WO2025041308A1 - Wireless communication system - Google Patents

Abstract

The objective of the present invention is to provide a wireless communication system that can implement compensation for wireless signals without distinguishing faults in the wireless signals by type. A wireless communication system according to the present disclosure comprises a transmitting station that transmits wireless signals and a receiving station that receives the wireless signals. The receiving station has a compensation circuit that includes a learning model. The receiving station is configured to execute: a process of using the learning model to process input signals and generate output signals, the input signals including received signals at the receiving station that originate from known bit sequences transmitted from the transmitting station; a process of training the learning model such that the output signals approach supervisory data comprising the known bit sequences; and a process of performing processing on a newly inputted input signal on the basis of the trained learning model to output an improved output signal.

Description

Wireless communication system

This disclosure relates to wireless communication systems.

In radio stations that transmit and receive radio signals, device failures such as IQ imbalance, carrier frequency offset, phase noise, and nonlinear distortion of amplifiers occur, which are known to degrade the quality of wireless communications. In addition, radio signals can experience failures such as channel fading during propagation, which further degrades the quality of communications.

Technologies have been proposed to estimate and compensate for such impairments. For example, Non-Patent Document 1 discloses an estimation and compensation technique for IQ imbalance.

S. Fouladifard, H. Shafiee, “Frequency offset estimation in OFDM systems in presence of IQ imbalance,” ICCS, pp. 214-218, 2002.

However, in conventional methods, the type of interference contained in the wireless signal was estimated separately and compensation was then performed. In some cases, compensation was performed using a separate compensation circuit for each type of interference.

In order to solve the above-mentioned problems, the present disclosure aims to provide a wireless communication system that can compensate for wireless signals without separating the types of impairments contained in the wireless signals.

An aspect of the present disclosure is
A transmitting station for transmitting a radio signal;
a receiving station for receiving the radio signal;
Equipped with
The receiving station,
A compensation circuit including a learning model is provided.
a process of processing an input signal including a received signal at the receiving station caused by a known bit sequence transmitted from the transmitting station using the learning model to generate an output signal;
A process of training the learning model so that the output signal approaches training data consisting of the known bit string;
A process of processing a newly input signal based on the learned learning model and outputting an improved output signal;
Preferably, the wireless communication system is configured to:

According to an aspect of the present disclosure, it is possible to provide a wireless communication system that can compensate for a wireless signal without separating the types of impairments contained in the wireless signal.

FIG. 1 is a diagram illustrating a failure that occurs in wireless communication. FIG. 1 is a diagram illustrating a method for estimating and compensating for a fault according to the prior art of the present disclosure. 1 is a configuration example of a wireless communication system according to a first embodiment of the present disclosure. 4 illustrates an example of the configuration of an input signal according to the first embodiment of the present disclosure. 4 is a diagram illustrating a data size of an input signal according to the first embodiment of the present disclosure. FIG. FIG. 10 is a diagram illustrating the data size of an output signal in TD and FD according to the first embodiment of the present disclosure. 13 is a configuration example of a receiving digital circuit according to a second embodiment of the present disclosure. 13 illustrates an example of the configuration of an input signal according to a second embodiment of the present disclosure. 13 is a configuration example of a receiving digital circuit according to a third embodiment of the present disclosure. 13 illustrates an example of the configuration of an input signal according to a third embodiment of the present disclosure. 13 illustrates a configuration example of a receiving digital circuit in a case where a compensation circuit receives LLR data according to a first modified example of the present disclosure. 13 illustrates a configuration example of a receiving digital circuit in a case where a compensation circuit receives bit string data according to a second modified example of the present disclosure.

Comparative Example
Here, a conventional technique will be described as a comparative example. Fig. 1 is a diagram for explaining a failure that occurs in wireless communication. A conventional wireless communication system 200 includes a transmitting station 210 and a receiving station 220.

The transmitting station 210 includes a transmitting digital circuit 211 and a transmitting analog circuit 212. The transmitting digital circuit 211 is a part that performs digital signal processing such as modulation and error correction coding on the data signal to be transmitted.

The transmitting analog circuit 212 includes a quadrature modulation circuit 214, a frequency conversion circuit 215, and a power amplification circuit 216. The quadrature modulation circuit 214 is a part that performs quadrature modulation on the analog signal. The frequency conversion circuit 215 is a part that converts the intermediate frequency band signal into a high-frequency radio signal 10. The power amplification circuit 216 is a part that amplifies the high-frequency radio signal 10. The antenna 217 is a part that transmits the amplified radio signal 10 to the receiving station 220.

The receiving station 220, like the transmitting station 210, includes a receiving digital circuit 221 and a receiving analog circuit 222. The frequency conversion circuit 225 of the receiving analog circuit 222 converts the high-frequency radio signal 10 received by the antenna 226 into an intermediate frequency band. The orthogonal demodulation circuit 224 is a part that performs analog orthogonal demodulation on the intermediate frequency band signal.

The receiving digital circuit 221 performs demodulation, error correction decoding, etc. on the digitized signal.

It is known that in the signal processing circuits provided in the transmitting station 210 and the receiving station 220, failures occur due to imperfections in the equipment, resulting in a deterioration in communication quality.

For example, in the quadrature modulation circuit 214 and the quadrature demodulation circuit 224, an IQ imbalance occurs in which the phase difference between the I component and the Q component is no longer 90 degrees. Also, in the frequency conversion circuit 215 and the frequency conversion circuit 225, phase noise occurs due to phase fluctuation, and a frequency offset occurs due to a frequency shift between the transmitting station 210 and the receiving station 220.

It is also known that the wireless signal 10 transmitted and received between the transmitting station 210 and the receiving station 220 is subject to interference such as channel fading, which deteriorates the communication quality.

2 is a diagram illustrating a method for estimating and compensating for impairments according to the prior art of the present disclosure. In the transmitting analog circuit 212 of the transmitting station 210, two types of impairments occur during signal processing of the wireless signal 10. These two types of impairments are later modeled as functions G _TX,1 and G _TX,2, respectively, in the receiving station 220.

Furthermore, the radio signal 10 transmitted from the transmitting station 210 is subject to impairments due to channel fading. These impairments are later modeled as a function H in the receiving station 220.

Furthermore, in the receiving analog circuit 222 of the receiving station 220, two types of impairments occur during signal processing of the received radio signal 10. These two types of impairments are later modeled as functions G _RX,1 and G _RX,2, respectively.

The receiving digital circuit 221 of the receiving station 220 includes an estimation circuit 21 and a compensation circuit 22.

The estimation circuit 21 estimates the impairments included in the input wireless signal 10 for each type of impairment based on a pilot signal having a predetermined pattern between the transmitting station 210 and the receiving station 220. That is, the estimation circuit 21 calculates functions G _TX,1 , G _TX,2 , H, G _RX,1 and G _RX,2 to estimate the impairments. Furthermore, the estimation circuit 21 calculates compensation weights based on the calculated functions.

The compensation circuit 22 compensates for the disturbance that occurs based on the compensation weight calculated by the estimation circuit 21. The compensation here also includes equalizing the disturbance that occurs in the wireless signal 10 due to channel fading. The compensation circuit 22 then transmits the compensated digital signal to the downstream circuit 23, which includes a demodulation circuit, a bit detection circuit, etc.

In this way, in conventional technology, the type of impairment contained in the wireless signal 10 was estimated individually and then compensation was performed.

The following describes embodiments of the present disclosure with reference to the drawings. The same or corresponding components are given the same reference numerals, and repeated descriptions may be omitted.

First embodiment
3 shows a configuration example of a wireless communication system 100 according to the first embodiment of the present disclosure. The wireless communication system 100 includes a transmitting station 110 and a receiving station 120. Note that the transmitting station 110 is similar to the transmitting station 210 of the related art described in FIG. 2, and therefore a description thereof will be omitted.

The receiving station 120 receives the radio signal 10 from the transmitting station 110. The receiving analog circuit 232 performs analog signal processing on the radio signal 10 and transmits it to the receiving digital circuit 231.

The time domain (TD) signal received by the receiving digital circuit 231 is branched into two. One of the branches is input directly to the compensation circuit 30. The other is FFT (Fast Fourier Transform) converted by the FFT circuit 20 into a frequency domain (FD) signal, and then input to the compensation circuit 30. Hereinafter, the signal input to the compensation circuit 30 is referred to as the input signal 50.

The machine learning circuit 31 of the compensation circuit 30 performs machine learning using a learning model on the input signal 50. The machine learning circuit 31 executes processing of the input signal 50, which includes a received signal caused by a known bit sequence transmitted from the transmitting station 110, using the learning model to generate an output signal 60. Furthermore, it executes a learning process to learn the learning model so that the output signal 60 approaches teacher data consisting of a known bit sequence.

In the learning process, learning is performed so that the output signal 60 approaches the teacher data based on an RMSE (Root Mean Squared Error) value or the like. There is no need to estimate each type of fault contained in the received signal. That is, there is no need to grasp the types of faults represented as functions G _TX,1 and G _TX,2 , etc., individually, as in the conventional technology. Note that the criterion in the learning process is not limited to RMSE, and may be any index indicating the difference between the output signal 60 and the teacher data. For example, MSE (Mean Squared Error), cross entropy error, etc., which are usually used as loss functions in neural networks, may be used.

The teacher data consisting of a known bit sequence is the input signal 50 in TD and FD that is generated from a known bit sequence in the transmitting station 110 and can be considered to have been unaffected by a disturbance that requires compensation until it reaches the compensation circuit 30 in the receiving station 120. However, if it is equivalent to this, the teacher data does not have to be generated from the wireless signal 10 actually transmitted from the transmitting station 110 to the receiving station 120. For example, the input signal 50 can be generated by arithmetic processing from a known bit sequence and used as the teacher data.

Furthermore, a received signal resulting from a known bit sequence is a received signal that is generated from a known bit sequence in the transmitting station 110 and can be considered to have been affected by a fault that requires compensation before it reaches the compensation circuit 30 of the receiving station 120. However, as long as it is equivalent to this, the received signal does not have to be a signal that the receiving station 120 actually received from the transmitting station 110. For example, it may be a signal that is generated by calculation from a known bit sequence based on a model of distortion of the received signal due to a fault that requires compensation.

Learning methods include well-known methods such as random forest, SVM (Support Vector Machine), K-nearest neighbor method, and neural network.

The compensation circuit 30 executes a compensation process that processes the newly input signal 50 based on the learning model obtained by the learning process and outputs an improved output signal 61. Note that the input signal 50 here is the input signal 50 that is generated from an arbitrary bit string in the transmitting station 110 and reaches the compensation circuit 30 of the receiving station 120.

FIG. 4 shows an example of the configuration of an input signal 50 according to the first embodiment of the present disclosure. FIG. 5 is a diagram illustrating the data size of the input signal 50 according to the first embodiment of the present disclosure. The input signal 50 is symbol point data 51 on the IQ coordinate plane of the wireless signal 10.

Since a symbol point exists for each OFDM subcarrier, the input signal 50 has a multi-dimensional array format as shown in Fig. 4. The dimensions here are a first dimension having N _F elements and representing the number of subcarriers, a second dimension having N _S elements and representing the number of OFDM symbols, and a third dimension having x elements and representing the coordinates of the symbol point. Note that here, the symbol points are expressed as real numbers, and the number of elements required for this is 2, but since symbol points exist in both TD and FD, x is 4.

FIG. 4 shows the input signal 50 when x is 4. However, the value of x is only an example and is not limited to this. For example, if the compensation circuit 30 is capable of accepting input in the form of complex numbers, x is not limited to this. Also, the first dimension may be the symbol block size instead of the number of subcarriers. Also, the second dimension may be the number of symbol blocks.

The data size of the input signal 50 is expressed as follows: Note that as the superscript indicates, the data size depends on the total number of elements in the multi-dimensional array of the input signal 50 (N _F ×N _S ×x).

FIG. 6 is a diagram illustrating the data size of the output signal 60 in TD and FD according to the first embodiment of the present disclosure. The output signal 60 may be in the form of a symbol point on the I-Q coordinate plane, like the input signal 50. Furthermore, it may be in the form of a bit log-likelihood ratio (hereinafter referred to as LLR) that indicates the reliability of the transmitted bit, or in the form of a bit string.

When the output signal 60 is in the form of symbol points on the IQ coordinate plane, the data size of the output signal 60 in TD is expressed as follows: As in the case of the input signal 50, the data size depends on the total number of elements in the multi-dimensional array (N _F ×N _S ×2).

Similarly, the data size of the output signal 60 in the FD domain is expressed as follows: Here again, the data size depends on the total number of elements of the multidimensional array (N _F ×N _S ×2).

On the other hand, if the output signal 60 is in the form of LLRs, the data size is expressed as follows: where B is the number of bits per symbol.

Furthermore, when the output signal 60 is a bit string, the data size is expressed as follows:

Note that the explanation in Figure 6 also applies to the improved output signal 61.

As described above, the present disclosure can provide a wireless communication system that can compensate for wireless signals without separating the types of impairments contained in the wireless signals.

The processing performed by the receiving station 120 may be executed by a program using a computer equipped with a CPU and memory and storing a program in the memory. Alternatively, the processing may be executed by a program using an integrated circuit such as an FPGA (Field Programmable Gate Array). The program may be provided by recording it on a storage medium, or may be provided via a network. This point is common to all of the following embodiments.

Note that the input signal 50 does not necessarily have to be in the form of a multidimensional array as long as it contains the above-mentioned information on the number of subcarriers, the number of OFDM symbols, and the coordinates of the symbol points. Furthermore, the input signal 50 may be processed on a frame-by-frame basis or on a symbol-by-symbol basis.

Embodiment 2
In this embodiment, the compensation circuit 30 receives a pilot signal 52 having a predetermined pattern between the transmitting station 110 and the receiving station 120. In the following, changes from the first embodiment will be described.

FIG. 7 shows an example configuration of a receiving digital circuit 231 according to a second embodiment of the present disclosure. In addition to the configuration example of the first embodiment, the receiving digital circuit 231 further includes a signal creation circuit 40. The signal creation circuit 40 creates a pilot signal 52 on the IQ coordinate plane. Furthermore, the signal creation circuit 40 adds the pilot signal 52 to the receiving signal derived from the wireless signal 10 input to the compensation circuit 30.

FIG. 8 shows an example of the configuration of an input signal 50 according to the second embodiment of the present disclosure. A pilot signal 52 corresponding to a symbol point is added to a received signal represented as symbol point data 51 on an I-Q coordinate plane. Therefore, the number of elements x in the third dimension representing the coordinates of the symbol point is 6. The pilot signal 52 is generated so as to correspond to the modulation multi-level number of the wireless signal 10. As a result, in the symbol point data 51 and the pilot signal 52, the number of elements in the first dimension representing the number of subcarriers matches the number of elements in the second dimension representing the number of symbols included in the subcarriers. For incompatible modulation multi-level numbers, the pilot signal 52 is subjected to processing such as zero padding.

The received signal includes a pilot data signal generated in the transmitting station 110. The pilot data signal generated in the transmitting station 110 is affected by interference before it reaches the compensation circuit 30 in the receiving station 120.

By including the original pilot signal 52 that is not affected by the impairment in the input signal 50, it is possible to grasp the difference with the pilot data signal included in the received signal. This makes it possible to further advance the learning of the learning model in the machine learning circuit 31, leading to a reduction in the learning time and an improvement in the learning accuracy. By improving the accuracy of the learning model, it is possible to improve the quality of the improved output signal 61 that is output by the compensation process.

Embodiment 3
In this embodiment, the compensation circuit 30 receives quality data 53 of the wireless signal 10. In the following, changes from the first embodiment will be described.

FIG. 9 shows an example configuration of a receiving digital circuit 231 according to the third embodiment of the present disclosure. In addition to the configuration example of the first embodiment, the receiving digital circuit 231 further includes a quality estimation circuit 70. The quality estimation circuit 70 estimates the quality of the received signal derived from the wireless signal 10. The quality here is, for example, a channel estimation value. Furthermore, the quality estimation circuit 70 adds quality data 53 to the received signal derived from the wireless signal 10 that is input to the compensation circuit 30. Note that, as in the second embodiment, the quality data 53 is also generated so as to correspond to the modulation multi-level number of the wireless signal 10.

FIG. 10 shows an example of the configuration of an input signal 50 according to the third embodiment of the present disclosure. A received signal is represented as symbol point data 51 on an IQ coordinate plane, and quality data 53 corresponding to the symbol point is added to the received signal.

By including quality data 53 in the input signal 50, the compensation circuit 30 can grasp the characteristics of the fault based on the quality. This allows the machine learning circuit 31 to further advance the learning of the learning model, resulting in improved learning accuracy. By improving the accuracy of the learning model, the quality of the improved output signal 61 output by the compensation process can also be improved.

Note that the present disclosure is not limited to the above-described embodiments, and various modifications can be made in the implementation stage without departing from the gist of the disclosure. Furthermore, the embodiments may be implemented in appropriate combination, in which case the combined effects can be obtained. For example, by combining the second and third embodiments, the compensation circuit 30 may be configured to accept all of the symbol point data 51, the pilot signal 52, and the quality data 53.

<Variation 1>
In the above description, the received signal included in the input signal 50 is the symbol point data 51 on the IQ coordinate plane. However, the format of the received signal does not have to be the symbol point data 51 format. For example, the received signal may be LLR data 54.

FIG. 11 shows a configuration example of the receiving digital circuit 231 according to the first modified example of the present disclosure, in which the compensation circuit 30 receives LLR data 54. The receiving digital circuit 231 here further includes an LLR calculation circuit 80 in addition to the configuration example in the first embodiment. The LLR calculation circuit 80 receives symbol point data 51 on the IQ coordinate plane. Furthermore, the LLR calculation circuit 80 calculates a bit log-likelihood ratio (hereinafter referred to as LLR) representing the reliability of the transmitted bit from the symbol point. Furthermore, the LLR calculation circuit 80 inputs the calculation result to the FFT transformation circuit 20 and the compensation circuit 30 as LLR data 54.

The FFT conversion circuit 20 accepts the LLR data 54 in TD, converts it to FD, and transmits it to the compensation circuit 30.

<Modification 2>
Furthermore, the received signal may be in the form of bit string data 55 .

FIG. 12 shows a configuration example of the receiving digital circuit 231 according to the second modified example of the present disclosure, in which the compensation circuit 30 accepts bit string data 55. The receiving digital circuit 231 here further includes an error correction decoding circuit 90 in addition to the configuration example in the first modified example. The error correction decoding circuit 90 performs error correction decoding on the LLR calculated by the LLR calculation circuit 80, and calculates a bit string. Furthermore, the error correction decoding circuit 90 inputs the calculation result as bit string data 55 to the FFT transformation circuit 20 and the compensation circuit 30.

The FFT conversion circuit 20 accepts the bit string data 55 in TD, converts it to FD, and transmits it to the compensation circuit 30.

The configurations of Modifications 1 and 2 make it possible to convert the format of the input signal 50 in advance so that it matches the format of the output signal 60. This eliminates the need to convert the format of the output signal 60 in the compensation circuit 30, making it possible to simplify the configuration of the compensation circuit 30. Note that although the embodiment has been described here in combination with the first embodiment, it may also be combined with other embodiments, in which case the effects of the combination can be obtained.
<Modification 3>
In addition, in cases where the learning accuracy of the machine learning circuit 31 is sufficient with only the input signal 50 in TD, the input signal 50 in FD is not necessary, and the FFT conversion circuit 20 is not necessarily required. This makes it possible to reduce the data size of the input signal 50, and to reduce the burden associated with the learning process and compensation process.

10 radio signal, 20 FFT conversion circuit, 21 estimation circuit, 22 compensation circuit, 23 subsequent circuit, 30 compensation circuit, 31 machine learning circuit, 40 signal creation circuit, 50 input signal, 51 symbol point data, 52 pilot signal, 53 quality data, 54 LLR data, 55 bit string data, 60 output signal, 61 improved output signal, 70 quality estimation circuit, 80 LLR calculation circuit, 90 error correction decoding circuit, 100 radio communication system, 110 transmitting station, 120 receiving station, 200 wireless communication system, 210 transmitting station, 211 transmitting digital circuit, 212 transmitting analog circuit, 214 quadrature modulation circuit, 215 frequency conversion circuit, 216 power amplifier circuit, 217 antenna, 220 receiving station, 221 receiving digital circuit, 222 receiving analog circuit, 224 quadrature demodulation circuit, 225 frequency conversion circuit, 226 antenna, 231 receiving digital circuit, 232 receiving analog circuit

Claims (4)

A transmitting station for transmitting a radio signal;
a receiving station for receiving the radio signal;
Equipped with
The receiving station,
A compensation circuit including a learning model is provided.
a process of processing an input signal including a received signal at the receiving station caused by a known bit sequence transmitted from the transmitting station using the learning model to generate an output signal;
A process of training the learning model so that the output signal approaches training data consisting of the known bit string;
A process of processing a newly input signal based on the learned learning model and outputting an improved output signal;
A wireless communication system configured to: The wireless communication system of claim 1, wherein the received signal resulting from the known bit sequence includes a time domain and a frequency domain. the form of the received signal resulting from the known bit sequence is one of three forms: a symbol point on an IQ coordinate plane of the wireless signal, a bit log-likelihood ratio indicating the reliability of a transmitted bit in the wireless signal, and a bit sequence obtained by performing error correction decoding on the wireless signal;
3. The wireless communication system according to claim 1, wherein the format of the output signal is also one of the three formats. The input signal includes a pilot signal having a predetermined pattern between the transmitting station and the receiving station;
quality data that is an estimation result of the quality of the received signal caused by the known bit sequence;
At least one of
Further learning of the learning model based on at least one of the pilot signal and the quality data;
A wireless communication system according to any one of claims 1 to 3.

Images