Reception device

Abstract

The present disclosure relates to a reception device. This reception device is used for receiving a data signal from a transmission device, the reception device comprising a processor and a memory in which is stored a program to be executed by the processor. The processor is configured to execute a process for generating a first dataset necessary for acquiring a trained model used in compensation, a process for performing machine learning based on the first dataset and acquiring a first trained model used for compensation, a process for performing de-noising for removing random noise of the data signal, a process for applying compensation based on the first trained model to the de-noised data signal, and a process for demodulating the compensated data signal.

Description

Receiving device

[0001] The present disclosure relates to a receiving device.

[0002] In order to improve communication quality, there is a technology that outputs the original signal by performing compensation based on the signal that has changed during the communication process. For example, Non-Patent Document 1 discloses a method for improving demodulation performance by performing likelihood calculation based on the signal affected by amplifier distortion and information on the distortion constellation. The distortion constellation used in this method is identified based on the device specifications of the amplifier installed in the transmitting station.

[0003] Fukuzono, Kuriyama, Uchida, Yoshioka, and Miyagi, “Log-Likelihood Ratio for Nonlinear Distortion Constellations in High-Frequency Band Single-Carrier Transmission,” IEICE Society Conference B-5-57, August 31, 2021.

[0004] However, with the above-mentioned method, when the effect of random noise on signal changes is large, there are problems such as an increase in the time required for compensation and a decrease in compensation accuracy.

[0005] In order to solve the above-mentioned problems, an object of the present disclosure is to provide a receiving device that can reduce the time required for compensation and improve the accuracy of compensation.

[0006] An aspect of the present disclosure is a receiving device that receives a data signal from a transmitting device, and is preferably configured to include a processor and a memory that stores a program to be executed by the processor, and to perform the following processes: generating a first dataset necessary to obtain a learning model to be used for compensation; performing machine learning based on the first dataset to obtain a first learning model to be used for compensation; performing denoising to remove random noise from the data signal; applying compensation to the denoised data signal based on the first learning model; and demodulating the compensated data signal.

[0007] According to aspects of the present disclosure, it is possible to reduce the time required for compensation and at the same time improve the accuracy of compensation.

[0008] 1 is a diagram illustrating a configuration example of a wireless communication system according to a first embodiment of the present disclosure; FIG. 2 is a diagram illustrating a hardware configuration of a receiving device according to a first embodiment of the present disclosure; FIG. 3 is a diagram illustrating changes in a data signal according to a first embodiment of the present disclosure;

[0009] 1 is a diagram illustrating a configuration example of a wireless communication system according to a first embodiment of the present disclosure. The wireless communication system 100 includes a transmitting device 20. The transmitting device 20 is a device that transmits a data signal to a receiving device 40.

[0010] The transmitting device 20 includes a data generating unit 22. The data generating unit 22 generates a data signal for information to be transmitted to the receiving device 40, and transmits the data signal to a modulating unit .

[0011] The modulator 24 modulates the data signal and transmits it to the RF circuit unit 26. The RF circuit unit 26 performs transmission processing on the modulated data signal. The transmission processing includes, for example, converting a digitally modulated data signal into an analog signal, quadrature modulation of an I signal and a Q signal, or amplifying the data signal using an amplifier. The RF circuit unit 26 transmits the data signal that has undergone transmission processing to the receiving device 40.

[0012] The RF circuit unit 42 of the receiving device 40 performs reception processing of the transmitted data signal. The reception processing includes, for example, amplifying the data signal using an amplifier, demodulating the data signal into I and Q signals, or digitizing a digitized data signal. The RF circuit unit 42 transmits the data signal that has undergone reception processing to the denoising unit 44.

[0013] The denoising unit 44 denoises the transmitted data signal. The denoising may be performed by a denoising autoencoder (DAE) or may be performed based on a second learning model acquired by second machine learning performed by a preprocessing device (described later). The second learning model will be described in detail later. This denoising can remove random noise associated with signal changes.

[0014] The denoising unit 44 transmits the denoised data signal to the compensation unit 46. The compensation unit 46 compensates the received data signal and transmits the compensated data signal to the demodulation unit 48. This compensation is performed based on a first learning model acquired by first machine learning performed in a preprocessing unit, which will be described later in detail.

[0015] The compensation described above is compensation for signal changes caused by transmission and reception processes. The cause of the signal changes is, for example, imperfections in the RF circuit. The imperfections in the RF circuit are, for example, nonlinear distortion, IQ imbalance, or phase noise of the amplifier.

[0016] The demodulation unit 48 demodulates the data signal compensated by the compensation units 46a and 46b.

[0017] The acquisition of a model required for compensation and denoising will now be described. The wireless communication system 100 includes a preprocessing unit 60. The preprocessing unit 60 has a dataset generation unit 66. The dataset generation unit 66 generates a second dataset required to acquire a learning model used for denoising, and transmits the second dataset to a machine learning unit 68.

[0018] This second data set is a data set that links data before and after a signal change. The second data set may include, for example, multiple sets of training data that link one frame of data signals that include the influence of random noise with one frame of data signals that have had the influence of random noise removed. Furthermore, this signal change is a signal change due to transmission processing and reception processing, and is, for example, due to random noise.

[0019] The machine learning unit 68 performs second machine learning based on the second dataset to obtain a second learning model. The machine learning unit 64 also transmits the second learning model to the denoising unit 44. As described above, the denoising unit 44 performs denoising based on the learning model. The machine learning unit 68 also transmits the second learning model to the dataset generation unit 62.

[0020] The data set generation unit 62 generates a first data set required to obtain a learning model to be used for compensation, and transmits it to the machine learning unit 64 .

[0021] This first data set is a data set that links data before and after a signal change. The first data set may include, for example, multiple pieces of training data that link one frame of data signals that include the influence of the imperfection of the RF circuit with one frame of data signals that have not been influenced by the imperfection of the RF circuit. Furthermore, this signal change is a signal change due to the transmission process and the reception process, and is caused, for example, by the imperfection of the RF circuit.

[0022] Note that the dataset generation unit 62 may perform denoising processing based on a second learning model when generating the first dataset. In this case, the dataset generation unit 62 first performs denoising processing based on the second learning model on the data signal. Then, the dataset generation unit 62 uses the data signal after the denoising processing as the first dataset. As a result, random noise in the first dataset can be removed, thereby improving the accuracy of the first machine learning described below.

[0023] The machine learning unit 64 performs first machine learning based on the first data set to obtain a first learning model. The machine learning unit 64 also transmits the first learning model to the compensation unit 46. As described above, the compensation unit 46 performs compensation based on the first learning model.

[0024] 2 is a diagram illustrating a hardware configuration of a receiving device according to the first embodiment of the present disclosure. Each function of the receiving device 40 may be partially or entirely configured by hardware such as a programmable logic device (PLD) or a field programmable gate array (FPGA), or may be configured as a program executed by a processor such as a CPU.

[0025] For example, the receiving device 40 can be realized using a computer and a program, and the program can be recorded on a storage medium or provided via a network.

[0026] 2, the receiving device 40 has an input unit 400, an output unit 401, a communication unit 402, a CPU 403, a memory 404, and an HDD 405 connected via a bus 406, and functions as a computer. The receiving device 40 is also capable of inputting and outputting data to and from a computer-readable storage medium 407.

[0027] The input unit 400 is, for example, a keyboard and a mouse, etc. The output unit 401 is, for example, a display device such as a display.

[0028] The communication unit 402 is, for example, a communication interface that communicates with a wireless device to be controlled.

[0029] The CPU 403 controls each component of the receiving device 40 and performs predetermined processing, etc. The memory 404 and HDD 405 store data, etc.

[0030] The storage medium 407 is capable of storing programs and the like that cause the receiving device 40 to execute the functions of the receiving device 40. Note that the architecture that configures the receiving device 40 is not limited to the example shown in FIG.

[0031] 3 is a diagram illustrating changes in a data signal according to the first embodiment of the present disclosure. First, the transmitting device 20 transmits a data signal 2. The data signal 2 changes due to signal changes caused by transmission processing and reception processing, and becomes a data signal 3. The signal changes are caused by, for example, random noise and imperfections in the RF circuit.

[0032] The receiving device 40 denoises the data signal 3. The data signal 3 becomes a data signal 4 by removing random noise.

[0033] Next, the receiving device 40 compensates for the data signal 4. The data signal 4 becomes the data signal 2a through the compensation. The data signal 2a is a data signal that matches the data signal 2. The receiving device 40 then demodulates the data signal 2a.

[0034] As described above, in the present disclosure, compensation is performed after the effects of random noise are removed in advance. As a result, compensation can be performed that targets signal changes caused by factors other than random noise. This reduces the time required for compensation and improves the accuracy of compensation.

[0035] In this disclosure, DAE is exemplified as a denoising technique, but the present disclosure is not limited thereto. Furthermore, the machine learning described in this disclosure is merely an example, and the network configuration, training data, and the like are not limited thereto. Furthermore, the configuration of the wireless communication system, and the like are not limited thereto.

[0036] Although machine learning has been exemplified as a compensation method in this disclosure, the compensation method is not limited to this.Furthermore, the system configuration or modulation method is not limited.

[0037] 2 Data signal 2a Data signal 3 Data signal 4 Data signal 20 Transmitting device 40 Receiving device

Claims

1. A receiving device that receives a data signal from a transmitting device, comprising a processor and a memory that stores a program to be executed by the processor, wherein the processor is configured to perform the following processes: generating a first data set necessary to obtain a learning model to be used for compensation; performing machine learning based on the first data set to obtain a first learning model to be used for compensation; denoising the data signal to remove random noise; applying compensation to the denoised data signal based on the first learning model; and demodulating the compensated data signal.

2. The receiving device according to claim 1, wherein the denoising is performed by a denoising auto encoder.

3. The receiving device of claim 1, wherein the processor is further configured to perform the following processes: generating a second dataset including multiple pieces of training data linking one frame of data signals that include the influence of random noise with one frame of data signals from which the influence of the random noise has been removed; and performing machine learning based on the second dataset to obtain a second learning model; and wherein the denoising is performed based on the second learning model.

4. The receiving device described in claim 3, wherein the process of generating the first data set includes a process of performing a denoising process on the data signal based on the second learning model, and a process of treating the data signal after the denoising process as the first data set.

Images