, based on the above design process, the amplitude range and output bit width of the linear quantizer can be reasonably limited while ensuring decoding performance, achieving a balance between hardware complexity and system performance.

[0080] As an example, the maximum mutual information algorithm can employ a linear programming approach that maximizes the mutual information of a nonlinear quantizer. This nonlinear quantizer can be designed for a discrete memoryless channel (DMC) with binary inputs, and its optimization objective is to maximize the mutual information between the quantizer's input and output. In other words, the quantizer's structure and parameter configuration are designed with maximizing the mutual information between the quantizer's input and output as the guiding principle, ensuring that the decoder obtains as much information as possible about the original transmitted bits.

[0081] For example, Brian M et al. proposed an optimization algorithm based on linear programming in "Quantization of Binary-Input Discrete Memoryless Channels" (IEEE, May 15, 2014), which can design the optimal LLR quantizer under binary-input DMC conditions. This type of quantizer typically exhibits a nonlinear structure in practical designs, which is beneficial for further enhancing information preservation capabilities. The aforementioned literature is incorporated herein by reference.

[0082] In the specific implementation of this application embodiment, based on the conditional transition probability P(y|x) obtained in step S201, and assuming the bit transmission probability P(x=0)=P(x=1)=0.5, the above-mentioned optimal nonlinear quantizer design method is adopted. Based on the linear programming algorithm, relevant parameters for linear quantizer design are effectively obtained, including the saturation level T and the upper bound of maximum mutual information. .

[0083] The maximum mutual information upper bound achievable by this nonlinear quantizer The calculation expression is as follows:

[0084] in, This represents the optimal quantizer structure. This refers to the mutual information between the transmitted bit x and the quantizer output z.

[0085] Therefore, this application provides a specific parameter calculation method. By combining the channel conditional probability distribution obtained from simulation, the optimal parameters for the linear quantizer are determined using the maximum mutual information algorithm of the nonlinear quantizer. This provides a theoretical basis for calculating the saturation level and upper bound of mutual information of the linear quantizer, and further guides the selection of the quantizer bit width and quantization factor.

[0086] Furthermore, based on the obtained upper bound of maximum mutual information The optimal bit width m required for the linear quantizer in the embodiments of this application can be determined.

[0087] As an example, determining the bit width for a linear quantizer may include the following steps: first, setting a mutual information performance threshold d; then, iterating through all possible bit width values ​​and calculating the corresponding linear quantizer mutual information value for each bit width m. The minimum bit width that can satisfy the mutual information performance threshold is selected as the final bit width of the linear quantizer.

[0088] In this example, the optimal bit width m of the linear quantizer is obtained using an exhaustive search method. Preferably, the value of the bit width m is limited to the candidate set M = [3, 4, ..., 9]. The smaller the selected value of m, the lower the hardware resources and power consumption required by the decoder, which is beneficial for system integration and energy efficiency optimization.

[0089] In the actual design process, the required bit width m should meet the following conditions:

[0090] ,

[0091] ,

[0092] in, This represents the maximum mutual information that a linear quantizer with a bit width of m can achieve;

[0093] This represents the upper bound of the maximum mutual information;

[0094] d represents the threshold factor, used to control the required threshold. near The degree of.

[0095] It should be understood that setting the threshold factor d can flexibly balance the performance and resource consumption of the linear quantizer, avoiding mutual information loss and decoding performance degradation caused by excessively low bit width. According to simulation results, the preferred value of the threshold factor d is d>0.9 to prevent performance degradation due to over-compression.

[0096] Figure 5The following is a pseudocode flowchart illustrating the calculation of the optimal bit width m for a linear quantizer, as shown in this application example. Specifically, the mutual information of the linear quantizer for each bit width in the candidate bit width set M is calculated one by one. And determine whether the mutual information value is greater than or equal to the performance threshold. If the condition is met, the traversal ends and the current m is selected as the optimal bit width; otherwise, the traversal continues to the next bit width until the traversal ends.

[0097] Through the above calculation process, we can ensure that the minimum quantizer bit width m is selected while meeting the mutual information performance requirements. This effectively reduces the hardware complexity and power consumption of the decoder, improves system energy efficiency, and ensures that the decoding performance is not significantly affected.

[0098] Optionally, if the required bit width m cannot be determined under the preset threshold factor d, the threshold factor d can be appropriately increased, and the above parameter search process can be re-executed to obtain the optimal bit width that meets the requirements. Furthermore, the final value of the threshold factor d can also be comprehensively evaluated in conjunction with system simulation results (such as actual performance indicators such as frame error rate FER), thereby avoiding waste of system resources or insufficient performance due to unreasonable parameter settings.

[0099] As an example, the mutual information value of a linear quantizer can be calculated by accumulating the partial mutual information of the output of the linear quantizer.

[0100] Optionally, the mutual information of the linear quantizer The calculation method can be found in patent application GB2434948A, which is incorporated herein by reference. In actual calculations, the partial mutual information corresponding to each output of the quantizer can be calculated step-by-step, and the contributions of each output can be accumulated to obtain the mutual information of the linear quantizer. .

[0101] Set up a linear quantizer The preimage of a specific output z is , in This represents the value space of the LLR signal output by the demodulator. Then, the linear quantizer... Output conditional probability The distribution can be represented as:

[0102]

[0103] For each quantizer output z, it provides partial conditional mutual information. It can be represented as:

[0104]

[0105] Finally, the linear quantizer Mutual information can be obtained through partial conditional mutual information of all possible outputs z. To perform accumulation, the specific expression is:

[0106]

[0107] The above calculation method can accurately evaluate the contribution of different quantization levels to the total mutual information, providing a theoretical basis for quantizer parameter optimization and performance evaluation. This method can effectively guide the design of linear quantizers, achieving a reasonable balance between decoding performance and hardware resource consumption.

[0108] Step S203: Determine the quantization factor of the linear quantizer based on the determined bit width and saturation level.

[0109] In this step, based on the saturation level T determined in step S202 and the selected bit width m, the quantization factor k of the linear quantizer is further calculated. The formula for calculating the quantization factor is as follows:

[0110]

[0111] in, Indicates the quantification factor. Indicates the selected bit width. This indicates the determined saturation level. Quantization factor. It is used to linearly map the amplitude of the input signal (such as the LLR value) to the integer range of the quantizer output, so as to achieve a reasonable correspondence between the input signal and the range of the quantizer output.

[0112] Specifically, the above formula linearly maps the saturation level T of the nonlinear quantizer to the saturation output value A of the linear quantizer, thereby achieving a reasonable allocation of the quantization interval. In other words, through this mapping, the quantization factor is utilized... It can convert the input signal (maximum amplitude T) into the maximum output integer value of the linear quantizer (by...). This decision effectively improves the utilization rate of the quantized signal and the decoding performance of subsequent decoders.

[0113] It should be noted that the saturation level T and the maximum mutual information The quantization factor k is theoretically optimized based on a Discrete Memoryless Channel (DMC) model. However, in real-world physical channel environments, certain memory effects may exist, such as multipath propagation, trailing, and colored noise. These factors can cause the reception of the current bit to be affected by the preceding and following bits or signals. Therefore, in practical applications, the calculated quantization factor k value may need to be fine-tuned based on system simulation or actual performance feedback to ensure optimal decoding performance and reasonable utilization of hardware resources.

[0114] Step S204: Perform linear quantization on the LLR signal using a quantization factor to serve as the input LLR signal for the decoder.

[0115] In this step, based on the quantization factor k determined in step S203, the LLR signal y output by the demodulator is linearly quantized to obtain the input signal z required by the decoder. The specific processing flow may include the following steps:

[0116] Multiply the LLR signal y by the quantization factor k to scale the signal amplitude and make it fit the output range of the quantizer.

[0117] The scaled values ​​are rounded, preferably using methods such as rounding to the nearest integer, to improve quantization accuracy and optimize decoding performance; and

[0118] The rounded result is subjected to saturation limiting to ensure that the resulting integer signal z is within a predetermined finite bit-width integer range (e.g., [-A, A]), thereby preventing overflow or distortion.

[0119] In summary, through steps S201 to S204, the first embodiment of this application provides a linear quantization method for the input LLR signal of a decoder. By obtaining the conditional probability distribution of the demodulator output LLR signal and combining it with a quantizer mutual information maximization algorithm, the saturation level and bit width of the optimal quantizer (usually a nonlinear quantizer) can be obtained. Furthermore, the above parameters can be used as the saturation level and output bit width of the linear quantizer, providing a basis for the subsequent optimization design of linear quantizer parameters (e.g., quantization factor). Thus, on the one hand, the embodiment of this application avoids the high dependence on system simulation and the risk of parameters getting trapped in local optima in traditional methods, effectively improving the accuracy and practicality of the quantization process; on the other hand, the embodiment of this application can achieve efficient linear quantization of the demodulator output LLR signal, reducing the bit width of the input LLR signal, thereby saving hardware resources and power consumption of the subsequent decoder.

[0120] [Second Embodiment]

[0121] The second embodiment of this application provides a method for generating a lookup table of linear quantization parameters for the input LLR signal of a decoder. The focus is on the offline design and parameter storage process of the method, and its core is to construct a lookup table (LUT) that can be called online by the communication system.

[0122] Unlike the first embodiment, which directly addresses real-time quantization processing of a single LLR signal, the main objective of this embodiment is to generate and store a parameter lookup table (LUT). Specifically, this application proposes a systematic LUT design method. Through offline simulation analysis of the conditional probability mass function (CPMF) of the demodulator output LLR signal under different channel conditions and modulation schemes, and combined with mutual information theory, the optimal quantization parameters (such as quantization factor K, bit width m, saturation level T, etc.) are calculated for each set of channel parameters and modulation schemes, and stored in the LUT.

[0123] Figure 6 A flowchart illustrating a method for generating a parameter lookup table for linear quantization of the input LLR signal of a decoder, according to a second embodiment of this application, is shown, including the following steps S601 to S603.

[0124] Step S601: Determine the communication system parameters used for the lookup table, and obtain the conditional probability quality function of the LLR signal output by the demodulator based on the communication system parameters.

[0125] In this step, the communication system parameters used for the lookup table are first determined. These parameters will serve as the index for the lookup table and include at least one or more of the following: modulation and coding scheme (MCS), channel type, and signal-to-noise ratio (SNR). After defining the dimensions of the lookup table, based on the aforementioned communication system parameters, a real or approximate communication channel model and demodulator model are used to obtain the conditional probability quality function P(y|x) of the LLR signal output by the demodulator through offline operation.

[0126] As an example, see Figure 7 The flowchart illustrates a method for obtaining a conditional probability mass function according to an example of this application, which may include the following steps S7001 to S7004:

[0127] Step S7001: Obtain the transmit bits output by the encoder and the corresponding LLR signal output by the demodulator;

[0128] Step S7002: According to the value of the transmitted bit, divide the LLR signal into two groups of data: one group with bit input 0 and the other group with bit input 1.

[0129] Step S7003: For each set of LLR data, count the frequency of each possible value of the LLR signal and normalize it to a probability, obtaining the conditional probability distribution of the LLR signal under the conditions of bit input 0 and bit input 1 respectively; and

[0130] Step S7004: Save the obtained conditional probability distribution as a conditional probability mass function.

[0131] Specifically, the conditional probability distribution T(z|x) of the quantizer output z given the input x can be expressed as:

[0132]

[0133] in:

[0134] The bit (0 or 1) represents the sending end;

[0135] This represents the value space (set of discrete real numbers) of the LLR signal output by the demodulator.

[0136] This represents the output space of the quantizer (a finite set of integers, such as [-A, A]).

[0137] This represents the probability distribution of the demodulator outputting an LLR signal of y when the transmitted bit is x;

[0138] This represents the probability that the LLR signal y is quantized into z (for a deterministic quantizer). (It is 1 at the corresponding z-axis and 0 elsewhere).

[0139] This represents the probability that the quantizer output is z given that the transmitted bit is x.

[0140] The above formula This is represented as the output of all possible LLRs, expressed in terms of... The weights represent the probabilities that y is quantized as z. By performing a weighted summation, we can obtain the conditional probability distribution of the quantizer output being z given an input bit value of x. .

[0141] During the simulation, the transmitted bit x output by the encoder and the corresponding LLR signal y output by the demodulator are acquired. It is crucial to ensure that the transmitted bit x and the LLR signal y are data aligned to guarantee the accuracy of the probability distribution.

[0142] During data processing, bits that have not been transmitted through the channel (such as shortened or punctured bits) can be removed to ensure the accuracy of the statistical results. It is understandable that for each set of fixed simulation conditions (such as a specific MCS and channel type), sufficient sample data needs to be collected to ensure that the probability distribution is statistically significant.

[0143] Step S602: Based on the conditional probability quality function, determine the quantization factor for the linear quantizer using the mutual information maximization algorithm.

[0144] In this step, the mutual information between the encoder output x and the LLR quantizer output z can be expressed as:

[0145]

[0146] in:

[0147] Indicates the bits sent from the sending end;

[0148] This represents the conditional probability that the quantizer output is z given that the transmitted bit is x;

[0149] This represents the marginal probability that the quantizer output is z.

[0150] Mutual Information Mutual information is used to measure the degree of information correlation between the transmitted bit x and the quantizer output z. The higher the mutual information value, the more information about the original transmitted bit x is retained in the quantized output z, and the more effectively the decoder can recover the original data, thereby improving decoding performance and overall system performance.

[0151] From a computational perspective, mutual information The calculation process is as follows: For all possible quantized output z and all possible input bits x, calculate the probability of outputting z under condition x (and multiply it by the probability of x), then take the logarithm of the ratio with the marginal probability of z, and finally sum all the results in a weighted sum.

[0152] The quantitative analysis of mutual information described above provides a theoretical basis for quantizer structure design and parameter optimization, ensuring that decoder performance and system reliability are maximized while reducing hardware complexity and power consumption.

[0153] As an example, the mutual information maximization algorithm can employ a linear programming method that maximizes mutual information using a nonlinear quantizer.

[0154] As an example, a nonlinear quantizer can be designed for discrete memoryless channels with binary inputs, maximizing the mutual information between the quantizer input and the quantizer output.

[0155] For detailed definitions and principles of the mutual information maximization algorithm and the nonlinear quantizer, please refer to the relevant description in the first embodiment, and therefore will not be repeated here.

[0156] In practical implementation, based on the above conditional probability distribution and assuming that the probability of sending bits 0 and 1 is equal, the saturation level T and maximum mutual information of the nonlinear quantizer can be obtained through the mutual information maximization algorithm of the nonlinear quantizer. This application uses the above parameters as the saturation level and upper bound of the mutual information performance of the linear quantizer, and uses them as a basis to guide the subsequent optimization design of the linear quantizer parameters.

[0157] As an example, see Figure 8 Determining the quantization factor for the linear quantizer may include the following steps S8001 to S8003:

[0158] Step S8001: Determine the parameters for the linear quantizer, including the saturation level and the upper bound of the maximum mutual information;

[0159] Step S8002: Determine the bit width for the linear quantizer based on the upper bound of the maximum mutual information; and

[0160] Step S8003: Determine the quantization factor of the linear quantizer based on the saturation level and the determined bit width.

[0161] As an example, determining the bit width for a linear quantizer may include: setting a mutual information performance threshold, calculating the mutual information value of the linear quantizer for each bit width by iterating through different bit widths in turn, and selecting the smallest bit width that can satisfy the mutual information performance threshold as the bit width of the linear quantizer.

[0162] As an example, the mutual information performance threshold is 90% to 100% of the upper bound of the maximum mutual information.

[0163] As an example, the traversal of different bit widths uses an exhaustive method, with the bit width ranging from 3 to 9.

[0164] In some embodiments, the formula for calculating the quantization factor is:

[0165]

[0166] in, As a quantification factor, For the determined bit width, This is the saturation level.

[0167] It should be understood that the specific methods and technical principles for determining the saturation level, the upper bound of the maximum mutual information, the mutual information performance threshold, the bit width, and the quantization factor can be found in the detailed description of the first embodiment mentioned above in this specification, and therefore will not be repeated here.

[0168] Through the above process, the parameters of the linear quantizer can be systematically determined based on theoretical analysis, avoiding reliance on extensive system simulations and effectively improving the accuracy and efficiency of parameter design. Simultaneously, by reasonably setting the mutual information performance threshold and bit width, decoding performance and hardware implementation complexity can be balanced, achieving an optimal balance between system performance and resource consumption.

[0169] Step S603: Store at least each set of communication system parameters and the corresponding quantization factor in a lookup table to generate the lookup table.

[0170] As an example, the contents of the lookup table can be selected from one or more of the following: communication system parameters, quantization factor, bit width, saturation level, and upper bound of maximum mutual information. Specifically, the field contents of the lookup table can be flexibly configured according to actual system requirements to adapt to the parameter management requirements of different application scenarios.

[0171] As an example, the lookup table can be saved as a data file or database, facilitating subsequent parameter retrieval and maintenance. It should be noted that this invention does not limit the dimensions of the lookup table; for different types of communication systems, the parameter dimensions and size of the lookup table can be flexibly selected according to actual needs and implementation methods. For example, for multi-modulation systems, modulation and coding schemes (MCS), signal-to-noise ratio (SNR), and channel characteristics (such as delay spread, multipath parameters, etc.) can be used as input indexes for the lookup table; in Wi-Fi systems, different MCS and channel types (such as AWGN, block fading B, frequency selective fading F, etc.) can be used to construct the lookup table; in ultra-wideband (UWB) systems, the lookup table can be designed by combining parameters such as channel type and receiver-estimated signal power.

[0172] In addition, in some practical applications, if channel state information (CSI) cannot be accurately obtained, other channel-related parameters (such as SNR) can be used instead of CSI as the input index of the lookup table to ensure the practicality of the lookup table and the system's adaptability.

[0173] As an example, lookup tables can be used during the online operation phase of the system to quickly obtain and apply the corresponding quantization factor based on the actual communication system parameters, thereby achieving efficient quantization processing of LLR signals and improving the overall system performance and response speed.

[0174] Figure 9 This invention illustrates a flowchart of a lookup table design method according to a second embodiment of the present application. This process is executed during the offline simulation stage and specifically includes the following steps S901 to S906:

[0175] Step 901: Determine the input dimensions of the lookup table. Input dimensions may include, but are not limited to, communication system parameters such as modulation and coding scheme (MCS), channel type, and signal-to-noise ratio (SNR).

[0176] Step 902: Set simulation parameters for different application scenarios and use them as input conditions for the lookup table.

[0177] Step 903: For each set of input parameters, run an offline simulation to collect the conditional probability distribution of the LLR signal output by the demodulator.

[0178] Step 904: Based on the collected conditional probability distribution, obtain the saturation threshold T of the nonlinear quantizer and the maximum mutual information I' it can provide through numerical calculation. This saturation threshold T serves as the basis for the subsequent parameter derivation of the linear quantizer.

[0179] Step 905: Based on the obtained maximum mutual information I', combined with the preset mutual information loss threshold, determine the bit width m required by the linear quantizer, so as to optimize hardware resource consumption while ensuring decoding performance.

[0180] Step 906: Based on the determined saturation threshold T and bit width m, calculate the output quantization factor k of the lookup table under this input parameter. This quantization factor k is used for subsequent linear quantization processing of the LLR signal.

[0181] Step 907: Store all the above input parameter combinations and their corresponding quantization factors k in a lookup table (LUT) to complete the offline design of the lookup table.

[0182] The lookup table can be flexibly expanded according to actual system requirements, including but not limited to parameters such as quantization factor k, bit width m, saturation threshold T, and maximum mutual information I', to support rapid parameter retrieval during online system operation and achieve efficient quantization processing of LLR signals. Compared with existing technologies that typically rely on system simulation frame error rate (FER) results to determine the quantizer lookup table, this embodiment effectively reduces the dependence on simulation through theoretical analysis and numerical calculation, avoiding the problem of parameters getting trapped in local optima, thereby significantly improving the accuracy and practicality of the quantization process and enhancing the robustness and scalability of the system.

[0183] In summary, through steps S601 to S603, the second embodiment of this application provides a method for generating a lookup table for linear quantization parameters of the decoder input LLR signal. By analyzing the conditional probability distribution of the demodulator output LLR signal under different communication system parameters through offline simulation, and combining mutual information theory, the optimal quantization parameters (such as quantization factor k, bit width m, saturation level T, etc.) are calculated for each set of communication system parameters and stored in the lookup table (LUT). During actual operation of the communication system, there is no need to recalculate the parameters in real time; instead, the most suitable quantization parameters for the current channel and modulation conditions can be quickly retrieved and applied through the lookup table, achieving efficient and accurate quantization of the LLR signal, thereby improving system operating efficiency and hardware scalability.

[0184] [Third Embodiment]

[0185] The third embodiment of this application provides a linear quantization method for the LLR signal output by a demodulator. This method specifically describes how, during the operation of an actual communication system, a pre-built lookup table (LUT) is invoked to achieve efficient, real-time linear quantization processing of the LLR signal.

[0186] Figure 10 A flowchart illustrating a linear quantization method for the output LLR signal of a demodulator according to a third embodiment of this application is shown, including the following steps S1001 to S1003.

[0187] Step S1001: Obtain the current communication system parameters.

[0188] As an example, communication system parameters include at least one or more of the following: modulation and coding scheme, channel type, and signal-to-noise ratio.

[0189] Step S1002: Select the corresponding quantization factor from the pre-built lookup table according to the communication system parameters.

[0190] As an example, the communication system can utilize the lookup table (LUT) generated and stored in the second embodiment of this application. The lookup table includes one or more combinations selected from communication system parameters, quantization factors, bit width, saturation level, and the upper bound of maximum mutual information. For different parameter combinations, the corresponding quantization factors have been pre-calculated and stored in the lookup table. When the communication system receives a new frame, it performs a lookup operation based on the current communication system parameters to quickly obtain the quantization factor that best matches the parameter combination, which is then applied to the subsequent linear quantization processing of the LLR signal.

[0191] It should be understood that the dimensions and content of the lookup table can be flexibly set according to the actual system requirements and are not restricted. For example, for multi-modulation systems, the lookup table can use different modulation and coding schemes (MCS), signal-to-noise ratios (SNR), and channel characteristics (such as delay spread, multipath parameters, etc.) as index fields. In Wi-Fi systems, lookup tables can be constructed based on different MCS and channel types (such as AWGN, block fading B, frequency selective fading F, etc.); in ultra-wideband (UWB) systems, the lookup table can be designed by combining the channel type and the signal power estimated by the receiver. For some practical applications, if channel state information (CSI) cannot be accurately obtained, other channel-related parameters (such as SNR) can also be used as input indexes for the lookup table to ensure the applicability of the lookup table and the system's adaptability.

[0192] By utilizing the quantization parameters stored in the lookup table, the corresponding quantization factor, bit width, and saturation threshold can be quickly retrieved and applied based on the current communication system parameters (such as modulation and coding scheme, channel type, signal-to-noise ratio, etc.), thereby simplifying the system design process, improving quantization efficiency and decoder performance, and meeting the actual system's requirements for low latency, high reliability, and hardware resource optimization.

[0193] Step S1003: Linearly quantize the LLR signal output by the demodulator using a quantization factor to serve as the input LLR signal for the decoder.

[0194] In this step, see Figure 11 The flowchart illustrates a linear quantization processing method according to an example of this application, including the following steps S1101 to S1103:

[0195] Step S1101: Multiply the LLR signal by the quantization factor to perform scaling;

[0196] Step S1102: Round the scaled value; and

[0197] Step S1103: Perform saturation limiting processing on the rounded result to generate an integer signal with a finite bit width.

[0198] As an example, Figure 12 This diagram illustrates the structure of an LLR quantizer 113 according to an embodiment of this application. The LLR quantizer 113 includes a parameter lookup table (LUT) 1201, a multiplier 1202, and a rounding and saturation processing unit 1203. The specific implementation process is as follows:

[0199] The modulation and coding scheme (MCS) and channel state information (CSI) are used as input indices for lookup table 1201, which outputs the corresponding quantization factor k based on the input indices. Subsequently, the quantization factor k is multiplied by the LLR signal y output by the demodulator, and the result is rounded and saturated to generate the final quantized output z.

[0200] This structure can perform adaptive quantization processing of LLR signals according to the actual communication system parameters, thereby improving decoder performance and optimizing hardware resource utilization.

[0201] As an example, saturation limiting processing of the rounded result includes:

[0202] The rounded result is limited to the interval [-A, A], where A is the maximum allowed integer value, determined by the bit width of the linear quantizer;

[0203] When the rounded result is greater than A, the generated finite-width integer signal is set to A.

[0204] When the rounded result is less than -A, the generated finite-width integer signal is set to -A; and

[0205] When the rounded result is between -A and A, the generated finite-width integer signal is the rounded result.

[0206] As an example, A is determined by the bit width of the linear quantizer, and

[0207]

[0208] in, is the maximum allowed integer value, and m is the bit width of the linear quantizer.

[0209] The above saturation limiting process ensures that the quantized output signal always remains within the predetermined integer range, avoiding overflow or distortion and effectively improving the system's stability and decoding performance.

[0210] As an example, Figure 13 A flowchart illustrating the online quantization method of the third embodiment of this application is shown. This method is executed during the practical application stage after the lookup table design is completed, and specifically includes the following steps S1301 to S1303:

[0211] Step 1301: During actual system operation, the configuration parameters and channel characteristics of the current received frame are obtained in real time. The parameters include, but are not limited to, modulation and coding scheme (MCS), channel type and signal-to-noise ratio (SNR).

[0212] Step 1302: Based on the acquired system parameters, retrieve and select the quantization factor k that matches the current system conditions from the pre-built lookup table (LUT). The lookup table may contain multiple items such as communication system parameters, quantization factor, bit width, saturation level, and upper bound of maximum mutual information, to achieve rapid retrieval and application of quantization parameters.

[0213] Step 1303: Perform linear quantization on the LLR signal y output by the demodulator. The specific operation is as follows: Z = rounding(y×k), where y is the original LLR signal, k is the quantization factor obtained from the lookup table, and Z is the quantized output.

[0214] Subsequently, Z is saturated and limited to generate the final finite-width integer output, according to the following rules:

[0215] If Z > A, then Z = A;

[0216] If Z < -A, then Z = -A;

[0217] If Z is within the interval [-A, A], then Z is the result after rounding.

[0218] Where A is the saturation value of the linear quantizer, determined by the quantizer's bit width m, typically... .

[0219] The above steps achieve the following technical effects: the original LLR signal y is usually a real number with a wide range; by selecting the quantization factor k through a lookup table, the original LLR signal y is scaled to a suitable quantization range, achieving a linear transformation of the signal amplitude; after the multiplication operation y × k, it is rounded to obtain the closest integer value Z; finally, through saturation limiting processing, it is ensured that the quantized output Z is always within the finite range [-A, A], serving as the input signal of the decoder and effectively preventing overflow or distortion.

[0220] As can be seen, the above operation "Z = rounding(y×k)" realizes the linear scaling and discretization of the continuous LLR signal y. Through a hardware-friendly operation process, it is transformed into an integer Z within a finite range, which significantly simplifies the hardware implementation and improves the real-time processing capability of the system.

[0221] For example: Suppose y=1.23, k=4, then y×k=4.92, and after rounding, Z=5. If A=7, then Z=5 is within the range and no amplitude limiting is needed.

[0222] In summary, through steps S1101 to S1103, the third embodiment of this application provides an efficient and low-complexity online quantization scheme. By using a lookup table, the device does not need to perform complex mutual information calculations and parameter optimizations; quantization can be completed with only one multiplication, rounding, and saturation operation. This is highly suitable for real-time communication systems with limited hardware resources, improving the system's flexibility, scalability, and practical application value.

[0223] It is worth noting that the technical solutions, technical details, and technical effects described in the first, second, and third embodiments of this application can be combined with and referenced in conjunction with each other. Identical or similar content will not be repeated in this document.

[0224] This application also provides electronic devices, see [link to relevant documentation]. Figure 14 The diagram shows a schematic of the structure of an electronic device according to an embodiment of the present application. The electronic device 14 includes a memory 1401, a processor 1402, and computer instructions 1403 stored in the memory and executable on the processor. When the computer instructions 1403 are executed by the processor 1402, they implement the methods described in the various embodiments of the present application.

[0225] In this embodiment, each module can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of the computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0226] This application also provides a computer program product that includes computer instructions, which, when executed by a processor, implement the methods described in the various embodiments of this application.

[0227] This application also provides a computer-readable storage medium storing computer program instructions thereon, which, when executed by a processor, can implement the methods described in the various embodiments of this application.

[0228] This application also provides a computer program instruction that, when executed by a processor, can implement the methods described in the various embodiments of this application.

[0229] For the implementation principles and beneficial effects of the electronic devices, computer program products, computer-readable storage media, and computer programs in the above embodiments, please refer to the foregoing descriptions of the linear quantization method for the input LLR signal of the decoder in the first embodiment of this application, the method for generating a parameter lookup table for linear quantization of the input LLR signal of the decoder in the second embodiment, and the linear quantization method for the output LLR signal of the demodulator in the second embodiment. Therefore, they will not be repeated here.

[0230] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic resistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0231] While various embodiments of various aspects of this application have been described for the purposes of this disclosure, they should not be construed as limiting the teachings of this disclosure to these embodiments. Features disclosed in one specific embodiment are not limited to that embodiment, but can be combined with features disclosed in different embodiments. For example, one or more features and / or operations of the method according to this application described in one embodiment can also be applied individually, in combination, or in whole in another embodiment. Those skilled in the art will understand that there are many more possible alternative implementations and variations, and various changes and modifications can be made to the above system without departing from the scope defined by the claims of this application.

Claims

1. A method for linear quantization of input LLR signals for a decoder, characterized in that, comprising the steps of: obtaining a conditional probability distribution of the output LLR signal of the demodulator based on current communication system parameters; determining a saturation level for a linear quantizer and a bit-width of the linear quantizer according to the conditional probability distribution using a maximum mutual information algorithm; determining a quantization factor of the linear quantizer based on the determined bit-width and the saturation level; and linearly quantizing the LLR signal using the quantization factor as an input LLR signal of a decoder.

2. The method of claim 1, wherein the communication system parameters comprise at least one or more of a modulation and coding scheme, a channel type, and a signal-to-noise ratio.

3. The method of claim 1, wherein the conditional probability distribution is obtained through offline system simulation.

4. The method of claim 1, wherein the maximum mutual information algorithm employs a linear programming method of mutual information maximization for a nonlinear quantizer.

5. The method of claim 4, wherein the nonlinear quantizer is a quantizer that maximizes mutual information between a quantizer input and a quantizer output for a binary-input discrete memoryless channel. determining a maximum mutual information upper bound for a linear quantizer according to the conditional probability distribution using a maximum mutual information algorithm; and 6. The method of claim 1, further comprising: determining a bit-width of the linear quantizer based on the maximum mutual information upper bound.

7. The method of claim 6, wherein determining the bit-width of the linear quantizer comprises: setting a mutual information performance threshold, calculating a mutual information value of the linear quantizer at each bit-width by traversing different bit-widths, and selecting a minimum bit-width that satisfies the mutual information performance threshold as the bit-width of the linear quantizer.

8. The method of claim 7, wherein the mutual information performance threshold is 90% to 100% of the maximum mutual information upper bound.

9. The method of claim 7, wherein the traversing of different bit-widths employs an exhaustive method, and the bit-widths take values between 3 and 9.

10. The method of claim 7, wherein the calculation of the mutual information value of the linear quantizer is based on an accumulation of partial mutual information of the output of the linear quantizer.

11. The method of claim 1, wherein a calculation formula of the quantization factor of the linear quantizer is:

12. The method of claim 1, wherein the linear quantization of the LLR signal using the quantization factor comprises: multiplying the LLR signal by the quantization factor for scaling; wherein quantization factor, the determined bit width, the saturation level. performing an integer operation on the scaled value; and performing a saturation clipping on the integer result to generate a finite bit-width integer signal. The computer program product comprises computer instructions that, when executed by a processor, implement the method of any one of claims 1 to 12. The computer readable storage medium stores computer instructions that, when executed by a processor, implement the method of any one of claims 1 to 12. ​ 13. A computer program product, characterised in that, ​ 14. A non-transitory computer readable storage medium, comprising: ​ 15. An electronic device, comprising: The electronic device includes a memory, a processor, and computer instructions stored on the memory and executable on the processor, the computer instructions, when executed by the processor, implementing the method according to any one of claims 1 to 12.

Images