KR20210133687A - Apparatus for transmitting data based on mimi-auto encoder - Google Patents

Abstract

Disclosed is a data transmission apparatus based on a multi-input multi-output (MIMO)-auto encoder to achieve optimal performance with simple system changes. According to one embodiment of the present invention, the data transmission apparatus comprises: a channel encoder encoding an input signal on the basis of a low density parity check (LDPC) method; a one-hot encoder outputting a one-hot vector by a one-hot encoding process of an LDPC-encoded bit signal; a learning-based modulator using a transmission signal learning model using the one-hot vector as an input and a modulated signal as an output to output the modulated signal corresponding to the one-hot vector; and a complex number mapping unit mapping the modulated signal into a complex format to output an MIMO-based transmission signal.

Description

MIMO-Auto Encoder Based Data Transmission Device {APPARATUS FOR TRANSMITTING DATA BASED ON MIMI-AUTO ENCODER}

The present disclosure relates to MIMO-based communication technology, and more particularly, to a method and apparatus for performing MIMO communication using machine learning.

In order to increase the transmission speed in a communication environment, particularly, in a wireless communication environment, a method of widening a transmission bandwidth or performing high-order modulation is used.

When the transmission bandwidth is widened, the transmission speed can be simply increased, but there is a compatibility problem with the existing method, and a decrease in network capacity may occur due to a decrease in the number of available channels. On the other hand, when high-order modulation is performed, the transmission rate does not increase in proportion to the modulation order, and as the order increases, problems in implementation become severe, and the coverage becomes narrow.

Furthermore, although spatial division multiplexing (SDM) is realized using multiple antennas as a technique for increasing the data rate, the SDM technique has a problem in that the complexity of an operation of separating and detecting parallel transmitted data is high.

In the SDM system, although the ML (maximum likelihood) detection method and the OSIC (Ordered Successive Interference Cancelation) detection method are used, the ML detection method and the OSIC detection method also increase in complexity depending on the system environment or have problems in detection efficiency, etc. can occur.

In recent years, as machine learning technology has become more popular, attempts have been made to transform or demodulate a signal by applying machine learning technology to a wireless communication system. However, in the technology applied to the wireless communication system, the study of simply applying machine learning to the modulation or demodulation method of the signal is mainly made.

However, since the communication system transmits and receives signals in consideration of both the channel encoder (decoder) and the modulator (demodulator), it is difficult to realize the actual channel optimization when the machine learning method is applied only to the modulation (demodulation) method. there is a problem.

The technical task of the present disclosure is to establish a wireless communication system and method that can realize optimal performance by simply changing the system by applying a machine learning-based MIMO Auto Encoder capable of transmitting and receiving multiple signals.

Another technical object of the present disclosure is to provide a method and apparatus for changing a bit probability value so that an output of a MIMO Auto Encoder is suitable for an LDPC decoder.

The technical problems to be achieved in the present disclosure are not limited to the technical problems mentioned above, and other technical problems not mentioned will be clearly understood by those of ordinary skill in the art to which the present disclosure belongs from the description below. will be able

According to an aspect of the present disclosure, a MIMO-Auto Encoder-based data transmission apparatus may be provided. The apparatus comprises: a channel encoder for encoding an input signal based on a low density parity check (LDPC) method; a one-hot encoder for outputting a one-hot vector by one-hot encoding the LDPC-encoded bit signal; A learning-based modulator that outputs a modulated signal corresponding to the one-hot vector using a transmission signal learning model with a one-hot vector as an input and a modulated signal as an output, and MIMO by mapping the modulated signal into a complex format It may include a complex number mapping unit for outputting a base transmission signal.

The features briefly summarized above with respect to the present disclosure are merely exemplary aspects of the detailed description of the present disclosure that follows, and do not limit the scope of the present disclosure.

According to the present disclosure, a wireless communication system and method including a machine learning-based MIMO Auto Encoder capable of transmitting and receiving multiple signals may be provided.

Also, according to the present disclosure, a method and apparatus for changing a bit probability value so that an output of a MIMO Auto Encoder is suitable for an LDPC decoder may be provided.

Effects obtainable in the present disclosure are not limited to the above-mentioned effects, and other effects not mentioned may be clearly understood by those of ordinary skill in the art to which the present disclosure belongs from the description below. will be.

1 is a block diagram illustrating a configuration of a MIMO-Auto Encoder based communication system according to an embodiment of the present disclosure.
2 is a diagram illustrating a one-hot encoding operation in a MIMO-Auto Encoder-based communication system according to an embodiment of the present disclosure.

Hereinafter, embodiments of the present disclosure will be described in detail with reference to the accompanying drawings so that those of ordinary skill in the art to which the present disclosure pertains can easily implement them. However, the present disclosure may be implemented in several different forms and is not limited to the embodiments described herein.

In describing an embodiment of the present disclosure, if it is determined that a detailed description of a well-known configuration or function may obscure the gist of the present disclosure, a detailed description thereof will be omitted. And, in the drawings, parts not related to the description of the present disclosure are omitted, and similar reference numerals are attached to similar parts.

In the present disclosure, when it is said that a component is "connected", "coupled" or "connected" with another component, it is not only a direct connection relationship, but also an indirect connection relationship in which another component exists in the middle. may also include. In addition, when a component is said to "include" or "have" another component, it means that another component may be further included without excluding other components unless otherwise stated. .

In the present disclosure, components that are distinguished from each other are for clearly explaining each characteristic, and do not necessarily mean that the components are separated. That is, a plurality of components may be integrated to form one hardware or software unit, or one component may be distributed to form a plurality of hardware or software units. Accordingly, even if not specifically mentioned, such integrated or distributed embodiments are also included in the scope of the present disclosure.

In the present disclosure, components described in various embodiments do not necessarily mean essential components, and some may be optional components. Accordingly, an embodiment composed of a subset of components described in an embodiment is also included in the scope of the present disclosure. In addition, embodiments including other components in addition to components described in various embodiments are also included in the scope of the present disclosure.

Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings.

1 is a block diagram illustrating a configuration of a MIMO-Auto Encoder based communication system according to an embodiment of the present disclosure.

Referring to FIG. 1 , a MIMO-Auto Encoder-based communication system according to an embodiment of the present disclosure uses an Auto Encoder to which an LDPC encoder is applied to perform a multi-signal generation and mapping function. It may include a receiver 150 for detecting and demodulating a signal received through the channel. Furthermore, multiple transmission/reception channels may exist between the transmitter 110 and the receiver 150, and signal transmission/reception may be processed in a MIMO environment including T transmit antennas and R receive antennas.

Hereinafter, detailed configurations of the transmitter 110 and the receiver 150 will be described.

First, the transmitter 110 may include a channel encoder 111 , a One-Hot encoder 112 , a learning-based modulator 113 , and a complex number mapping unit 114 .

The channel encoder 111 may encode an input signal based on a Low Density Parity Check (LDPC) method to output an LDPC-encoded bit signal. In this case, the LDPC-encoded bit signal may be configured in a binary bit format.

The one-hot encoder 112 may perform one-hot encoding processing on the LDPC-encoded bit signal, and output a changed result into a one-hot vector. A specific operation of the one-hot encoding process will be described in detail with reference to FIG. 2 below.

크기의 벡터를 출력할 수 있다. The learning-based modulator 113 may include a transmission signal learning model, and the transmission signal learning model may include a fully connected network including a deep neural network (DNN) function and a non-linear activation function. In this case, the Activation function may include a non-linear function. For example, the Activation function may include ReLU, tanh, sigmoid, and the like. Furthermore, the learning-based modulator 113 is

You can output a vector of magnitude.

In particular, the learning-based modulator 113 may perform learning on the transmission signal learning model, and may set the input of the transmission signal learning model as a one-hot vector and set the modulation signal as label data. The learning-based modulator 113 may control learning of the learning model so that a difference value between the estimated value of the modulated signal estimated through the transmission signal learning model and the label data (modulated signal) is minimized.

크기의 벡터를 입력받고, 이를 무선 통신시스템의 송신 신호에 적합한 I/Q 데이터로 변환할 수 있다. 이때, 복소수 맵핑부(114)는 이웃하는 2개의 입력값을 이용하여 I/Q 데이터를 구성할 수 있다. 예를 들어, 복소수 맵핑부(114)에는

크기의 벡터가 입력될 수 있는데, 복소수 맵핑부(114)는 제1 및 제2벡터의 입력 데이터를 사용하여 제1출력 데이터를 구성할 수 있다. 이와 같은 방법을 통해, 복소수 맵핑부(114)는

차원의 실수 형태 입력 데이터를

차원의 복소수 데이터로 변환하여 출력할 수 있다. The complex number mapping unit 114 is outputted from the learning-based modulator 113 .

It is possible to receive a vector of magnitude and convert it into I/Q data suitable for a transmission signal of a wireless communication system. In this case, the complex number mapping unit 114 may configure I/Q data using two neighboring input values. For example, the complex number mapping unit 114 has

A vector of magnitude may be input, and the complex number mapping unit 114 may configure first output data using input data of the first and second vectors. Through this method, the complex number mapping unit 114

dimensional real number input data

It can be converted into dimensional complex data and output.

차원의 복소수 데이터를 정규화(Normalization)하여 출력할 수 있다. 즉, 복소수 맵핑부(114)는 전송 신호의 파워가 1이 되도록 하거나, 특정 무선통신 시스템이 전송하고자 하는 송신 전력을 기준으로 입력 신호의 전력을 제어할 수 있다. 이렇게 Power Normalization이 수행된 신호는 MIMO 채널을 거쳐 R개의 수신 안테나에 전달될 수 있다. Furthermore, the complex number mapping unit 114

The dimensional complex data can be normalized and output. That is, the complex number mapping unit 114 may set the power of the transmission signal to 1 or may control the power of the input signal based on the transmission power desired to be transmitted by a specific wireless communication system. The signal on which the power normalization has been performed may be transmitted to the R reception antennas through the MIMO channel.

The signal output from the above-described transmitter 110 may be transmitted to the receiver 150 through the wireless communication network 130 .

On the other hand, the receiving device 150, the I/Q separation unit 151, the learning-based demodulator 152, the classifier (Classification) 153, One-Hot-based LLR (Log Likelihood Ratio) processing unit 154, and a channel A decoder 155 may be included.

The I/Q separation unit 151 may perform an I/Q separation operation to fit the structure of the learning-based demodulator 152 . The I/Q separation operation may be an operation of dividing one complex number into two real numbers. In general, since it is difficult to process complex numbers in machine learning-based functions, it is necessary to convert data in the form of complex numbers into real numbers. Accordingly, the I/Q separation unit 151 may output data converted into two real numbers in response to each complex data input, and thereby convert the R complex numbers input through the R reception antennas to 2 R pieces. It can be converted to a real vector and output.

The learning-based demodulator 152 is a component corresponding to the learning-based modulator 113 of the transmission device 110 and may include a received signal learning model. In this case, the received signal learning model may be configured to include a plurality of neural networks composed of a fully connected network and an activation function.

Furthermore, the learning-based demodulator 152 may control learning of the received signal learning model. That is, the learning-based demodulator 152 can be trained so that the Fully Connected Network of the received signal learning model can reduce the difference between the signal transmitted from the transmitter 110 and the signal demodulated from the receiver 150 . In this case, the difference value between the two signals (the transmitted signal and the demodulated signal) may be calculated by the classifier 153, for example, may be calculated based on a cross-entropy operation. In this case, the cross-entropy operation can be exemplified by Equation 1 below.

는 수신장치(150)의 기계학습 함수를 나타내며

는 수신장치(150) 기계학습 함수의 계수를 나타낸다. 분류기(153)는, 수학식 1에서와 같이 Negative Log-Likelihood 기반의 Cross-Entropy 값이 점차 줄어들 수 있도록 기계학습 함수의 계수를 최적화하여 학습기반 복조기(152)에 구비되는 학습모델, 예컨대, Fully Connected Network에 대한 학습을 수행할 수 있다. in the above expression

represents the machine learning function of the receiver 150

denotes the coefficients of the machine learning function of the receiving device 150 . The classifier 153 optimizes the coefficients of the machine learning function so that the negative log-likelihood-based cross-entropy value can be gradually reduced as in Equation 1, and the learning model provided in the learning-based demodulator 152, for example, Fully You can learn about Connected Network.

Data output through the receiver 150 may be configured as a Softmax output of the same length as the One-hot vector configured by the One-Hot encoder 112 of the transmitter 110 . In consideration of this, the One-Hot-based LLR processing unit 154 may perform Log-Likelihood Ratio conversion processing on the data provided from the classifier 153 to be suitable for the input of the channel decoder 155 .

The channel decoder 155 may perform LDPC decoding and finally output channel-decoded bit data.

Hereinafter, detailed operations of the above-described MIMO-Auto Encoder-based communication system will be described with a focus on the operation processing operation.

로 설정될 수 있다. 각 안테나에서 이진 비트에 해당하는 신호를 전송하므로, 2개의 전송 안테나를 갖는 MIMO 시스템의 전송을 위한 one-hot 벡터는 도 2와 같이 예시할 수 있다. First, the channel encoder 111 encodes an input signal based on a Low Density Parity Check (LDPC) method, and then outputs it as an LDPC-encoded binary bit. The binary bits output in this way are input to the One-Hot encoder 112 . One-Hot encoder 112 converts binary bits into One-Hot vectors with information that can be transmitted through T transmit antennas. For example, considering the same transmission efficiency as a transmission system having two transmit antennas using QPSK modulation, the length of the One-Hot vector is

can be set to Since each antenna transmits a signal corresponding to a binary bit, a one-hot vector for transmission of a MIMO system having two transmit antennas may be exemplified as shown in FIG. 2 .

The one-hot vector may be input to the learning-based modulator 113, and the learning-based modulator 113 may include a Fully Connected Neural Network and an Activation function, and output data in a format exemplified in Equation 2 below. can do.

는 간략하게 표현된 기계학습 기반의 딥러닝 함수이고 이때 사용되는 함수의 계수는

로 나타낼 수 있다. 또한,

은 학습기반 변조기(113)의 출력 신호를 나타내며,

개의 원소를 구비하는 신호 벡터의 형태로 구성될 수 있다.here,

is a simplified machine learning-based deep learning function, and the coefficients of the function used in this case are

can be expressed as In addition,

represents the output signal of the learning-based modulator 113,

It may be configured in the form of a signal vector having elements.

개의 원소를 구비하는 신호 벡터는 복소수 맵핑부(114)에 입력될 수 있으며, 복소수 맵핑부(114)는 입력된 데이터 중에서 이웃하는 두개의 값을 조합하여 I/Q 데이터(real term, imaginary term)로 mapping 처리할 수 있다. 복소수 맵핑부(114)는 하기의 수학식 3의 연산을 I/Q 데이터의 매핑을 수행할 수 있다.A signal output from the learning-based modulator 113, for example,

A signal vector including elements may be input to the complex number mapping unit 114, and the complex number mapping unit 114 combines two neighboring values among the input data to obtain I/Q data (real term, imaginary term). Mapping can be processed with . The complex number mapping unit 114 may perform I/Q data mapping by the operation of Equation 3 below.

는 t번째 매핑 처리된 결과를 나타내며, 입력값인

과

가 I/Q 형태로 매핑된 결과를

로 구성할 수 있다. here,

represents the t- th mapping process, and the input value

class

is mapped to the I/Q form.

can be configured as

Furthermore, the complex number mapping unit 114 may control the power of a transmission symbol in a baseband by normalizing the result of the mapping process (Normalization). Power control of the transmission symbol may be performed through the operation of Equation 4 below.

In an embodiment of the present disclosure, it has been exemplified that the complex number mapping unit 114 controls the power of the transmission symbol through the operation of Equation 4, but the present disclosure is not limited thereto, and those of ordinary skill in the art It can be changed in various ways by the person.

Signal output from the complex mapping unit 114 is illustrated that includes a may be transmitted to the receiving apparatus 150 via a MIMO channel and the AWGN channel, this time, the reception apparatus 150 is the R receive antennas and T Transmission symbols may be input to R receiving terminals through the above-described channels. A signal transmitted to the receiver 150 in the above-described environment may be expressed as in Equation 5 below.

는 t번째 전송 안테나로부터 r번째 수신 안테나까지의 채널을 나타내며,

은 r번째 수신 안테나의 가우시안 잡음(AWGN: Additive White Gaussian Noise)을 나타낸다. here,

denotes the channel from the t- th transmit antenna to the r- th receive antenna,

is Gaussian noise (AWGN) of the r-th receiving antenna.

는 기계학습 기반의 Neural Network 구조에 맞도록 특정한 규칙으로 정규화된 신호이므로, 전술한 기계학습 기반의 Neural Network 구조에 맞춰 복원할 필요가 있다. 그러나, 전술한 기계학습 기반의 Neural Network는 복소수 데이터 처리에 적합하지 않으므로, 수신된 신호를 real값과 imaginary값으로 분리할 필요가 있다. 전술한 바를 고려하여, I/Q분리부(151)는 수신된 신호를 real값과 imaginary값으로 분리할 수 있으며, 이를 하기의 수학식 6의 연산을 통해 수행할 수 있다. The signal received by the receiving device 150, that is,

Since is a signal normalized with a specific rule to fit the machine learning-based neural network structure, it is necessary to restore it according to the above-described machine learning-based neural network structure. However, since the aforementioned machine learning-based neural network is not suitable for complex data processing, it is necessary to separate the received signal into a real value and an imaginary value. In consideration of the above, the I/Q separation unit 151 may separate the received signal into a real value and an imaginary value, and this may be performed through the operation of Equation 6 below.

는 r번쩨 수신된 신호

의 real값이고,

는 imaginary 값을 나타낸다. N_y는 I/Q분리부(151)가 출력하는 데이터 포맷을 나타낸다.in Equation 6

is the rth received signal

is the real value of

represents an imaginary value. N _y represents a data format output by the I/Q separator 151 .

The learning-based demodulator 152 performs machine learning-based demodulation by using the data output from the I/Q separation unit 151 as input data. At this time, the operation corresponding to each neuron can be expressed as a fully connected (FC) function, and the activation function operation is performed to process the neuron output as a non-linear function. Activation function can use various existing functions. For example, it is possible to use existing functions such as ReLU, tanh, and sigmoid for the Activation function. Such a procedure can be expressed as in Equation 7 below.

는 간략하게 표현된 딥러닝 함수이고 이때 사용되는 함수의 계수는

로 표현된다. 또한, 출력신호인

은 기계학습 과정을 통해 추정된 신호이다. 수신 신호가 다중 채널을 통과하여 수신된 경우, 또는 시간/주파수 옵셋이 발생한 경우에는 이전에 전송된 심볼이 이후 전송된 심볼에 지속적으로 영향을 주게 된다. 이로 인해, 하나의 시점에서 수신된 신호를 통해 전송된 신호를 추정하게 되면 정확도가 감소하게 된다. 따라서, 다수개의 수신신호를 사용하여 하나의 전송된 신호를 검출하는 연산이 필요하며, 하나의 신호 검출을 위해 다수개의 수신 신호를 사용하는 것이 용이하도록 CNN(Convolutional Neural Network) 및 RNN(Recurrent Neural Network) 등의 수신신호 학습모델을 구성할 수 있다.

is a simplified deep learning function, and the coefficients of the function used in this case are

is expressed as Also, the output signal

is the signal estimated through the machine learning process. When a received signal is received through multiple channels, or when a time/frequency offset occurs, the previously transmitted symbol continuously affects the subsequently transmitted symbol. For this reason, when estimating the signal transmitted through the signal received at one point in time, the accuracy is reduced. Therefore, an operation of detecting one transmitted signal using a plurality of received signals is required, and a Convolutional Neural Network (CNN) and a Recurrent Neural Network (RNN) to facilitate using a plurality of received signals for detecting one signal is required. ), etc., can be configured as a received signal learning model.

과 전송된

값을 비교하여 차이값이 점차 줄어드는 방향으로 학습을 수행할 수 있다. 일반적으로 이러한 차이값을 훈련에 따른 cost 또는 cost function이라고 하는데, cost function 계산을 위해서는 MSE(Mean Square Error) 방식 또는 Cross Entropy 방식이 사용될 수 있다. In consideration of the above, the learning-based demodulator 152 estimates a one-hot vector through the received signal learning model. In the learning process of the received signal learning model, the output value through the estimated learning model

and transmitted

By comparing values, learning can be performed in a direction in which the difference value gradually decreases. In general, such a difference value is called a cost or cost function according to training. In order to calculate the cost function, a Mean Square Error (MSE) method or a Cross Entropy method may be used.

와

값을 업데이트할 수 있다. 계수를 업데이트 하는 방법은 일반적인 SGD(Stochastic Gradient Decent) 방식 또는 Adam optimization 방식 등이 사용될 수 있다. 본 개시의 일 실시예에서, 계산된 차이값을 기반으로 송신단 및 수신단을 모두 학습하므로 주어진 무선 MIMO 채널에 적합한 신호 변조 뿐만 아니라 복조까지 시스템 전반을 모두 효과적으로 학습하는 것이 가능하다. 따라서, 본 개시의 일 실시예에 따른 통신 시스템은 송신단 및 수신단의 학습모델을 채널 특성에 맞도록 함께 업데이트하는 방식이므로 채널 환경에 적합한 형태로 전송신호가 생성되고 그에 따라 다중수신 신호의 검출 및 복조가 용이한 방향으로 수신단이 구축될 수 있다. 기계학습 함수의 계수인

및

값은 특정 반복횟수만큼 반복학습이 수행된 후 update가 완료되거나, 특정 기준값 이하의 에러값이 도출되면 학습을 중단하고, 학습을 통해 구축된 학습모델을 사용하여 신호의 변조 및 복조를 수행할 수 있다. In the Auto Encoder-based multiplexing/receiving system, the coefficient of the machine learning function is

Wow

You can update the value. As a method of updating the coefficients, a general SGD (Stochastic Gradient Decent) method or an Adam optimization method may be used. In an embodiment of the present disclosure, since both the transmitting end and the receiving end are learned based on the calculated difference value, it is possible to effectively learn the entire system from signal modulation to demodulation as well as signal modulation suitable for a given wireless MIMO channel. Therefore, in the communication system according to an embodiment of the present disclosure, since the learning models of the transmitting end and the receiving end are updated together to match the channel characteristics, the transmission signal is generated in a form suitable for the channel environment, and accordingly, the detection and demodulation of the multiple reception signal The receiving end can be built in a direction that is easy to use. Coefficients of machine learning functions

and

The value is updated after repeated learning is performed for a certain number of repetitions, or when an error value below a certain reference value is derived, learning is stopped, and signal modulation and demodulation can be performed using the learning model built through learning. have.

In addition, the final derived result is the output value of the Softmax function having the same length as the input One-Hot vector. The output value of the Softmax function can be expressed as Equation 8 below.

는 Softmax 함수의 k번째 입력이며 M은 One-Hot벡터의 길이와 동일하다. 따라서, Softmax의 출력값은 One-Hot 벡터의 각 비트들이 1이 될 확률을 나타낸다. 그러므로, One-Hot 기반 LLR 처리부(154)는 상기 값을 이용하여 LLR값을 계산할 수 있으며, 예컨대, 하기의 수학식 9의 연산을 통해 LLR값을 계산할 수 있다.

is the kth input of the Softmax function, and M is equal to the length of the One-Hot vector. Therefore, the output value of Softmax represents the probability that each bit of the One-Hot vector becomes 1. Therefore, the One-Hot-based LLR processing unit 154 may calculate the LLR value by using the value, for example, may calculate the LLR value through the operation of Equation 9 below.

는 전송 심볼 x의 i번째 비트를 나타낸다. 본 개시의 일 실시예에서, Auto Encoder 기반 송수신 시스템의 출력은 채널 등화가 완료된 Softmax값이다. 따라서, 제안된 Auto Encoder 기반의 시스템을 위한 LLR 출력은 하기의 수학식 10과 같이 나타낼 수 있다.

denotes the i-th bit of the transmission symbol x. In an embodiment of the present disclosure, the output of the Auto Encoder-based transmission/reception system is a Softmax value on which channel equalization is completed. Accordingly, the LLR output for the proposed Auto Encoder-based system can be expressed as Equation 10 below.

Since the Softmax output value is a negative Log-Likelihood value, the One-Hot-based LLR processing unit 154 may use the Softmax output value as it is to generate the LLR value through the operation of Equation 11 below.

Thereafter, the channel decoder 155 may perform LDPC decoding to finally output channel-decoded bit data.

The steps of a method or algorithm described in connection with the embodiments disclosed herein may be directly implemented in hardware, a software module executed by a processor, or a combination of the two. A software module may reside in a storage medium (ie, memory and/or storage) such as RAM memory, flash memory, ROM memory, EPROM memory, EEPROM memory, registers, hard disk, a removable disk, a CD-ROM. An exemplary storage medium is coupled to the processor, the processor capable of reading information from, and writing information to, the storage medium. Alternatively, the storage medium may be integral with the processor. The processor and storage medium may reside within an application specific integrated circuit (ASIC). The ASIC may reside within the user terminal. Alternatively, the processor and storage medium may reside as separate components within the user terminal.

Example methods of the present disclosure are expressed as a series of operations for clarity of description, but this is not intended to limit the order in which the steps are performed, and if necessary, each step may be performed simultaneously or in a different order. In order to implement the method according to the present disclosure, other steps may be included in addition to the illustrated steps, steps may be excluded from some steps, and/or other steps may be included except for some steps.

Various embodiments of the present disclosure do not list all possible combinations, but are intended to describe representative aspects of the present disclosure, and the details described in various embodiments may be applied independently or in combination of two or more.

In addition, various embodiments of the present disclosure may be implemented by hardware, firmware, software, or a combination thereof. For implementation by hardware, one or more Application Specific Integrated Circuits (ASICs), Digital Signal Processors (DSPs), Digital Signal Processing Devices (DSPDs), Programmable Logic Devices (PLDs), Field Programmable Gate Arrays (FPGAs), general purpose It may be implemented by a processor (general processor), a controller, a microcontroller, a microprocessor, and the like.

The scope of the present disclosure includes software or machine-executable instructions (eg, operating system, application, firmware, program, etc.) that cause an operation according to the method of various embodiments to be executed on a device or computer, and such software or and non-transitory computer-readable media in which instructions and the like are stored and executed on a device or computer.

Claims (1)

In the MIMO-based data transmission device,
A channel encoder for encoding an input signal based on a Low Density Parity Check (LDPC) method;
One-hot encoder that outputs one-hot vector by one-hot encoding processing of LDPC-encoded bit signal;
a learning-based modulator for outputting a modulated signal corresponding to the one-hot vector by using a transmission signal learning model with the one-hot vector as an input and a modulated signal as an output;
and a complex number mapping unit that maps the modulated signal to a complex format and outputs a MIMO-based transmission signal.

Images