Channel-free deep learning-based synchronization method and device

Abstract

The present disclosure may provide a channel-free deep learning-based synchronization technology. According to the present disclosure, a method of a first communication node may be provided, the method comprising the steps of: generating a first base sequence by using a Zadoff-Chu sequence; generating a second base sequence by using the first base sequence; and generating a synchronization signal sequence by performing a half / half distributed concatenation of the first base sequence and the second base sequence.

Description

Channel-Free Deep Learning-Based Synchronization Method and Device

[0001] The present disclosure relates to a channel-free deep learning-based synchronization technology, and more particularly, to a channel-free deep learning-based synchronization technology that can provide accurate timing by combining synchronization based on signal design and synchronization based on deep learning.

[0002] Advances in information and communication technology (ICT) can lead to the development of various wireless communication technologies. Representative wireless communication technologies include LTE (long term evolution), NR (new radio), and 6G (6th Generation), all of which are defined by the 3rd Generation Partnership Project (3GPP) standards. LTE can be one of the 4th Generation (4G) wireless communication technologies, and NR can be one of the 5th Generation (5G) wireless communication technologies.

[0003] In order to process the rapidly increasing amount of wireless data following the commercialization of 4G communication systems (e.g., communication systems supporting LTE), 5G communication systems (e.g., communication systems supporting NR) that use a higher frequency band (e.g., a frequency band higher than 6 GHz) than the frequency band of the 4G communication system (e.g., a frequency band below 6 GHz) may be considered. 5G communication systems may support enhanced Mobile Broadband (eMBB), Ultra-Reliable and Low Latency Communication (URLLC), and massive Machine Type Communication (mMTC).

[0004] Communication systems may use synchronization techniques based on signal design for physical layer synchronization. Such synchronization techniques may struggle to support devices utilizing artificial intelligence, which will become widespread in the near future.

[0005] The purpose of the present disclosure to solve the above problems is to provide a channel-free deep learning-based synchronization method and device that can provide accurate timing by combining synchronization based on signal design and synchronization based on deep learning.

[0006] A channel-free deep learning synchronization method according to a first embodiment of the present disclosure for achieving the above object may include, as a method of a first communication node, a step of generating a first base sequence using a Zadoff-Chu sequence; a step of generating a second base sequence using the first base sequence; and a step of generating a synchronization signal sequence by performing half / half distributed concatenation of the first base sequence and the second base sequence.

[0007] Here, the second base sequence may be a sequence having an opposite polarity to the first base sequence.

[0008] Here, the step of mapping modulation symbols generated by modulating the above-mentioned synchronization signal sequence to 2(M+1) subcarriers; and the step of transmitting a synchronization signal composed of the mapped modulation symbols to a second communication node, wherein M may be a natural number.

[0009] Here, the step of mapping modulation symbols generated by modulating the synchronization signal sequence to 2(M+1) subcarriers may include the step of sequentially mapping the first base sequence constituting the synchronization signal sequence in the frequency domain to odd subcarrier indices of the 2(M+1) subcarriers; and the step of sequentially mapping the second base sequence constituting the synchronization signal sequence to even subcarrier indices of the 2(M+1) subcarriers.

[0010] Here, at least one null subcarrier may be placed around the (2M+1) subcarriers or between the 2(M+1) subcarriers.

[0011] Here, the step of generating a synchronization signal sequence by performing half / half distributed concatenation of the first base sequence and the second base sequence may include the steps of: generating a first partial base sequence and a second partial base sequence of the first base sequence; generating a third partial base sequence and a fourth partial base sequence of the second base sequence; arranging elements of the first partial base sequence in an upper available subcarrier group by spacing them out by one subcarrier, and arranging elements of the third partial base sequence between elements of the first partial base sequence; and performing half / half distributed concatenation in an lower available subcarrier group by spacing elements of the second partial base sequence by one subcarrier, and arranging elements of the fourth partial base sequence between elements of the second partial base sequence, thereby generating the synchronization signal sequence.

[0012] Meanwhile, a channel-free deep learning synchronization method according to a second embodiment of the present disclosure for achieving the above purpose may include, as a method of a second communication node, a step of receiving a synchronization signal based on a synchronization signal sequence formed by a first base sequence and a second base sequence from a first communication node; a step of calculating a cross-correlation value for the received synchronization signal; a step of providing the cross-correlation value as an input of a deep learning model; a step of estimating a synchronization sample point by the deep learning model; and a step of obtaining a synchronization signal using the synchronization sample point.

[0013] Here, the synchronization signal may be based on the synchronization signal sequence formed by half / half distributed concatenation of the first base sequence and the second base sequence.

[0014] Here, the first base sequence is based on a Zadov-Chu sequence, and the second base sequence may have a polarity opposite to that of the first base sequence.

[0015] Here, the deep learning model includes a convolution layer and a fully connected layer, and the convolution layer includes a ReLU (rectified linear unit) activation function, and the fully connected layer may include a softmax activation function.

[0016] Here, the deep learning model includes a first convolutional layer, a second convolutional layer, and a fully connected layer, and the first convolutional layer and the second convolutional layer include a ReLU activation function, and the fully connected layer may include a softmax activation function.

[0017] Here, the step of estimating a synchronous sample point by the deep learning model may include a step of estimating candidate synchronous sample points using a convolution layer of the deep learning model; and a step of estimating the synchronous sample point from the candidate synchronous sample points using a fully connected layer of the deep learning model.

[0018] Meanwhile, a channel-free deep learning synchronization device according to a third embodiment of the present disclosure for achieving the above purpose includes, as a second communication node, at least one processor, wherein the at least one processor can cause the second communication node to receive a synchronization signal based on a synchronization signal sequence formed by a first base sequence and a second base sequence from a first communication node; calculate a cross-correlation value for the received synchronization signal; provide the cross-correlation value as an input of a deep learning model; estimate a synchronization sample point by the deep learning model; and obtain a synchronization signal using the synchronization sample point.

[0019] Here, the synchronization signal may be based on the synchronization signal sequence formed by half / half distributed concatenation of the first base sequence and the second base sequence.

[0020] Here, the first base sequence is based on a Zadov-Chu sequence, and the second base sequence may have a polarity opposite to that of the first base sequence.

[0021] Here, the deep learning model includes a convolution layer and a fully connected layer, and the convolution layer includes a ReLU (rectified linear unit) activation function, and the fully connected layer may include a softmax activation function.

[0022] Here, in order to estimate the synchronous sample point by the deep learning model, the at least one processor can cause the second communication node to estimate candidate synchronous sample points using a convolutional layer of the deep learning model; and to estimate the synchronous sample point from the candidate synchronous sample points using a fully connected layer of the deep learning model.

[0023] According to the present disclosure, a receiving device can utilize the cross-correlation value used in a signal design-based synchronization method as input data for a deep learning model. As a result, the receiving device can prevent timing synchronization performance degradation due to differences in wireless channels when applying the wireless channel model used in offline learning and the online real-world learning model. Furthermore, the receiving device can achieve high timing synchronization accuracy under strong phase distortion environments by using a synchronization method based on a combination of signal design and deep learning.

[0024] Figure 1 is a conceptual diagram illustrating an embodiment of a communication system.

[0025] Figure 2 is a block diagram illustrating an embodiment of a communication node constituting a communication system.

[0026] Figure 3 is a graph showing the detection error rate versus the signal-to-noise ratio (SNR) of a receiving antenna using a tapped delay line-A (TDL-A).

[0027] Fig. 4 is a graph showing the detection error rate for the SNR of a receiving antenna using a tapped delay line-B (TDL-B).

[0028] Figure 5 is a conceptual diagram showing the direction of development of synchronization technology.

[0029] Figure 6 is a conceptual diagram for explaining embodiments of a synchronization signal in a communication system.

[0030] Figure 7 is a conceptual diagram decomposing the first synchronization signal and the NR (new radio) synchronization signal in the time domain.

[0031] Figure 8 is a conceptual diagram showing examples of deep learning model structures and parameters.

[0032] Figure 9 is a conceptual diagram showing examples of deep learning model structures and parameters.

[0033] Figure 10 is a graph showing examples of performance evaluation.

[0034] Figure 11 is a graph showing examples of performance evaluation.

[0035] Fig. 12 is a flowchart illustrating embodiments of a channel-free deep learning-based synchronization method.

[0036] This disclosure may be subject to various modifications and various embodiments. Specific embodiments are illustrated and described in detail in the drawings. However, this is not intended to limit the disclosure to specific embodiments, but rather to encompass all modifications, equivalents, and alternatives falling within the spirit and technical scope of the disclosure.

[0037] While terms such as "first" and "second" may be used to describe various components, these components should not be limited by these terms. These terms are used solely to distinguish one component from another. For example, without departing from the scope of the present disclosure, a first component could be referred to as a "second component," and similarly, a second component could also be referred to as a "first component." The term "and / or" includes a combination of multiple related items described herein or any of multiple related items described herein.

[0038] In embodiments of the present disclosure, “at least one of A and B” may mean “at least one of A or B” or “at least one of combinations of one or more of A and B.” Furthermore, in embodiments of the present disclosure, “at least one of A and B” may mean “at least one of A or B” or “at least one of combinations of one or more of A and B.”

[0039] When a component is referred to as being "connected" or "connected" to another component, it should be understood that it may be directly connected or connected to that other component, but that there may be other components intervening. Conversely, when a component is referred to as being "directly connected" or "connected" to another component, it should be understood that there are no other components intervening.

[0040] In the present disclosure, a phrase including “if (e.g., when ~)” can be expressed as a phrase including “based on (e.g., based on ~)” or a phrase including “in response to (e.g., in response to ~)”. In other words, a phrase including “if ~)” can be interpreted as being identical or similar to a phrase including “based on” or a phrase including “in response to”.

[0041] The terminology used in this disclosure is only used to describe specific embodiments and is not intended to limit the present disclosure. The singular expression includes the plural expression unless the context clearly indicates otherwise. In this disclosure, it should be understood that the terms "comprises" or "has" indicate the presence of a feature, number, step, operation, component, part, or combination thereof described in the specification, but do not preclude the presence or addition of one or more other features, numbers, steps, operations, components, parts, or combinations thereof.

[0042] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by a person of ordinary skill in the art to which this disclosure pertains. Terms defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and shall not be interpreted in an idealized or overly formal sense unless explicitly defined herein.

[0043] A communication system to which embodiments according to the present disclosure are applied will be described. The communication system to which embodiments according to the present disclosure are applied is not limited to the scope described below, and embodiments according to the present disclosure can be applied to various communication systems. Here, the term "communication system" may be used interchangeably with "communication network."

[0044] Throughout the specification, the network may include, for example, wireless internet such as WiFi (wireless fidelity), mobile internet such as WiBro (wireless broadband internet) or WiMax (world interoperability for microwave access), 2G mobile communication networks such as GSM (global system for mobile communication) or CDMA (code division multiple access), 3G mobile communication networks such as WCDMA (wideband code division multiple access) or CDMA2000, 3.5G mobile communication networks such as HSDPA (high speed downlink packet access) or HSUPA (high speed uplink packet access), 4G mobile communication networks such as LTE (long term evolution) or LTE-Advanced, and 5G mobile communication networks.

[0045] Throughout the specification, a terminal may refer to a mobile station, a mobile terminal, a subscriber station, a portable subscriber station, a user equipment, an access terminal, etc., and may include all or part of the functions of a terminal, a mobile station, a mobile terminal, a subscriber station, a portable subscriber station, a user equipment, an access terminal, etc.

[0046] Here, a desktop computer, laptop computer, tablet PC, wireless phone, mobile phone, smart phone, smart watch, smart glass, e-book reader, portable multimedia player (PMP), portable game console, navigation device, digital camera, digital multimedia broadcasting (DMB) player, digital audio recorder, digital audio player, digital picture recorder, digital picture player, digital video recorder, digital video player, etc. capable of communicating with the terminal can be used.

[0047] Throughout the specification, a base station may also refer to an access point, a radio access station, a node B, an evolved node B, a base transceiver station, a mobile multihop relay (MMR)-BS, etc., and may include all or part of the functions of a base station, an access point, a radio access station, a node B, an eNodeB, a base transceiver station, an MMR-BS, etc.

[0048] Hereinafter, preferred embodiments of the present disclosure will be described in more detail with reference to the attached drawings. In order to facilitate an overall understanding in describing the present disclosure, identical reference numerals will be used for identical components in the drawings, and redundant descriptions of identical components will be omitted.

[0049] Figure 1 is a conceptual diagram illustrating an embodiment of a communication system.

[0050] Referring to FIG. 1, a communication system (100) may include a plurality of communication nodes (110-1, 110-2, 110-3, 120-1, 120-2, 130-1, 130-2, 130-3, 130-4, 130-5, 130-6). Here, the communication system may be referred to as a "communication network." Each of the plurality of communication nodes may support at least one communication protocol. For example, each of the plurality of communication nodes may support a communication protocol based on CDMA (code division multiple access), a communication protocol based on WCDMA (wideband CDMA), a communication protocol based on TDMA (time division multiple access), a communication protocol based on FDMA (frequency division multiple access), a communication protocol based on OFDM (orthogonal frequency division multiplexing), a communication protocol based on OFDMA (orthogonal frequency division multiple access), a communication protocol based on SC (single carrier)-FDMA, a communication protocol based on NOMA (non-orthogonal multiple access), a communication protocol based on SDMA (space division multiple access), etc. Each of the plurality of communication nodes may have the following structure.

[0051] Figure 2 is a block diagram illustrating an embodiment of a communication node constituting a communication system.

[0052] Referring to FIG. 2, a communication node (200) may include at least one processor (210), a memory (220), and a transceiver (230) that is connected to a network and performs communication. In addition, the communication node (200) may further include an input interface device (240), an output interface device (250), a storage device (260), etc. Each component included in the communication node (200) may be connected by a bus (270) to perform communication with each other. However, each component included in the communication node (200) may be connected through an individual interface or an individual bus centered around the processor (210), rather than a common bus (270). For example, the processor (210) may be connected to at least one of the memory (220), the transceiver (230), the input interface device (240), the output interface device (250), and the storage device (260) through a dedicated interface.

[0053] The processor (210) can execute program commands stored in at least one of the memory (220) and the storage device (260). The processor (210) may refer to a central processing unit (CPU), a graphics processing unit (GPU), or a dedicated processor in which methods according to embodiments of the present invention are performed. Each of the memory (220) and the storage device (260) may be configured with at least one of a volatile storage medium and a non-volatile storage medium. For example, the memory (220) may be configured with at least one of a read-only memory (ROM) and a random access memory (RAM).

[0054] Referring again to FIG. 1, the communication system (100) may include a plurality of base stations (110-1, 110-2, 110-3, 120-1, 120-2) and a plurality of user equipment (UEs) (130-1, 130-2, 130-3, 130-4, 130-5, 130-6). Each of the first base station (110-1), the second base station (110-2), and the third base station (110-3) may form a macro cell. Each of the fourth base station (120-1) and the fifth base station (120-2) may form a small cell. The fourth base station (120-1), the third UE (130-3), and the fourth UE (130-4) may be within the coverage of the first base station (110-1). The second UE (130-2), the fourth UE (130-4), and the fifth UE (130-5) may be within the coverage of the second base station (110-2). The fifth base station (120-2), the fourth UE (130-4), the fifth UE (130-5), and the sixth UE (130-6) may be within the coverage of the third base station (110-3). The first UE (130-1) may be within the coverage of the fourth base station (120-1). The sixth UE (130-6) may be within the coverage of the fifth base station (120-2).

[0055] Here, each of the plurality of base stations (110-1, 110-2, 110-3, 120-1, 120-2) may be referred to as a NodeB, an evolved NodeB, a BTS (base transceiver station), a radio base station, a radio transceiver, an access point, an access node, a road side unit (RSU), a DU (digital unit), a CDU (cloud digital unit), a RRH (radio remote head), a RU (radio unit), a TP (transmission point), a TRP (transmission and reception point), a relay node, etc. Each of the plurality of UEs (130-1, 130-2, 130-3, 130-4, 130-5, 130-6) may be referred to as a terminal, an access terminal, a mobile terminal, a station, a subscriber station, a mobile station, a portable subscriber station, a node, a device, etc.

[0056] Each of the plurality of communication nodes (110-1, 110-2, 110-3, 120-1, 120-2, 130-1, 130-2, 130-3, 130-4, 130-5, 130-6) can support cellular communication (e.g., long term evolution (LTE), LTE-A (advanced) as defined in the 3rd generation partnership project (3GPP) standard). Each of the plurality of base stations (110-1, 110-2, 110-3, 120-1, 120-2) can operate in a different frequency band or can operate in the same frequency band. Each of the plurality of base stations (110-1, 110-2, 110-3, 120-1, 120-2) can be connected to each other via an ideal backhaul or a non-ideal backhaul, and can exchange information with each other via the ideal backhaul or the non-ideal backhaul. Each of the plurality of base stations (110-1, 110-2, 110-3, 120-1, 120-2) can be connected to a core network (not shown) via an ideal backhaul or a non-ideal backhaul. Each of the plurality of base stations (110-1, 110-2, 110-3, 120-1, 120-2) can transmit a signal received from the core network to the corresponding UE (130-1, 130-2, 130-3, 130-4, 130-5, 130-6), and can transmit a signal received from the corresponding UE (130-1, 130-2, 130-3, 130-4, 130-5, 130-6) to the core network.

[0057] Each of the plurality of base stations (110-1, 110-2, 110-3, 120-1, 120-2) can support OFDMA-based downlink transmission and SC-FDMA-based uplink transmission. In addition, each of the plurality of base stations (110-1, 110-2, 110-3, 120-1, 120-2) can support MIMO (multiple input multiple output) transmission (e.g., single user (SU)-MIMO, multi user (MU)-MIMO, massive MIMO, etc.), CoMP (coordinated multipoint) transmission, carrier aggregation transmission, transmission in an unlicensed band, device to device (D2D) communication (or, ProSe (proximity services), etc.). Here, each of the plurality of UEs (130-1, 130-2, 130-3, 130-4, 130-5, 130-6) can support base stations (110-1, 110-2, 110-3, 120-1, 120-2) and can perform operations supported by base stations (110-1, 110-2, 110-3, 120-1, 120-2).

[0058] Meanwhile, in a communication system, a terminal can perform physical layer synchronization estimation by performing cross-correlation based on a signal design defined as in mathematical expression 1 for physical layer synchronization.

[0059]

[0060] can be a coefficient for normalizing the cross-correlation value and can be an integer. R xy(n) can be a cross-correlation function for sample time n, and n can be an integer. y(m+n) can be a received signal received from a transmitter at a terminal at time m+n. m can be an element index of a time-domain synchronization signal, and can be an integer. N can be the total number of elements of the synchronization signal, and can be an integer. x*(m) can be a conjugate signal of x(m). For example, the conjugate signal (a+jb) of (a+jb) * can be a-jb. The transmitter can transmit a synchronization signal x(m), and the terminal can receive the synchronization signal from the transmitter.

[0061] Phase distortion caused by various factors can affect timing accuracy during synchronization at the terminal. Factors affecting phase distortion include carrier frequency offset (CFO), phase noise (PhN), the Doppler effect, power amplifier nonlinearity, and sampling jitter. Phase distortion caused by CFO and PhN can significantly affect the received signal, as shown in Equation 2. The received signal can be affected by CFO and PhN. This can deteriorate cross-correlation characteristics and significantly degrade synchronization accuracy.

[0062]

[0063] can mean the CFO component normalized to SCS (subcarrier spacing). φ n can mean the PhN component that changes depending on n.

[0064] Figure 3 is a graph showing the detection error rate versus the signal-to-noise ratio (SNR) of a receiving antenna using a tapped delay line-A (TDL-A).

[0065] Referring to FIG. 3, the transmitter can transmit a new radio primary synchronization signal (NR PSS), an alternative radio primary synchronization signal (AR PSS), a beyond radio primary synchronization signal 1 (BR PSS1), a beyond radio primary synchronization signal 2 (BR PSS2), etc. to the receiver. The receiver can receive the NR PSS, AR PSS, BR PSS1, BR PSS2, etc. from the transmitter and perform cross-correlation to estimate timing. The NR PSS, AR PSS, BR PSS1, BR PSS2, etc. can experience strong phase distortion. Timing estimation performance can be greatly degraded due to such strong phase distortion. β can represent a line bandwidth. η can be a normalized SCS.

[0066] On line 301, the normalized SCS can be 480 kHz, the normalized CFO component to the SCS can be 0.67, the line width can be 0.4, and the synchronization signal can be NR PSS. On line 302, the normalized SCS can be 120 kHz, the normalized CFO component to the SCS can be 0.67, the line width can be 0.4, and the synchronization signal can be NR PSS. On line 303, the normalized SCS can be 120 kHz, the normalized CFO component to the SCS can be 0.67, the line width can be 0.0, and the synchronization signal can be BR PSS1. The normalized SCS on line 304 can be 120 kHz, the normalized CFO component to SCS can be 0.67, the line width can be 0.0, and the synchronization signal can be BR PSS2.

[0067] Fig. 4 is a graph showing the detection error rate for the SNR of a receiving antenna using a tapped delay line-B (TDL-B).

[0068] Referring to FIG. 4, the transmitter can transmit NR PSS, AR PSS, BR PSS1, BR PSS2, etc. to the receiver. The receiver can receive NR PSS, AR PSS, BR PSS1, BR PSS2, etc. from the transmitter and perform cross-correlation to estimate timing. NR PSS, AR PSS, BR PSS1, BR PSS2, etc. can experience strong phase distortion. Timing estimation performance can be greatly degraded due to such strong phase distortion. β can represent the line width. η can be the normalized SCS. The normalized SCS in line 401 can be 120 kHz, the normalized CFO component to the SCS can be 0.67, the line width can be 0.0, and the synchronization signal can be AR PSS. On line 402, the normalized SCS can be 480 kHz, the normalized CFO component to the SCS can be 0.67, the line width can be 0.0, and the synchronization signal can be AR PSS. On line 403, the normalized SCS can be 120 kHz, the normalized CFO component to the SCS can be 0.67, the line width can be 0.4, and the synchronization signal can be NR PSS. On line 404, the normalized SCS can be 480 kHz, the normalized CFO component to the SCS can be 0.67, the line width can be 0.4, and the synchronization signal can be NR PSS. On line 405, the normalized SCS can be 120 kHz, the normalized CFO component to the SCS can be 0.67, the line width can be 0.0, and the synchronization signal can be BR PSS1. The normalized SCS on line 406 can be 120 kHz, the normalized CFO component to SCS can be 0.67, the line width can be 0.0, and the synchronization signal can be BR PSS2. The normalized SCS on line 407 can be 480 kHz, the normalized CFO component to SCS can be 0.67, the line width can be 0.4, and the synchronization signal can be BR PSS1.The normalized SCS on line 408 can be 480 kHz, the normalized CFO component to SCS can be 0.67, the line width can be 0.4, and the synchronization signal can be BR PSS2.

[0069] Referring to FIGS. 3 and 4, BR PSS1 and BR PSS2 can provide better synchronization estimation performance than NR PSS. BR PSS1 and BR PSS2 may be synchronization signals developed based on signal design. A communication system may use BR PSS1 and BR PSS2 to prevent synchronization estimation performance degradation due to strong phase distortion. Synchronization signals based on signal designs such as BR PSS1 and BR PSS2 can provide much higher timing accuracy than NR PSS. Robots using artificial intelligence may become widespread in the near future. In order to support robots using artificial intelligence that will become widespread in the future, a communication system may require even higher and more precise timing accuracy under a strong phase distortion environment.

[0070] Before the introduction of the 5G standard, phase-distortion-tolerant synchronization signal designs used machine learning-based estimation to improve timing accuracy. However, since the introduction of the 5G standard, phase-distortion-tolerant synchronization signal designs have shifted to improving timing accuracy by using deep learning (DL) models, which can replace traditional cross-correlation estimators.

[0071] However, deep learning model-based synchronization technology cannot fundamentally solve performance degradation caused by high memory usage, complexity, and wireless channel mismatch (e.g., mismatch between wireless channels during offline and online learning) for timing detection. Deep learning technology is poised to achieve remarkable advancements in the near future. Given these circumstances, deep learning model-based synchronization technology could address complexity and high memory usage in the near future. It is expected that most communication devices will utilize deep learning model-based synchronization technology in the near future.

[0072] Deep learning-based synchronization technology can address wireless channel mismatch issues by equipping a receiver with multiple deep learning models trained on various wireless channels. This deep learning-based synchronization technology can monitor wireless channel characteristics in real time and apply appropriate deep learning models. Notably, this deep learning-based synchronization technology can align with the requirements of the 3GPP standard.

[0073] Figure 5 is a conceptual diagram showing the direction of development of synchronization technology.

[0074] Referring to Figure 5, research on synchronization technologies in the coming years is expected to focus on both signal design-based synchronization and deep learning-based synchronization. Synchronization technologies for 6G or beyond may shift to integrated synchronization research, which generates signals robust to phase distortion based on signal design and uses these signals as input to deep learning models to address wireless channel mismatch issues. Integrated synchronization research may require proactive research to address wireless channel mismatch issues.

[0075] Therefore, the method of the present disclosure can propose a signal design-based synchronization technique that can improve timing synchronization accuracy, as well as a deep learning-based synchronization technique. Furthermore, the method of the present disclosure can propose a combined technique of signal design-based synchronization technique and deep learning-based synchronization technique that can address wireless channel mismatch issues and improve timing synchronization estimation performance.

[0076] The present disclosure proposes a physical layer synchronization method and device that can design a synchronization signal based on signal design that improves timing synchronization accuracy in order to solve the problems of the prior art described above. The present disclosure proposes a deep learning-based physical layer synchronization method and device that can improve timing accuracy. The purpose of the present disclosure is to provide a physical layer synchronization method and communication device that combine a signal design-based synchronization technology and a deep learning-based synchronization technology that can solve the problem of wireless channel mismatch and improve timing synchronization estimation performance.

[0077] In the present disclosure, a wireless device may be referred to as a UE (user equipment), a device that transmits a signal to or receives a signal from a UE may be referred to as a TRP (transmission and reception point), a device that manages a TRP may be referred to as a base station (BS), and an area managed by a BS may be referred to as a cell. The present disclosure may describe specific methods, procedures, and devices based on the configurations of a UE, a TRP, a BS, and a cell. However, this is for convenience of explanation and may not be limited thereto, and all possible configurations that are read within the concepts of the methods, procedures, and devices of the present disclosure may be included in the scope of the present disclosure. It should be noted that a UE may also be referred to as an MS (mobile station) or a terminal. While the configuration descriptions of the present disclosure may mostly refer to synchronization as timing synchronization, it should be noted that the present disclosure is not limited thereto, and frequency synchronization may also be included in the present disclosure.

[0078] The first synchronization signal (SS1) sequences P proposed in the present disclosure u (m) can be as in mathematical expression 3. The first synchronization signal sequences are the first base sequences b u (m) and second base sequences It can be composed of half / half distributed concatenation. Half / half distributed concatenation can mean composing a synchronization signal by repeatedly arranging or distributed concatenation of the first base sequence and the second base sequence.

[0079] The half-and-half distributed sequence of the first and second base sequences may not simply mean attaching the second base sequence after the first base sequence, but may also mean cross-distributing half of each of the first and second base sequences. In other words, the half-and-half distributed sequence of the first and second base sequences may mean splitting the two base sequences into front and back halves and cross-distributing them.

[0080] For example, the transmitter transmits the first partial base sequence of the first base sequence and the second part bass sequence can generate a third partial base sequence of the second base sequence. and the fourth part bass sequence can be created.

[0081] The transmitter may place elements of the first partial base sequence in the upper available subcarrier group, spaced by one subcarrier, and place elements of the third partial base sequence between the elements of the first partial base sequence. The transmitter may perform half / half distributed concatenation, spaced by one subcarrier, and place elements of the fourth partial base sequence between the elements of the second partial base sequence, to generate a synchronization signal sequence (see FIG. 6).

[0082] In this way, distributed concatenation can mean connecting by distributing according to a certain rule, and can be a method of dividing sequences into parts and interleaving or distributing them. The first base sequence can be a Zadoff-Chu base sequence. The second base sequences can be modified base sequences of the first base sequences. The second base sequence can be a modified Zadoff-Chu base sequence. The second base sequences can be generated from the first base sequences by multiplying the first base sequences by a negative number. In other words, the second base sequences silver It could be. can mean the largest integer less than the real number W.

[0083]

[0084] In mathematical expression 3, u may be a physical identity (PID). m may denote an index of the first synchronization signal sequences and may be a positive integer. The number of the first synchronization signal sequences may be 2M, and M may be a positive integer. m may be 0≤m≤2M. P u (M) can be 0. Accordingly, the first synchronization signal sequences can use one less subcarrier than the NR SS (synchronization signal) sequences. The first base sequences can be as shown in the following mathematical expression 4. In mathematical expression 4, μ u may be an index of a ZC (Zadoff-Chu) sequence, and the index of the ZC sequence may be, for example, 1, but may not be limited thereto.

[0085]

[0086] Figure 6 is a conceptual diagram for explaining embodiments of a synchronization signal in a communication system.

[0087] Referring to FIG. 6, a communication system may include a plurality of communication nodes. A first communication node may generate a first synchronization signal and transmit it to a second communication node. The first communication node may generate the first synchronization signal by first synchronization signal sequences P u (m) can be used to generate. m can mean the index of the first synchronization signal sequences and can be a positive integer. The number of the first synchronization signal sequences can be 2M, and M can be a positive integer. m can be 0≤m≤2M. P u (M) can be 0.

[0088] The first synchronization signal sequences are the first base sequences b u (m) and second base sequences It can be composed of. Specifically, each of the first synchronization signal sequences can be composed of one first base sequence or one second base sequence. The first communication node can generate the first base sequences and the second base sequences to generate the first synchronization signal sequences. The first communication node can map the generated first base sequences and the second base sequences to subcarriers in the frequency domain. The number of subcarriers can be a positive integer N. The index k of the subcarriers can be greater than or equal to 0 and less than or equal to N-1. In other words, k can be 0, 1, or N-1. The first communication node can generate the first base sequences and the second base sequences based on a physical identifier u. N can be a size of an inverse fast Fourier transform (IFFT).

[0089] The first communication node transmits the first synchronization signal sequences P u (m) can be modulated and mapped to 2(M+1) subcarriers represented by index k (k=0, 1, to N-1). The first synchronization signal sequences P u Each of the first synchronization signal sequences (P) constituting (m)u (0), P u (1), inland, P u (2M)) can be mapped to each subcarrier having a corresponding index. The first synchronization signal sequences P u (m) are mapped to 2(M+1) subcarriers (i.e., the first synchronization signal sequences P u (N subcarriers to which the (m) modulated modulation symbols are mapped) may be included in the first subcarrier group.

[0090] The two (M+1) subcarriers constituting the first subcarrier group may be adjacent or spaced apart from each other in the frequency domain. Although FIG. 6 illustrates a case where at least some of the two (M+1) subcarriers constituting the first subcarrier group are arranged adjacent to each other, this is merely an example for convenience of explanation and may not be limited thereto. For example, the first subcarrier group may be composed of two (M+1) subcarriers that are spaced apart from each other. In other words, the first subcarrier group may be composed of N subcarriers that are not adjacent to each other.

[0091] One or more null subcarriers may be arranged around the (2M+1) subcarriers that constitute the first subcarrier group, or between the two (M+1) subcarriers. A signal may not be carried on a null subcarrier. In other words, a modulation symbol may not be assigned to a null subcarrier. A null subcarrier may have a value of 0. A null subcarrier may correspond to a gap subcarrier, a direct current (DC) subcarrier, etc. A null subcarrier may be arranged to easily identify each subcarrier.

[0092] A null subcarrier may be arranged at the front end and / or the back end of the first subcarrier group in the frequency domain. For example, a null subcarrier may be arranged at the front end of a subcarrier corresponding to a subcarrier index NM-1, and / or at the back end of a subcarrier corresponding to a subcarrier index M. In addition, one or more null subcarriers may be arranged between two (M+1) subcarriers constituting the first subcarrier group. For example, a null subcarrier may be arranged at the front end and / or the back end of one or more subcarriers (hereinafter, referred to as center subcarriers) located at the center of the two (M+1) subcarriers constituting the first subcarrier group. Alternatively, the first subcarrier group may be divided into a plurality of subgroups, each including one or more subcarriers. Null subcarriers may be arranged at the front end and / or the back end of each of the subgroups.

[0093] The time-domain signal of the first synchronization signal may be a complex-valued signal with periodic zeros. Compared to 5G NR, which is a complex-valued signal with periodic zeros for each time sample, the first synchronization signal may have periodic zeros. Accordingly, the synchronization estimation complexity at the receiver may be reduced.

[0094] As the length of a sequence increases, the cross-correlation property (e.g., high peak or low sideload) may improve. On the other hand, as the length of the sequence increases, the phase distortion due to CFO / PhN may also increase. As described above, the cross-correlation property and phase distortion may conflict with each other as the length of the sequence increases. As a solution to this, the first communication node may consider combining two short sequences to generate a long sequence in the time domain. When the first communication node combines two short sequences to generate a long sequence, the cross-correlation property may not be sacrificed, and the sequence may be generated that is robust to the phase distortion of CFO / PhN.

[0095] A half / half distributed concatenation of the first base sequence and the second base sequence in the same frequency domain as the first synchronization signal may be a combination of two short-length sequences in the time domain. The frequency domain signal of the first synchronization signal may be generated by interleaving the first base sequence and the second base sequence.

[0096] Figure 7 is a conceptual diagram decomposing the first synchronization signal and the NR (new radio) synchronization signal in the time domain.

[0097] Referring to Fig. 7, the first synchronization signal may be in the form of a superposition of two different components. In other words, the first synchronization signal is x generated based on the first base sequence. o and x generated based on the second base sequence e It can be a nested form of components composed of repetitions.

[0098] Cross-correlation coefficient between the first synchronization signal and its corresponding received signal r[τ] can be as shown in Equation 5. τ can be sufficiently small to be very close to the peak.

[0099]

[0100] Cross-correlation coefficient in mathematical formula 5 can be as follows: Mathematical formula 6.

[0101]

[0102] Mathematical expression 7 can be established by distributed aggregation.

[0103]

[0104] The absolute value of the cross-correlation coefficient can be as shown in the following mathematical expression 8. Referring to mathematical expression 8, the first synchronization signal is It can be robust to hardware damage due to the positive effect of sidelobe suppression. Sidelobe can be, for example, the absolute value of the cross-correlation coefficient when τ≠0.

[0105]

[0106] Referring to Equations 5 and 8, short sequences may be less sensitive to hardware damage such as CFO / PhN with severe phase distortion changes.

[0107] Mathematical expression 9 is x e Wow x o can represent the size of x as in mathematical formula 9 e Wow x o The sidelobe can be effectively suppressed while being shorter than the time domain signal x3 of the NR SS. Accordingly, the first synchronization signal can be robust to phase distortion.

[0108]

[0109] As previously discussed, the first sync signal is not a repeating form of a short sequence, but it can have the robustness against phase distortion that is an advantage of short signal repetition. The first sync signal can be a full-length sync signal rather than a half-length sync signal with suppressed sidelobes.

[0110] Meanwhile, deep learning-based synchronization methods can achieve excellent timing synchronization accuracy under strong phase distortion. The input matrix R of the deep learning model t can be modeled as a matrix that reconstructs a complex number into real and imaginary components using mathematical expression 10. In order to solve the problem of significantly reduced reliability under high phase distortion, the present disclosure can use the received signal itself as input data of a deep learning model instead of the output of cross-correlation produced in a signal design-based synchronization method.

[0111] For each training example t, (1×N s,l ) can be a received signal vector. l can be, for example, l=0, 1, or 13, and can be a time domain symbol index. N s,l can be the total number of time samples of symbol l. One of the total symbols can include a transmission (Tx) synchronization signal (e.g., NR SS) for the proposed deep-learning (PDL). One frame is the total number of samples N that constitute all symbols. f , in other words, for example, N s,0 +N s,1 +… +N s,13 may require.

[0112]

[0113] In Equation 10, A can represent the total number of receiving antennas. Re{r] and Im[r] can represent the real and imaginary component vectors of vector r, respectively. Considering the total number of receiving antennas, the input matrix R for each training example t t Is (In other words, 2A×N f (the real space of the matrix) can be.

[0114] Figure 8 is a conceptual diagram showing examples of deep learning model structures and parameters.

[0115] Referring to Figure 8, the deep learning model can be a fully convolutional network (FCN) model, consisting of three convolutional (Conv) layers. The first and second layers can use the rectified linear unit (ReLU) activation function. The final layer can provide linear output and allow for a wide output range. The parameters of each layer can be as shown in Table 1.

[0116] Layer parameters Output format Input 2×N First layer kernel (kernal) = (16,1,35,2) 16×1×N Second layer kernel (kernal) = (16,16,35,1) 16×1×N Third layer kernel (kernal) = (1,16,25,1) 1×1×N

[0117] Figure 9 is a conceptual diagram illustrating examples of a deep learning model structure and parameters. Referring to Figure 9, the deep learning model is a proposed deep learning model, which can be a convolutional neural network (CNN) model consisting of two convolutional (Conv) layers and one fully connected layer. The first and second layers can use the ReLU activation function. The third layer can use the softmax function. The parameters of each layer can be as shown in Table 2.

[0118] Layer parameters Output format Input 4×960 First layer Kernel = (32,1,65,4) 32×1×960 Second layer Kernel = (32,32,65,1) 32×1×960 Third layer Size = 960 1×960

[0119] Among various algorithms used for deep learning-based synchronization, Convolutional Neural Networks (CNNs) may be particularly advantageous for synchronization estimation. This may be because the convolutional operation performed in the convolutional layer is very similar to the cross-correlation performed in signal design-based synchronization. Therefore, a convolutional layer composed of filters with the same window size can serve as a cross-correlation function. Furthermore, a larger number of filters may have the advantage of analyzing correlations with more learned signals. According to the proposed deep learning model, the output of the first layer can be as shown in the following mathematical expression 11. The output of the first layer can be the output of the ith field at the kth input lag for each training example t.

[0120]

[0121] Filter i can be i=0, 1, or F-1. Input lag k can be k=0, 1, or N. f -1 can be input channel c is 0, l c1 -1 could be l c1 can be the number of input channels. The kernel column m can be m=0, 1, or K. l -1 can be. Kernel row n is n = 0, or K h1 -1 could be K h1 can be the height of the filter kernel. k' can be the length of the filter kernel, and k'=-2N, -2N-1, to, -1, to, -1, N f , N f + 1, or N f It can be +2N. For k' can be 0. is input to the proposed deep learning model of Fig. 8 and has 2A×N real values ​​that are back propagated. f It can mean a matrix. (R)a,b can mean the (ab)th element of matrix R. can represent a weight tensor. F can be the number of filters in the first convolutional layer. b (1) can mean a bias term added to all filters. ReLU is a nonlinear activation function max(2,0) It may be an output that has passed through.

[0122] According to the proposed deep learning model, the output of the second layer can be as shown in the following mathematical expression 12. The output of the second layer can be the output of the ith field at the kth input lag for each training example t.

[0123]

[0124] Filter i can be i=0, 1, or F-1. Input lag k can be k=0, 1, or N. f -1 can be input channel c is 0, l c2 -1 could be l c2 can be the number of input channels. The kernel column m can be m=0, 1, or K. l -1 can be. Kernel row n is n=0, or K h2 -1 could be K h2 can be the height of the filter kernel. k' can be the length of the filter kernel, and k'=-2N, -2N-1, to, -1, to, -1, N f , N f + 1, or N f It can be +2N. For k' can be 0. can represent a weight tensor. F can be the number of filters in the first convolutional layer. b (2) can mean a bias term added to all filters. ReLU is a nonlinear activation function max(x,0) It may be an output that has passed through.

[0125] The third layer can be processed using the softmax function after the output of the second layer is processed by the FC layer. The computational complexity of the i-th convolutional layer is, for example, I ci K l K hi FN f N f It can be, and the memory usage is (I ci Kl K hi +1)N f It can be. The computational complexity of the FC layer is FN f N f It can be, and memory usage is (FN f +1)N f It could be.

[0126] Loss function of the proposed deep learning model can be as shown in the following mathematical expression 13. Since the loss function has no priority on the index of the synchronization point, the cross entropy loss can be reflected as shown in mathematical expression 13. In mathematical expression 13, y(t) can mean a one-hot vector at the first sample synchronization point. y(t) can mean an exact sample synchronization point among the time-domain samples of one frame in the training example t. may mean the index probability at the first sample synchronization point estimated or predicted by the PDL model.

[0127]

[0128] The hyperparameters configured for offline training of a deep learning model can be as shown in Table 3. The hyperparameters can utilize PyTorch's deep learning capabilities.

[0129] Parameter settings (settings) Optimizer / learning rate ADAM / 0.001 SNR used during training / epoch / mini-batch size -2 dB / 100 / 128 Learn rate drop period / learn rate drop factor 3 / 0.6

[0130] Table 4 can be an evaluation parameter for comparing complexity and memory usage.

[0131] Evaluation parameter settings (settings) Carrier frequency / SCSη / speed 70GHz / 120KHz / 3Km / h Normalized CFO / Normalized PhNβ(2 / 3) / (0.4) channel models TDL-A, TDL-C, TDL- / dRMS(root mean-square) delay spread [ns] 16, 150, 200, 300

[0132] In Table 5, SCS can refer to the spacing between subcarriers. CFO can refer to the offset between the transmit and receive subcarriers. Normalized CFO may mean CFO [Hz] / SCS. PhN may be flicker noise generated by an oscillator. Normalized PhN β may mean PhN linewidth / SCS. The channel model TDL-A may mean an urban microcell and a rural environment with non-line-of-sight (NLoS) root mean-square (RMS) delay spreads of 16 ns and 150 ns, respectively. TDL-C may mean a suburban macrocell and a rural environment with NLoS RMS delay spreads of 16 ns and 200 ns, respectively. TDL-D may mean an urban microcell and a rural environment with LoS RMS delay spreads of 16 ns and 300 ns, respectively. The wireless environment of the channel models may apply two-branch reception (Rx) equal gain combined diversity (i.e., A = 2). The wireless environment of the channel models can be assumed to maintain full system load by uniformly utilizing all possible subcarriers for channels other than the synchronization signal. The frame synchronization sample point (i.e., the starting point of the synchronization signal) for each input data example can be assumed to be randomly set. Figure 10 is a graph illustrating examples of performance evaluation.

[0133] Figure 10 is a graph showing examples of performance evaluation.

[0134] Referring to Fig. 10, a performance evaluation graph can show performance evaluation results in a short delay spread environment. The performance evaluation graph can use the detection error rate (DER) as a performance evaluation index. If the synchronization point defined in the DER is located within 1 / 4 of the CP (cyclic prefix) interval, it can be considered a success. Additionally, if the synchronization point estimated in the DER is not located within 1 / 4 of the CP interval, it can be considered an error. In the performance evaluation graph, 'PDL: V→W' can refer to a synchronization method when a PDL model trained offline with channel model 'V' is applied online to channel model 'W'. In the performance evaluation graph, 'PDL: V→W' can be said to correspond to the DER performance evaluation of "wireless channel mismatch" between offline and online.

[0135] Line 1001 can represent the DER of the synchronization signal by the synchronization method when the PDL model is applied to the channel model TDL-A in LTE SS. Line 1002 can represent the DER of the synchronization signal by the synchronization method when the PDL model is applied to the channel model TDL-D in LTE SS. Line 1003 can represent the DER of the synchronization signal by the synchronization method when the model of DL

[0011] is applied to the channel model TDL-C. Line 1004 can represent the DER of the synchronization signal by the synchronization method when the PDL model trained with the channel model TDL-A offline is applied to the channel model TDL-C online as PDL: TDL-A→TDL-C.

[0136] Figure 11 is a graph showing examples of performance evaluation.

[0137] Referring to Figure 11, the performance evaluation graph can show performance evaluation results in a long-latency spread environment. The performance evaluation graph can use DER as a performance evaluation index. In the performance evaluation graph, "PDL: V→W" can refer to a synchronization method when a PDL model trained offline with channel model "V" is applied online to channel model "W." In the performance evaluation graph, "PDL: V→W" can be said to correspond to the DER performance evaluation of "wireless channel mismatch" between offline and online.

[0138] Line 1101 may represent the DER of a synchronization signal by a synchronization method when the PDL model is applied to the channel model TDL-C in LTE SS. Line 1102 may represent the DER of a synchronization signal by a synchronization method when the PDL model is applied to the channel model TDL-A in LTE SS. Line 1103 may represent the DER of a synchronization signal by a synchronization method when the PDL model is applied to the channel model TDL-D in LTE SS. Line 1104 may represent the DER of a synchronization signal by a synchronization method when the PDL model is applied to the channel model TDL-A. Line 1105 may represent the DER of a synchronization signal by a synchronization method when the DL

[0011] model is applied to the channel model TDL-A. Line 1106 can represent the DER of the synchronization signal by the synchronization method in the case where the PDL model trained offline with the channel model TDL-A is applied to the channel model TDL-C online, as in PDL: TDL-A→TDL-C. Line 1107 can represent the DER of the synchronization signal by the synchronization method in the case where the PDL model is applied to the channel model TDL-C. Line 1108 can represent the DER of the synchronization signal by the synchronization method in the case where the PDL model trained offline with the channel model TDL-D is applied to the channel model TDL-C online, as in PDL: TDL-D→TDL-C.

[0139] Referring to Figures 10 and 11, LTE SS and NR SS may not function effectively even with an increase in SNR. On the other hand, the synchronization signal by PDL and the synchronization signal by DL without channel mismatch and SS1 can overcome phase distortion and consistently show excellent timing accuracy. In the absence of channel mismatch, all PDL cases can show better performance than SS1 and DL. In the presence of channel mismatch, 'PDL: TDL-D→TDL-A' and 'PDL: TDL-D→C' may perform worse in timing accuracy than the SS1 case based on all signal designs. However, 'PDL: TDL-D→TDL-A' and 'PDL: TDL-D→C' can still show better performance than NR SS, suggesting the possibility of overcoming channel mismatch. Table 5 shows the computational complexity and memory usage.

[0140] Scheme Complexity Memory Usage SS1 / SS2 PDL

[0141] In Table 5, the computational complexity of timing estimation can only consider the number of multiplication operations. Memory usage can consider the total number of parameters required for timing estimation. Referring to Table 5, the signal design-based SS1 has a complexity that is approximately 4F times lower and a FN factor of approximately 100% lower than the deep learning-based PDL. f The ship can have a small memory footprint.

[0142] Meanwhile, the synchronization signal generated by combining signal design-based and deep learning-based signals can achieve excellent timing synchronization accuracy under strong phase distortion.

[0143] Fig. 12 is a flowchart illustrating embodiments of a channel-free deep learning-based synchronization method.

[0144] Referring to FIG. 12, the transmitter can transmit all possible synchronization signals that are robust to phase distortion. The receiver can receive all possible synchronization signals that are robust to phase distortion from the transmitter. The receiver can perform cross-correlation on the received signals from the transmitter as in Equation 14 (S1201).

[0145]

[0146] In mathematical expression 14 may be a coefficient for normalization of the cross-correlator. a may be a receiving antenna index. For example, a may be 0, 1. P(m) may be a synchronization signal transmitted from a transmitter for timing synchronization estimation. In the method of the present disclosure, only all possible synchronization signals that are robust to phase distortion may be included in the scope of the present disclosure. p*(m) may be a conjugate signal of p(m). N may be the length of the synchronization signal. y(n',a) may be a received signal received by the synchronization estimator. R py is the complex output of the cross-correlator, R py An example of an input matrix of (4×960) whose elements are real and imaginary components for receiving antenna a of (n,a) can be. In the input matrix of (4×960), 4 can mean real and imaginary components for receiving antennas 0 and 1, and 960 can mean N f It can mean the number of samples in one received frame.

[0147] The receiver can estimate or predict synchronous sample points using the cross-correlation values ​​of the cross-correlator as input to a deep learning model. For example, the deep learning model can be a CNN model consisting of one convolutional layer and one fully connected layer. The first layer can use a ReLU activation function. The second layer can use a softmax function. The receiver can estimate or predict first synchronous sample points using one convolutional layer and the ReLU activation function (S1202). The first synchronous sample points can be candidate synchronous sample points. The receiver can estimate or predict second synchronous sample points using the first synchronous sample points as input using a fully connected layer and a softmax function (S1203). The second synchronous sample point can be a final synchronous sample point.

[0148] In this way, the receiver can estimate the final sync sample point by applying ReLU in the convolutional layer and softmax in the fully connected layer. The size of the singularity in the convolutional layer can be 1920, and in the fully connected layer, the size of the singularity can be 960, the total number of samples in a single frame. The convolutional layer can learn about more wireless channels and phase distortion. As a result, the fully connected layer can improve the accuracy of the final sync sample estimation.

[0149] The deep learning model used in the receiver could be, for example, a CNN model consisting of three convolutional layers. The first and second layers could use the ReLU activation function. Another example is a CNN model consisting of two convolutional layers and one fully connected layer. The first and second layers could use the ReLU activation function. The third layer could use the softmax function.

[0150] Unlike deep learning-based synchronization methods, signal design and deep learning-based synchronization methods can use the cross-correlation values ​​used in signal design-based synchronization methods as input data. The reason for this may be to prevent the degradation of timing synchronization performance due to the difference in wireless channels when applying the wireless channel model used in offline learning and the online actual learning model, as mentioned above. At this time, it should be noted that not all types of synchronization signals can be transmitted from the transmitter and used for synchronization at the receiver, but rather, a synchronization signal that has been theoretically and performance-wise proven to be robust to phase distortion can be applied and transmitted at the transmitter, and the receiver can perform timing synchronization by receiving this.

[0151] According to the physical layer synchronization method and device based on signal design, deep learning, or a combination of signal design and deep learning of the present disclosure, high timing synchronization accuracy can be achieved under a strong phase distortion environment.

[0152] The operations of the method according to the embodiments of the present disclosure can be implemented as a computer-readable program or code on a computer-readable recording medium. A computer-readable recording medium includes any type of recording device that stores information readable by a computer system. Furthermore, a computer-readable recording medium can be distributed across network-connected computer systems, allowing the computer-readable program or code to be stored and executed in a distributed manner.

[0153] Additionally, the computer-readable recording medium may include hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, flash memory, etc. The program instructions may include not only machine language codes produced by a compiler, but also high-level language codes that can be executed by a computer using an interpreter, etc.

[0154] While some aspects of the present disclosure have been described in the context of a device, they may also represent a description of a corresponding method, wherein a block or device corresponds to a method step or a feature of a method step. Similarly, aspects described in the context of a method may also be described as a corresponding block or item or a feature of a corresponding device. Some or all of the method steps may be performed by (or using) a hardware device, such as, for example, a microprocessor, a programmable computer, or an electronic circuit. In some embodiments, at least one or more of the most important method steps may be performed by such a device.

[0155] In embodiments, a programmable logic device (e.g., a field-programmable gate array) may be used to perform some or all of the functions of the methods described herein. In embodiments, the field-programmable gate array may operate in conjunction with a microprocessor to perform one of the methods described herein. In general, the methods are preferably performed by some hardware device.

[0156] Although the present disclosure has been described with reference to preferred embodiments thereof, it will be understood by those skilled in the art that various modifications and changes may be made to the present disclosure without departing from the spirit and scope of the present disclosure as set forth in the claims below.

Claims

1. As a method of the first communication node, A step of generating a first base sequence using a Zadoff-Chu sequence; A step of generating a second base sequence using the first base sequence; and A step of generating a synchronization signal sequence by performing half / half distributed concatenation of the first base sequence and the second base sequence, Method of the first communication node.

2. In claim 1, The second base sequence is a sequence having an opposite polarity to the first base sequence. Method of the first communication node.

3. In claim 1, A step of mapping modulation symbols generated by modulating the above synchronization signal sequence to 2(M+1) subcarriers; and A step of transmitting a synchronization signal composed of the above mapped modulation symbols to a second communication node, The above M is a natural number, Method of the first communication node.

4. In claim 3, The step of mapping the modulation symbols generated by modulating the above-mentioned synchronization signal sequence to 2(M+1) subcarriers is as follows: A step of sequentially mapping the first base sequence constituting the synchronization signal sequence in the frequency domain to odd subcarrier indices of the 2(M+1) subcarriers; and A method of a first communication node, comprising a step of sequentially mapping the second base sequence constituting the synchronization signal sequence to even subcarrier indices of the 2(M+1) subcarriers.

5. In claim 3, At least one null subcarrier is placed around the (2M+1) subcarriers or between the 2(M+1) subcarriers. Method of the first communication node.

6. In claim 1, The step of generating a synchronization signal sequence by distributing and combining the first base sequence and the second base sequence in half is as follows: A step of generating a first partial base sequence and a second partial base sequence of the first base sequence; A step of generating a third partial base sequence and a fourth partial base sequence of the second base sequence; A step of arranging elements of the first partial base sequence in the upper available subcarrier group with one subcarrier spaced apart, and arranging elements of the third partial base sequence between elements of the first partial base sequence; and Including a step of generating the synchronization signal sequence by performing half-and-half distributed aggregation in which elements of the second partial base sequence are spaced by one subcarrier in the lower available subcarrier group and elements of the fourth partial base sequence are placed between the elements of the second partial base sequence. Method of the first communication node.

7. As a method of the second communication node, A step of receiving a synchronization signal based on a synchronization signal sequence formed by a first base sequence and a second base sequence from a first communication node; A step of calculating a cross-correlation value for the received synchronization signal; A step of providing the above cross-correlation value as input to a deep learning model; A step of estimating a synchronous sample point by the above deep learning model; and A step of obtaining a synchronization signal using the above synchronization sample point, Method of the second communication node.

8. In claim 7, The above synchronization signal is based on the synchronization signal sequence formed by half / half distributed concatenation of the first base sequence and the second base sequence. Method of the second communication node.

9. In claim 7, The first base sequence is based on the Zadov-Chu sequence, and the second base sequence has an opposite polarity to the first base sequence. Method of the second communication node.

10. In claim 7, The above deep learning model includes a convolutional layer and a fully connected layer, wherein the convolutional layer includes a ReLU (rectified linear unit) activation function, and the fully connected layer includes a softmax activation function. Method of the second communication node.

11. In claim 7, The deep learning model includes a first convolutional layer, a second convolutional layer, and a fully connected layer, wherein the first convolutional layer and the second convolutional layer include a ReLU activation function, and the fully connected layer includes a softmax activation function. Method of the second communication node.

12. In claim 7, The step of estimating the synchronous sample point by the above deep learning model is, A step of estimating candidate synchronous sample points using a convolutional layer of the above deep learning model; and A step of estimating the synchronous sample point from the candidate synchronous sample points using a fully connected layer of the deep learning model, Method of the second communication node.

13. As a second communication node, comprising at least one processor, wherein said at least one processor comprises said second communication node; Receive a synchronization signal based on a synchronization signal sequence formed by a first base sequence and a second base sequence from a first communication node; Calculating a cross-correlation value for the received synchronization signal; The above cross-correlation values ​​are provided as input to the deep learning model; Estimating the synchronous sample point by the above deep learning model; and Causing the synchronization signal to be acquired using the above synchronization sample point, Second communication node.

14. In claim 13, The above synchronization signal is based on the synchronization signal sequence formed by half / half distributed concatenation of the first base sequence and the second base sequence. Second communication node.

15. In claim 13, The first base sequence is based on the Zadov-Chu sequence, and the second base sequence has an opposite polarity to the first base sequence. Second communication node.

16. In claim 13, The above deep learning model includes a convolutional layer and a fully connected layer, wherein the convolutional layer includes a ReLU (rectified linear unit) activation function, and the fully connected layer includes a softmax activation function. Second communication node.

17. In claim 13, In order to estimate the synchronous sample point by the deep learning model, the at least one processor is configured to cause the second communication node to: Estimating candidate synchronous sample points using the convolutional layer of the above deep learning model; and Causing the candidate synchronous sample points to be estimated from the fully connected layer of the deep learning model, Second communication node.

Images