Decoding method and device using noisy embedding in wireless mobile communication system

Abstract

The present disclosure provides a decoding method and device using noisy embedding in a wireless mobile communication system. According to an embodiment of the present disclosure, a method performed by a decoding device in a wireless communication system comprises the operations of: receiving a first signal with noise; generating at least one random noise on the basis of a scaling factor; generating at least one second signal by applying the at least one random noise to the first signal; and outputting a codeword by performing at least one neural network decoding on the at least one second signal.

Description

Decoding method and device using noisy embedding in wireless mobile communication systems

[0001] The present disclosure relates to a decoding method and device using noise embedding in a receiving device in a wireless communication system.

[0002] With the development of digital technology, electronic devices are available in various forms, such as smart phones, tablet personal computers, and personal digital assistants (PDAs).

[0003] As artificial intelligence (AI) technology advances, electronic devices can apply AI technology to provide a variety of AI services. Electronic devices can provide AI services by providing software components (e.g., services, functions, or programs) that process user-requested tasks and provide customized services (e.g., customized information based on user voice commands) based on AI and voice recognition technologies.

[0004] Machine learning is a field related to artificial intelligence, developing algorithms and technologies that enable computers to learn. Deep learning refers to a set of machine learning algorithms that attempt to achieve a high level of abstraction (the task of extracting only the essential information from large amounts of complex data) through a combination of nonlinear transformation methods.

[0005] Meanwhile, key technologies for various wireless service standards, including the New Radio (NR) standard, a recent fifth-generation data transmission method, enhanced Mobile Broad-Band (eMBB), Ultra Reliable Low Latency Communication (URLLC), and the Internet of Things (IoT), are attracting significant attention. Accordingly, active research is being conducted on decoding methods for linear block codes (LBCs) used for error correction in wireless communication systems or storage devices, including flash memory.

[0006] The above information may be provided as background information to aid in understanding this document. None of the above is claimed to be prior art related to this document or can be used to determine prior art.

[0007] The present disclosure aims to provide a decoding method and device using noisy embedding in a wireless mobile communication system.

[0008] The technical problems to be achieved in the present disclosure are not limited to those mentioned above, and other technical problems not mentioned can be considered by a person having ordinary skill in the art from the various embodiments of the present disclosure described below.

[0009] According to one embodiment, a method of a decoding device in a wireless communication system includes: receiving a first signal having noise; generating at least one random noise based on a scaling factor; applying the at least one random noise to the first signal to generate at least one second signal; and performing at least one neural network decoding on the at least one second signal to output a codeword.

[0010] According to one embodiment, a storage medium storing at least one computer-readable instruction, wherein the at least one instruction, when executed by at least a part of at least one processor (120) of an electronic device, causes the electronic device to perform at least one operation, the at least one operation including: receiving a first signal having noise; generating at least one random noise based on a scaling factor; applying the at least one random noise to the first signal to generate at least one second signal; and performing at least one neural network decoding on the at least one second signal to output a codeword.

[0011] According to one embodiment, an electronic device comprises at least one processor (120); and a memory (130) storing at least one instruction, wherein the at least one instruction, when executed by at least a portion of the at least one processor (120), causes the electronic device to perform at least one operation, the at least one operation including: receiving a first signal having noise; generating at least one random noise based on a scaling factor; generating at least one second signal by applying the at least one random noise to the first signal; and performing at least one neural network decoding on the at least one second signal to output a codeword.

[0012] By the present disclosure, a code decoding device using noise embedding can be configured, and the performance of the decoding device can be improved by performing real-time learning using input signals with noise embedding applied as learning data.

[0013] In connection with the description of the drawings, the same or similar reference numerals may be used for the same or similar components.

[0014] FIG. 1 is a block diagram of an electronic device within a network environment according to one embodiment of the present disclosure.

[0015] FIG. 2 illustrates a block diagram of an electronic device according to one embodiment of the present disclosure.

[0016] FIG. 3 illustrates a block diagram of a decoder according to one embodiment of the present disclosure.

[0017] FIG. 4 illustrates the structure of a neural network-based decoder with scalable noisy embedding applied according to one embodiment of the present disclosure.

[0018] FIG. 5 illustrates the structure of a scalable noisy embedding included in a decoder according to one embodiment of the present disclosure.

[0019] FIG. 6 illustrates the structure of a decoder including multiple sub-neural network-based decoders according to one embodiment of the present disclosure.

[0020] FIG. 7 is a diagram illustrating a real-time learning operation for the output of a scalable noisy embedding included in a decoder of a decoder according to one embodiment of the present disclosure and a decoding result.

[0021] FIG. 8 is a flowchart illustrating the operation of a decoder according to one embodiment of the present disclosure.

[0022] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings so that those skilled in the art can easily implement the present disclosure. However, the present disclosure may be implemented in various different forms and is not limited to the embodiments described herein. In connection with the description of the drawings, the same or similar reference numerals may be used for identical or similar components. Furthermore, in the drawings and related descriptions, descriptions of well-known functions and configurations may be omitted for clarity and conciseness.

[0023] FIG. 1 is a block diagram of an electronic device (101) within a network environment (100) according to various embodiments. Referring to FIG. 1, in the network environment (100), the electronic device (101) may communicate with the electronic device (102) via a first network (198) (e.g., a short-range wireless communication network), or may communicate with at least one of the electronic device (104) or the server (108) via a second network (199) (e.g., a long-range wireless communication network). In one embodiment, the electronic device (101) may communicate with the electronic device (104) via the server (108). According to one embodiment, the electronic device (101) may include a processor (120), a memory (130), an input module (150), an audio output module (155), a display module (160), an audio module (170), a sensor module (176), an interface (177), a connection terminal (178), a haptic module (179), a camera module (180), a power management module (188), a battery (189), a communication module (190), a subscriber identification module (196), or an antenna module (197). In some embodiments, the electronic device (101) may omit at least one of these components (e.g., the connection terminal (178)), or may have one or more other components added. In some embodiments, some of these components (e.g., the sensor module (176), the camera module (180), or the antenna module (197)) may be integrated into one component (e.g., the display module (160)).

[0024] The processor (120) may, for example, execute software (e.g., a program (140)) to control at least one other component (e.g., a hardware or software component) of the electronic device (101) connected to the processor (120) and perform various data processing or operations. According to one embodiment, as at least a part of the data processing or operations, the processor (120) may store commands or data received from other components (e.g., a sensor module (176) or a communication module (190)) in a volatile memory (132), process the commands or data stored in the volatile memory (132), and store result data in a non-volatile memory (134). According to one embodiment, the processor (120) may include a main processor (121) (e.g., a central processing unit or an application processor) or an auxiliary processor (123) (e.g., a graphics processing unit, a neural processing unit (NPU), an image signal processor, a sensor hub processor, or a communication processor) that can operate independently or together with the main processor (121). For example, when the electronic device (101) includes the main processor (121) and the auxiliary processor (123), the auxiliary processor (123) may be configured to use less power than the main processor (121) or to be specialized for a given function. The auxiliary processor (123) may be implemented separately from the main processor (121) or as a part thereof.

[0025] The auxiliary processor (123) may control at least a portion of functions or states associated with at least one component (e.g., a display module (160), a sensor module (176), or a communication module (190)) of the electronic device (101), for example, on behalf of the main processor (121) while the main processor (121) is in an inactive (e.g., sleep) state, or together with the main processor (121) while the main processor (121) is in an active (e.g., application execution) state. In one embodiment, the auxiliary processor (123) (e.g., an image signal processor or a communication processor) may be implemented as a part of another functionally related component (e.g., a camera module (180) or a communication module (190)). In one embodiment, the auxiliary processor (123) (e.g., a neural network processing unit) may include a hardware structure specialized for processing artificial intelligence models. The artificial intelligence models may be generated through machine learning. This learning can be performed, for example, on the electronic device (101) itself where the artificial intelligence model is executed, or can be performed through a separate server (e.g., server (108)). The learning algorithm can include, for example, supervised learning, unsupervised learning, semi-supervised learning, or reinforcement learning, but is not limited to the examples described above. The artificial intelligence model can include multiple artificial neural network layers.The artificial neural network may be one of a deep neural network (DNN), a convolutional neural network (CNN), a recurrent neural network (RNN), a restricted Boltzmann machine (RBM), a deep belief network (DBN), a bidirectional recurrent deep neural network (BRDNN), a deep Q-network, or a combination of two or more of the above, but is not limited to the examples described above. In addition to, or alternatively to, a hardware structure, an artificial intelligence model may include a software structure.

[0026] The memory (130) can store various data used by at least one component (e.g., processor (120) or sensor module (176)) of the electronic device (101). The data can include, for example, software (e.g., program (140)) and input data or output data for commands related thereto. The memory (130) can include volatile memory (132) or non-volatile memory (134).

[0027] The program (140) may be stored as software in memory (130) and may include, for example, an operating system (142), middleware (144), or an application (146).

[0028] The input module (150) can receive commands or data to be used in a component of the electronic device (101) (e.g., a processor (120)) from an external source (e.g., a user) of the electronic device (101). The input module (150) can include, for example, a microphone, a mouse, a keyboard, a key (e.g., a button), or a digital pen (e.g., a stylus pen).

[0029] The audio output module (155) can output audio signals to the outside of the electronic device (101). The audio output module (155) can include, for example, a speaker or a receiver. The speaker can be used for general purposes, such as multimedia playback or recording playback. The receiver can be used to receive incoming calls. In one embodiment, the receiver can be implemented separately from the speaker or as part of the speaker.

[0030] The display module (160) can visually provide information to an external party (e.g., a user) of the electronic device (101). The display module (160) may include, for example, a display, a holographic device, or a projector and a control circuit for controlling the device. According to one embodiment, the display module (160) may include a touch sensor configured to detect a touch, or a pressure sensor configured to measure the intensity of a force generated by the touch.

[0031] The audio module (170) can convert sound into an electrical signal, or vice versa, convert an electrical signal into sound. According to one embodiment, the audio module (170) can acquire sound through the input module (150), output sound through the sound output module (155), or an external electronic device (e.g., electronic device (102)) (e.g., speaker or headphone) directly or wirelessly connected to the electronic device (101).

[0032] The sensor module (176) can detect the operating status (e.g., power or temperature) of the electronic device (101) or the external environmental status (e.g., user status) and generate an electrical signal or data value corresponding to the detected status. According to one embodiment, the sensor module (176) can include, for example, a gesture sensor, a gyro sensor, a barometric pressure sensor, a magnetic sensor, an acceleration sensor, a grip sensor, a proximity sensor, a color sensor, an IR (infrared) sensor, a biometric sensor, a temperature sensor, a humidity sensor, or an illuminance sensor.

[0033] The interface (177) may support one or more designated protocols that may be used to directly or wirelessly connect the electronic device (101) with an external electronic device (e.g., the electronic device (102)). In one embodiment, the interface (177) may include, for example, a high definition multimedia interface (HDMI), a universal serial bus (USB) interface, an SD card interface, or an audio interface.

[0034] The connection terminal (178) may include a connector through which the electronic device (101) may be physically connected to an external electronic device (e.g., electronic device (102)). According to one embodiment, the connection terminal (178) may include, for example, an HDMI connector, a USB connector, an SD card connector, or an audio connector (e.g., a headphone connector).

[0035] The haptic module (179) can convert electrical signals into mechanical stimuli (e.g., vibration or movement) or electrical stimuli that a user can perceive through tactile or kinesthetic sensations. According to one embodiment, the haptic module (179) can include, for example, a motor, a piezoelectric element, or an electrical stimulation device.

[0036] The camera module (180) can capture still images and videos. According to one embodiment, the camera module (180) may include one or more lenses, image sensors, image signal processors, or flashes.

[0037] The power management module (188) can manage power supplied to the electronic device (101). According to one embodiment, the power management module (188) can be implemented as, for example, at least a part of a power management integrated circuit (PMIC).

[0038] A battery (189) may power at least one component of the electronic device (101). In one embodiment, the battery (189) may include, for example, a non-rechargeable primary battery, a rechargeable secondary battery, or a fuel cell.

[0039] The communication module (transmitter / receiver) (190) may support the establishment of a direct (e.g., wired) communication channel or a wireless communication channel between the electronic device (101) and an external electronic device (e.g., electronic device (102), electronic device (104), or server (108)), and the performance of communication through the established communication channel. The communication module (190) may operate independently from the processor (120) (e.g., application processor) and may include one or more communication processors that support direct (e.g., wired) communication or wireless communication. According to one embodiment, the communication module (190) may include a wireless communication module (192) (e.g., a cellular communication module, a short-range wireless communication module, or a global navigation satellite system (GNSS) communication module) or a wired communication module (194) (e.g., a local area network (LAN) communication module, or a power line communication module). Among these communication modules, the corresponding communication module can communicate with an external electronic device (104) via a first network (198) (e.g., a short-range communication network such as Bluetooth, wireless fidelity (WiFi) direct, or infrared data association (IrDA)) or a second network (199) (e.g., a long-range communication network such as a legacy cellular network, a 5G network, a next-generation communication network, the Internet, or a computer network (e.g., a LAN or WAN)). These various types of communication modules can be integrated into a single component (e.g., a single chip) or implemented as multiple separate components (e.g., multiple chips). The wireless communication module (192) can verify or authenticate the electronic device (101) within a communication network such as the first network (198) or the second network (199) by using subscriber information (e.g., an international mobile subscriber identity (IMSI)) stored in the subscriber identification module (196).

[0040] The wireless communication module (192) can support 5G networks and next-generation communication technologies following the 4G network, such as NR access technology (new radio access technology). The NR access technology can support high-speed transmission of high-capacity data (eMBB (enhanced mobile broadband)), minimization of terminal power and connection of multiple terminals (mMTC (massive machine type communications)), or high reliability and low latency (URLLC (ultra-reliable and low-latency communications)). The wireless communication module (192) can support, for example, a high-frequency band (e.g., mmWave band) to achieve a high data transmission rate. The wireless communication module (192) can support various technologies for securing performance in a high-frequency band, such as beamforming, massive multiple-input and multiple-output (MIMO), full dimensional MIMO (FD-MIMO), array antenna, analog beam-forming, or large scale antenna. The wireless communication module (192) can support various requirements specified in the electronic device (101), an external electronic device (e.g., the electronic device (104)), or a network system (e.g., the second network (199)). According to one embodiment, the wireless communication module (192) can support a peak data rate (e.g., 20 Gbps or more) for eMBB realization, a loss coverage (e.g., 164 dB or less) for mMTC realization, or a U-plane latency (e.g., 0.5 ms or less for downlink (DL) and uplink (UL), or 1 ms or less for round trip) for URLLC realization.

[0041] The antenna module (197) can transmit or receive signals or power to or from an external device (e.g., an external electronic device). In one embodiment, the antenna module (197) may include an antenna including a radiator formed of a conductor or a conductive pattern formed on a substrate (e.g., a PCB). In one embodiment, the antenna module (197) may include a plurality of antennas (e.g., an array antenna). In this case, at least one antenna suitable for a communication method used in a communication network, such as the first network (198) or the second network (199), may be selected from the plurality of antennas, for example, by the communication module (190). A signal or power may be transmitted or received between the communication module (190) and an external electronic device via the at least one selected antenna. In some embodiments, in addition to the radiator, another component (e.g., a radio frequency integrated circuit (RFIC)) may be additionally formed as a part of the antenna module (197).

[0042] According to various embodiments, the antenna module (197) may form a mmWave antenna module. In one embodiment, the mmWave antenna module may include a printed circuit board, an RFIC disposed on or adjacent a first side (e.g., a bottom side) of the printed circuit board and capable of supporting a designated high-frequency band (e.g., a mmWave band), and a plurality of antennas (e.g., an array antenna) disposed on or adjacent a second side (e.g., a top side or a side side) of the printed circuit board and capable of transmitting or receiving signals in the designated high-frequency band.

[0043] At least some of the above components can be interconnected and exchange signals (e.g., commands or data) with each other via a communication method between peripheral devices (e.g., a bus, GPIO (general purpose input and output), SPI (serial peripheral interface), or MIPI (mobile industry processor interface)).

[0044] According to one embodiment, commands or data may be transmitted or received between the electronic device (101) and an external electronic device (104) via a server (108) connected to a second network (199). Each of the external electronic devices (102 or 104) may be the same or a different type of device as the electronic device (101). According to one embodiment, all or part of the operations executed in the electronic device (101) may be executed in one or more of the external electronic devices (102, 104, or 108). For example, when the electronic device (101) is to perform a certain function or service automatically or in response to a request from a user or another device, the electronic device (101) may, instead of or in addition to executing the function or service itself, request one or more external electronic devices to perform the function or at least a part of the service. One or more external electronic devices that receive the request may execute at least a portion of the requested function or service, or an additional function or service related to the request, and transmit the result of the execution to the electronic device (101). The electronic device (101) may process the result as is or additionally and provide it as at least a portion of a response to the request. For this purpose, cloud computing, distributed computing, mobile edge computing (MEC), or client-server computing technology may be used, for example. The electronic device (101) may provide an ultra-low latency service by using distributed computing or mobile edge computing, for example. In another embodiment, the external electronic device (104) may include an Internet of Things (IoT) device. The server (108) may be an intelligent server utilizing machine learning and / or a neural network. According to one embodiment, the external electronic device (104) or the server (108) may be included in the second network (199).The electronic device (101) can be applied to intelligent services (e.g., smart home, smart city, smart car, or healthcare) based on 5G communication technology and IoT-related technology.

[0045] Electronic devices according to various embodiments disclosed in this document may take various forms. Electronic devices may include, for example, portable communication devices (e.g., smartphones), computer devices, portable multimedia devices, portable medical devices, cameras, wearable devices, or home appliances. Electronic devices according to embodiments of this document are not limited to the aforementioned devices.

[0046] The various embodiments of this document and the terminology used herein are not intended to limit the technical features described in this document to a specific embodiment, but should be understood to include various modifications, equivalents, or substitutes of the embodiment. In connection with the description of the drawings, similar reference numerals may be used for similar or related components. The singular form of a noun corresponding to an item may include one or more of the item, unless the context clearly indicates otherwise. In this document, each of the phrases "A or B", "at least one of A and B", "at least one of A or B", "A, B, or C", "at least one of A, B, and C", and "at least one of A, B, or C" can include any one of the items listed together in the corresponding phrase among those phrases, or all possible combinations thereof. Terms such as "first," "second," or "first" or "second" may be used merely to distinguish one component from another, and do not limit the components in any other respect (e.g., importance or order). When a component (e.g., a first component) is referred to as "coupled" or "connected" to another (e.g., a second component), with or without the terms "functionally" or "communicatively," it means that the component can be connected to the other component directly (e.g., wired), wirelessly, or through a third component.

[0047] The term "module" used in various embodiments of this document may include a unit implemented in hardware, software, or firmware, and may be used interchangeably with terms such as logic, logic block, component, or circuit. A module may be an integral component, or a minimum unit or portion of such a component that performs one or more functions. For example, according to one embodiment, a module may be implemented in the form of an application-specific integrated circuit (ASIC).

[0048] Various embodiments of the present document may be implemented as software (e.g., a program (140)) including one or more instructions stored in a storage medium (e.g., an internal memory (136) or an external memory (138)) readable by a machine (e.g., an electronic device (101)). For example, a processor (e.g., a processor (120)) of the machine (e.g., an electronic device (101)) may call at least one instruction among the one or more instructions stored from the storage medium and execute it. This enables the machine to operate to perform at least one function according to the at least one called instruction. The one or more instructions may include code generated by a compiler or code executable by an interpreter. The machine-readable storage medium may be provided in the form of a non-transitory storage medium. Here, 'non-transitory' simply means that the storage medium is a tangible device and does not contain signals (e.g., electromagnetic waves), and the term does not distinguish between cases where data is stored semi-permanently or temporarily on the storage medium.

[0049] In the following description, an electronic device may include a receiving device that receives a signal (or data) from a transmitting device of a wireless communication system. For example, the electronic device may include a base station that receives a signal (or data) from a terminal. For example, the base station may be at least one of a Node B, a BS (base station), an eNB (eNode B), or a gNB (gNode B). For example, the electronic device may include a terminal that receives a signal (or data) from the base station. For example, the terminal may be at least one of a UE (user equipment), an MS (mobile station), a cellular phone, a smartphone, a computer, or a multimedia system capable of performing a communication function.

[0050] FIG. 2 illustrates a block diagram of an electronic device according to one embodiment of the present disclosure.

[0051] FIG. 2 illustrates a block diagram of an electronic device according to one embodiment of the present disclosure.

[0052] According to one embodiment, the electronic device (101) may include and / or execute a frontend module (210) and / or a service processing module (220). The frontend module (210) and / or the service processing module (220) may be executed by, for example, the processor (120), or may be included as at least a part of the processor (120) or another entity. At least some of the operations performed by the frontend module (210) and / or the service processing module (220) in the present disclosure may be understood to be performed by, for example, the processor (120) and / or another entity under the control of the processor (120).

[0053] The front-end module (210) can perform at least one operation for data exchange with, for example, external electronic devices (106a, 106b, 106c, ..., 106n). For example, the front-end module (210) can provide data that can configure a user interface (UI) that can input user input from the external electronic devices (106a, 106b, 106c, ..., 106n). The front-end module (210) can provide processing for a user request to the service processing module (220). The service processing module (220) can perform a service using the user request and may also be referred to as a back-end module. The service processing module (220) can provide a response corresponding to the user request to the front-end module (210). The front-end module (210) can provide a response received from the service processing module (220) to an external electronic device (106a, 106b, 106c, ..., 106n).

[0054] The service processing module (220) may include, for example, a user request confirmation module (221), an optimization module (222), an AI model management module (224), a policy management module (226), and / or a service execution module (227).

[0055] The user request verification module (221) can verify information associated with a user request provided from an external electronic device (106a, 106b, 106c, ..., 106n). The information associated with the user request can be expressed as, for example, the number of user requests over a certain period of time and / or the size of the user request, but there is no limitation thereto. For example, the user request verification module (221) can count the number of user requests provided from the external electronic device (106a, 106b, 106c, ..., 106n) and / or monitor the size of the content included in the user request (for example, text and / or graphic objects, but there is no limitation thereto), but there is no limitation on the type of information associated with the user request and / or the verification method.

[0056] The optimization module (222) can provide at least one optimal number of instances for each of at least one AI model linked to the service.

[0057] The AI ​​model management module (224) can store and / or manage (e.g., including but not limited to, adding, deleting, and / or updating) AI models associated with a service.

[0058] The policy management module (226) may, for example, store and / or manage (e.g., including but not limited to, adding, deleting, and / or updating) acceptable response times. For example, the management device (104) may verify (e.g., receive or determine) at least one input for determining an acceptable response time and provide it to the electronic device (101). For example, an administrator may input information regarding an acceptable response time for a given service into the management device (104), but this is exemplary and there is no limitation on the manner in which the acceptable response time is verified.

[0059] The service execution module (227) may execute, for example, at least one instance group (231, 232, 233) corresponding to at least one AI model linked to the service. The service execution module (227) may execute, for example, at least one instance group (231, 232, 233) corresponding to at least one AI model according to the optimal number of instances provided by the optimization module (222). The service execution module (227) may process a user request based on the executed at least one instance group (231, 232, 233) and provide a response according to the processing result to an external electronic device (106a, 106b, 106c, ..., 106n) through the front-end module (210). According to one embodiment, when there are multiple instances being executed, the user request may be distributedly processed by the multiple instances.

[0060] For decoding, a neural network based on the sum-product algorithm can be used. This constructs a DNN neural network based on the parity check matrix of error-correcting codes, and layer-by-layer operations in the neural network follow the formula of the sum-product algorithm. For example, a syndrome test is performed on even-numbered layers, and decoding is terminated if the syndrome test passes.

[0061] In an error correction code transformer (ECCT), a mask can be generated based on the parity check matrix of the error correction code and utilized for self-attention of the transformer. The transformer estimates noise present in the input signal and determines the transmitted codeword by removing the noise present in the input signal.

[0062] In decoding of error correction codes based on reverse diffusion, decoding can be performed by reducing noise in an input signal as much as possible by repeatedly performing the error correction code transformer operation.

[0063] The sum-and-multiply algorithm described above can achieve a certain level of error correction performance only if it performs a minimum number of parity iterations. Their neural networks must have a number of layers equal to twice the number of iterations. Because they are very deep neural networks with tens to hundreds of hidden layers, they are complex to train and have similar or greater latency than existing algorithms. Furthermore, because existing technologies operate based on existing iterative decoding algorithms, they have the disadvantage of being difficult to vary the number of nodes within the neural network to improve its inference performance. A deeper structure can be used.

[0064] The transformer architecture based on the error-correcting code transformer and decoding of error-correcting codes based on inverse diffusion described above offers greater structural flexibility compared to decoders using neural networks based on the sum-product algorithm. Because the dimensions and number of nodes of the transformer can be freely varied, the size of the neural network can be increased to improve performance. However, as the size of the AI ​​(artificial intelligence) model used increases, the number of parameters increases proportionally. Furthermore, parameters have different values ​​depending on the code type, length, and code rate. Therefore, various communication systems require storing all parameters corresponding to the various code types, lengths, and code rates. Increasing the decoder size to achieve superior error-correcting performance and thus the number of parameters, can lead to a proportional increase in storage space, which can increase overhead.

[0065] This disclosure proposes an ensemble neural network decoder with improved error correction performance. This decoder adds scalable noisy embedding to the neural network decoder, and determines the final decoding result based on the decoding results of each neural network decoder. The proposed decoder architecture allows for more flexible modification of the neural network structure and offers improved error correction performance. Furthermore, the proposed decoder achieves improved performance with a small neural network size. Therefore, the proposed decoder can utilize storage memory resources more efficiently with a small neural network size, and can perform learning with less latency and resources when real-time learning is required. Furthermore, the scalable noisy embedding can create a more diverse training data set during real-time learning, thereby improving real-time learning performance.

[0066] FIG. 3 illustrates a block diagram of a decoder according to one embodiment of the present disclosure.

[0067] Referring to FIG. 3, the decoder proposed in the present disclosure may include a processing unit (300), a storage unit (310), and a training unit (320).

[0068] In one embodiment, the processing unit (300) may include at least one scalable noisy embedding layer and at least one neural network-based decoder. An example of a detailed structure of the processing unit (300) is described in FIG. 4. In one embodiment, each decoder may have the same structure or different structures. In one embodiment, the parameter values ​​constituting the neural network of the at least one neural network-based decoder may be fixed and used, and may be set by receiving model parameters (350) from the storage unit (310). In one embodiment, the parameter values ​​may include at least one of a weight and a bias value for a model constituting the neural network. In one embodiment, the parameters applied to each decoder may have the same value or different values.

[0069] In one embodiment, the processing unit (300) may input a modulated symbol with noise (330) and output a code word (335). In one embodiment, the decoder may remove a CRC (cyclic redundancy check) portion from the output code word (335) and perform an operation (CRC detachment and codeblock desegmentation) (305) of reconnecting the divided code blocks.

[0070] In one embodiment, if the decoder succeeds in decoding, the decoder may transfer training data (340) including input signals with noisy embedding and the results of the neural network decoder to the storage unit (310) for real-time learning. In one embodiment, if real-time learning is not performed, this process may be omitted.

[0071] In one embodiment, the storage unit (310) may include a memory device that stores parameters to be applied to the decoder of the processing unit (300). In one embodiment, the memory may store one or more parameter sets, and may transmit a corresponding parameter set to the processing unit (300) included in the decoder according to the input of a control signal. In one embodiment, the storage unit (310) may include a memory device that stores a learning data set for real-time learning. In one embodiment, the storage unit (310) may transmit the learning data set to the learning unit when a certain amount of data or more is collected. In one embodiment, when the decoder does not perform real-time learning, the memory device that stores the learning data set and the transmission process may be omitted. In one embodiment, the storage unit (310) may store code information for a modulated symbol with noise (330). In one embodiment, the code information may include a code length or a code rate for the input signal. In one embodiment, the code information may include modulation information.

[0072] The learning unit (320) is configured as a device for real-time parameter learning and can perform learning by receiving learning data (340) from the storage unit (310). In one embodiment, the learning unit (320) can transfer the learned model parameters to the storage unit (310) when learning is completed. In one embodiment, if the decoder does not use real-time learning, the learning unit (320) can be excluded from the structure.

[0073] FIG. 4 illustrates the structure of a neural network-based decoder with scalable noisy embedding applied according to one embodiment of the present disclosure.

[0074] FIG. 4 is a drawing showing an example of a detailed structure of a processing unit (300) illustrated in FIG. 3, according to one embodiment of the present disclosure.

[0075] Referring to FIG. 4, the processing unit (300) may include a scalable noisy embedding unit (400), a neural decoder (410), and a scaling factor calculation unit (420).

[0076] In one embodiment, the noisy embedding performed in the scalable noisy embedding unit (400) can be used to increase the size of the data set by changing the data set during learning and to prevent overfitting. The noisy embedding can be deactivated during inference (e.g., decoding), but in the present disclosure, the noisy embedding can be activated. Noise can be considered as an element that must be removed because it has a negative effect on inference, but virtually generated noise can be used as a way to improve decoding performance.

[0077] In one embodiment, the neural decoder (410) may include at least one of a deep neural network (DNN), a convolutional neural network (CNN), a recurrent neural network (RNN), or a transformer.

[0078] In one embodiment, a noisy input signal (430) may be input to at least one of a scalable noisy embedding unit (400) and a scaling factor calculation unit (420). In one embodiment, the scalable noisy embedding unit (400) may perform noisy embedding on the noisy input signal (430). In one embodiment, the scalable noisy embedding unit (400) may perform noisy embedding by applying a scaling factor (435) calculated and output by a scaling factor calculation unit (420) to the noisy input signal (430).

[0079] In one embodiment, the scaling factor calculation unit (420) can calculate the scaling factor (435) by estimating noise present in the input signal through the output of the neural decoder (410). In one embodiment, the scaling factor calculation unit (420) can calculate and derive the scaling factor (435) using mathematical expression 2 described below. In one embodiment, the output signal to which noisy embedding is applied in the scalable noise embedding unit (400) can be input to the neural decoder (410), decoded, and then output as a result the decoded output signal (440).

[0080] FIG. 5 illustrates the structure of a scalable noisy embedding unit included in a decoder according to one embodiment of the present disclosure.

[0081] FIG. 5 is a diagram illustrating an example of a detailed structure of a scalable noise embedding unit (400) illustrated in FIG. 4 according to one embodiment of the present disclosure.

[0082] In one embodiment, the scalable noise embedding unit (400) may include a random noise generation unit (500). In one embodiment, the scalable noise embedding unit (400) may receive a scaling factor (510) as input and adjust the size of the generated noise. Mathematical expression 1 below shows an example of a random noise generation method.

[0083] [Mathematical Formula 1]

[0084]

[0085] In one embodiment, the above represents a scaling factor (510), and GN is a complex random vector whose real and imaginary parts may have a standard normal distribution. In one embodiment, a randomly generated noise vector N (520) has a real and imaginary part with a mean of 0 and a standard deviation of can have a normal distribution.

[0086] In one embodiment, the length of the generated noise vector N (520) may be the same as the length of the input (IN) signal (530) of the scalable noise embedding unit (400). Accordingly, the generated noise (e.g., Gaussian random noise)) vector N (520) and the input (IN) signal (530) may be added 1:1 to derive the output (OUT) signal (540). (OUT = IN + N).

[0087] For example, the vector of the input (IN) signal (530) In this case, the noise vector N (520) that can be generated is a vector of the same length as the vector of the input (IN) signal (530). It can be randomly generated as, and the generated noise vectors N0, N1, N2, ... have a mean of 0 and a standard deviation of It can form a normal distribution.

[0088] As explained above, the reason why noise size adjustment is necessary is because the added noise size needs to be adjusted according to the noise size in the original input signal (530). If the noise added through noisy embedding is too small compared to the noise present in the input signal, it will be ineffective, and if it is too large, the signal-to-noise ratio (SNR) of the signal will be too low, which can have a negative effect on decoding. Therefore, appropriate size adjustment according to the noise in the input signal is necessary.

[0089] To this end, the present disclosure proposes a method for calculating a scaling factor (510) by estimating noise present in an input signal through the output of a neural decoder. In one embodiment, calculating the scaling factor (510) described below may be performed in the scaling factor calculation unit (420) illustrated in FIG. 4.

[0090] Mathematical expression 2 below is the input signal y and the decoded codeword. An example of a formula for estimating noise in an input signal using is shown.

[0091] [Equation 2]

[0092]

[0093] In one embodiment, the scaling factor calculation unit (420) calculates the average power of noise in the input signal based on the noise estimated from the above mathematical expression 2. ) can be calculated. In one embodiment, a scaling factor can be calculated so that the power of noise generated by noisy embedding is smaller than the power of noise in the input signal. Mathematical expression 3 shows an example of a formula for calculating the scaling factor.

[0094] [Equation 3]

[0095]

[0096] In one embodiment, the above D may refer to a value for generating the size of noise applied through noise embedding as a real number greater than 1 smaller than the estimated noise in the original signal. In one embodiment, the value of D may be determined based on the performance of the decoder. For example, the value of D may be determined as a value applied when the decoding performance is the best while experimentally changing the value of D and testing the decoding performance. For example, the value of D may be an exponent of 2 (2) for the convenience of implementation. n ) can be used. If expressed as an exponent of 2, the division operation can be implemented as a shift operation.

[0097] In one embodiment, if the receiver has a signal-noise ratio (SNR) estimator, the noise magnitude information estimated by the SNR estimator can be utilized. In this case, the noise estimation using the codeword described above can be omitted.

[0098] In one embodiment, the scaling factor (510) may be directly input as a fixed value. In this case, the noise estimation and scaling factor calculation operations described above may be omitted.

[0099] FIG. 6 illustrates the structure of a decoder including multiple sub-neural network-based decoders according to one embodiment of the present disclosure.

[0100] FIG. 6 illustrates an example of a detailed structure of a processing unit (300) illustrated in FIG. 3, according to one embodiment of the present disclosure, and illustrates a structure including a plurality (L) of sub-neural network-based decoders.

[0101] Referring to FIG. 6, the processing unit (300) may include at least one of L scalable noisy embedding units (600, ..., 605), a scaling factor calculation unit (610), L neural decoders (620, ..., 625), L parity check units (630, ..., 635), and a path selection unit (640).

[0102] In one embodiment, a noisy input signal (650) may be input to L scalable noisy embedding units (600, ..., 605). In one embodiment, the operations leading to each scalable noisy embedding unit (600, ..., 605), a neural decoder (620, ..., 625), and a parity check unit (630, ..., 635) may be defined as one path, and the index of each path may be represented as i.

[0103] In one embodiment, a noise vector N generated from a scalable noisy embedding part (600, ..., 605) of path index i i may have the same length as the length of the input signal (650). For example, the vector of the input (IN) signal (650) If , the noise vector N that can be generated i is a vector of the same length as the vector of the input (IN) signal (650). It can be randomly generated as, and the generated noise vectors N0, N1, N2, ... have a mean of 0 and a standard deviation of can have a normal distribution.

[0104] In one embodiment, the scalable noisy embedding part of the i-th path among L scalable noisy embedding parts (600, ..., 605) i There are L scaling factorsi Scaling factor among (i=0, ..., L-1)(673, ..., 675) i A scaled noise can be applied to the input signal based on the value. The scaling factor applied to the scalable noise embedding unit (600, ..., 605) of the above path index i i (673, ..., 675) can be applied as a value obtained by multiplying the path index (path index, i) by the scaling factor (670) value, which is the output value of the scaling factor calculation unit (610).

[0105] In one embodiment, for a path with a path index of 0 (660), the scaling factor (670) is multiplied by 0, so the scaling factor 0 (673) Since it becomes 0, it is as if noisy embedding is not applied, so the scaling noisy embedding part (600) of the corresponding path can be omitted.

[0106] In one embodiment, a neural decoder (620, ..., 625) may input a noisy embedded signal and output a decoded result. Thereafter, the decoder may perform a parity check in a parity check unit (630, ..., 635) to confirm whether the decoding has been performed correctly. Mathematical equation 4 below is an example of a formula for performing a parity check.

[0107] [Equation 4]

[0108]

[0109] In one embodiment, the syndrome is the parity check result in the above mathematical expression 4. If the vector is 0, it can be determined that the decoding is successful. In one embodiment, if the parity check result in the parity check unit (630, ..., 635) is that all paths have failed to decode (i.e., ), the scalable noise embedding unit (600, ..., 605) can be used again. In one embodiment, at this time, the noise applied to the noisy embedding in the scalable noise embedding unit (600, ..., 605) is generated again and may have a different pattern than before. In one embodiment, the decoder may repeat this process a predefined number of times ( ) can be repeated as many times as (690).

[0110] In one embodiment, if the parity check result in the parity check unit (630, ..., 635) shows that one or more paths have been successfully decoded, the decoder may transmit the successful decoding result to the path selection unit (640) and terminate the iteration. In one embodiment, the path with the path index of '0' is not affected by the noisy embedding, so the result is not affected even if it is repeatedly performed. Accordingly, in one embodiment, the decoder may not perform the noisy embedding of the path index 0 from the second iteration.

[0111] In one embodiment, the path selection unit (640) may perform a CRC (cyclic redundancy check) check based on the decoding results received from each path. In one embodiment, if there are multiple results that pass the CRC check, the path selection unit (640) may select and output the decoding result of the path with the smallest path index (i.e., the path with the highest reception SNR).

[0112] In one embodiment, the scaling factor calculation unit (610) calculates the input signal (650) and the decoded codeword as described above. (680) can be used to estimate the noise in the input signal and calculate the scaling factor.

[0113] In one embodiment, the scaling factor calculation unit (610) may also consider the number of paths (=L) (667) when calculating the scaling factor. Mathematical expression 5 below shows an example of a formula for calculating the scaling factor by the scaling factor calculation unit (610).

[0114] [Equation 5]

[0115]

[0116] Referring to mathematical expression 5, the scaling factor is inversely proportional to the number of paths (L), and as the number of paths (L) increases, the output value of the scaling factor calculation unit (610) decreases. In one embodiment, the above D may refer to a value for generating the size of noise applied through noise embedding to be smaller than the estimated noise in the original signal as a real number greater than 1. In one embodiment, the value of D may be determined based on the performance of the decoder. For example, the value of D may be determined as a value applied when the decoding performance is the best while experimentally changing the value of D and testing the decoding performance. For example, the value of D may be determined as an exponent of 2 (2) for the convenience of implementation. n ) can be used. If expressed as an exponent of 2, the division operation can be implemented as a shift operation. In one embodiment, the value of D can be set to 4 or 8 considering performance. In one embodiment, when determining the value of D, it can be considered that when the value of D is set to 16, the change in the input is small, so there is a tendency for the decoders of all paths to fail to decode all of them. In one embodiment, on the contrary, when D is set to 2, the size of the noise applied to paths with large path indices increases, so that the decoders of the corresponding paths do not succeed in decoding, and thus it may be difficult to see the performance gain due to the increase in the number of paths.

[0117] In one embodiment, referring to FIG. 6, since L neural decoders (620, ..., 625) all operate independently, parallelization is possible, so that there may be no increase in latency due to an increase in the number of paths. In one embodiment, a fixed-value scaling factor may be used without calculating the scaling factor in the scaling factor calculation unit (610). In one embodiment, even if a fixed-value scaling factor is used, a value obtained by multiplying the scaling factor by the path index becomes the input of each scalable noise embedding unit (600, ..., 605) of the path index i, so that noise of different sizes may be embedded in each path.

[0118] FIG. 7 is a diagram illustrating a real-time learning operation for the output of a scalable noisy embedding included in a decoder of a decoder according to one embodiment of the present disclosure and a decoding result.

[0119] Figure 7 illustrates a detailed structure of a decoder for real-time learning. The decoder may include multiple (L) sub-neural network-based decoders and a storage unit capable of storing learning data.

[0120] Referring to FIG. 7, the decoder may include at least one of L scalable noisy embedding units (700, ..., 705), a scaling factor calculation unit (710), L neural decoders (720, ..., 725), L parity check units (730, ..., 735), a path selection unit (740), and a storage unit (790).

[0121] In one embodiment, when the decoder performs real-time learning, the decoding result (780) and input signals for the paths that have passed through the CRC in the path selection step of the path selection unit (740) can be transferred to the storage unit (790) and stored. At this time, the input signal corresponding to the stored L data (785) can include L signals that have passed through the scalable noise embedding unit (700, ..., 705). In this case, not only the signals that have succeeded in decoding among the L signals but also the signals that have failed in decoding are included, so that various learning data can be secured. In one embodiment, the stored input signal can only store the original signal (750) that has not passed through the scalable noise embedding unit (700, ..., 705).

[0122] In one embodiment, the decoder can generate a learning data set from stored input signals and decoding results, and perform model parameter learning in the learning unit according to a control signal. The learned parameters can be transferred to the storage unit (790), so that the model parameters stored in the storage unit (790) can be updated with the learned parameters. In one embodiment, the learning and parameter update process is performed according to the control signal and can be performed periodically or aperiodically.

[0123] In one embodiment, the training data may include a modulation symbol of a codeword, where the modulation may include a modulation method such as QPSK, 16QAM, 256QAM, and 1024QAM. The label (785) stored in the storage unit (790) may vary depending on the target that each neural decoder (620, ..., 625) intends to estimate. For example, when estimating a codeword, the label (785) may be the codeword, and when estimating noise, the label (785) may be a noise value. In one embodiment, the loss function, which is the error between the label (785) and the estimation result, may be in the form of a mean squared error (MSE) or a cross entropy function. In one embodiment, the learning may find model parameters that minimize the loss function using a gradient descent method. In one embodiment, the model parameters may include at least one of weight and bias values ​​for a model constituting a neural network. In one embodiment, a scalable noisy embedding unit (700, ..., 705) may be applied or omitted during learning.

[0124] FIG. 8 is a flowchart illustrating the operation of a decoder according to one embodiment of the present disclosure.

[0125] The decoder can perform at least one operation as illustrated in FIG. 8, and in one embodiment, can perform some operations partially by omitting some operations.

[0126] In step 800, the decoder can obtain code information. In step 805, the decoder can set the parameters of the neural network decoder included in the decoder. In step 810, the decoder can set the value of the variable I to be used when repeatedly performing noisy embedding to '1'. In step 815, the decoder can perform scalable noisy embedding on the input signal.

[0127] In step 820, the decoder can decode the codeword using a neural network decoder. In step 825, the decoder can perform a parity check to verify the validity of the decoded codeword. In step 830, the decoder can determine whether the number of valid codewords confirmed through the parity check is greater than 0. In one embodiment, if the number of valid codewords is 0, the decoder determines in step 835 that the value of the variable I is I MAX It can be identified as the same.

[0128] In one embodiment, the value of the parameter I is I MAX If not, I+1 can be substituted for the value of the above parameter I. Afterwards, the decoder can proceed to step 815 and perform the operation described above.

[0129] In step 830, if the decoder determines that the number of valid codewords confirmed through the parity check is greater than 0, the decoder may select one codeword from among the valid results in step 850. In one embodiment, if there are multiple valid codewords, the decoder may select the codeword corresponding to the lowest path index. Thereafter, in step 860, the decoder may estimate noise from the valid results. In step 865, the decoder may calculate a scaling factor based on the noise estimate. In step 870, the decoder may perform a calculation of multiplying the scaling factor by the path index. In one embodiment, the decoder may reflect the calculation result of multiplying the scaling factor by the path index in the scaling factor applied to the scalable noise embedding in step 815. In one embodiment, if the scaling factor is input externally or is a predetermined value, the feedback path (steps 860, 865, and 870) for obtaining the scaling factor may be omitted.

[0130] Below, the results of measuring the performance of the method proposed in this disclosure are described.

[0131] Table 1 shows the bit error rate (BER) performance as the number of neural decoders applied to the method proposed in this disclosure increases. Simulations were conducted at signal-to-noise ratios (SNRs) of 4, 5, and 6 dB, and the neural decoder used an error correction code transformer (ECCT) decoder. It can be seen that BER performance gradually improves as the number of neural decoders increases.

[0132] BER performance of the proposed decoder in the disclosure of the code neural network decoder 4 dB 5 dB 6 dB Polar (64, 48) 11.73.E-03 2.12.E-04 1.53.E-05 21.70.E-031.98.E-04 1.45.E-05 41.67.E-031.87.E-04 1.36.E-05 81.59.E-031.75.E-04 1.20.E-05 161.44.E-031.51.E-04 9.62.E-06 321.15.E-031.03.E-04 5.11.E-06

[0133] BER performance change according to the number of paths

[0134] Table 2 below compares the BER performance of the decoder proposed in this disclosure with that of a conventional ECCT decoder. Simulations were performed at SNRs of 4, 5, and 6 dB, and the ECCT decoder was configured with 32 ECCTs with noisy embedding. A parity check was performed based on the output results of each noisy embedded ECCT, and the decoding result with the smallest path index among the paths that passed the parity check was selected as the final result. Table 2 confirms that the proposed method can improve performance.

[0135] BER performance of CodeECCT decoder BER performance of decoder proposed in this disclosure 4 dB5 dB6 dB4 dB5 dB6 dBPolar(64, 32)9.21.E-047.95.E-054.46.E-066.54.E-045.91.E-051.91.E-06Polar(64, 48)1.73.E-032.12.E-041.53.E-051.15.E-031.03.E-045.11.E-06LDPC(49, 24)2.18.E-031.65.E-045.56.E-061.30.E-038.91.E-051.69.E-06BCH(63, 36)7.75.E-031.29.E-031.12.E-046.61.E-038.29.E-044.18.E-05BCH(63, 45)3.70.E-034.14.E-041.79.E-053.28.E-033.07.E-041.01.E-05BCH(63, 51)3.48.E-033.74.E-041.65.E-052.61.E-032.36.E-046.64.E-06

[0136] Performance comparison with ECCT decoder

[0137] This same effect can be applied to other AI-based decoders, achieving the same performance improvements. Table 3 below compares BER performance when applying a decoder based on a diffusion model with a transformer structure.

[0138] BER performance of a decoder based on a diffusion model with a code transformer structure BER performance of the decoder proposed in this disclosure 4 dB5 dB6 dB4 dB5 dB6 dBPolar(64, 32) 1.82.E-0 37.4 1.E-0 5 2.79.E-0 6 1.62.E-0 4 7.62.E-0 6 8.60.E-0 7Polar(64, 48) 2.48.E-0 32.64.E-0 4 1.70.E-0 5 7.81.E-0 4 1.86.E-0 5 6.10.E-0 6LDPC(49, 24)2.24.E-031.60.E-046.66.E-069.09.E-049.96.E-061.23.E-06BCH(63, 36)5.57.E-038.25.E-044.73.E-051.84.E-031.02.E-047.75.E-06BCH(63, 45)3.66.E-033.56.E-041.17.E-059.77.E-047.29.E-052.23.E-07BCH(63, 51)4.84.E-036.82.E-043.99.E-051.90.E-031.69.E-041.68.E-06

[0139] Performance comparison with a decoder based on a diffusion model with a transformer structure.

[0140] While AI-based decoders can improve decoding performance by increasing the size of their neural networks, this also increases the number of parameters they comprise, placing a burden on latency and resources, both of which are crucial factors in real-time learning. The proposed decoder architecture allows for the application of the same neural network to multiple AI-based decoders, thereby improving decoding performance even with relatively small neural networks without increasing the size of the neural network, making it advantageous for real-time learning.

[0141] This is also advantageous from the perspective of storing model parameters for neural networks. Various communication systems, including New Radio (NR), use channel coding with various code lengths and code rates. In the case of a neural network-based decoder, memory space is required to store the parameters because different optimal parameters must be applied depending on the code length and code rate. In the present disclosure, multiple neural network decoders having the same structure and parameters can be used, and in this case, while achieving superior performance compared to existing inventions with the same number of parameters, the number of parameters that must be stored may be reduced compared to existing invention structures with similar performance.

[0142] Table 4 shows the number of parameters compared to an ECCT decoder with similar performance. The ECCT decoder has 6 transformer layers, while the number of transformer layers in the present disclosure is 2. For example, the performance of the noisy embedding-based decoder proposed in the present disclosure with 84,129 model parameters is similar to that of the ECCT decoder with 1,205,697 model parameters for the Polar (64, 48) code. Therefore, the noisy embedding-based decoder proposed in the present disclosure can achieve similar performance with about 7% of the memory required for parameter storage compared to the conventional one for the Polar (64, 48) code.

[0143] Code① Number of parameters (number) that need to be stored in the ECCT decoder② Number of parameters (number) that need to be stored in the decoder proposed in this disclosure② / ① (%)Polar(64, 48)1,205,69784,1296.98%LDPC (49, 24)1,203,95182,6716.87%BCH (63, 45)1,205,80784,1436.98%BCH (63, 51)1,204,66183,5736.94%

[0144] Comparison of the number of parameters required to store for similar performance

[0145] The scalable noisy embedding layer proposed in this disclosure is effective in solving the problem of generalization performance degradation due to overfitting by increasing the training data set during real-time learning, and also allows for rapid acquisition of a training data set. In the case of decoders that do not include a scalable noisy embedding layer, only one data set can be generated from a single received signal. The training data set generated in this way consists only of decodable cases. Retraining using already decodable training data does not improve decoding performance. Furthermore, these data are close to the minimum point in terms of loss. When the gradient value is close to the minimum point of loss, the amount of parameter change during training is very small, making them inefficient as training data.

[0146] On the other hand, the scalable noisy embedding layer proposed in this disclosure can create various types of received signals by adding noise of various sizes to the received signal during the decoding process. For example, in the case of 32 paths, 32 data can be created from a single received signal. Among these, there are cases where decoding is successful and cases where decoding fails. The decoder can improve its decoding performance as it progresses through learning even in cases where decoding fails during real-time learning.

[0147] For example, if the batch size is 128, at least 128 training data are required for one training, but if the scalable noisy embedding layer is not applied, at least 128 receiver operations are required to secure training data, but in the case of a decoder including 32 paths due to the multiple noisy embedding layers proposed in the present disclosure, 128 training data can be secured with only at least 4 receiver operations.

[0148] In one embodiment, the device proposed in this disclosure can be used to decode linear block codes. The device proposed in this disclosure may include a processing unit that receives a signal containing noise and errors as input and outputs a codeword, a storage unit that stores the model structure and parameters of the decoder, and a learning unit for real-time parameter learning.

[0149] In one embodiment, the processing unit may include a scalable noisy embedding layer, a plurality of neural network-based decoders, a selector for selecting one of the outputs of the decoders, and a scaling factor calculation unit.

[0150] In one embodiment, the scalable noisy embedding layer can add noise of adjustable size to the input signal. In one embodiment, the neural network-based decoder can include at least one of a convolutional neural network (CNN) layer, a fully connected neural network (FC) layer, a recurrent neural network (RNN) layer, or a transformer layer. In one embodiment, the scaling factor calculator can perform the operations of estimating noise in the input signal based on the decoding result, calculating the average power of the estimated noise, and calculating the scaling factor. In one embodiment, the processing unit can perform the operation of repeating a series of processes when the decoding is not successful, and the operation of terminating the iteration when the decoding is successful.

[0151] In one embodiment, the storage unit may include a memory for storing the model structure and parameters and a memory for storing a training data set. In one embodiment, the learning unit may receive the training data set as input and learn model parameters. In one embodiment, if decoding is successful for real-time learning, the processing unit may store the output signals of the scalable noisy embedding layer and the codeword resulting from the decoding in the storage unit.

[0152] According to one embodiment, the method according to one embodiment disclosed in this document may be provided as included in a computer program product. The computer program product may be traded as a product between a seller and a buyer. The computer program product may be distributed in the form of a machine-readable storage medium (e.g., compact disc read-only memory (CD-ROM)), or may be distributed online (e.g., downloaded or uploaded) via an application store (e.g., Play Store™) or directly between two user devices (e.g., smart phones). In the case of online distribution, at least a portion of the computer program product may be temporarily stored or temporarily generated in a machine-readable storage medium, such as the memory of a manufacturer's server, an application store's server, or an intermediary server.

[0153] According to one embodiment, each component (e.g., a module or a program) of the above-described components may include one or more entities, and some of the entities may be separated and arranged in other components. According to one embodiment, one or more components or operations of the aforementioned components may be omitted, or one or more other components or operations may be added. Alternatively or additionally, a plurality of components (e.g., a module or a program) may be integrated into a single component. In this case, the integrated component may perform one or more functions of each of the plurality of components identically or similarly to those performed by the corresponding component among the plurality of components prior to the integration. According to one embodiment, the operations performed by a module, program, or other component may be executed sequentially, in parallel, iteratively, or heuristically, or one or more of the operations may be executed in a different order, omitted, or one or more other operations may be added.

Claims

1. In electronic devices, At least one processor (120); and Contains a memory (130) storing at least one instruction, wherein said at least one instruction, when executed by at least a part of said at least one processor (120), causes said electronic device to perform at least one operation; At least one of the above actions: An act of receiving a first signal with noise; An operation that generates at least one random noise based on a scaling factor; An operation of generating at least one second signal by applying at least one random noise to the first signal; and An electronic device characterized by comprising: an operation of performing at least one neural network decoding on at least one second signal to output a codeword.

2. In paragraph 1, An electronic device, characterized in that the scaling factor is determined based on the first signal and the codeword.

3. In paragraph 1, An operation of estimating noise of the first signal based on the output code word; An operation of calculating the average power of the above estimated noise; and An electronic device characterized by comprising an operation of determining the scaling factor based on the average power.

4. In the first paragraph, the operation of performing at least one neural network decoding on at least one second signal to output a codeword; An operation of performing a parity check on at least one codeword output by performing at least one neural network decoding to determine at least one valid codeword; and An electronic device characterized by comprising: an operation of selecting one valid codeword when there are two or more valid codewords; 5. In the first paragraph, if the neural network decoding for at least one second signal fails: An operation of generating at least one random noise based on the above scaling factor; An operation of generating at least one third signal by applying at least one random noise to the first signal; and An electronic device characterized by comprising: an operation of performing at least one neural network decoding on at least one third signal to output a codeword.

6. In paragraph 5, Decoding of the I+1 signal, which generates and applies random noise I times to the above first signal, fails, and I is a predetermined I max An electronic device characterized by comprising: an operation for terminating decryption; 7. In paragraph 1, An operation of storing the code word and the at least one second signal; and An electronic device characterized by comprising: an operation for determining model parameters for a neural network decoder based on the stored codeword and the at least one second signal; 8. In a method of a decryption device in a wireless communication system, An act of receiving a first signal with noise; An operation that generates at least one random noise based on a scaling factor; An operation of generating at least one second signal by applying at least one random noise to the first signal; and A method characterized by comprising: an operation of performing at least one neural network decoding on at least one second signal to output a codeword.

9. In paragraph 8, A method characterized in that the scaling factor is determined based on the first signal and the codeword.

10. In paragraph 8, An operation of estimating noise of the first signal based on the output code word; An operation of calculating the average power of the above estimated noise; and A method characterized by comprising: an operation of determining the scaling factor based on the average power; 11. In the 8th paragraph, the operation of performing at least one neural network decoding on at least one second signal to output a codeword; An operation of performing a parity check on at least one codeword output by performing at least one neural network decoding to determine at least one valid codeword; and A method characterized by comprising: an operation of selecting one valid codeword when there are two or more valid codewords.

12. In the 8th paragraph, if the neural network decoding for at least one second signal fails: An operation of generating at least one random noise based on the above scaling factor; An operation of generating at least one third signal by applying at least one random noise to the first signal; and A method characterized by comprising: an operation of performing at least one neural network decoding on at least one third signal to output a codeword.

13. In paragraph 12, Decoding of the I+1 signal, which generates and applies random noise I times to the above first signal, fails, and I is a predetermined I max A method characterized by comprising: an action of terminating decryption; 14. In a storage medium storing at least one computer-readable instruction, the at least one instruction, when executed by at least a part of at least one processor (120) of an electronic device, causes the electronic device to perform at least one operation, At least one of the above actions: An act of receiving a first signal with noise; An operation that generates at least one random noise based on a scaling factor; An operation of generating at least one second signal by applying at least one random noise to the first signal; and A storage medium characterized by comprising an operation of performing at least one neural network decoding on at least one second signal to output a codeword.

15. In paragraph 14, A storage medium, characterized in that the scaling factor is determined based on the first signal and the codeword.

Images