阅读说明：本技术 削波信号的重构 (Reconstruction of clipped signals ) 是由 尼古劳斯·科洛瓦基斯 托马斯·埃里克森 迈克尔·科尔德雷 马茨·维贝里 于 2019-04-23 设计创作，主要内容包括：使用被包含在来自模数(ADC)转换的削波样本中的信息来提高接收机性能,例如通过减少ADC由于其数据分辨率限制而导致的削波失真。这提供了优于已有解决方案的优势,由于已有解决方案丢弃了削波样本中的信息,因此已有解决方案执行效果欠佳。(Information contained in clipped samples from analog-to-digital (ADC) conversion is used to improve receiver performance, for example by reducing clipping distortion of the ADC due to its data resolution limitations. This provides an advantage over existing solutions, which perform poorly because they discard information in the clipped samples.)

1. A method (400) implemented by a receiver (100), the method comprising:

receiving (s402) a signal y;

sampling (s404) y, thereby producing a sample set;

quantizing (s404) each sample of the set of samples to produce a quantized received signal r, wherein quantizing each sample of the set of samples comprises clipping at least M of the samples, wherein M > 0, such that r comprises M clipped samples;

obtaining (s406) information representative of the clipped and unclipped samples;

using (s412) the probability density function of y and information representative of the clipped and unclipped samples to obtain a probability density function g (x) of unknown samples in y, the unknown samples having been clipped conditional on a quantized received vector r; and

modifying r (s414) by replacing, for each clipping sample value in r, the clipping sample value with an expected value corresponding to the clipping sample value, wherein the expected value is based on g (x), thereby producing a reconstructed received signal.

2. The method of claim 1, further comprising: decoding the reconstructed received signal.

3. The method of claim 1 or 2, further comprising: training data is received and a probability density function for y is obtained using the received training data.

4. The method of any of claims 1 to 3, further comprising: a corresponding expected value is determined.

5. The method of claim 4, wherein determining the corresponding expected value comprises:

initializing (s501) a first vector mu⁽⁰⁾Wherein, mu⁽⁰⁾Is given by the value mu₁ ⁽⁰⁾，μ₂ ⁽⁰⁾，μ₃ ⁽⁰⁾，...，μ_M ⁽⁰⁾A vector of components;

defining (s502) a first termination condition δ and a second termination condition J;

based on mu⁽⁰⁾And G (x) to calculate (s506) a second vector μ⁽¹⁾(ii) a And

determining (s508) whether | | | μ⁽¹⁾-μ⁽⁰⁾| is less than δ or the index value J is equal to J.

6. The method of claim 5, wherein μ is based⁽⁰⁾And G (x) to calculate μ⁽¹⁾The method comprises the following steps:

calculating mu₁ ⁽¹⁾＝E(x₁|μ₂ ⁽⁰⁾，μ₃ ⁽⁰⁾，...，μ_M ⁽⁰⁾) Wherein, E (x)₁|μ₂ ⁽⁰⁾，μ₃ ⁽⁰⁾，...，μ_M ⁽⁰⁾) Is at a given mu₂ ⁽⁰⁾，μ₃ ⁽⁰⁾，...，μ_M ⁽⁰⁾And an expected value of a first clipped sample in a case; and calculating mu₂ ⁽¹⁾＝E(x₂|μ₁ ⁽¹⁾，μ₃ ⁽⁰⁾，...，μ_M ⁽⁰⁾) Wherein, E (x)₂|μ₁ ⁽¹⁾，μ₃ ⁽⁰⁾，...，μ_M ⁽⁰⁾) Is at a given mu₁ ⁽¹⁾，μ₃ ⁽⁰⁾，...，μ_M ⁽⁰⁾Andthe expected value of the second clipped sample in the case.

7. The method of claim 6, wherein modifying r by replacing each clipped sample in r with its corresponding expected value comprises:

as determination of | | mu⁽¹⁾-μ⁽⁰⁾As a result of | | being less than δ or J ═ J, the first clipped sample value in r is replaced with μ₁ ⁽¹⁾And replacing the second clipped sample value in r with μ₂ ⁽¹⁾。

8. A receiver apparatus (100), the receiver apparatus (100) being configured to:

sampling the received signal y, thereby generating a sample set;

quantizing each sample in the set of samples to produce a quantized received signal r, wherein quantizing each sample in the set of samples comprises clipping at least M of the samples, wherein M > 0, such that r comprises M clipped samples;

obtaining information representative of the clipped and unclipped samples;

using the probability density function of y and information representative of the clipped and unclipped samples to obtain a probability density function g (x) of unknown samples in y, the unknown samples having been clipped conditioned on a quantized received vector r; and

modifying r by replacing, for each clipping sample value in r, the clipping sample value with an expected value corresponding to the clipping sample value, wherein the expected value is based on g (x), thereby producing a reconstructed received signal.

9. The receiver apparatus of claim 8, further comprising: decoding the reconstructed received signal.

10. The receiver apparatus of claim 8 or 9, further comprising: training data is received and a probability density function for y is obtained using the received training data.

11. The receiver apparatus of any of claims 8 to 10, further comprising: a corresponding expected value is determined.

12. The receiver device of claim 11, wherein determining the corresponding expected value comprises:

initializing a first vector mu⁽⁰⁾Wherein, mu⁽⁰⁾Is given by the value mu₁ ⁽⁰⁾，μ₂ ⁽⁰⁾，μ₃ ⁽⁰⁾，...，μ_M ⁽⁰⁾A vector of components;

defining a first termination condition delta and a second termination condition J;

based on mu⁽⁰⁾And G (x) to calculate a second vector mu⁽¹⁾(ii) a And

determine if | | mu⁽¹⁾-μ⁽⁰⁾| is less than δ or the index value J is equal to J.

13. The receiver apparatus of claim 12, wherein μ -based⁽⁰⁾And G (x) to calculate μ⁽¹⁾The method comprises the following steps:

calculating mu₁ ⁽¹⁾＝E(x₁|μ₂ ⁽⁰⁾，μ₃ ⁽⁰⁾，...，μ_M ⁽⁰⁾) Wherein, E (x)₁|μ₂ ⁽⁰⁾，μ₃ ⁽⁰⁾，...，μ_M ⁽⁰⁾) Is at a given mu₂ ⁽⁰⁾，μ₃ ⁽⁰⁾，...，μ_M ⁽⁰⁾And first in the case ofAn expected value of clipped samples; and calculating mu₂ ⁽¹⁾＝E(x₂|μ₁ ⁽¹⁾，μ₃ ⁽⁰⁾，...，μ_M ⁽⁰⁾) Wherein, E (x)₂|μ₁ ⁽¹⁾，μ₃ ⁽⁰⁾，...，μ_M ⁽⁰⁾) Is at a given mu₁ ⁽¹⁾，μ₃ ⁽⁰⁾，...，μ_M ⁽⁰⁾Andthe expected value of the second clipped sample in the case.

14. The receiver device of claim 13, wherein modifying r by replacing each clipped sample in r with its corresponding expected value comprises:

as determination of | | mu⁽¹⁾-μ⁽⁰⁾As a result of | | being less than δ or J ═ J, the first clipped sample value in r is replaced with μ₁ ⁽¹⁾And replacing the second clipped sample value in r with μ₂ ⁽¹⁾。

15. A computer program comprising instructions which, when executed by processing circuitry (602) of a receiver apparatus (600), cause the receiver apparatus to:

sampling the received signal y, thereby generating a sample set;

quantizing each sample in the set of samples to produce a quantized received signal r, wherein quantizing each sample in the set of samples comprises clipping at least M of the samples, wherein M > 0, such that r comprises M clipped samples;

obtaining information representative of the clipped and unclipped samples;

using the probability density function of y and information representative of the clipped and unclipped samples to obtain a probability density function g (x) of unknown samples in y, the unknown samples having been clipped conditioned on a quantized received vector r; and

modifying r by replacing, for each clipping sample value in r, the clipping sample value with an expected value corresponding to the clipping sample value, wherein the expected value is based on g (x), thereby producing a reconstructed received signal.

16. A carrier containing the computer program of claim 15, wherein the carrier is one of an electronic signal, an optical signal, a radio signal, and a computer readable storage medium.

Technical Field

Embodiments related to reconstruction of clipping signals, including reconstruction of clipping signals in large-scale multiple-input multiple-output (MIMO) systems, are disclosed.

Background

Clipping is a form of distortion that limits signal values above or below a certain threshold. In practice, clipping may be necessary due to system limitations (e.g., to avoid over-modulating the audio transmitter). In a discrete system, it may be caused unintentionally due to data resolution limitations (e.g., when the sample exceeds a maximum that can be represented), or intentionally when the process of signal value limitation is simulated.

Clipping is a non-linear operation and introduces frequency components that are not present in the original signal. In the digital domain, these new components are reflected into the baseband when their frequency exceeds the nyquist limit, resulting in aliasing.

Massive MIMO systems are now a mature technology that forms the backbone of fifth generation (5G)3GPP mobile networks. In the case of using massive MIMO, the number of antennas of a Base Station (BS) is increased by several orders of magnitude compared to a conventional multi-antenna system, with the goal of achieving significant gains, e.g., higher capacity and energy efficiency.

In a conventional multi-antenna BS, each Radio Frequency (RF) port is connected to a pair of high resolution analog-to-digital converters (ADCs) (typically, in-phase and quadrature signal components are quantized with a resolution of over 10 bits). Extending this architecture to massive MIMO with hundreds or thousands of active antenna elements would result in prohibitively high power consumption and hardware costs. The hardware complexity and power consumption of the ADC increases approximately exponentially in the number of quantization bits. Therefore, an effective solution to keep power consumption and system cost within desired limits is to reduce the accuracy of the ADC (e.g., up to 8 bits). An additional motivation for reducing the resolution of the employed ADCs is to limit the amount of data that must be transmitted over the link connecting the RF components (also known as Radio Units (RUs)) and the baseband processing unit (BBU), which may be located remotely from the RUs.

The ADC can be modeled as two processes: sampling and quantizing. Sampling converts a continuous time-varying voltage signal into a discrete-time signal-i.e., a sequence of real numbers. Quantization replaces each real number with an approximation from a limited set of ranges of discrete values and performs clipping when the input is out of the supported range to limit the output to that range. The error introduced by such clipping is called overload distortion. Within the limits of the support range, the amount of separation between selectable output values of the quantizer is referred to as its granularity, and the error introduced by this separation is referred to as granularity distortion. The quantizer design typically determines a suitable balance between granular distortion and overload distortion. For a given supported number of possible output values, reducing the average granular distortion may involve increasing the average overload distortion, and vice versa.

Disclosure of Invention

Certain challenges exist. For example, overload distortion can severely affect the quality of a digital signal by corrupting the data represented by the digital signal. In fact, even a very low percentage of clipped samples can result in significant overload distortion. In multi-user MIMO (MU-MIMO) systems, overload distortion can lead to inaccurate Channel State Information (CSI), and it degrades data estimation at the BS and/or at User Equipment (UE) providing network access by the BS.

The present disclosure proposes to utilize information contained in clipped samples from the ADC conversion to improve receiver performance, for example by reducing clipping distortion of the ADC due to its data resolution limitations. This provides an advantage over existing solutions, which perform poorly because they discard information in the clipped samples.

Thus, in one aspect, a method is provided for reconstructing clipped samples by exploiting the correlation between clipped and unclipped samples and thereby reducing overload distortion. In one embodiment, the method comprises: a signal y is received and sampled, resulting in a sample set. The method further comprises the following steps: quantizing each sample in the set of samples to produce a quantized received signal r, wherein quantizing each sample in the set of samples comprises clipping at least M of the samples, wherein M > 0, such that r comprises M clipped samples. The method further comprises the following steps: obtaining information representing clipped samples and unclipped samples, and obtaining a probability density function g (x) of unknown samples in y, which have been clipped conditioned on a quantized received vector r, using the probability density function of y and the information representing clipped samples and unclipped samples. The method further comprises the following steps: modifying r by replacing, for each clipping sample value in r, the clipping sample value with an expected value corresponding to the clipping sample value, wherein the expected value is based on g (x), thereby producing a reconstructed received signal.

In another aspect, there is provided a receiver apparatus configured to: sampling the received signal y, thereby generating a sample set; quantizing each sample in the set of samples to produce a quantized received signal r, wherein quantizing each sample in the set of samples comprises clipping at least M of the samples, wherein M > 0, such that r comprises M clipped samples; obtaining information representative of clipped samples and unclipped samples; using the probability density function of y and information representing clipped and unclipped samples to obtain a probability density function g (x) of unknown samples in y, the unknown samples having been clipped conditioned on a quantized received vector r; and modifying r by replacing, for each clipping sample value in r, the clipping sample value with an expected value corresponding to the clipping sample value, wherein the expected value is based on g (x), thereby producing a reconstructed received signal.

Drawings

The accompanying drawings, which are incorporated herein and form a part of the specification, illustrate various embodiments.

Fig. 1 shows a Clipping Aware (CA) receiver according to an embodiment.

Fig. 2A is a graph illustrating the improvement provided by a CA-MMSE receiver, according to an embodiment.

Fig. 2B is a graph illustrating the improvement provided by a CA-MMSE receiver, according to an embodiment.

Fig. 3A is a graph illustrating the improvement provided by a CA-MMSE receiver, according to an embodiment.

Fig. 3B is a graph illustrating the improvement provided by a CA-MMSE receiver, in accordance with an embodiment.

Fig. 4 is a flow chart illustrating a process according to an embodiment.

Fig. 5 is a flow chart illustrating a process according to an embodiment.

Fig. 6 is a block diagram of an apparatus according to an embodiment.

Fig. 7 is a schematic block diagram of an apparatus according to an embodiment.

Detailed Description

Fig. 1 shows the components of a Minimum Mean Square Error (MMSE) based clipping-aware receiver 100(CA-MMSE receiver 100). The CA-MMSE receiver 100 comprises a quantizer 102 that receives an input signal y and produces a quantized output r. The quantizer 102 (e.g., ADC) also outputs information (C) indicating which samples of y have been clipped. For example, set C represents the index of the clipping element of the received vector r. The set G represents the indices of the unclipped elements of the received vector r.

The CA-MMSE receiver 100 also includes a reconstructor 104 that reconstructs clipped samples conditioned on the quantized received vector rFor example, at a given observed received vectorAnd in this case, the reconstructor 104 reconstructs the clipping sample values by replacing each clipping sample value in r with an expected value corresponding theretoTo reconstruct the clipped samples This method is optimal because it minimizes the unknown received signalAnd estimation thereofMean square error between. Mathematically speaking:

CA-MMSE receiver 100 also includes a decoder 106 that decodes the information output by reconstructor 104 (i.e.,wherein the content of the first and second substances,) And decoding is carried out.

1.1 System model

Consider a single-cell MU-MIMO system that includes a BS equipped with N antennas communicating with K single-antenna UEs, and assumes that the BS and UEs are fully synchronized and run a Time Division Duplex (TDD) protocol with universal frequency reuse. The Nx1 received vector at the BS is:

where H is an NxN small-scale channel coefficient matrix between the K UEs and the BS. Further, x is a Kx1 vector of independent unit power symbols simultaneously transmitted by K UEs, and the average transmission power per UE is ρ. Finally, n is Additive White Gaussian Noise (AWGN).

1.2 quantization of complex valued vectors

The in-phase and quadrature components of the received signal at each antenna are individually quantized by a b-bit resolution ADC. More precisely, we model the ADCs as symmetrical uniform quantizers with step size Δ, and each ADC is characterized by a set of L-2 b quantization levelsWherein

Furthermore, we define a set of L +1 quantization thresholds So that ∞ τ₀＜τ₁＜…＜τ_L-1＜τ_LIs ∞ and

a practical technique for controlling the signal amplitude (or equivalently, the quantization step size Δ) to achieve a proper balance between granular distortion and overload distortion is to use Automatic Gain Control (AGC). Next, we define a non-linear quantizer mapping function that describes the joint operation of the 2N b-bit ADCs at the BS. For convenience, we first define the Cartesian productAnd let y_nAnd r_nThe nth elements of the N x1 vectors y and r, respectively. The quantizer mapping function may then be formed from the functionDescribed, the function maps the received continuous value signal y to the quantized output r as follows: if it is not And isThen r is_n＝l_k+jl_m. Thus, the quantized received signal r can be written as

Where a is the NxN diagonal matrix of Automatic Gain Control (AGC) that scales the received signal y. In addition to this, the present invention is,andrepresenting the real and imaginary parts of the signal, respectively.

1.3 augmented (augmented) real-valued representation

Since the ADC quantizes the real and imaginary parts of the signal separately (or quantizes the in-phase and quadrature components separately), channel estimation and data detection should allow the real and imaginary parts of the received signal to be processed separately. Thus, the complex-valued problem (1) can be conveniently converted to an equivalent augmented real-valued representation using the following definitions:

the quantized signal can be written as

For ease of markup, we use y, r, and x to denote y, respectively, for the rest of the document_R、r_RAnd x_R。

We only focus on the reconstruction of the clipped samples of the quantized received signal r, while the samples that are not clipped (i.e. within the granularity region of the quantizer) remain unchanged after quantization. For ease of notation, we first define the following two sets of indices that quantize the received vector r, i.e.,and

i.e. collectionsIndex representing a clipping element of a received vector r, andthe index in (1) represents the element of r that belongs to the granularity region of the quantizer. Then, we define a vectorAndit is to be noted that it is preferable that,andrepresenting the quantized received vector of the observed clipped samples and the vector located within the granularity region of the quantizer, respectively. Vector, on the other handAndrepresenting unknown continuous signals, which are quantized to produce vectorsAnd

next, we introduce the proposed clipping recovery receiver, which is conditioned on quantizing the received vector r to reconstruct only the clipped samplesGiven the observed received vectorAndthe proposed receiver replaces each clipped sample value in r with its corresponding oneTo reconstruct clipped samplesThis method is optimal because it minimizes the unknown received signalAnd estimation thereofMean square error between. Mathematically, we have the following receivers:

therefore, in determining the expectation in (2), we first need the a posteriori probability density function(in FIG. 5, it is also denoted G (x), where x is indicated byAs condition unknown continuous sample y_C) The posterior probability density function produces all feasible valuesThe probability of the quantized vector r is observed. Next, in view of the fact that the received continuous signal follows a gaussian distribution, we give a more detailed description of the proposed receiver.

Suppose the mth of the vector rThe element has been clipped. I.e. r_m＝l_jWhere j is ∈ {0, L-1 }. In addition, letRepresenting a vectorM unknown consecutive samples y_mEstimate (expectation). Therefore, according to the desired definition in (2), we have:

wherein, a posterioriIs the interval [ alpha ', beta']Inner unilateral truncated normal distribution. Now, let α'_mAnd beta'_mThe mth elements of the vectors α 'and β', respectively. Note that the interval [ α'_m，β′_m]Representing the unknown continuous value y given its clipped observation sample_mAnd thus its expected value is calculated within this interval. By vectorConditional, unknown vectorMay be (a) left truncated, i.e., α'_m＝τ₀- ∞ and beta'_m＝τ₁Or (b) right truncation, i.e.,. alpha'_m＝τ_L-1And beta'_m＝τ_L＝∞。

It is worth mentioning that: the elements of the received continuous vectors are correlated and thus by observationAlso give us information about unknown clipped samplesThe information of (1). This correlation is captured by the covariance matrix of the posterior p.d.f. in (3).

The integral in expression (3) cannot be evaluated in closed form, so we can implement the estimator by means of numerical integration. However, for a large number of clipped samples, this becomes impractical. Therefore, we next provide a low complexity iterative algorithm to approximate the mean value E { y } in (3)_C|r_C，r_G}. The algorithm relies on a single-sided truncated normal distributionA closed form one-dimensional condition expectation, where M represents the cardinality of the clipped samples.

Specifically, the jth iteration of this process returns a vectorAnd it is as follows:

the process is iteratively repeated until a numberBelow a constant δ or the maximum number of iterations J has been exceeded. Next, we generate a closed form expression for the above expected values.

For convenience of notation, we define vectorsWherein the vectorHas been removed. Each expectation of the iterative algorithm is then given by the following closed form expression, which is the average of the Univariate truncated normal distribution (see n.l. johnson, s.kotz, n.balakrishnan, "Continuous Univariate Distributions", 2 nd edition, volume 1, 1994, willi):

wherein the content of the first and second substances,and isIn addition to this, the present invention is,is a probability density function of a standard normal distribution,is its cumulative distribution function and erfc (-) is the complementary error function. Parameter(s)Andis given by the following formula

Wherein the (N-1) x (N-1) matrix is sigma_-i-iIs formed by removing the ith row and ith column from Σ, and a (N-1) × 1 vector σ_-iIs the ith column of Σ after removal of the ith element, and (N-1) × 1 average vector μ_-iIs obtained after removing the i-th element from μ. In addition, the variance σ_iiCorresponding to the ith diagonal element of the covariance matrix sigma. Finally, the parameters μ and Σ are given by the following formulas, respectively

Wherein the content of the first and second substances,and

with respect to computational complexity, iterative algorithms require evaluation of simple closed form formulas. The main computational burden is due to the computation of the M inverse matrices that appear in (5) and (6).

The proposed algorithm is a deterministic approximation of the Gibbs (Gibbs) sampler, where the randomly generated samples of the Gibbs sampler are replaced by the mean of the corresponding conditional distribution. It is worth mentioning that deterministic approximations of gibbs samplers have also been used for semi-supervised Hyperspectral Unmixing problems (see, e.g., k.e. themelis and a.a.rontogranians and k.d.koutrocus, "a Novel hierarchy Bayesian Approach for Sparse semi-conditioned Hyperspectral ultrasonication", IEEE journal signal processing, volume 60, phase 2, pages 585-599, month 2 2012).

If the matrix ∑ -¹Is greater than all the terms of its corresponding row, then the mapping in expression (3) isIn relation toA union of norms (and, therefore, it converges to a unique fixed solution).

We now use the proposed receiver to estimate the Mean Square Error (MSE) on a massive MU-MIMO uplink system, where each RF port at the BS is equipped with a resolution limited ADC. The AGC is selected to minimize the Mean Square Error (MSE) between the non-quantized received vector y and the quantized vector r.

On this basis, by assuming 3-bit, 4-bit and 6-bit resolution ADCs, on average 10%, 2% and 0.25% of the samples in the received vector r are clipped, respectively.

Finally, we assume that the entries of the channel matrix H are independent anddistributed and defining the average signal-to-noise ratio as

First, in fig. 2A and 2B, we visualize the improvement of clipping-aware MMSE (CA-MMSE) receiver on the average granularity per sample and the overload distortion. Recall that CA-MMSE performs reconstruction only on clipped samples, while the granular samples remain unchanged and thus there is no improvement on granular distortion. However, CA-MMSE brings a significant improvement in overload distortion. For example, in the case of b-4 bits (fig. 2A) and high signal-to-noise ratio (SNR) (i.e., 20dB), it reduces the overload distortion by 83% compared to the quantized non-perceptual case (QU) where quantized samples are not reconstructed. Indeed, when the resolution of the ADC is increased to B-6 bits (fig. 2B), then the gain of CA-MMSE for overload distortion is increased to 95%. The reason is that in ADCs with higher resolution, the granularity distortion is lower and therefore the CA-MMSE can reconstruct clipped samples with higher accuracy. The latter result is particularly important because it means that even a small proportion of clipped samples can lead to significant overload distortion (recall that when b is 6 bits, the percentage of clipped samples is very low, i.e. 0.25%). Further, note that the overload distortion of the quantized perceptual MMSE (QA-MMSE) (reconstructed for both clipped samples and samples located within the granularity region) is similar to that of the CA-MMSE, indicating the efficiency of the latter for clipped sample reconstruction.

Finally, note that in the case of b-4 bits and high SNR, the QA-MMSE produces a 37% lower granularity distortion than the CA-MMSE, while when the resolution of the ADC is increased to b-6 bits, its granularity gain is negligible as expected. Although QA-MMSE is superior to CA-MMSE in terms of granular distortion, it does not yield a substantial improvement in data estimation. The reason is that overload distortion dominates over per-sample granularity distortion, and therefore it is more important to compensate for overload distortion.

We now turn our attention to the convergence of the iterative algorithm and confirm that the use of the CA-MMSE receiver 100 instead of the more computationally expensive QA-MMSE (i.e.,is sufficient. In FIG. 3A and FIG. 3AIn B, estimating symbolsMSE of (a) is shown as a function of the average SNR and ADC resolution for different numbers of BS antennas. More precisely, we compare the CA-MMSE implemented by the iterative algorithm with the gibbs sampling (GS-AC-MMSE) and QA-MMSE based clipping-aware receiver. We note that if b ═ 4 bits (fig. 3A), the CA-MMSE converges to GS-AC-MMSE after only 15 iterations. Note, however, that the CA-MMSE converges to the GS-CA-MMSE faster when the number of BS antennas is relatively small. The reason is that in the case of N-16, the clipping samples to be estimated are much fewer than in the case of N-64. Thus, the fewer the number of BS antennas, the faster the iterative CA-MMSE receiver. Also, convergence is faster when the number of bits in the ADC is larger. Thus, convergence already occurs at 5 iterations in the case of B-6 bits (fig. 3B).

Furthermore, it is worth mentioning that after 15 iterations the CA-MMSE is almost identical to the optimal QA-MMSE receiver, confirming that it is sufficient to employ the CA-MMSE and reconstruct only the clipped samples, while the unclipped samples in the granularity region of the quantizer can remain unchanged. This result is particularly important because it means that we can achieve near optimal performance by using a receiver of lower complexity.

Fig. 4 is a flow diagram illustrating a process 400 according to an embodiment. Process 400 may begin at step s 402.

Step s402 includes: receiver 100 receives signal y.

Step s404 includes: receiver 100 samples y to produce a set of samples and then quantizes each sample in the set of samples to produce a quantized received signal r, wherein quantizing each sample in the set of samples comprises clipping at least M of the samples, wherein M > 0, such that r comprises M clipped samples.

Step s406 includes: the receiver 100 obtains information representing clipped samples and unclipped samples. For example, a list of clipped samples and/or unclipped samples is obtained.

Step s412 includes: receiver 100 uses the probability density function of y and information representing clipped and unclipped samples to obtain a probability density function g (x) of unknown samples in y that have been clipped from received quantization vector r, where x ═ x (x ═ x)₁，...，x_M) Is shown being clipped to the vector r_cThe previous unknown value.

Step s414 includes: the receiver 100 modifies r by, for each of the clipping sample values in r, replacing the clipping sample value with an expected value corresponding to the clipping sample value, wherein the expected value is based on g (x), thereby producing a reconstructed received signal.

Process 400 may also include steps s408, s410, and s 416. Steps s408 and s410 respectively include: training data is received, and a probability density function for y is obtained using the received training data. Step s416 comprises: to pairAnd decoding is carried out.

Fig. 5 is a flow diagram illustrating a process 500 for implementing step s414 according to an embodiment. Process 500 may begin at step s 501.

Step s501 includes: initializing the expected value of the sample that has been clipped in y, i.e., μ⁽⁰⁾＝(μ₁ ⁽⁰⁾，μ₂ ⁽⁰⁾，...μ_M ⁽⁰⁾). Where the initial values are equal to their corresponding clipping values.

That is, in step s501, the first vector μ is initialized⁽⁰⁾Wherein the first vector is formed by the value mu₁ ⁽⁰⁾，μ₂ ⁽⁰⁾，μ₃ ⁽⁰⁾，...，μ_M ⁽⁰⁾And (4) forming.

Step s502 includes defining termination conditions δ and J.

Step s504 includes setting j to 1.

Step s506 comprises determining μ^(j)Which is a vector of expected values. Determining mu^(j)The method comprises the following steps: for i-1 to M, each element μ of the vector is calculated_i ^(j)Wherein M is equal to clipTotal number of samples. I.e. determining μ^(j)The method comprises the following steps:

μ₁ ^(j)＝E(x₁|μ₂ ^(j-1)，μ₃ ^(j-1)，...，μ_M ^(j-1))

μ₂ ^(j)＝E(x₂|μ₁ ^(j)μ₃ ^(j-1)，...，μ_M ^(j-1))

μ_M ^(j)＝E(x_M|μ₁ ^(j)，μ₂ ^(j)，…，μ_M-1 ^(j)).

E(x₁|μ₂ ^(j-1)，μ₃ ^(j-1)，...，μ_M ^(j-1)) Is based on Expected values of the first clipped sample of (a); e (x)₂|μ₁ ^(j)，μ₃ ^(j-1)，...，μ_M ^(j-1)) Is based onThe expected value of the second clipped sample of (a); …, respectively; and E (x)_M|μ₁ ^(j)，μ₃ ^(j)，…，μ_M ^(j)) Is based on Is calculated for the mth clip sample.

Determining mu^(j)Thereafter, it is determined in step s508 whether:

||μ^(j)-μ^(j-1)| l < δ or J ═ J.

If | | | mu^(j)-μ^(j-1)If J is true, then the process proceeds to step s510, otherwise the process proceeds to step s509, where J is incremented by 1. After step s509, the process returns to step s 506.

Step s510 comprises setting upFor example, as the determination | | | μ⁽¹⁾-μ⁽⁰⁾Replacing the first clipping sample value in r with mu as a result of | | being less than delta₁ ⁽¹⁾Replacing the second clipped sample value in r with μ₂ ⁽¹⁾,.., and replacing the mth clipping sample value in r with μ_M ⁽¹⁾。

Fig. 6 is a block diagram of a receiver device 600 according to some embodiments. Receiver apparatus 600 may be used to implement receiver 100. As shown in fig. 6, receiver apparatus 600 may include: a Processing Circuit (PC)602, which may include one or more processors (P)655 (e.g., a general purpose microprocessor and/or one or more other processors, such as an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), etc.), which may be co-located in a single housing or single data center, or may be geographically distributed (i.e., receiver device 600 may be a distributed computing device); a network interface 648, including a transmitter (Tx)645 and a receiver (Rx)647, for enabling apparatus 600 to transmit and receive data to and from other nodes connected to network 110 (e.g., an Internet Protocol (IP) network), network interface 648 connected to network 110; and a local storage unit (also referred to as a "data storage system") 608, which may include one or more non-volatile storage devices and/or one or more volatile storage devices. In embodiments where the PC 602 includes a programmable processor, a Computer Program Product (CPP)641 may be provided. CPP 641 includes a Computer Readable Medium (CRM)642, which CRM 642 stores a Computer Program (CP)643 including Computer Readable Instructions (CRI) 644. CRM 642 may be a non-transitory computer readable medium, such as a magnetic medium (e.g., hard disk), an optical medium, a storage device (e.g., random access memory, flash memory), and so forth. In some embodiments, the CRI 644 of the computer program 643 is configured such that, when executed by the PC 602, the CRI causes the apparatus 600 to perform the steps described herein (e.g., the steps described herein with reference to the flow diagrams). In other embodiments, the apparatus 600 may be configured to perform the steps described herein without the need for code. That is, for example, the PC 602 may be composed of only one or more ASICs. Thus, the features of the embodiments described herein may be implemented in hardware and/or software.

Fig. 7 is a schematic block diagram of a receiver apparatus 600 according to some other embodiments. Receiver apparatus 600 includes one or more modules 700, each implemented in software. The module 700 provides the functionality of the apparatus 600 described herein (e.g., the steps described above, e.g., with respect to fig. 4 and/or fig. 5).

While various embodiments of the present disclosure have been described herein, it should be understood that they have been presented by way of example only, and not limitation. Thus, the breadth and scope of the present disclosure should not be limited by any of the above-described exemplary embodiments. Generally, all terms used herein are to be interpreted according to their ordinary meaning in the relevant art, unless explicitly given and/or otherwise implied by the context in which they are used. All references to "a/an/the element, device, component, means, step, etc" are to be interpreted openly as referring to at least one instance of the element, device, component, means, step, etc., unless explicitly stated otherwise. Any combination of the above-described elements in all possible variations thereof is encompassed by the disclosure unless otherwise indicated herein or otherwise clearly contradicted by context.

Additionally, although the processes described above and shown in the figures are shown as a series of steps, they are for illustration purposes only. Accordingly, it is contemplated that some steps may be added, some steps may be omitted, the order of steps may be rearranged, and some steps may be performed in parallel. That is, the steps of any method disclosed herein do not have to be performed in the exact order disclosed, unless one step must be explicitly described as being after or before another step and/or implicitly one step must be after or before another step.