Communication signal type identification method and device, computer device, and storage medium

Abstract

The application relates to a communication signal type identification method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring a plurality of channel state information corresponding to a communication signal of a terminal device at a target sampling point and a base station, each channel state information corresponding to an antenna of the base station; obtaining initial channel impulse response information corresponding to each channel state information according to each channel state information; obtaining a channel map of the target sampling point according to the initial channel impulse response information corresponding to each channel state information; inputting the channel map of the target sampling point into a type identification model, and acquiring an identification result output by the type identification model, the identification result being used for determining a propagation type of the communication signal of the target sampling point, and the type identification model being generated after being trained by taking historical channel maps and type labels at different sampling points as a sample set. The method can improve the efficiency and real-time performance of the identification result.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to a method, apparatus, computer equipment, storage medium, and computer program product for identifying the type of communication signal. Background Technology

[0002] In real-world environments, such as urban settings, buildings, and underground tunnels, obstacles like walls, people, and doors can block and reflect wireless signals, resulting in non-line-of-sight (NLOS) propagation. Compared to line-of-sight (LOS) propagation, NLOS propagation involves additional time and distance, leading to signal arrival angle deviations and additional power losses. NLOS propagation issues exist in many applications, such as positioning, direction finding, ranging, and inter-base station clock calibration. When NLOS propagation exists between terminal devices (e.g., User Equipment) and base stations, directly using NLOS propagation measurement information without any preprocessing will significantly reduce positioning or calibration accuracy. Therefore, accurately identifying NLOS propagation is crucial for wireless communication systems that rely on LOS propagation.

[0003] In traditional technologies, the identification of non-line-of-sight propagation and line-of-sight propagation types based on artificial intelligence is often carried out by extracting statistical feature values ​​of channel state information and inputting them into machine learning models. However, the method based on spectral input has high computational complexity, resulting in low signal identification efficiency and difficulty in meeting the real-time requirements of actual systems, thus making it difficult to use in actual deployment environments. Summary of the Invention

[0004] Therefore, it is necessary to provide a communication signal type identification method, apparatus, computer equipment, computer-readable storage medium, and computer program product that can improve signal identification efficiency in response to the above-mentioned technical problems.

[0005] In a first aspect, this application provides a method for identifying the type of communication signal, including:

[0006] The terminal device acquires multiple channel state information corresponding to the communication signal between the terminal device and the base station at the target sampling point, and each channel state information corresponds to one antenna of the base station.

[0007] Based on each channel state information, the initial channel impulse response information corresponding to each channel state information is obtained;

[0008] Based on the initial channel impulse response information corresponding to each channel state information, the channel map of the target sampling point is obtained;

[0009] The channel map of the target sampling point is input into the type recognition model, and the recognition result output by the type recognition model is obtained. The recognition result is used to determine the propagation type of the communication signal at the target sampling point. The type recognition model is generated after training with historical channel maps and type labels at different sampling points as sample sets.

[0010] In one embodiment, obtaining the channel map of the target sampling point based on the initial channel impulse response information corresponding to each channel state information includes:

[0011] Obtain the amplitude information of each initial channel impulse response;

[0012] The amplitude information of each initial channel impulse response is processed by time delay normalization and amplitude normalization to obtain the normalized initial channel impulse response information;

[0013] Denoising is performed on multiple normalized initial channel impulse response information to obtain multiple target channel impulse response information;

[0014] The channel impulse response information of the multiple target channels is spliced ​​together to obtain the channel map of the target sampling points.

[0015] In one embodiment, the step of performing time delay normalization and amplitude normalization processing on the amplitude information of each initial channel impulse response information to obtain normalized initial channel impulse response information includes:

[0016] Based on the amplitude information of each initial channel impulse response information, determine the maximum amplitude value of each initial channel impulse response information and the sampling point corresponding to the maximum amplitude value of each initial channel impulse response information;

[0017] Based on the sampling point corresponding to the maximum amplitude value of each initial channel impulse response information, the sampling point corresponding to the maximum amplitude value of each initial channel impulse response information is cyclically shifted to the first sampling point to normalize the time delay of each initial channel impulse response information.

[0018] The amplitude information of each initial channel impulse response is divided by the maximum amplitude value to normalize the amplitude of each initial channel impulse response.

[0019] In one embodiment, the step of concatenating the plurality of target channel impulse response information to obtain the channel map of the target sampling points includes:

[0020] Obtain the transmission channel corresponding to each target channel impulse response information; the transmission channel is used to transmit communication signals between the terminal device at the target sampling point and any antenna of the base station;

[0021] By splicing the target channel impulse response information corresponding to the target sampling point along the transmission channel, a channel map corresponding to the target sampling point is obtained.

[0022] In one embodiment, the channel map of the target sampling points is input into the type recognition model, and the recognition result output by the type recognition model is obtained, including:

[0023] Feature values ​​are extracted from the channel spectrum of the input target sampling points to obtain the feature matrix;

[0024] Flattening the feature matrix yields a one-dimensional feature vector of the channel spectrum;

[0025] The recognition result is determined based on the preset weights and bias matrices, as well as the one-dimensional feature vector.

[0026] In one embodiment, before inputting the channel map of the target sampling point into the type identification model and obtaining the identification result output by the type identification model, the method further includes:

[0027] Obtain historical communication signals from different sampling points and type labels corresponding to the different sampling points respectively. The type labels are used to indicate whether the propagation type of the historical communication signal is line-of-sight propagation or non-line-of-sight propagation.

[0028] Channel estimation is performed on the historical communication signals at the different sampling points to generate a set of historical channel state information.

[0029] Based on the set of historical channel state information, a historical channel map of the different sampling points is generated;

[0030] The type recognition model is trained using the historical channel maps of the different sampling points and the type labels corresponding to the different sampling points as sample sets.

[0031] Secondly, this application also provides a communication signal type identification device, comprising:

[0032] The first acquisition module is used to acquire multiple channel state information corresponding to the communication signal between the terminal device and the base station at the target sampling point, and each channel state information corresponds to one antenna of the base station.

[0033] The conversion module is used to obtain the initial channel impulse response information corresponding to each channel state information based on each channel state information.

[0034] The channel map generation module is used to obtain the channel map of the target sampling point based on the initial channel impulse response information corresponding to each channel state information.

[0035] The identification result generation module is used to input the channel map of the target sampling point into the type identification model and obtain the identification result output by the type identification model. The identification result is used to determine the propagation type of the communication signal of the target sampling point. The type identification model is generated after training with historical channel maps and type labels at different sampling points as sample sets.

[0036] In one embodiment, the type recognition model includes a lightweight residual network model, the lightweight residual network including residual modules, a flattening layer and a fully connected layer, the residual modules and the fully connected layer being connected through the flattening layer;

[0037] The residual module is used to extract the feature values ​​of the channel spectrum to obtain a feature matrix, and input the feature matrix into the flattening layer.

[0038] The flattening layer is used to flatten the feature matrix to obtain a one-dimensional feature vector of the channel spectrum, and the one-dimensional feature vector is input to the fully connected layer.

[0039] The fully connected layer is used to determine the recognition result based on the preset weights and bias matrices, as well as the one-dimensional feature vector.

[0040] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method.

[0041] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method.

[0042] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the above-described method.

[0043] The aforementioned communication signal type identification method, apparatus, computer equipment, storage medium, and computer program product first acquire multiple channel state information corresponding to the communication signal between the terminal device and the base station at the target sampling point, with each channel state information corresponding to one antenna of the base station; secondly, based on each channel state information, initial channel impulse response information corresponding to each channel state information is obtained; thirdly, based on the initial channel impulse response information corresponding to each channel state information, the channel spectrum of the target sampling point is obtained; finally, the channel spectrum of the target sampling point is input into a type identification model, and the identification result output by the type identification model is obtained. The identification result is used to determine the propagation type of the communication signal at the target sampling point. The type identification model is generated after training with historical channel spectra and type labels at different sampling points as a sample set. This method performs low-complexity computational processing on multiple channel state information to obtain the channel spectrum of the target sampling point. By inputting the channel spectrum of the target sampling point into a pre-trained type identification model, the computational processing of the channel spectrum of the target sampling point by the type identification model is reduced, thereby obtaining the identification result output by the type identification model and improving the efficiency and real-time performance of the identification result. Attached Figure Description

[0044] To more clearly illustrate the technical solutions in the embodiments or related technologies of this application, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is an application environment diagram of a communication signal type identification method in one embodiment;

[0046] Figure 2 This is a flowchart illustrating a method for identifying the type of communication signal in one embodiment;

[0047] Figure 3 This is a flowchart illustrating the process of obtaining the channel map of the target sampling point based on the initial channel impulse response information corresponding to each channel state information in one embodiment.

[0048] Figure 4 This is a waveform diagram of the initial channel impulse response information TN_CIR after normalization and noise reduction truncation processing in one embodiment, propagated under LOS and NLOS.

[0049] Figure 5 This is a schematic diagram of the channel map in one embodiment;

[0050] Figure 6This is a flowchart illustrating the process of performing time delay normalization and amplitude normalization on the amplitude information of each initial channel impulse response in one embodiment.

[0051] Figure 7 This is a flowchart illustrating the process of stitching together multiple target channel impulse response information to obtain a channel map of target sampling points in one embodiment.

[0052] Figure 8 This is a flowchart illustrating the process of inputting the channel map of the target sampling point into a type recognition model and obtaining the recognition result output by the type recognition model in one embodiment.

[0053] Figure 9 This is a schematic diagram of a lightweight residual network in one embodiment;

[0054] Figure 10 This is a schematic diagram of the residual module structure in one embodiment;

[0055] Figure 11 In another embodiment, a flowchart is shown before the channel map of the target sampling point is input into the type recognition model and the recognition result output by the type recognition model is obtained;

[0056] Figure 12 This is a structural block diagram of a communication signal type identification device in one embodiment;

[0057] Figure 13 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0059] The communication signal type identification method provided in this application embodiment can be applied to, for example... Figure 1In the application environment shown, terminal device 102 communicates with base station 104 via a network. Terminal device 102 acquires multiple channel state information corresponding to the communication signal between the terminal device and the base station at the target sampling point, with each channel state information corresponding to one antenna of the base station. Based on each channel state information, it obtains the initial channel impulse response information corresponding to each channel state information. Based on the initial channel impulse response information corresponding to each channel state information, it obtains the channel map of the target sampling point. The channel map of the target sampling point is input into a type recognition model, and the recognition result output by the type recognition model is obtained. The recognition result is used to determine the propagation type of the communication signal at the target sampling point. The type recognition model is generated after training with historical channel maps and type labels at different sampling points as a sample set. Terminal device 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can be smart speakers, smart TVs, smart air conditioners, smart in-vehicle devices, etc. Portable wearable devices can be smartwatches, smart bracelets, head-mounted devices, etc. The communication signal type identification method involved in the embodiments of this application can be applied to a communication signal type identification device, which can be the aforementioned terminal device, or a base station or other device.

[0060] In one exemplary embodiment, such as Figure 2 As shown, a method for identifying the type of communication signal is provided, which can be applied to... Figure 1 Taking terminal device 102 as an example, the explanation includes the following steps S202 to S208. Wherein:

[0061] Step S202: Obtain multiple channel state information corresponding to the communication signal between the terminal device and the base station at the target sampling point. Each channel state information corresponds to one antenna of the base station.

[0062] A base station, or public mobile communication base station, is a type of radio station that transmits and receives radio signals within a defined radio coverage area, communicating with terminal devices through a mobile communication switching center. The primary function of a base station is to provide wireless coverage, enabling wireless signal transmission between wired communication networks and wireless terminals. Base stations use antennas to send and receive messages. One base station can have one or more antennas.

[0063] An antenna is a transducer that converts guided waves propagating on a transmission line into electromagnetic waves propagating in an unbounded medium (usually free space), or vice versa. It is a component in wireless equipment used to transmit or receive electromagnetic waves. Engineering systems such as radio communication, broadcasting, television, radar, navigation, electronic countermeasures, remote sensing, and radio astronomy—anything that uses electromagnetic waves to transmit information—rely on antennas. Furthermore, antennas are also needed for non-signal energy radiation in the transmission of energy using electromagnetic waves. Generally, antennas are reversible, meaning the same antenna can be used as both a transmitting and receiving antenna. The basic characteristic parameters of the same antenna are identical whether it is used for transmitting or receiving.

[0064] Channel state information (CSA) is a channel attribute of a communication link, describing the signal attenuation factors along each transmission path, such as signal scattering, environmental fading (multipath fading or shadowing fading), and power decay of distance. CSA enables communication systems to adapt to current channel conditions, ensuring high-reliability and high-speed communication in multi-antenna systems.

[0065] Optionally, the terminal device includes a user equipment (UE). The communication signals between the terminal device and the base station at the target sampling point are acquired. These communication signals can be signals transmitted from the UE to the base station at the target sampling point, and received by the base station's antenna; or they can be signals transmitted from the base station's antenna to the UE, and received by the UE at the target sampling point. The communication signals can be either line-of-sight or non-line-of-sight propagation. The target sampling point where the UE is located can be mobile or stationary. Channel estimation is performed on the communication signals between the UE at the target sampling point and multiple antennas of the base station to obtain multiple channel state information entries, each corresponding to one antenna of the base station. The terminal device acquires the multiple channel state information entries corresponding to the communication signals between the terminal device and the base station at the target sampling point.

[0066] Optionally, the terminal device acquires multiple channel state information corresponding to the communication signals between the terminal device and the base station at the target sampling point as a dataset, with each channel state information corresponding to one antenna of the base station. The dataset consists of the location of the base station, the unique identifier of the base station antenna, and the channel estimation matrix of the communication signals between the user equipment at the sampling point. Specifically, it is assumed that the number of antennas in the base station is N. ant The antenna receives signals and performs channel estimation to obtain Channel State Information (CSI), and then constructs a dataset. in Let represent the channel estimation matrix between the m-th antenna of the base station at the p-th location and the user equipment.

[0067] Step S204: Based on each channel state information, obtain the initial channel impulse response information corresponding to each channel state information.

[0068] Channel impulse response information refers to the signal energy information of a communication signal arriving at the receiver after different times (different propagation paths result in different propagation times). In other words, it reflects the influence on the communication signal during transmission through the channel, demonstrating the channel's signal processing characteristics. It should be noted that both channel state information and channel impulse response information are representations of channel signal information, and both provide the same information.

[0069] Optionally, the terminal device performs an inverse fast fourier transform (IFFT) on each channel state information in the dataset according to formula (1), thereby converting the channel state information into the initial channel impulse response information corresponding to each channel state information.

[0070] CIR(n) = IFFT(CSI(k)) (1)

[0071] Where CIR(n) represents the initial channel impulse response information, and n is the sampling point number; CSI(k) represents the channel state information, and k is the subcarrier number; IFFT represents the inverse fast Fourier transform.

[0072] Step S206: Obtain the channel map of the target sampling point based on the initial channel impulse response information corresponding to each channel state information.

[0073] Among them, the channel spectrum is used to indicate the channel impulse response information of the sample signal between the target sampling point and the base station of the terminal device. It can be a one-dimensional channel spectrum or a multi-dimensional channel spectrum.

[0074] Optionally, the terminal device has multiple signal transmission channels between the target sampling point and the base station. Accordingly, based on the target sampling point where the terminal device is located, the terminal device sorts and combines the channel impulse response information of the sample signals in the multiple channel transmission channels according to certain rules, thereby constructing a channel map.

[0075] Step S208: Input the channel map of the target sampling point into the type recognition model and obtain the recognition result output by the type recognition model. The recognition result is used to determine the propagation type of the communication signal at the target sampling point. The type recognition model is generated after training with historical channel maps and type labels at different sampling points as sample sets.

[0076] The type recognition model is generated after training on a sample set based on historical channel maps and type labels from different sampling points. The recognition results are used to determine the propagation type of the communication signal at the target sampling point; the propagation type can be either line-of-sight (LOS) or non-line-of-sight (NLOS). The type label indicates the truth value label of the communication signal between the target sampling point where the terminal device is located and the base station. The terminal device marks the LOS / NLOS truth values ​​between the terminal device and the base station at different P sampling points as a label set Y = {Ind}. p}, p = 1, 2, 3…P, where Ind p = (y1, y2), used to distinguish between LOS and NLOS. Without loss of generality, for example, when the terminal device propagates between a sampling point and the base station in LOS mode, Indp = (1, 0), and when it propagates in NLOS mode, Indp = (0, 1).

[0077] Optionally, the terminal device can construct a type recognition model based on convolutional neural networks (CNNs), recurrent neural networks (RNNs), connection-temporal classification (CTC) algorithms, or lightweight neural network models and other machine learning models. The terminal device constructs a sample set based on historical channel maps and type labels from different sampling points, and inputs the sample set into the type recognition model to train the model.

[0078] Furthermore, the terminal device inputs the channel map of the target sampling point into the trained type recognition model to obtain the propagation type of the communication signal at the target sampling point. The propagation type can be LOS or NLOS. The target sampling point can be a base station or a terminal device.

[0079] Furthermore, the terminal device acquires real-time CSI based on the base station or the terminal device as sampling points and constructs a channel map. This channel map is then input into a type recognition model deployed on the base station or terminal device to perform LOS / NLOS identification on the real-time measured CSI. The type recognition model outputs... assumed Represents the probability of LOS. Representing the NLOS probability, LOS / NLOS is determined according to formula (2):

[0080]

[0081] When the output result is When the LOS probability is greater than the NLOS probability, the signal propagation type is considered to be LOS; when the output result is... If the LOS probability is less than the NLOS probability, then the signal propagation type is considered to be NLOS.

[0082] In the aforementioned communication signal type identification method, firstly, multiple channel state information corresponding to the communication signal between the terminal device and the base station at the target sampling point is acquired, with each channel state information corresponding to one antenna of the base station; secondly, based on each channel state information, initial channel impulse response information corresponding to each channel state information is obtained; thirdly, based on the initial channel impulse response information corresponding to each channel state information, the channel map of the target sampling point is obtained; finally, the channel map of the target sampling point is input into a type identification model, and the identification result output by the type identification model is obtained. The identification result is used to determine the propagation type of the communication signal at the target sampling point. The type identification model is generated after training with historical channel maps and type labels at different sampling points as a sample set. This method performs low-complexity computation on multiple channel state information to obtain the channel map of the target sampling point. By inputting the channel map of the target sampling point into a pre-trained type identification model, the computational processing of the channel map of the target sampling point by the type identification model is reduced, thereby obtaining the identification result output by the type identification model and improving the efficiency and real-time performance of the identification result.

[0083] In one exemplary embodiment, such as Figure 3 As shown, the channel map of the target sampling point is obtained based on the initial channel impulse response information corresponding to each channel state information, including steps S302 to S306.

[0084] in:

[0085] Step S302: Obtain the amplitude information of each initial channel impulse response.

[0086] Among them, amplitude information is information that indicates the stability of the initial channel impulse response information.

[0087] Optionally, the initial channel impulse response information corresponds to a set of sampling points, and the amplitude of all sampling points in the initial channel impulse response information is taken as the amplitude of the channel impulse response information. That is, each initial channel impulse response information corresponds to a set of amplitudes, and the number of amplitudes of each initial channel impulse response information is consistent with the number of corresponding sampling points. The terminal device obtains the amplitude information of each initial channel impulse response information. Specifically, it is calculated according to formula (3):

[0088] CIR amp (n)=abs(CIR(n)) (3)

[0089] Where abs(·) represents the amplitude, CIR amp(n) represents the amplitude information of the initial channel impulse response.

[0090] Step S304: Perform time delay normalization and amplitude normalization on the amplitude information of each initial channel impulse response information to obtain the normalized initial channel impulse response information.

[0091] Optionally, in order to eliminate random jitter in time delay and instability in channel impulse response information, the terminal device performs time delay normalization and amplitude normalization processing on the amplitude information of each initial channel impulse response information to obtain the normalized initial channel impulse response information.

[0092] After normalization, the influence of unstable factors is eliminated from the initial channel impulse response information. It should be noted that the order of time delay normalization and amplitude normalization of the amplitude information of each initial channel impulse response is not limited in the processing of the above initial channel impulse response information.

[0093] Step S306: Denoise the multiple normalized initial channel impulse response information to obtain multiple target channel impulse response information.

[0094] After obtaining the normalized initial signal impulse response information, to eliminate the noise component of each channel impulse response information, it is necessary to determine the number of sampling points for each initial signal impulse response signal based on preset sampling points, truncate the initial channel impulse response information, and obtain the target channel impulse response information. It should be noted that the value of this sampling point needs to be determined based on the actual usage environment. The basic principle is to remove noise sampling points to the greatest extent possible while preserving the signal, thereby reducing noise redundancy and lowering the input dimension of the subsequent type recognition model.

[0095] Optionally, the terminal device truncates the initial channel impulse response information after multiple normalization processes. Specifically, the truncation method is shown in formula (4):

[0096] TN_CIR = CIR apm_norm (1:N) cut (4)

[0097] Where, N cut CIR indicating reservation apm_norm The number of sampling points is a pre-set fixed value; the specific value needs to be determined based on the actual usage environment. (CIR) amp_norm TN_CIR represents the initial channel impulse response information after normalization; TN_CIR represents the truncated normalized channel impulse response, which is also the target channel impulse response information.

[0098] Based on the sample index, the initial channel impulse response information (CIR) after normalization and denoising truncation from the target sampling point is determined, i.e., the waveform of TN_CIR under LOS and NLOS propagation is as follows. Figure 4 As shown in the figure, there is a significant difference in TN_CIR between LOS and NLOS. The signal energy is more concentrated under LOS propagation with a clear main peak, while the signal energy is dispersed under NLOS propagation, exhibiting a multi-peak state. This difference can provide key information for subsequent type identification models.

[0099] Step S308: The impulse response information of multiple target channels is spliced ​​together to obtain the channel spectrum of the target sampling points.

[0100] Optionally, the terminal device acquires multiple target channel impulse response information between the target sampling point where the terminal device is located and multiple antennas of the base station, and splices the multiple target channel impulse response information according to rules to obtain the channel map of the target sampling point where the terminal device is located.

[0101] Furthermore, the terminal device acquires the target channel impulse response information between the target sampling point where the terminal device is located and antennas 1, 2, ..., n. This target channel impulse response information is then combined according to the antenna arrangement order to form a channel map of the target sampling point, such as... Figure 5 As shown.

[0102] In this embodiment, the target channel impulse response information is obtained by performing time delay normalization, amplitude normalization, and denoising on the amplitude information of each initial channel impulse response. This target channel impulse response information is then stitched together to form a channel map of the target sampling points. The low-complexity normalization and denoising processes eliminate the time delay and instability between the initial channel impulse response information. The truncation process ensures the validity of each initial channel impulse response information and eliminates redundant data. The target channel impulse response information is then stitched together to form a channel map, which is used as input samples for the type recognition model. This reduces the amount of data processed by the type recognition model and improves its real-time performance and efficiency.

[0103] In one exemplary embodiment, such as Figure 6 As shown, the amplitude information of each initial channel impulse response is processed by time delay normalization and amplitude normalization to obtain the normalized initial channel impulse response information, including steps S602 to S606. Wherein:

[0104] Step S602: Based on the amplitude information of each initial channel impulse response information, determine the maximum amplitude value of each initial channel impulse response information and the sampling point corresponding to the maximum amplitude value of each initial channel impulse response information.

[0105] Optionally, the terminal device determines the maximum amplitude value (max(CIR)) of each initial channel impulse response information based on the amplitude information of each initial channel impulse response information. amp And the sampling points (n) corresponding to the maximum amplitude value of each initial channel impulse response information. max ).

[0106] Furthermore, the terminal device determines the maximum amplitude value based on the amplitude information of each initial channel impulse response, denoted as max(CIR). amp ), where max(·) represents taking the maximum value.

[0107] Furthermore, to eliminate random jitter in the time delay, the terminal device selects the maximum amplitude from the amplitudes corresponding to multiple sampling points of the initial channel impulse response information, and determines the sampling point corresponding to the maximum amplitude value of the initial channel impulse response information as the sampling point corresponding to the maximum amplitude value. The sampling point n corresponding to the maximum amplitude value is determined as follows. max The process is characterized by formula (5):

[0108] n max =argmax(CIR amp (n)) (5)

[0109] Among them, CIR amp This represents the amplitude information of the initial channel impulse response, and argmax(·) represents taking the CIR. amp The location of the sampling point corresponding to the maximum amplitude value.

[0110] Step S604: Based on the sampling point corresponding to the maximum amplitude value of each initial channel impulse response information, the sampling point corresponding to the maximum amplitude value of each initial channel impulse response information is cyclically shifted to the first sampling point to normalize the time delay of each initial channel impulse response information.

[0111] Optionally, the terminal device cyclically shifts the sampling points corresponding to the maximum amplitude values ​​of each initial channel impulse response information to the n1th sampling point, based on the sampling points corresponding to the maximum amplitude values ​​of each initial channel impulse response information, in order to eliminate the time delay between each initial channel impulse response information. The cyclic shift operation can be characterized by formula (6):

[0112] CIR amp =cirshift(CIR) amp -n max +n1) (6)

[0113] Where cirshift(x(n), n0) means cyclically shifting the sequence x(n) to the right by n0 points, n max This represents the sampling point corresponding to the maximum amplitude value. The parameter n1 can be selected according to the actual application environment. In order to ensure CIR after the cyclic displacement operation... amp The integrity of the main diameter is usually greater than 1.

[0114] Step S606: Divide the amplitude information of each initial channel impulse response information by the maximum amplitude value to normalize the amplitude of each initial channel impulse response information.

[0115] Optionally, to eliminate channel impulse response instability, the terminal device obtains the amplitude of each initial channel impulse response and the maximum amplitude value of each initial channel impulse response, then divides the amplitude of each initial channel impulse response by the maximum amplitude value to obtain the amplitude normalization result of each initial channel impulse response. The amplitude normalization process is characterized by formula (7):

[0116] CIR amp_norm =(CIR) amp / max(CIR amp (7)

[0117] Where max(CIR) amp () represents the maximum amplitude value of each initial channel impulse response information, CIR amp The CIR represents the amplitude value of each initial channel impulse response. amp_norm This represents the amplitude normalization result of the initial channel impulse response information.

[0118] After normalization, the influence of unstable factors is eliminated from the initial channel impulse response information. It should be noted that the order of time delay normalization and amplitude normalization of the amplitude information of each initial channel impulse response is not limited in the processing of the above initial channel impulse response information.

[0119] In one exemplary embodiment, such as Figure 7 As shown, multiple target channel impulse response information are spliced ​​together to obtain the channel spectrum of the target sampling points, including steps S702 to S704. Wherein:

[0120] Step S702: Obtain the transmission channel corresponding to each target channel impulse response information; the transmission channel is used to transmit communication signals between the terminal device at the target sampling point and any antenna of the base station.

[0121] The transmission channel is used to transmit communication signals between the terminal device at the target sampling point and any antenna of the base station.

[0122] Optionally, the terminal device acquires the transmission channel corresponding to each target channel impulse response information. For example, when the terminal device is located at target sampling point p, the target channel impulse response information corresponds to the 1st to the mth antennas of the base station. Accordingly, the terminal device acquires one target channel impulse response information corresponding to a transmission channel between the target sampling point p where the terminal device is located and the 1st antenna; one target channel impulse response information corresponding to a transmission channel between the target sampling point p where the terminal device is located and the 2nd antenna; and one target channel impulse response information corresponding to a transmission channel between the target sampling point p where the terminal device is located and the mth antenna.

[0123] Step S704: The target channel impulse response information corresponding to the target sampling point is spliced ​​together with the transmission channel as the dimension to obtain the channel spectrum corresponding to the target sampling point.

[0124] Optionally, the terminal device splices the target channel impulse response information corresponding to the target sampling point according to the arrangement order of the transmission channels, using the transmission channel as the dimension, to form a structure with N ant A one-dimensional spectrum of channels, with dimension (N) ant ×1×N cut ), where N ant N represents the number of base station antennas; cut CIR indicating reservation apm_norm The number of sampling points, also known as the number of channels in the type recognition model, means that there is a one-to-one correspondence between the transmission channels of the terminal device between the target sampling point and any antenna of the base station and the channels in the type recognition model. This yields the channel map corresponding to the target sampling point, as shown in the example below. Figure 5 As shown.

[0125] In this embodiment, by concatenating the impulse response information of multiple target channels corresponding to the target sampling point along the transmission channel dimension, a one-dimensional channel map corresponding to the target sampling point can be generated. This preserves the time-domain information of the initial CIR, preventing the loss of key information, and fully utilizes the spatial-domain information of multiple antennas, providing rich LOS / NLOS implicit features for subsequent network identification. By inputting the one-dimensional channel map into the type identification model, the computational load of the type identification model on the one-dimensional channel map is reduced, thereby improving the real-time performance and efficiency of the category identification model.

[0126] In one exemplary embodiment, such as Figure 8 As shown, the channel map of the target sampling point is input into the type recognition model, and the recognition result output by the type recognition model is obtained, including steps S802 to S806.

[0127] in:

[0128] Step S802: Extract feature values ​​from the channel spectrum of the input target sampling points to obtain the feature matrix.

[0129] The type recognition model includes a lightweight residual network model, which consists of a residual block, a flattened layer, and a fully connected layer (Fc). The residual block and the fully connected layer are connected through the flattened layer, as shown below. Figure 9 As shown. The residual module is used to extract the feature values ​​of the channel spectrum to obtain the feature matrix, and input the feature matrix into the flattening layer; the flattening layer is used to flatten the feature matrix to obtain a one-dimensional feature vector of the channel spectrum, and input the one-dimensional feature vector into the fully connected layer; the fully connected layer is used to determine the identification results of the two propagation types, LOS and NLOS, according to the preset weights and bias matrices, and the one-dimensional feature vector; the fully connected layer contains two neurons, the output of the first neuron is the probability of predicting LOS, and the output of the second neuron is the probability of predicting NLOS.

[0130] Among them, the residual module is as follows Figure 10 As shown, the model includes two one-dimensional convolutional layers. The first convolutional layer is followed by a batch normalization layer and a first rectified linear unit (ReLU). After the first ReLU, there is an adder and a second ReLU. The adder is used to add the input data of layer 0 of the lightweight residual network model to the output data of layer 4. The specific number of layers is shown in Table 1. Both the first and second ReLUs are used as activation functions for each one-dimensional convolutional layer.

[0131] Table 1 Parameter settings for each layer of the lightweight residual network

[0132]

[0133] Optionally, the terminal device inputs the channel map of the target sampling point into the type identification model. The terminal device performs one-dimensional convolution on the channel map of the input target sampling point through the residual module to extract feature values ​​used to distinguish LOS / NLOS propagation, thereby obtaining a feature matrix.

[0134] Step S804: Flatten the feature matrix to obtain a one-dimensional feature vector of the channel spectrum.

[0135] Optionally, the terminal device flattens the feature matrix through a flattening layer, transforming it into a one-dimensional feature vector to obtain a one-dimensional feature vector of the channel spectrum, and then transmits it to the fully connected layer of the model.

[0136] Step S806: Determine the recognition result based on the preset weights and bias matrices, as well as the one-dimensional feature vector.

[0137] Optionally, the fully connected layer is connected to the flat layer, and the terminal device determines the recognition result based on the preset weight and bias matrix between the flat layer and the fully connected layer, as well as the one-dimensional feature vector output by the flat layer.

[0138] The connection between the planar layer and the fully connected layer is achieved through formula (8):

[0139] Y = W*X + B (8)

[0140] Where W and B represent the weight and bias matrices between the flat layer and the fully connected layer, respectively, and X is the one-dimensional feature vector output by the flat layer. It is the output of the fully connected layer, that is, the output of the lightweight network model.

[0141] For LOS / NLOS classification problems, cross-entropy is used as the loss function, as shown in formulas (9) and (10).

[0142]

[0143]

[0144] Where N is the number of samples, and yi is the type label. It is a lightweight residual network output.

[0145] Optionally, the terminal device will output a lightweight residual network. The cross-entropy of the corresponding type label yi is used as the loss function to determine whether the loss function value meets the target accuracy. When the loss function value does not meet the target accuracy, the weights and bias matrices in the lightweight residual network are updated multiple times. The lightweight residual network outputs the updated weights and bias matrices multiple times. Until the loss function value meets the target accuracy.

[0146] The weights between the flattened layer and the fully connected layer are updated using formula (11):

[0147]

[0148] The bias matrix between the flat layer and the fully connected layer is updated using formula (12):

[0149]

[0150] Where lr is the learning rate, W old and W new These are the old and updated weights, B. old and B newThese are the old and the updated biases, respectively.

[0151] In each training phase, the network weights and biases are updated iteratively multiple times until the network loss function value meets the target accuracy.

[0152] While meeting the target accuracy, the signal propagation type is determined based on the output of the lightweight residual network. For example, the output of the lightweight residual network... assumed Represents the probability of LOS. Representing the NLOS probability, LOS / NLOS is determined according to formula (2):

[0153]

[0154] When the output result is When the LOS probability is greater than the NLOS probability, the signal propagation type is considered to be LOS; when the output result is... If the LOS probability is less than the NLOS probability, then the signal propagation type is considered to be NLOS.

[0155] In this embodiment, the channel map of the target sampling point is input into the type recognition model. After obtaining the recognition result, the one-dimensional channel map is processed using a low-complexity type recognition model. Moreover, the type recognition model has a small computational load, which can reduce the system's memory overhead and improve the efficiency of result output.

[0156] In one exemplary embodiment, such as Figure 11 As shown, before inputting the channel map of the target sampling point into the type recognition model and obtaining the recognition result output by the type recognition model, steps S1102 to S1108 are included. Wherein:

[0157] Step S1102: Obtain historical communication signals from different sampling points and type labels corresponding to different sampling points. The type labels are used to indicate whether the propagation type of the historical communication signals is line-of-sight propagation or non-line-of-sight propagation.

[0158] Among them, the terminal device marks the LOS / NLOS truth values ​​between the terminal device and the base station at different P sampling points as a type label set Y = {Ind p}, p = 1, 2, 3…P, where Ind p = (y1, y2), used to distinguish between LOS and NLOS. Without loss of generality, for example, when the terminal device propagates between a sampling point and the base station in LOS mode, Indp = (1, 0), and when it propagates in NLOS mode, Indp = (0, 1).

[0159] Optionally, the terminal device acquires multiple historical communication signals corresponding to the communication signals between the terminal device and the base station at different sampling points as a dataset. The dataset consists of the location of the base station, the unique identifier of the base station antenna, and the channel estimation matrix of the historical communication signals between the user equipment at the sampling points.

[0160] Furthermore, the historical communication signals of the terminal device at different sampling points are obtained, as well as the type labels corresponding to different sampling points. The category label of the historical communication signal can be LOS or NLOS. If the category label is LOS, it can be recorded as Indp = (1, 0); if the category label is NLOS, it can be recorded as Indp = (0, 1).

[0161] Step S1104: Perform channel estimation on the historical communication signals at different sampling points to generate a set of historical channel state information.

[0162] Optionally, the terminal device performs channel estimation on the historical communication signals between different sampling points and the base station antenna to generate a set of historical channel state information (CSI).

[0163] Step S1106: Generate historical channel maps for different sampling points based on the historical channel state information set.

[0164] Optionally, the terminal device performs an Inverse Fast Fourier Transform (IFFT) on each historical channel state information in the historical channel state information set to convert the historical channel state information into historical initial channel impulse response information corresponding to each historical channel state information; obtains the amplitude information of each historical initial channel impulse response information; performs time delay normalization and amplitude normalization processing on the amplitude information of each historical initial channel impulse response information to obtain normalized historical initial channel impulse response information; performs noise reduction processing on multiple normalized historical initial channel impulse response information to obtain multiple historical target channel impulse response information; and splices the multiple historical target channel impulse response information to generate historical channel maps at different sampling points.

[0165] Step S1108: Use the historical channel maps of different sampling points and the type labels corresponding to different sampling points as sample sets to train the type recognition model.

[0166] Optionally, the terminal device obtains historical channel maps of different sampling points and the corresponding type labels for each sampling point as a training sample set. The historical channel maps are then input into the initial type recognition model, i.e., the initial lightweight network model, to obtain the output of the initial lightweight network model. Determine whether the loss function value meets the target accuracy; where the loss function is the output of a lightweight network model based on historical channel maps from different sampling points. It is constructed by using type labels corresponding to different sampling points.

[0167] When the loss function value does not meet the target accuracy, the weights and bias matrices in the lightweight residual network are updated multiple times. The initial lightweight residual network outputs the updated weights and bias matrices multiple times. Training continues until the loss function value meets the target accuracy, at which point the training ends, resulting in a lightweight network model that has been trained.

[0168] In this embodiment, by using historical channel maps of different sampling points and type labels corresponding to different sampling points as sample sets, a type recognition model is trained, and a trained type recognition model can be obtained. Using the type recognition model to identify signal types improves the accuracy of recognition.

[0169] The validity of this application is demonstrated through two experimental tests. The experiments are based on a 5G positioning system operating in a sub-6GHz frequency band, using a 5G Sounding Reference Signal (SRS) as the positioning signal. The SRS bandwidth is configured to 100MHz, occupying 1632 subcarriers with a subcarrier spacing of 60kHz. A 5G radio remote unit equipped with a 4-element uniform linear array antenna is used as the receiving device during the experiments. In this experiment, the CIR formula is... amp =cirshift(CIR) amp -n max +n i The parameter n1 in the formula is set to 16, and the formula TN_CIR = CIR is used. apm-norm (1:N) cut The parameter N in ) cut Set to 150. As shown in Table 2, the model parameters of the lightweight residual network are 1.314K and the total floating-point operations are 18K.

[0170] Table 2

[0171] 1.314K 18K

[0172] In an exemplary embodiment, to verify the effectiveness of this application under static conditions for the user equipment (UE), 310 different sampling points for LOS and 310 different sampling points for NLOS were set up in a typical obstacle environment (including pillars, walls, vehicles, etc.) in an indoor parking lot. Each sampling point collected 10 frames of data. The training set was constructed using the aforementioned 3100 frames of LOS and 3100 frames of NLOS data. One week later, 100 different sampling points for LOS and 100 different sampling points for NLOS were randomly set up again, with 100 frames of data collected at each sampling point. The test set was constructed using the aforementioned 10000 frames of LOS and 10000 frames of NLOS data. After 15 iterations of learning, the final recognition accuracy of the lightweight residual network is shown in Table 3, with an LOS recognition accuracy of 0.93 and an NLOS recognition accuracy of 0.90.

[0173] Table 3

[0174] 0.93 0.90

[0175] In another exemplary embodiment, to verify the effectiveness of this application under UE movement conditions, in a typical obstacle environment of an indoor parking lot (including pillars, walls, etc.), the UE moves at a non-uniform speed of 1-3 m / s, with a data acquisition rate of 40 ms / frame, collecting a total of 17,979 frames of LOS data and 18,033 frames of NLOS data, forming the training set. Under the same conditions on the same day, an additional 11,908 frames of LOS data and 22,687 frames of NLOS data were collected, forming test set 1. Two months later, after significant changes in the distribution of parked vehicles in the parking lot and multiple power cycles of the positioning system, the UE was again moved at a speed of 1-3 m / s, and 11,885 frames of LOS data and 21,304 frames of NLOS data were collected at a acquisition rate of 40 ms / frame, forming test set 2. The recognition accuracy of the lightweight residual network trained on the training set on test set 1 and test set 2 is shown in Table 4. The LOS recognition accuracy of test set 1 is 0.924 and the NLOS recognition accuracy is 0.910; the LOS recognition accuracy of test set 2 is 0.911 and the NLOS recognition accuracy is 0.917.

[0176] Table 4

[0177] Test set 1 0.924 0.910 Test Set 2 0.911 0.917

[0178] The above recognition accuracy is the recall rate in classification problems, i.e.

[0179] LOS recognition accuracy = Number of correctly identified LOS / Total number of LOS;

[0180] NLOS recognition accuracy = Number of correctly identified NLOS / Total number of NLOS;

[0181] The above experiments demonstrate that in a typical indoor parking environment, a LOS / NLOS recognition accuracy of over 90% can be achieved regardless of whether the terminal is stationary or moving. Even when the distribution of obstacles in the parking lot changes or the device is powered on or off, the recognition accuracy remains almost unchanged, indicating that this application has strong generalization ability to environmental and hardware changes. Furthermore, the lightweight residual network of this application has only 18K model parameters, which helps save memory overhead, and the floating-point operation amount is only 1.314K, far less than commonly used deep networks, making it suitable for systems with high real-time requirements.

[0182] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0183] Based on the same inventive concept, this application also provides a communication signal type identification device for implementing the communication signal type identification method described above. The solution provided by this device is similar to the implementation described in the above method; therefore, the specific limitations in one or more communication signal type identification device embodiments provided below can be found in the limitations of the communication signal type identification method described above, and will not be repeated here.

[0184] In one exemplary embodiment, such as Figure 12 As shown, a communication signal type identification device is provided, comprising: a first acquisition module 1201, a conversion module 1202, a channel map generation module 1203, and an identification result generation module 1204, wherein:

[0185] The first acquisition module 1201 is used to acquire multiple channel status information corresponding to the communication signal between the terminal device and the base station at the target sampling point, and each channel status information corresponds to one antenna of the base station.

[0186] The conversion module 1202 is used to obtain the initial channel impulse response information corresponding to each channel state information based on each channel state information.

[0187] The channel map generation module 1203 is used to obtain the channel map of the target sampling point based on the initial channel impulse response information corresponding to each channel state information.

[0188] The identification result generation module 1204 is used to input the channel map of the target sampling point into the type identification model and obtain the identification result output by the type identification model. The identification result is used to determine the propagation type of the communication signal at the target sampling point. The type identification model is generated after training with historical channel maps and type labels at different sampling points as sample sets.

[0189] In one exemplary embodiment, the type recognition model includes a lightweight residual network model, which includes residual modules, flat modules, and fully connected modules, with the residual modules and fully connected modules connected through the flat modules.

[0190] The residual module is used to extract feature values ​​from the channel spectrum of the input target sampling points to obtain the feature matrix.

[0191] The flattening module is used to flatten the feature matrix to obtain a one-dimensional feature vector of the channel spectrum.

[0192] The flat module is used to determine the recognition result based on the preset weights and bias matrices, as well as the one-dimensional feature vector.

[0193] In one exemplary embodiment, the channel map generation module 1203 includes:

[0194] The amplitude information acquisition unit is used to acquire the amplitude information of each initial channel impulse response.

[0195] The processing unit is used to perform time delay normalization and amplitude normalization on the amplitude information of each initial channel impulse response information to obtain the normalized initial channel impulse response information.

[0196] The denoising unit is used to denoise multiple normalized initial channel impulse response information to obtain multiple target channel impulse response information.

[0197] The splicing unit is used to splice multiple target channel impulse response information to obtain the channel spectrum of the target sampling points.

[0198] In one exemplary embodiment, the processing unit includes:

[0199] The amplitude sampling point determination subunit is used to determine the maximum amplitude value of each initial channel impulse response information and the sampling point corresponding to the maximum amplitude value of each initial channel impulse response information based on the amplitude information of each initial channel impulse response information.

[0200] The delay normalization subunit is used to cyclically shift the sampling point corresponding to the maximum amplitude value of each initial channel impulse response information to the first sampling point, so as to normalize the delay of each initial channel impulse response information.

[0201] The amplitude normalization sub-unit is used to divide the amplitude value of each initial channel impulse response information by the maximum amplitude value in order to normalize the amplitude value of each initial channel impulse response information.

[0202] In one exemplary embodiment, the splicing unit includes:

[0203] The transmission channel acquisition subunit is used to acquire the transmission channel corresponding to each target channel impulse response information; the transmission channel is used to transmit communication signals between the terminal device at the target sampling point and any antenna of the base station.

[0204] The splicing subunit is used to splice the target channel impulse response information corresponding to the target sampling point along the transmission channel dimension to obtain the channel map corresponding to the target sampling point.

[0205] In one exemplary embodiment, the recognition result generation module 1204 includes:

[0206] The feature matrix determination unit is used to extract feature values ​​from the channel spectrum of the input target sampling points to obtain the feature matrix.

[0207] The one-dimensional eigenvector determination unit is used to flatten the eigenma matrix to obtain the one-dimensional eigenvector of the channel spectrum.

[0208] The recognition result determination unit is used to determine the recognition result based on the preset weight and bias matrix and the one-dimensional feature vector.

[0209] Each module in the aforementioned communication signal type identification device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0210] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 13As shown, this computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores communication signal type identification data. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When executed by the processor, the computer program implements a communication signal type identification method.

[0211] Those skilled in the art will understand that Figure 13 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0212] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0213] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0214] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0215] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0216] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0217] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for identifying the type of communication signal, characterized in that, The method includes: The terminal device acquires multiple channel state information corresponding to the communication signal between the terminal device and the base station at the target sampling point, and each channel state information corresponds to one antenna of the base station. Based on each channel state information, the initial channel impulse response information corresponding to each channel state information is obtained; Obtain the amplitude information of each initial channel impulse response; The amplitude information of each initial channel impulse response is processed by time delay normalization and amplitude normalization to obtain the normalized initial channel impulse response information; Denoising is performed on multiple normalized initial channel impulse response information to obtain multiple target channel impulse response information; Obtain the transmission channel corresponding to each target channel impulse response information; the transmission channel is used to transmit communication signals between the terminal device at the target sampling point and any antenna of the base station; The target channel impulse response information corresponding to the target sampling point is concatenated according to the arrangement order of the transmission channels, using the transmission channels as the dimension, to obtain the channel map corresponding to the target sampling point; wherein, the channel map is a one-dimensional channel map with dimension N. ant ×1×N cut N ant N represents the number of base station antennas; cut CIR indicating reservation apm_norm Number of sampling points; CIR apm_norm This represents the initial channel impulse response information after normalization. The channel map of the target sampling point is input into the type recognition model, and the recognition result output by the type recognition model is obtained. The recognition result is used to determine the propagation type of the communication signal at the target sampling point. The type recognition model is generated after training with historical channel maps and type labels at different sampling points as sample sets.

2. The method according to claim 1, characterized in that, The step of performing time delay normalization and amplitude normalization processing on the amplitude information of each initial channel impulse response information to obtain normalized initial channel impulse response information includes: Based on the amplitude information of each initial channel impulse response information, determine the maximum amplitude value of each initial channel impulse response information and the sampling point corresponding to the maximum amplitude value of each initial channel impulse response information; Based on the sampling point corresponding to the maximum amplitude value of each initial channel impulse response information, the sampling point corresponding to the maximum amplitude value of each initial channel impulse response information is cyclically shifted to the first sampling point to normalize the time delay of each initial channel impulse response information. The amplitude information of each initial channel impulse response is divided by the maximum amplitude value to normalize the amplitude of each initial channel impulse response.

3. The method according to claim 1, characterized in that, The step of inputting the channel map of the target sampling point into the type recognition model and obtaining the recognition result output by the type recognition model includes: Feature values ​​are extracted from the channel spectrum of the input target sampling points to obtain a feature matrix; Flattening the feature matrix yields a one-dimensional feature vector of the channel spectrum; The recognition result is determined based on the preset weight and bias matrix, as well as the one-dimensional feature vector.

4. The method according to any one of claims 1-3, characterized in that, Before inputting the channel map of the target sampling point into the type identification model and obtaining the identification result output by the type identification model, the method further includes: Obtain historical communication signals from different sampling points and type labels corresponding to the different sampling points respectively. The type labels are used to indicate whether the propagation type of the historical communication signal is line-of-sight propagation or non-line-of-sight propagation. Channel estimation is performed on the historical communication signals at the different sampling points to generate a set of historical channel state information. Based on the set of historical channel state information, a historical channel map of the different sampling points is generated; The type recognition model is trained using the historical channel maps of the different sampling points and the type labels corresponding to the different sampling points as sample sets.

5. A communication signal type identification device, characterized in that, The device includes: The first acquisition module is used to acquire multiple channel state information corresponding to the communication signal between the terminal device and the base station at the target sampling point, and each channel state information corresponds to one antenna of the base station. The conversion module is used to obtain the initial channel impulse response information corresponding to each channel state information based on each channel state information. A channel map generation module is used to acquire the amplitude information of each initial channel impulse response (ITR); perform time delay normalization and amplitude normalization on the amplitude information of each initial ITR to obtain normalized initial ITR; perform noise reduction on multiple normalized initial ITRs to obtain multiple target ITRs; acquire the transmission channel corresponding to each target ITR; the transmission channel is used to transmit communication signals between the terminal device at the target sampling point and any antenna of the base station; and concatenate the target ITRs corresponding to the target sampling point with the transmission channel as the dimension to obtain the channel map corresponding to the target sampling point; wherein, the channel map is a one-dimensional channel map with dimension N. ant ×1×N cut N ant N represents the number of base station antennas; cut CIR indicating reservation apm_norm Number of sampling points; CIR apm_norm This represents the initial channel impulse response information after normalization. The identification result generation module is used to input the channel map of the target sampling point into the type identification model and obtain the identification result output by the type identification model. The identification result is used to determine the propagation type of the communication signal of the target sampling point. The type identification model is generated after training with historical channel maps and type labels at different sampling points as sample sets.

6. The apparatus according to claim 5, characterized in that, The type identification model includes a lightweight residual network model, which includes residual modules, flat modules, and fully connected modules. The residual modules and the fully connected modules are connected through the flat modules. The residual module is used to extract feature values ​​from the channel spectrum of the input target sampling points to obtain a feature matrix; The flattening module is used to flatten the feature matrix to obtain a one-dimensional feature vector of the channel spectrum; The fully connected module is used to determine the recognition result based on the preset weights and bias matrices, as well as the one-dimensional feature vector.

7. The apparatus according to claim 5, characterized in that, The channel map generation module is further configured to determine the maximum amplitude value of each initial channel impulse response information and the sampling point corresponding to the maximum amplitude value of each initial channel impulse response information based on the amplitude information of each initial channel impulse response information. Based on the sampling point corresponding to the maximum amplitude value of each initial channel impulse response information, the sampling point corresponding to the maximum amplitude value of each initial channel impulse response information is cyclically shifted to the first sampling point to normalize the time delay of each initial channel impulse response information. The amplitude information of each initial channel impulse response is divided by the maximum amplitude value to normalize the amplitude of each initial channel impulse response.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 4.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 4.

Images