Model determination method and apparatus, semantic quantization method and apparatus, and communication device and storage medium

Abstract

The present disclosure relates to the technical field of communications, and in particular to a model determination method and apparatus, a semantic quantization method and apparatus, and a communication device and a storage medium. The model determination method comprises: determining a training sample set; on the basis of the training sample set, training a first initial model and a second initial model, wherein the first initial model when training is stopped is used as a semantic quantization model, the second initial model when training is stopped is used as a sensing model, an input of the semantic quantization model comprises channel measurement data, an output thereof comprises predicted semantic quantization data, an input of the sensing model comprises the predicted semantic quantization data, and an output thereof comprises a predicted sensing result corresponding to the channel measurement data. The present disclosure is conducive to ensuring that predicted semantic quantization data outputted by a semantic quantization model can accurately reflect the characteristics of channel measurement data inputted into the semantic quantization model, thereby further ensuring the accuracy of the predicted semantic quantization data being used by a sensing model to predict a sensing result of the channel measurement data.

Description

Model determination, semantic quantization methods and apparatus, communication equipment and storage media Technical Field

[0001] This disclosure relates to the field of communication technology, and more specifically, to model training methods, semantic quantization methods, model training devices, semantic quantization devices, communication equipment, and storage media. Background Technology

[0002] Integrated Sensing and Communication (ISAC) technology enables sensing services based on mobile communication infrastructure, fully leveraging the advantages of mobile communication networks to meet sensing needs in various service scenarios, and improving communication performance through sensing capabilities. However, some technical challenges remain to be addressed during the sensing process.

[0003] Summary of the Invention

[0004] Embodiments of this disclosure provide methods and apparatus for model determination and semantic quantization, as well as communication devices and storage media, to address technical problems in the related art.

[0005] According to a first aspect of the present disclosure, a model determination method is proposed, the method comprising: determining a training sample set, wherein the samples in the training sample set include channel measurement data corresponding to a sensing reference signal, and the labels of the samples include sensing results corresponding to the channel measurement data; training a first initial model and a second initial model based on the training sample set, wherein the first initial model at the end of training is used as a semantic quantization model, and the second initial model at the end of training is used as a sensing model; wherein the input of the semantic quantization model includes the channel measurement data, the output of the semantic quantization model includes predicted semantic quantization data corresponding to the channel measurement data, the input of the sensing model includes the predicted semantic quantization data, and the output of the sensing model includes the predicted sensing results corresponding to the channel measurement data.

[0006] According to a second aspect of the present disclosure, a semantic quantization method is proposed, executed by a first device, wherein the first device is equipped with the semantic quantization model, the method comprising: determining first channel measurement data by measuring a sensing reference signal; and inputting the first channel measurement data into the semantic quantization model to determine first predicted semantic quantization data corresponding to the first channel measurement data.

[0007] According to a third aspect of the present disclosure, a model training method is proposed, executed by a first training device. The method includes: determining a training sample set, wherein the samples in the training sample set include channel measurement data corresponding to a sensing reference signal, and the labels of the samples include sensing results corresponding to the channel measurement data; training a first initial model based on the training sample set, wherein the first initial model at the end of training is used as a semantic quantization model; wherein the input of the semantic quantization model includes the channel measurement data, and the output of the semantic quantization model includes predicted semantic quantization data corresponding to the channel measurement data.

[0008] According to a fourth aspect of the present disclosure, a model training method is proposed, executed by a second training device, the method comprising: receiving predicted semantic quantization data output by a first initial model trained by a first training device; training a second initial model based on the predicted semantic quantization data; wherein the input of the second initial model includes the predicted semantic quantization data.

[0009] According to a fifth aspect of the present disclosure, a model determination apparatus is provided, the apparatus comprising: a processing module configured to determine a training sample set, wherein the samples in the training sample set include channel measurement data corresponding to a sensing reference signal, and the labels of the samples include sensing results corresponding to the channel measurement data; training a first initial model and a second initial model based on the training sample set, wherein the first initial model at the end of training is used as a semantic quantization model, and the second initial model at the end of training is used as a sensing model; wherein the input of the semantic quantization model includes the channel measurement data, the output of the semantic quantization model includes predicted semantic quantization data corresponding to the channel measurement data, the input of the sensing model includes the predicted semantic quantization data, and the output of the sensing model includes predicted sensing results corresponding to the channel measurement data.

[0010] According to a sixth aspect of the present disclosure, a model training apparatus is provided, comprising: a processing module configured to determine a training sample set, wherein the samples in the training sample set include channel measurement data corresponding to a sensing reference signal, and the labels of the samples include sensing results corresponding to the channel measurement data; training a first initial model based on the training sample set, wherein the first initial model at the end of training is used as a semantic quantization model; wherein the input of the semantic quantization model includes the channel measurement data, and the output of the semantic quantization model includes predicted semantic quantization data corresponding to the channel measurement data.

[0011] According to a seventh aspect of the present disclosure, a model training apparatus is provided, comprising: a receiving module configured to receive predicted semantic quantization data output by a first initial model trained by a first training device; and a processing module configured to train a second initial model based on the predicted semantic quantization data; wherein the input of the second initial model includes the predicted semantic quantization data.

[0012] According to an eighth aspect of the present disclosure, a semantic quantization apparatus is provided, the apparatus comprising: a processing module configured to determine first channel measurement data by measuring a sensing reference signal; and inputting the first channel measurement data into a semantic quantization model to determine first predicted semantic quantization data corresponding to the first channel measurement data.

[0013] According to a ninth aspect of the present disclosure, a communication device is provided, comprising: one or more processors; wherein the communication device is configured to perform the model training method described in the first aspect, and / or the semantic quantization method described in the second aspect.

[0014] According to a tenth aspect of the present disclosure, a storage medium is provided that stores instructions, which, when executed on a communication device, cause the communication device to perform the model training method described in the first aspect and / or the semantic quantization method described in the second aspect.

[0015] According to the eleventh aspect of the present disclosure, a program product is provided, which, when executed by a communication device, causes the communication device to perform the model training method described in the first aspect and / or the semantic quantization method described in the second aspect.

[0016] According to embodiments of this disclosure, a semantic quantization model can be obtained by training a first initial model on a training sample set using algorithms such as machine learning and deep learning. Since the semantic quantization model is an AI model trained using algorithms such as machine learning and deep learning, it does not quantize channel measurement data according to fixed rules when performing semantic quantization. Instead, it fully considers the characteristics of the channel measurement data and quantizes the data based on these characteristics, thereby predicting appropriate semantic quantization results for channel measurement data with different characteristics.

[0017] Therefore, it is beneficial to ensure that the predicted semantic quantization data output by the semantic quantization model can accurately reflect the characteristics of the channel measurement data input to the semantic quantization model. Based on this, the accuracy of the predicted semantic quantization data used in the perception model to predict the perception results of the channel measurement data can be further guaranteed. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this disclosure, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1A is a schematic diagram of the architecture of a communication system according to an embodiment of the present disclosure.

[0020] Figure 1B is a schematic diagram illustrating a sensing process according to an embodiment of the present disclosure.

[0021] Figure 1C is a schematic diagram illustrating another sensing process according to an embodiment of the present disclosure.

[0022] Figure 2 is a schematic flowchart illustrating a model determination method according to an embodiment of the present disclosure.

[0023] Figure 3 is a schematic diagram illustrating a predictive perception result according to an embodiment of the present disclosure.

[0024] Figure 4 is a schematic diagram of a semantically quantized interaction according to an embodiment of the present disclosure.

[0025] Figure 5 is a schematic block diagram of a model determining apparatus according to an embodiment of the present disclosure.

[0026] Figure 6 is a schematic block diagram illustrating a semantic quantization device according to an embodiment of the present disclosure.

[0027] Figure 7A is a schematic diagram of the structure of the communication device proposed in an embodiment of this disclosure.

[0028] Figure 7B is a schematic diagram of the chip structure proposed in an embodiment of this disclosure. Detailed Implementation

[0029] Embodiments of this disclosure provide methods and apparatus for model determination and semantic quantization, as well as communication devices and storage media.

[0030] In a first aspect, embodiments of this disclosure propose a model determination method, the method comprising: determining a training sample set, wherein the samples in the training sample set include channel measurement data corresponding to a sensing reference signal, and the labels of the samples include sensing results corresponding to the channel measurement data; training a first initial model and a second initial model based on the training sample set, wherein the first initial model at the end of training is used as a semantic quantization model, and the second initial model at the end of training is used as a sensing model; wherein the input of the semantic quantization model includes the channel measurement data, the output of the semantic quantization model includes predicted semantic quantization data corresponding to the channel measurement data, the input of the sensing model includes the predicted semantic quantization data, and the output of the sensing model includes the predicted sensing results corresponding to the channel measurement data.

[0031] In the above embodiments, a semantic quantization model can be obtained by training the first initial model based on a training sample set using algorithms such as machine learning and deep learning. Since the semantic quantization model is an AI model trained using machine learning and deep learning algorithms, it does not quantize channel measurement data according to fixed rules when performing semantic quantization. Instead, it fully considers the characteristics of the channel measurement data and quantizes the data based on these characteristics, thereby predicting appropriate semantic quantization results for channel measurement data with different characteristics.

[0032] Therefore, it is beneficial to ensure that the predicted semantic quantization data output by the semantic quantization model can accurately reflect the characteristics of the channel measurement data input to the semantic quantization model. Based on this, the accuracy of the predicted semantic quantization data used in the perception model to predict the perception results of the channel measurement data can be further guaranteed.

[0033] In conjunction with some embodiments of the first aspect, in some embodiments, during the training of a first initial model and a second initial model based on the training sample set, the difference between the predicted perception result and the actual perception result is used as a loss function.

[0034] In conjunction with some embodiments of the first aspect, in some embodiments, the conditions for stopping training include at least one of the following: the result of the loss function is within a threshold range; the number of training epochs is greater than or equal to a threshold number.

[0035] In conjunction with some embodiments of the first aspect, in some embodiments, the predicted perception result includes at least one of the following: distance perception result; speed perception result; angle perception result.

[0036] In conjunction with some embodiments of the first aspect, in some embodiments, the semantic quantization model is deployed on a first device, and / or the perception model is deployed on a second device; the first device is configured to determine the channel measurement data by receiving a perception reference signal, and to send the predicted semantic quantization data to the second device.

[0037] In conjunction with some embodiments of the first aspect, in some embodiments, the channel measurement data includes at least one of the following: relevant information of a sensing reference signal; a channel state information matrix corresponding to the sensing reference signal; a complex number corresponding to the channel state information matrix; spectral information of the channel corresponding to the sensing reference signal; and point cloud information of the channel corresponding to the sensing reference signal.

[0038] In conjunction with some embodiments of the first aspect, in some embodiments, the relevant information of the sensing reference signal includes at least one of the following: the amplitude of the sensing reference signal; the phase of the sensing reference signal; the I-channel information of the sensing reference signal; and the Q-channel information of the sensing reference signal.

[0039] In conjunction with some embodiments of the first aspect, in some embodiments, the spectral information includes at least one of the following: time delay spread spectral information; Doppler spectral information; micro-Doppler spectral information; angular spectral information; and signal intensity spectral information.

[0040] Secondly, embodiments of this disclosure propose a model training method, executed by a first training device. The method includes: determining a training sample set, wherein the samples in the training sample set include channel measurement data corresponding to a sensing reference signal, and the labels of the samples include sensing results corresponding to the channel measurement data; training a first initial model based on the training sample set, wherein the first initial model at the end of training is used as a semantic quantization model; wherein the input of the semantic quantization model includes the channel measurement data, and the output of the semantic quantization model includes predicted semantic quantization data corresponding to the channel measurement data.

[0041] In conjunction with some embodiments of the second aspect, in some embodiments, the method further includes: sending the predicted semantic quantization data output by the first initial model to a second training device, for the second device to train the second initial model; wherein the second initial model at the end of training serves as a perception model, the input of the perception model includes the predicted semantic quantization data, and the output of the perception model includes the predicted perception result corresponding to the channel measurement data.

[0042] Thirdly, embodiments of this disclosure propose a model training method executed by a second training device, the method comprising: receiving predicted semantic quantization data output by a first initial model trained by a first training device; training a second initial model based on the predicted semantic quantization data; wherein the input of the second initial model includes the predicted semantic quantization data.

[0043] Fourthly, embodiments of this disclosure propose a semantic quantization method, executed by a first device, wherein the first device is equipped with the semantic quantization model, the method comprising: determining first channel measurement data by measuring a sensing reference signal; and inputting the first channel measurement data into the semantic quantization model to determine first predicted semantic quantization data corresponding to the first channel measurement data.

[0044] In conjunction with some embodiments of the fourth aspect, in some embodiments, the method further includes: sending the first predicted semantic quantization data to a second device, wherein the second device is equipped with the perception model, and the perception model determines a first predicted perception result corresponding to the first channel measurement data based on the first predicted semantic quantization data.

[0045] Fifthly, embodiments of this disclosure propose a model determination apparatus, the apparatus comprising: a processing module configured to determine a training sample set, wherein the samples in the training sample set include channel measurement data corresponding to a sensing reference signal, and the labels of the samples include sensing results corresponding to the channel measurement data; training a first initial model and a second initial model based on the training sample set, wherein the first initial model at the end of training is used as a semantic quantization model, and the second initial model at the end of training is used as a sensing model; wherein the input of the semantic quantization model includes the channel measurement data, the output of the semantic quantization model includes predicted semantic quantization data corresponding to the channel measurement data, the input of the sensing model includes the predicted semantic quantization data, and the output of the sensing model includes predicted sensing results corresponding to the channel measurement data.

[0046] In a sixth aspect, embodiments of this disclosure propose a model training apparatus, the apparatus comprising: a processing module configured to determine a training sample set, wherein the samples in the training sample set include channel measurement data corresponding to a sensing reference signal, and the labels of the samples include sensing results corresponding to the channel measurement data; training a first initial model based on the training sample set, wherein the first initial model at the end of training is used as a semantic quantization model; wherein the input of the semantic quantization model includes the channel measurement data, and the output of the semantic quantization model includes predicted semantic quantization data corresponding to the channel measurement data.

[0047] In a seventh aspect, embodiments of this disclosure provide a model training apparatus, the apparatus comprising: a receiving module configured to receive predictive semantic quantization data output by a first initial model trained by a first training device; and a processing module configured to train a second initial model based on the predictive semantic quantization data; wherein the input of the second initial model includes the predictive semantic quantization data.

[0048] Eighthly, embodiments of this disclosure provide a semantic quantization apparatus, the apparatus comprising: a processing module configured to determine first channel measurement data by measuring a sensing reference signal; and inputting the first channel measurement data into a semantic quantization model to determine first predicted semantic quantization data corresponding to the first channel measurement data.

[0049] In a ninth aspect, embodiments of this disclosure provide a communication device comprising: one or more processors; wherein the communication device is configured to perform a model training method according to any one of the first aspect and optional embodiments thereof, and / or a semantic quantization method according to any one of the second aspect and optional embodiments thereof.

[0050] In a tenth aspect, embodiments of this disclosure provide a storage medium storing instructions that, when executed on a communication device, cause the communication device to perform a model training method according to any one of the first aspect and optional embodiments of the first aspect, and / or a semantic quantization method according to any one of the second aspect and optional embodiments of the second aspect.

[0051] Eleventhly, embodiments of this disclosure provide a program product that, when executed by a communication device, causes the communication device to perform the model training method of any one of the first aspect and optional embodiments of the first aspect, and / or the semantic quantization method of any one of the second aspect and optional embodiments of the second aspect.

[0052] In a twelfth aspect, embodiments of this disclosure provide a computer program that, when run on a computer, causes the computer to perform the model training method of any one of the first aspect and any one of the optional embodiments of the first aspect, and / or the semantic quantization method of any one of the second aspect and any one of the optional embodiments of the second aspect.

[0053] Understandably, the aforementioned model determination, semantic quantization device, communication equipment, communication system, storage medium, program product, and computer program are all used to execute the methods proposed in the embodiments of this disclosure. Therefore, the beneficial effects that can be achieved can be referred to the beneficial effects in the corresponding methods, and will not be repeated here.

[0054] This disclosure provides a model determination and semantic quantization method and apparatus, a communication device, and a storage medium. In some embodiments, the terms "model determination" and "semantic quantization method" can be used interchangeably with "information processing method" and "communication method," and the terms "model determination" and "semantic quantization apparatus" can be used interchangeably with "information processing apparatus" and "communication apparatus," and the terms "information processing system" and "communication system" can be used interchangeably.

[0055] This disclosure is not exhaustive, but merely illustrative of some embodiments, and is not intended to limit the scope of protection of this disclosure. Unless otherwise specified, each step in a particular embodiment can be implemented as an independent embodiment, and the steps can be arbitrarily combined. For example, a solution after removing some steps in a particular embodiment can also be implemented as an independent embodiment, and the order of the steps in a particular embodiment can be arbitrarily interchanged. Furthermore, the optional implementation methods in a particular embodiment can be arbitrarily combined; moreover, the embodiments can be arbitrarily combined, for example, some or all steps of different embodiments can be arbitrarily combined, and a particular embodiment can be arbitrarily combined with the optional implementation methods of other embodiments.

[0056] In each of the disclosed embodiments, unless otherwise specified or in case of logical conflict, the terminology and / or descriptions of the embodiments are consistent and can be referenced by each other. Technical features in different embodiments can be combined to form new embodiments based on their inherent logical relationships.

[0057] The terminology used in the embodiments of this disclosure is for the purpose of describing particular embodiments only and is not intended to limit the scope of this disclosure.

[0058] In the embodiments of this disclosure, unless otherwise stated, elements expressed in the singular, such as “a,” “an,” “the,” “the,” “the,” “the,” “the,” “the,” “this,” etc., may mean “one and only one,” or “one or more,” “at least one,” etc.

[0059] For example, when using articles such as "a", "an", and "the" in translation, the noun following the article can be understood as either a singular or a plural form.

[0060] In the embodiments disclosed herein, "multiple" refers to two or more.

[0061] In some embodiments, the terms “at least one of”, “one or more”, “a plurality of”, “multiple”, etc., may be used interchangeably.

[0062] In some embodiments, the notation "at least one of A and B", "A and / or B", "A in one case, B in another", "in response to one case A, in response to another case B", etc., may include the following technical solutions depending on the situation: in some embodiments, A (execute A regardless of B); in some embodiments, B (execute B regardless of A); in some embodiments, execution is selected from A and B (A and B are selectively executed); in some embodiments, A and B (both A and B are executed). The same applies when there are more branches such as A, B, C, etc.

[0063] In some embodiments, the notation "A or B" may include the following technical solutions, depending on the situation: in some embodiments, A (execution of A regardless of B); in some embodiments, B (execution of B regardless of A); in some embodiments, execution is selected from A and B (A and B are selectively executed). The same applies when there are more branches such as A, B, C, etc.

[0064] The prefixes such as "first" and "second" in the embodiments of this disclosure are only for distinguishing different descriptive objects and do not constitute restrictions on the position, order, priority, number or content of the descriptive objects. For the description of the descriptive objects, please refer to the description in the claims or the context of the embodiments. The use of prefixes should not constitute unnecessary restrictions.

[0065] For example, if the descriptive object is "field," then the ordinal numbers preceding "field" in "first field" and "second field" do not restrict the position or order of the "fields." "First" and "second" do not restrict whether the "fields" they modify are in the same message, nor do they restrict the order of "first field" and "second field." Similarly, if the descriptive object is "level," then the ordinal numbers preceding "level" in "first level" and "second level" do not restrict the priority between "levels." Furthermore, the number of descriptive objects is not limited by ordinal numbers; there can be one or more. For example, in "first device," the number of "devices" can be one or more. In addition, objects modified by different prefixes can be the same or different. For example, if the descriptive object is "device," then "first device" and "second device" can be the same device or different devices, and their types can be the same or different. Similarly, if the descriptive object is "information," then "first information" and "second information" can be the same information or different information, and their content can be the same or different.

[0066] In some embodiments, “including A,” “containing A,” “for indicating A,” and “carrying A” can be interpreted as directly carrying A or indirectly indicating A.

[0067] In some embodiments, the terms “in response to…”, “in response to determining…”, “in the case of…”, “when…”, “if…”, “if…”, etc., can be used interchangeably.

[0068] In some embodiments, the terms “greater than,” “greater than or equal to,” “not less than,” “more than,” “more than or equal to,” “not less than,” “higher than,” “higher than or equal to,” “not lower than,” and “above” can be used interchangeably, as can the terms “less than,” “less than or equal to,” “not greater than,” “less than,” “less than or equal to,” “not more than,” “lower than,” “lower than or equal to,” “not higher than,” and “below”.

[0069] In some embodiments, devices, etc., can be interpreted as physical or virtual, and their names are not limited to the names recorded in the embodiments. Terms such as “device”, “equipment”, “circuit”, “network element”, “node”, “function”, “unit”, “section”, “system”, “network”, “chip”, “chip system”, “entity”, and “subject” can be used interchangeably.

[0070] In some embodiments, "network" can be interpreted as devices included in a network (e.g., access network devices, core network devices, etc.).

[0071] In some embodiments, the terms "access network device (AN device)," "radio access network device (RAN device)," "base station (BS)," "radio base station," "fixed station," "node," "access point," "transmission point (TP)," "reception point (RP)," "transmission / reception point (TRP)," "panel," "antenna panel," "antenna array," "cell," "macro cell," "small cell," "femto cell," "pico cell," "sector," "cell group," "serving cell," "carrier," "component carrier," and "bandwidth part (BWP)" can be used interchangeably.

[0072] In some embodiments, the terms "terminal", "terminal device", "user equipment (UE)", "user terminal", "mobile station (MS)", "mobile terminal (MT)", "subscriber station", "mobile unit", "subscriber unit", "wireless unit", "remote unit", "mobile device", "wireless device", "wireless communication device", "remote device", "mobile subscriber station", "access terminal", "mobile terminal", "wireless terminal", "remote terminal", "handset", "user agent", "mobile client", and "client" can be used interchangeably.

[0073] In some embodiments, access network devices, core network devices, or network devices can be replaced by terminals. For example, embodiments of this disclosure can also be applied to structures where communication between access network devices, core network devices, or network devices and terminals is replaced by communication between multiple terminals (e.g., device-to-device (D2D), vehicle-to-everything (V2X), etc.). In this case, the structure can also be configured such that the terminal has all or part of the functions of the access network device. Furthermore, terms such as "uplink" and "downlink" can be replaced with terms corresponding to communication between terminals (e.g., "sidelink"). For example, uplink channel, downlink channel, etc., can be replaced with sidelink channel, and uplink link, downlink, etc., can be replaced with sidelink link.

[0074] In some embodiments, the terminal may be replaced by an access network device, a core network device, or a network device. In this case, the access network device, core network device, or network device may also be configured to have all or some of the functions of the terminal.

[0075] In some embodiments, the acquisition of data, information, etc., may comply with the laws and regulations of the country where the location is situated.

[0076] In some embodiments, data, information, etc., may be obtained with the user's consent.

[0077] Furthermore, each element, each row, or each column in the table of this disclosure can be implemented as an independent embodiment, and any combination of any element, any row, or any column can also be implemented as an independent embodiment.

[0078] Figure 1A is a schematic diagram of the architecture of a communication system according to an embodiment of the present disclosure.

[0079] As shown in Figure 1A, the communication system 100 includes a first device 101 and a second device 102. For example, the first device can be a reference signal receiving device, and the second device can be a model inference device.

[0080] For example, the types of the first and second devices include, but are not limited to, terminals and network devices.

[0081] In some embodiments, the network device includes at least one of the following: an access network device and a core network device.

[0082] In some embodiments, terminal 101 includes, but is not limited to, at least one of the following: mobile phone, wearable device, Internet of Things device, car with communication function, smart car, tablet computer, computer with wireless transceiver function, virtual reality (VR) terminal device, augmented reality (AR) terminal device, wireless terminal device in industrial control, wireless terminal device in self-driving, wireless terminal device in remote medical surgery, wireless terminal device in smart grid, wireless terminal device in transportation safety, wireless terminal device in smart city, and wireless terminal device in smart home.

[0083] In some embodiments, the access network device is, for example, a node or device that connects a terminal to a wireless network. The access network device may include, but is not limited to, at least one of the following in a 5G communication system: evolved Node B (eNB), next-generation eNB (ng-eNB), next-generation Node B (gNB), node B (NB), home node B (HNB), home evolved node B (HeNB), radio backhaul device, radio network controller (RNC), base station controller (BSC), base transceiver station (BTS), base band unit (BBU), mobile switching center, base station in a 6G communication system, open RAN, cloud RAN, base station in other communication systems, and access node in a Wi-Fi system.

[0084] In some embodiments, a core network device may be a single device comprising one or more network elements, or it may be multiple devices or a group of devices, each comprising all or part of the aforementioned one or more network elements. Network elements may be virtual or physical. The core network may include, for example, at least one of an Evolved Packet Core (EPC), a 5G Core Network (5GCN), or a Next Generation Core (NGC).

[0085] In some embodiments, the technical solutions of this disclosure can be applied to the Open RAN architecture. In this case, the interfaces between or within access network devices involved in the embodiments of this disclosure can be transformed into internal interfaces of Open RAN. The processes and information interactions between these internal interfaces can be implemented by software or programs.

[0086] In some embodiments, the access network device may be composed of a central unit (CU) and a distributed unit (DU). The CU may also be called a control unit. The CU-DU structure can separate the protocol layer of the access network device. Some of the protocol layer functions are centrally controlled by the CU, while the remaining part or all of the protocol layer functions are distributed in the DU and centrally controlled by the CU. However, this is not the only possibility.

[0087] It is understood that the communication system described in this disclosure is for the purpose of more clearly illustrating the technical solutions of this disclosure, and does not constitute a limitation on the technical solutions proposed in this disclosure. As those skilled in the art will know, with the evolution of system architecture and the emergence of new business scenarios, the technical solutions proposed in this disclosure are also applicable to similar technical problems.

[0088] The following embodiments of this disclosure can be applied to the communication system 100 shown in FIG1A, or to some of the main bodies, but are not limited thereto. The main bodies shown in FIG1A are illustrative. The communication system may include all or some of the main bodies in FIG1A, or it may include other main bodies outside of FIG1A. The number and form of each main body are arbitrary. Each main body may be physical or virtual. The connection relationship between the main bodies is illustrative. The main bodies may not be connected or may be connected. The connection can be in any way, it can be a direct connection or an indirect connection, it can be a wired connection or a wireless connection.

[0089] The embodiments disclosed herein can be applied to Long Term Evolution (LTE), LTE-Advanced (LTE-A), LTE-Beyond (LTE-B), SUPER 3G, IMT-Advanced, 4th generation mobile communication system (4G), 5th generation mobile communication system (5G), 5G new radio (NR), Future Radio Access (FRA), New-Radio Access Technology (RAT), New Radio (NR), New radio access (NX), Future generation radio access (FX), Global System for Mobile communications (GSM), CDMA2000, Ultra Mobile Broadband (UMB), IEEE 802.11 (Wi-Fi), IEEE 802.16 (WiMAX), and IEEE 802.20, Ultra-Wideband (UWB), Bluetooth (a registered trademark), Public Land Mobile Network (PLMN) networks, Device-to-Device (D2D) systems, Machine-to-Machine (M2M) systems, Internet of Things (IoT) systems, Vehicle-to-Everything (V2X) systems, systems utilizing other communication methods, and next-generation systems built upon them, etc. Furthermore, multiple systems can be combined (e.g., a combination of LTE or LTE-A with 5G).

[0090] In some embodiments, Integrated Sensing and Communication (ISAC) is a crucial research direction in wireless communication, applicable to 5G-A (5G-Advanced) and future 6G mobile communications. ISAC technology enables sensing services based on existing mobile communication infrastructure, fully leveraging the advantages of mobile communication networks to meet sensing needs across various service scenarios. Furthermore, it enhances communication performance through sensing capabilities, thereby facilitating numerous application services such as detection, localization and tracking, environmental reconstruction and target imaging, and gesture and posture recognition.

[0091] In some embodiments, for integrated communication and sensing applications, channel modeling of the sensing system can be performed based on 3GPP channel modeling standards to support the evaluation of sensing technology solutions. For example, for scenarios such as highways, various estimation algorithms for parameters such as target speed, distance, and angle have been proposed and designed, including the MUSIC (Multiple Signal Classification) algorithm and the ESPRIT (Estimation of Signal Parameters via Rotational Invariance Technique) algorithm. These algorithms can achieve relatively good target sensing results in some integrated communication and sensing simulation scenarios.

[0092] In some embodiments, the integration of communication and sensing based on AI (Artificial Intelligence) technology is a potential application in future 6G wireless communication systems. By leveraging the powerful computing capabilities of dedicated devices such as GPUs (Graphics Processing Units), AI sensing models can be trained on datasets of a certain form to optimize aspects such as sensing accuracy, computational complexity, and applicability. This promotes the deep application of communication and sensing integration technology, supports more intelligent business applications and services, and thus helps to drive the development of communication, sensing, computing, and intelligence integration.

[0093] In some embodiments, compared to traditional communication technologies, semantic communication aims to extract key feature information from raw data, compress data while preserving important semantic information, thereby reducing data transmission overhead, and ensuring that the receiving end can complete data reconstruction with high accuracy, thus ensuring communication transmission quality. Implementing a semantic encoder at the sending end and a semantic decoder at the receiving end in a semantic communication system based on AI models is a common approach. Using neural network models to autonomously learn key semantic feature information from data based on large-scale training data, data compression and reconstruction are completed, achieving low-overhead and efficient data transmission.

[0094] In some embodiments, communication sensing integration may include six sensing modes, from Option 1 to Option 6. Mono-static indicates self-transmission and self-reception, where the sender and receiver are the same; Bi-static indicates cross-station transmission and reception, where the sender and receiver are different, as detailed below:

[0095] Option 1: Self-transmission and self-reception. Signals are sent through a terminal, such as a user (e.g., a person), and then the reflected echoes are received through the terminal (e.g., a car) to perceive information such as distance, speed, and angle.

[0096] Option 2: Self-transmission and self-reception. Through network devices, such as users (e.g., people), signals are sent and reflected echoes are received to perceive information such as distance, speed, and angle.

[0097] Option 3: Cross-site transceiver, from network device to terminal, for example, the network device sends a signal, the terminal (e.g., a car) receives the signal, and the UE (e.g., the car) calculates the channel matrix based on the received signal to perceive information such as the distance, speed, and angle of targets in the environment, such as users (e.g., people).

[0098] Option 4: Cross-site transceiver, from terminal to network device, for example, the terminal (e.g., a car) sends a signal, the network device receives the signal, and the network device calculates the channel matrix based on the received signal to sense information such as the distance, speed, and angle of targets in the environment, such as users (e.g., people);

[0099] Option 5: Cross-site transmission and reception, from network device to network device, for example, network device #1 sends a signal and network device #2 receives the signal. Network device #2 calculates the channel matrix based on the received signal to sense information such as the distance, speed, and angle of targets in the environment, such as users (e.g., people).

[0100] Option 6: Cross-station transceiver, from terminal (e.g., a car) to terminal (e.g., a car). For example, terminal #1 (e.g., a car) sends a signal, and terminal #2 (e.g., a car) receives the signal. Terminal #2 (e.g., a car) calculates a channel matrix based on the received signal to perceive information such as the distance, speed, and angle of targets in the environment, such as users (e.g., people).

[0101] In some embodiments, the perceived target may include: drones, people in indoor and outdoor scenes, cars in outdoor scenes such as highways, automated guided vehicles in indoor scenes such as factories, and targets that pose a danger on roads or railways. Sensor integration allows the receiver to calculate and obtain information such as the distance, speed, and angle of the perceived target in the environment based on the received signals. The perceived target can be the receiver itself or other objects.

[0102] In some embodiments, in the integrated sensing simulation scenario, there are sensing target terminals, network devices, and obstructions, which may be other terminals or clutter. Therefore, the sensing channel model mainly considers four types of signals: sensing LOS (Line of Sight) path, sensing NLOS (Non-Line of Sight) path, clutter LOS path, and clutter NLOS path. Furthermore, in some embodiments, the impact of channel noise also needs to be considered.

[0103] The LOS path for sensing: The path between the sensing signal transmitter and the sensing target is the LOS path, and the path between the sensing target and the sensing signal receiver is also the LOS path.

[0104] NLOS path: The path between the sensing signal transmitter and the sensing target is the NLOS path, and the path between the sensing target and the sensing signal receiver is also the NLOS path.

[0105] Clutter LOS path: The LOS path is between the transmitting end of the sensing signal and the scattering clusters in the environment, and also between the scattering clusters in the environment and the receiving end of the sensing signal.

[0106] Clutter NLOS path: The NLOS path is between the transmitting end of the sensing signal and the scattering clusters in the environment, and also between the scattering clusters in the environment and the receiving end of the sensing signal.

[0107] In some embodiments, the main information for target sensing is the channel state information matrix H at the receiver. This matrix is ​​a superposition of the target sensing channel, clutter channel, and noise, reflecting the channel environment in which the sensing signal is located and the state information of the target in that environment. The dimensions of the channel state information matrix H are related to the specific parameter settings in the actual sensing scenario, and the information in the data on different dimensions can reflect different types of state information of the target.

[0108] This disclosure illustrates a common scenario for receiver channel state information matrix data and sensing information types:

[0109] The channel state information matrix H typically contains three dimensions: subcarriers, OFDM (Orthogonal Frequency Division Multiplexing) symbols, and receiver antenna ports. Specifically, the channel matrix H represents channel state information and is a complex matrix of size M × N × P, where M is the number of OFDM symbols, N is the number of subcarriers, and P is the number of receiver antenna ports.

[0110] Wherein, in the channel matrix H, the element H at the position of the m-th OFDM symbol, the n-th subcarrier, and the p-th receive antenna port is... m,n,p It can be represented as a complex number C m,n,p H represents the influence on the amplitude and phase of a signal as it propagates from the transmitting antenna to the receiving antenna. m,n,p =C m,n,p =a m,n,p +i×b m,n,p

[0111] Among them, a m,n,p C m,n,p The real part, b m,n,p C m,n,p The imaginary part.

[0112] In some embodiments, based on the received signal model, the phase shifts in the three dimensions of the channel matrix do not affect each other, and different types of perception results of the target can be obtained according to the phase changes in the three dimensions. That is, the target distance can be estimated based on the phase difference caused by the time delay between subcarriers, the target velocity can be estimated based on the phase difference caused by the Doppler effect between OFDM symbols, and the target angle can be estimated based on the phase change between the received signals of multiple antenna ports.

[0113] Figure 1B is a schematic diagram illustrating a sensing process according to an embodiment of the present disclosure.

[0114] As shown in Figure 1B, perception algorithms (such as 3D-MUSIC, 3D-ESPRIT, etc.) can obtain perception results such as distance d, velocity v, and angle θ of the perceived target based on the channel state information matrix H.

[0115] Although existing sensing algorithms can achieve high accuracy in sensing information such as target distance, speed, and angle in some synesthetic applications, they still have some problems, such as the following three points:

[0116] (1) The applicability of perception algorithms to certain scenarios is limited. Due to factors such as clutter and noise in the scene, the performance of some perception algorithms will be significantly reduced. In the case of multiple targets, the direct path of non-perceived targets will have a certain impact on the perception results of perceived targets; when the distance of perceived targets is far, the impact of noise will be greater.

[0117] (2) The computational complexity of the perception algorithm is high. In some cases, in order to obtain a high-accuracy perception result, the perception algorithm is too complex, which is not conducive to the practical application of the perception algorithm.

[0118] (3) The accuracy needs to be improved in the case of multi-target perception. In actual application scenarios of integrated sensing, existing multi-target perception algorithms usually have a certain error in estimating the number of sensing targets, resulting in a large error in the algorithm estimation results.

[0119] In some embodiments, the strong learning and modeling capabilities of neural network models can be utilized to train and obtain an AI perception model based on a certain amount of data. This model can be used to estimate the number of targets in a multi-target perception scenario and to perceive information such as the distance, speed, and angle of each target.

[0120] Figure 1C is a schematic diagram illustrating another sensing process according to an embodiment of the present disclosure.

[0121] As shown in Figure 1C, the channel state information matrix H can be directly used as the input data for the AI ​​perception model to train an AI perception model. This AI perception model can output perception results such as the target's distance d, velocity v, and angle θ.

[0122] Although AI neural network models can effectively overcome the main problems of traditional target perception algorithms and achieve high perception accuracy, it is still necessary to consider the computational complexity of the models and the overhead of data reporting in the deployment and application of AI solutions in actual systems.

[0123] In some embodiments, the device on which the AI ​​perception model is deployed may include, for example, a mobile terminal, a car, or a network device, such as a network-side device dedicated to implementing the sensing function unit.

[0124] It should be noted that the device for acquiring channel measurement data for AI perception (e.g., the first device, also known as a reference signal receiving device) and the device for completing perception applications based on the AI ​​perception model (e.g., the second device, also known as a model inference device) may be the same or different, and this disclosure does not limit them.

[0125] In some embodiments, when the AI ​​sensing model is deployed on a terminal, the terminal can acquire channel measurement data (e.g., channel matrix H) through a self-transmitting and self-receiving method and then directly input it into the AI ​​sensing model to obtain the sensing results. However, when the AI ​​sensing model is deployed on a network device, the network device first needs to acquire channel measurement data for sensing reported by the terminal or base station, and then input the channel measurement data into the AI ​​sensing model to obtain the sensing results.

[0126] Because AI perception models are neural network models, they require significant computing power and are often deployed on network devices with abundant computing resources. In such cases, reporting channel measurement data for perception will result in substantial channel resource consumption overhead, and the computational complexity of the AI ​​perception model will also significantly impact the energy consumption and latency of the model application.

[0127] In some embodiments, to address the aforementioned issue of high overhead, data quantization can be performed before data transmission between devices. For example, continuous values ​​can be mapped to a finite set of discrete values ​​to simplify data representation and reduce the number of bits. The effects achievable by using data quantization in wireless communication systems include, but are not limited to:

[0128] (1) Reduce data size and processing complexity, and significantly improve transmission efficiency;

[0129] (2) Quantified data requires less storage space, which helps to store more data on storage devices;

[0130] (3) In low-bandwidth or high-latency network environments, quantization can reduce the amount of data, which helps to reduce transmission latency.

[0131] (4) The quantized data has a simpler representation, which can reduce computational complexity;

[0132] (5) Quantification can increase the ambiguity of data, thereby improving the privacy and security of data.

[0133] However, the problem with the aforementioned quantization methods lies in the fact that, based on relevant technologies, the quantization processing of channel measurement data used for sensing, such as the channel state information matrix, mainly employs methods such as uniform quantization and non-uniform quantization. These quantization methods convert high-precision floating-point numbers and other types of data in the original data into low-precision data represented by bitstream data according to fixed rules. However, these fixed-rule quantization methods may not conform to the actual characteristics of the data to be quantized in some cases, resulting in the loss of certain key data feature information, which in turn adversely affects the perception accuracy of the AI ​​perception model.

[0134] For example, in channel measurement data that requires quantization, the data at a certain location is 0.501, and the data type is a floating-point number, requiring 32 bits for precise representation. Quantizing the data 0.501 converts it into a value represented by a relatively small number of bits (e.g., 4 bits, 8 bits, etc.), such as 0.5.

[0135] Since uniform quantization, non-uniform quantization, and other methods quantize data according to fixed rules, the features will be different for different data.

[0136] Taking uniform quantization as an example, for data A, each bit of the value is of equal importance. For example, after quantization, the value is represented by n1 bits. Thus, before quantization, each bit of the value corresponds to the same number of bits in n1 bits. Therefore, before quantization, each bit of the value can be represented with the same precision. This quantization result can basically meet the characteristics of data A.

[0137] However, for data B, the importance of one bit (or multiple bits) differs from the importance of the other bits. Therefore, the value of the more important bit needs to be represented with relatively higher precision. Thus, when quantized into a value represented by n2 bits (which may be equal to or different from n1), the value of the more important bit requires a relatively larger number of bits compared to the other bits.

[0138] Because uniform quantization quantizes data using fixed rules, quantizing data B using this method results in each bit of the original value corresponding to the same number of bits in n² bits. This fails to meet the requirement of representing the relatively important bit with a relatively larger number of bits. The resulting quantization, when input into a subsequent AI perception model, will negatively impact the model's perception accuracy.

[0139] Taking non-uniform quantization as an example, for data C, the quantized value is represented by n3 bits, where the i-th bit in data C corresponds to j bits in the n3 bits. For data C, each bit is of equal importance, and it is assumed that the quantization result basically conforms to the characteristics of data C.

[0140] However, for data D, the importance of the i-th bit is relatively low compared to the other bits. Therefore, the i-th bit, which is relatively less important, needs to be represented with relatively low precision. Thus, when quantized into a value represented by n4 bits (which may be equal to or different from n3), the i-th bit, which is relatively less important, requires relatively fewer bits compared to the other bits.

[0141] Because non-uniform quantization quantizes data using fixed rules, quantizing data D using this method results in the i-th bit still corresponding to j bits out of n4 bits before quantization. This fails to meet the requirement of representing the relatively less important bit with fewer bits. The resulting quantization, when input into a subsequent AI perception model, will negatively impact the model's perception accuracy.

[0142] Firstly, embodiments of this disclosure provide a model determination method. Figure 2 is a schematic flowchart illustrating a model determination method according to an embodiment of this disclosure.

[0143] The model determination method shown in this embodiment can be executed by a training device. For example, the type of training device can be a communication device. For example, the training device can be a first device, or it can be a second device, or it can be a third device other than the first and second devices.

[0144] The first device, the second device, and the third device can be, for example, a terminal, a network device, a server, etc., and this disclosure does not limit them.

[0145] As shown in Figure 2, the model determination method may include the following steps:

[0146] In step S201, the training sample set is determined.

[0147] In some embodiments, the samples in the training sample set include channel measurement data corresponding to the sensing reference signal, and the labels of the samples include the sensing results corresponding to the channel measurement data.

[0148] In step S202, the first initial model and the second initial model are trained based on the training sample set. The first initial model at the end of training is used as the semantic quantization model, and the second initial model at the end of training is used as the perceptual model.

[0149] In some embodiments, the input of the semantic quantization model includes channel measurement data, and the output of the semantic quantization model includes predicted semantic quantization data corresponding to the channel measurement data. The input of the perception model includes predicted semantic quantization data, and the output of the perception model includes predicted perception results corresponding to the channel measurement data.

[0150] It should be noted that the embodiment shown in Figure 2 can be implemented independently or in combination with at least one other embodiment in this disclosure. The specific choice can be made as needed, and this disclosure does not limit the scope.

[0151] In some embodiments, a training sample set can first be constructed. The samples in the training sample set may include channel measurement data, and the labels of the samples include the semantic quantization results corresponding to the channel measurement data, which may be referred to as semantic quantization data.

[0152] In some embodiments, an initial model for training the semantic quantization model can be determined, for example, it may be referred to as a first initial model, and an initial model for training the perceptual model can be determined, for example, it may be referred to as a second initial model. The first initial model and the second initial model can then be trained based on a training sample set. For example, the first initial model and / or the second initial model can be neural network models. For example, the training process can be implemented based on supervised learning or unsupervised learning; this disclosure is not limited in this regard.

[0153] The first and second initial models are trained based on the training sample set. The first initial model at the end of training can be used as the semantic quantization model, and the second initial model at the end of training can be used as the perceptual model. The resulting semantic quantization model can perform feature extraction and data compression on the input channel measurement data, outputting semantic quantization data predicted from the channel measurement data, which can be called predicted semantic quantization data. For example, the predicted semantic quantization data can contain key semantic features from the channel measurement data, and can be represented as a bit data stream.

[0154] In some embodiments, the predicted semantic quantization data can be used to predict the sensing results corresponding to the channel measurement data. For example, the predicted semantic quantization data can be input into the sensing model, and the sensing model can predict the sensing results corresponding to the channel measurement data based on the predicted semantic quantization data, which is called the predicted sensing results.

[0155] According to embodiments of this disclosure, a semantic quantization model can be obtained by training a first initial model on a training sample set using algorithms such as machine learning and deep learning. Since the semantic quantization model is an AI model trained using algorithms such as machine learning and deep learning, it does not quantize channel measurement data according to fixed rules when performing semantic quantization. Instead, it fully considers the characteristics of the channel measurement data and quantizes the data based on these characteristics, thereby predicting appropriate semantic quantization results for channel measurement data with different characteristics.

[0156] Therefore, it is beneficial to ensure that the predicted semantic quantization data output by the semantic quantization model can accurately reflect the characteristics of the channel measurement data input to the semantic quantization model. Based on this, the accuracy of the predicted semantic quantization data used in the perception model to predict the perception results of the channel measurement data can be further guaranteed.

[0157] For example, when the semantic quantization model determined in the embodiments of this disclosure is used to quantize the above data B, and the output predicted semantic quantization data is n2 bits, it is beneficial to ensure that the value of the bit with relatively high importance in the channel measurement data corresponds to a relatively large number of bits in the n2 bits, relative to the values ​​of other bits, thereby satisfying the need to represent the value of the bit with relatively high importance in a relatively large number of bits.

[0158] It should be noted that the semantic quantization model and the perception model obtained from training can be validated by constructing a validation set or tested by constructing a test set. The specific choice can be made as needed, and this disclosure will not elaborate further.

[0159] In some embodiments, during the training of a first initial model and a second initial model based on a training sample set, the difference between the predicted perception result and the actual perception result is used as a loss function.

[0160] In some embodiments, the conditions for stopping training include at least one of the following:

[0161] The result of the loss function is within the threshold range;

[0162] The number of training cycles is greater than or equal to the number threshold.

[0163] In some embodiments, a loss function can be constructed, for example, the loss function is the difference between the predicted perceived result and the actual perceived result. The actual perceived result can be, for example, the label value of a sample in the sample set, or a manually set value.

[0164] For example, the condition for stopping training could be that the loss function is within a threshold range. For instance, the loss function can be processed using stochastic gradient descent, which calculates the gradient of the loss function and updates the model parameters in the direction of the negative gradient to minimize the loss function. With each training epoch, the loss function generally tends to decrease. We can determine whether the loss function in each training epoch is within the threshold range. If it is not, we continue to the next training epoch until the loss function is within the threshold range (e.g., the loss function is within the threshold range once or multiple times). In this case, the model's prediction results are already quite close to the sample label values, meeting the prediction accuracy requirements, and therefore training can be stopped.

[0165] For example, the condition for stopping training could be that the number of training epochs is greater than or equal to a threshold. For instance, a training epoch can go from inputting samples to the first initial model, to adjusting the parameter positions of the first and second initial models. The number of training epochs can be recorded. When the number of training epochs is greater than or equal to the threshold, the model training process has already taken a relatively long time, so training can be stopped to avoid excessive overhead.

[0166] It should be noted that the conditions for stopping training are not limited to the two shown in the above embodiments, and may include other conditions, which are not limited in this disclosure. For example, a perception accuracy condition may be included, wherein perception accuracy can be characterized as the difference between the predicted perception result and the target label result, wherein the target label result may not be associated with the sample, for example, it may be a manually set value.

[0167] In some embodiments, the predicted perception result includes at least one of the following: distance perception result; speed perception result; angle perception result.

[0168] It should be noted that the predicted perception results are not limited to the aforementioned perception results such as distance, speed, and angle, and may be extended as needed. This disclosure does not limit this.

[0169] In some embodiments, the channel measurement data includes at least one of the following:

[0170] Relevant information about the sensing reference signal;

[0171] The channel state information matrix corresponding to the sensing reference signal;

[0172] The complex number corresponding to the channel state information matrix;

[0173] Spectral information of the channel corresponding to the sensing reference signal;

[0174] Point cloud information of the channel corresponding to the sensing reference signal.

[0175] For example, point cloud information can be drawn using information such as the propagation path contained in the statistical sensing reference signal.

[0176] In some embodiments, the spectral information includes at least one of the following: time delay spread spectral information; Doppler spectral information; micro-Doppler spectral information; angular spectral information; and signal intensity spectral information.

[0177] For example, spectral information can be estimated based on data in each dimension of the channel state information matrix using a sensing algorithm (which can be selected as needed, and this disclosure is not limited thereto). For example, this spectral information contains information on multiple paths or multiple motion modes, each of which can be reflected by independent spectral lines or parameters.

[0178] In some embodiments, the relevant information of the sensed reference signal includes at least one of the following:

[0179] Sensing the amplitude of the reference signal;

[0180] Sensing the phase of the reference signal;

[0181] I-channel information of the sensing reference signal;

[0182] The Q-path information of the sensing reference signal.

[0183] In some embodiments, the semantic quantization model is deployed on a first device, and / or the perception model is deployed on a second device; the first device is configured to determine channel measurement data by receiving a perception reference signal, and to send the predicted semantic quantization data to the second device.

[0184] It should be noted that the first device and the second device can be different devices or the same device. In the case that the first device and the second device are the same device, the first device does not need to perform the operation of sending the predicted semantic quantization data to the second device.

[0185] It should be noted that the semantic quantization model and the perception model can be trained on a communication device or on different devices, and this disclosure does not limit this. The following examples describe the case where the semantic quantization model and the perception model are trained on different devices.

[0186] Secondly, embodiments of this disclosure propose a model training method, executed by a first training device. The method includes: determining a training sample set, wherein the samples in the training sample set include channel measurement data corresponding to a sensing reference signal, and the labels of the samples include sensing results corresponding to the channel measurement data; training a first initial model based on the training sample set, wherein the first initial model at the end of training is used as a semantic quantization model; wherein the input of the semantic quantization model includes the channel measurement data, and the output of the semantic quantization model includes predicted semantic quantization data corresponding to the channel measurement data.

[0187] In some embodiments, a training sample set can first be constructed. The samples in the training sample set may include channel measurement data, and the labels of the samples include the semantic quantization results corresponding to the channel measurement data, which may be referred to as semantic quantization data.

[0188] In some embodiments, an initial model, such as a first initial model, may be determined for training the semantic quantization model. The first initial model can then be trained based on a training sample set. For example, the first initial model may be a neural network model. The training process may be implemented based on supervised learning or unsupervised learning; this disclosure is not limited in this regard.

[0189] The first initial model is trained based on the training sample set. This first initial model, at the end of training, can serve as the semantic quantization model. The resulting semantic quantization model can perform feature extraction and data compression on the input channel measurement data, outputting semantically quantized data predicted from the channel measurement data; this can be termed predicted semantically quantized data. For example, the predicted semantically quantized data can contain key semantic features from the channel measurement data, and can be represented as a bitstream.

[0190] In some embodiments, the predicted semantic quantization data can be used to predict the sensing result corresponding to the channel measurement data. For example, the predicted semantic quantization data can be input into the sensing model trained by the second training device. The sensing model can predict the sensing result corresponding to the channel measurement data based on the predicted semantic quantization data, which is called the predicted sensing result.

[0191] In some embodiments, the model training method further includes: sending the predicted semantic quantization data output by the first initial model to a second training device, for the second device to train the second initial model; wherein the second initial model at the end of training is used as a perception model, the input of the perception model includes the predicted semantic quantization data, and the output of the perception model includes the predicted perception result corresponding to the channel measurement data.

[0192] In some embodiments, the first training device may interact with the second training device during the training of the first initial model in order to collaborate with the second training device in training the model.

[0193] For example, during the training process, the first training device can send the predicted semantic quantization data output by the first initial model to the second training device. The second training device can train the second initial model based on the predicted semantic quantization data, for example, by inputting the predicted semantic quantization data into the second initial model. The second initial model can output the predicted sensing result of the channel measurement data corresponding to the predicted semantic quantization data.

[0194] For example, during the training process, the first training device can send the label values ​​corresponding to the samples input to the first initial model to the second training device. The second training device can determine the loss function based on the predicted perception results and label values ​​output by the second initial model, and then determine whether to stop the training process based on the loss function.

[0195] For example, the second training device can feed back the predicted perception results output by the second initial model to the first training device during the training process. The first training device can determine the loss function based on the predicted perception results and label values ​​fed back by the second initial model, and then determine whether to stop the training process based on the loss function.

[0196] Thirdly, embodiments of this disclosure propose a model training method executed by a second training device, the method comprising: receiving predicted semantic quantization data output by a first initial model trained by a first training device; training a second initial model based on the predicted semantic quantization data; wherein the input of the second initial model includes the predicted semantic quantization data.

[0197] In some embodiments, the training of the second initial model, at the point where training stops, can serve as a sensing model, and the output of the sensing model can include the predicted sensing results corresponding to the measurement channel data.

[0198] In some embodiments, the second training device may interact with the first training device during the training of the second initial model in order to collaborate with the first training device in training the model.

[0199] For example, during the training process, the second training device can receive the predicted semantic quantization data output by the first initial model in the first training device. The second training device can train the second initial model based on the predicted semantic quantization data, for example, by inputting the predicted semantic quantization data into the second initial model. The second initial model can output the predicted sensing result of the channel measurement data corresponding to the predicted semantic quantization data.

[0200] For example, during the training process, the second training device can receive the label values ​​corresponding to the samples input to the first initial model sent by the first training device. The second training device can determine the loss function based on the predicted perception results and label values ​​output by the second initial model, and then determine whether to stop the training process based on the loss function.

[0201] For example, the second training device can feed back the predicted perception results output by the second initial model to the first training device during the training process. The first training device can determine the loss function based on the predicted perception results and label values ​​fed back by the second initial model, and then determine whether to stop the training process based on the loss function.

[0202] It should be noted that the concepts in the embodiments involved in the third aspect (such as training stop, channel measurement data, etc.) can be referred to the related content in the embodiments involved in the second aspect above, and will not be repeated here.

[0203] Fourthly, embodiments of this disclosure propose a semantic quantization method. For example, the semantic quantization method is performed by a first device, in which a semantic quantization model is deployed, and the method includes: determining first channel measurement data by measuring a sensing reference signal; and inputting the first channel measurement data into the semantic quantization model to determine first predicted semantic quantization data corresponding to the first channel measurement data.

[0204] In some embodiments, the semantic quantization model is determined based on the following: determining a training sample set, wherein the samples in the training sample set include channel measurement data corresponding to a sensing reference signal, and the labels of the samples include sensing results corresponding to the channel measurement data; training a first initial model and a second initial model based on the training sample set, wherein the first initial model at the end of training is used as the semantic quantization model, and the second initial model at the end of training is used as the sensing model; wherein the input of the semantic quantization model includes the channel measurement data, the output of the semantic quantization model includes predicted semantic quantization data corresponding to the channel measurement data, the input of the sensing model includes the predicted semantic quantization data, and the output of the sensing model includes the predicted sensing results corresponding to the channel measurement data.

[0205] In some embodiments, the semantic quantization model is determined based on the following: determining a training sample set, wherein the samples in the training sample set include channel measurement data corresponding to a sensing reference signal, and the labels of the samples include sensing results corresponding to the channel measurement data; training a first initial model based on the training sample set, and the first initial model at the end of training is used as the semantic quantization model; wherein the input of the semantic quantization model includes the channel measurement data, and the output of the semantic quantization model includes predicted semantic quantization data corresponding to the channel measurement data.

[0206] In some embodiments, the semantic quantization method further includes: sending first predicted semantic quantization data to a second device, wherein the second device is equipped with a perception model, and the perception model determines a first predicted perception result corresponding to the first channel measurement data based on the first predicted semantic quantization data.

[0207] It should be noted that the concepts in the embodiments involved in the fourth aspect (such as training stop, channel measurement data, etc.) can be referred to the related content in the embodiments involved in the second aspect above, and will not be repeated here.

[0208] Figure 3 is a schematic diagram illustrating a predictive perception result according to an embodiment of the present disclosure.

[0209] As shown in Figure 3, based on the trained semantic quantization model and perception model, the semantic quantization model can be deployed on the first device and the semantic quantization model can be deployed on the second device.

[0210] For example, the first device can be called a channel data measurement device, used to acquire channel measurement data.

[0211] For example, the second device can be called a model inference device, which is used to predict channel measurement data to obtain predictive sensing results.

[0212] In some embodiments, the semantic quantization model deployed in the channel data measurement device can perform operations such as feature extraction and compression on the channel measurement data input to the semantic quantization model, and predict the semantic quantization result corresponding to the channel measurement data, which can be referred to as predicted semantic quantization data.

[0213] In some embodiments, a sensing model is deployed in the model inference device. The channel data measurement device can send the predicted semantic quantization data to the model inference device. The model inference device can use the predicted semantic quantization data as input to the sensing model and perform sensing based on this input to predict the sensing result corresponding to the channel measurement data, which may be called the predicted sensing result, including but not limited to sensing results such as distance, angle, and speed.

[0214] Figure 4 is a schematic diagram of a semantically quantized interaction according to an embodiment of the present disclosure.

[0215] As shown in Figure 4, based on the trained semantic quantization model and perception model, the semantic quantization model can be deployed on the first device and the semantic quantization model can be deployed on the second device.

[0216] For example, the first device can be called a reference signal receiving device, and the second device can be called a model inference device.

[0217] The reference signal transmitting device can transmit reference signals, which may include at least one of the following: Channel State Information Reference Signal (CSI-RS), Synchronization Signal Block (SSB), and Sounding Reference Signal (SRS).

[0218] The reference signal receiving device can receive a reference signal, determine channel measurement data based on the received reference signal, input the channel measurement data into a semantic quantization model, and predict the semantic quantization result corresponding to the channel measurement data, which can be called predicted semantic quantization data. Furthermore, the reference signal receiving device can send the predicted semantic quantization data to a model inference device.

[0219] The model inference device can input predicted semantic quantization data into the perception model to predict the perception results corresponding to the channel measurement data, including but not limited to perception results such as distance, angle, and speed. For example, the predicted perception distance can be the distance between the reference signal transmitting device and the reference signal receiving device; the predicted perception angle can be the angle between the reference signal transmitting device and the reference signal receiving device; and the predicted perception speed can be the speed between the reference signal transmitting device and the reference signal receiving device.

[0220] The optional implementations of the first aspect and the optional embodiments of the first aspect can be found in the optional implementations of the embodiments shown in FIG2 and other related parts of the embodiments involved in FIG2, which will not be repeated here.

[0221] In some embodiments, the names of information, etc., are not limited to the names described in the embodiments. Terms such as "information", "message", "signal", "signaling", "report", "configuration", "indication", "instruction", "command", "channel", "parameter", "domain", "field", "symbol", "symbol", "codebook", "codeword", "codepoint", "bit", "data", "program", and "chip" can be used interchangeably.

[0222] In some embodiments, the terms "synchronization signal (SS)," "synchronization signal block (SSB)," "reference signal (RS)," "pilot," and "pilot signal" can be used interchangeably.

[0223] In some embodiments, “get,” “obtain,” “receive,” “transmit,” “bidirectional transmission,” and “send and / or receive” can be used interchangeably and can be interpreted as receiving from other entities, obtaining from protocols, obtaining from higher layers, obtaining through self-processing, or autonomous implementation, among other meanings.

[0224] In some embodiments, terms such as “send,” “transmit,” “report,” “distribute,” “transfer,” “bidirectional transmission,” “send and / or receive” can be used interchangeably.

[0225] In some embodiments, terms such as "certain," "preset," "default," "set," "indicated," "a certain," "any," and "first" can be used interchangeably. "Certain A," "preset A," "default A," "set A," "indicated A," "a certain A," "any A," and "first A" can be interpreted as A pre-defined in a protocol or the like, or as A obtained through setting, configuration, or instruction, or as specific A, a certain A, any A, or first A, but are not limited thereto.

[0226] Corresponding to the aforementioned embodiments of the model determination method and semantic quantization method, this disclosure also provides embodiments of the model determination apparatus and the semantic quantization apparatus.

[0227] Figure 5 is a schematic block diagram of a model determination device according to an embodiment of the present disclosure. As shown in Figure 5, the model determination device includes a processing module 501.

[0228] In some embodiments, the processing module is configured to determine a training sample set, wherein the samples in the training sample set include channel measurement data corresponding to a sensing reference signal, and the labels of the samples include sensing results corresponding to the channel measurement data; train a first initial model and a second initial model based on the training sample set, wherein the first initial model at the end of training is used as a semantic quantization model, and the second initial model at the end of training is used as a sensing model; wherein the input of the semantic quantization model includes the channel measurement data, the output of the semantic quantization model includes predicted semantic quantization data corresponding to the channel measurement data, the input of the sensing model includes the predicted semantic quantization data, and the output of the sensing model includes predicted sensing results corresponding to the channel measurement data.

[0229] In some embodiments, during the training of the first initial model and the second initial model based on the training sample set, the difference between the predicted perception result and the actual perception result is used as a loss function.

[0230] In some embodiments, the conditions for stopping training include at least one of the following: the result of the loss function is within a threshold range; or the number of training cycles is greater than or equal to a threshold.

[0231] In some embodiments, the predicted perception result includes at least one of the following: distance perception result; speed perception result; angle perception result.

[0232] In some embodiments, the semantic quantization model is deployed on a first device, and / or the perception model is deployed on a second device; the first device is configured to determine the channel measurement data by receiving a perception reference signal, and to send the predicted semantic quantization data to the second device.

[0233] In some embodiments, the channel measurement data includes at least one of the following: relevant information of the sensing reference signal; a channel state information matrix corresponding to the sensing reference signal; a complex number corresponding to the channel state information matrix; spectral information of the channel corresponding to the sensing reference signal; and point cloud information of the channel corresponding to the sensing reference signal.

[0234] In some embodiments, the relevant information of the sensing reference signal includes at least one of the following: the amplitude of the sensing reference signal; the phase of the sensing reference signal; the I-channel information of the sensing reference signal; and the Q-channel information of the sensing reference signal.

[0235] In some embodiments, the spectral information includes at least one of the following: time delay spread spectral information; Doppler spectral information; micro-Doppler spectral information; angular spectral information; and signal intensity spectral information.

[0236] Figure 6 is a schematic block diagram illustrating a semantic quantization device according to an embodiment of the present disclosure. For example, the semantic quantization device may be disposed in a first device. As shown in Figure 6, the model determination device includes: a processing module 601 and a sending module 602.

[0237] In some embodiments, the processing module is configured to determine first channel measurement data by measuring a sensing reference signal; and input the first channel measurement data into a semantic quantization model based on any one of the first aspect and optional embodiments of the first aspect to determine first predicted semantic quantization data corresponding to the first channel measurement data.

[0238] In some embodiments, the sending module is configured to send the first predicted semantic quantization data to a second device, wherein the second device is equipped with the perception model, and the perception model determines a first predicted perception result corresponding to the first channel measurement data based on the first predicted semantic quantization data.

[0239] The embodiments of this disclosure also propose a model training device. For example, the model training device can be set in a first training device. For example, the type of the first training device can be a communication device. For example, the first training device can be the first device, the second device, or the third device in the above embodiments, or it can be other devices.

[0240] In some embodiments, the model training apparatus includes: a processing module configured to determine a training sample set, wherein the samples in the training sample set include channel measurement data corresponding to a sensing reference signal, and the labels of the samples include sensing results corresponding to the channel measurement data; to train a first initial model based on the training sample set, wherein the first initial model at the end of training is used as a semantic quantization model; wherein the input of the semantic quantization model includes the channel measurement data, and the output of the semantic quantization model includes predicted semantic quantization data corresponding to the channel measurement data.

[0241] In some embodiments, the model training apparatus further includes: a sending module configured to send the predicted semantic quantization data output by the first initial model to a second training device, for the second device to train the second initial model; wherein the second initial model at the end of training is used as a perception model, the input of the perception model includes the predicted semantic quantization data, and the output of the perception model includes the predicted perception result corresponding to the channel measurement data.

[0242] The embodiments of this disclosure also propose a model training device. For example, the model training device can be set in a second training device. For example, the type of the second training device can be a communication device. For example, the second training device can be the first device, the second device, or the third device in the above embodiments, or it can be other devices.

[0243] In some embodiments, the model training apparatus includes: a receiving module configured to receive predictive semantic quantization data output by a first initial model trained by a first training device; and a processing module configured to train a second initial model based on the predictive semantic quantization data; wherein the input of the second initial model includes the predictive semantic quantization data.

[0244] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0245] This disclosure also provides an apparatus for implementing any of the above methods. For example, an apparatus is provided that includes units or modules for implementing the steps performed by the terminal in any of the above methods. Alternatively, another apparatus is provided that includes units or modules for implementing the steps performed by a network device (e.g., an access network device, a core network functional node, a core network device, etc.) in any of the above methods.

[0246] It should be understood that the division of units or modules in the above device is only a logical functional division. In actual implementation, they can be fully or partially integrated into a single physical entity, or they can be physically separated. Furthermore, the units or modules in the device can be implemented by a processor calling software: for example, the device includes a processor connected to a memory containing instructions. The processor calls the instructions stored in the memory to implement any of the above methods or to implement the functions of the units or modules in the above device. The processor can be, for example, a general-purpose processor, such as a Central Processing Unit (CPU) or a microprocessor, and the memory can be internal or external to the device. Alternatively, the units or modules in the device can be implemented in the form of hardware circuits. The functionality of some or all of the units or modules can be achieved through the design of these hardware circuits, which can be understood as one or more processors. For example, in one implementation, the hardware circuit is an application-specific integrated circuit (ASIC). The functionality of some or all of the units or modules is achieved through the design of the logical relationships between the components within the circuit. In another implementation, the hardware circuit can be implemented using a programmable logic device (PLD). Taking a field-programmable gate array (FPGA) as an example, it can include a large number of logic gates. The connection relationships between the logic gates are configured through configuration files, thereby achieving the functionality of some or all of the units or modules. All units or modules of the above device can be implemented entirely through processor-called software, entirely through hardware circuits, or partially through processor-called software with the remaining parts implemented through hardware circuits.

[0247] In this embodiment, the processor is a circuit with signal processing capabilities. In one implementation, the processor can be a circuit with instruction read and execute capabilities, such as a Central Processing Unit (CPU), a microprocessor, a graphics processing unit (GPU) (which can be understood as a microprocessor), or a digital signal processor (DSP). In another implementation, the processor can implement certain functions through the logical relationships of hardware circuits. The logical relationships of the aforementioned hardware circuits are fixed or reconfigurable. For example, the processor is a hardware circuit implemented using an application-specific integrated circuit (ASIC) or a programmable logic device (PLD), such as an FPGA. In a reconfigurable hardware circuit, the process of the processor loading a configuration document and configuring the hardware circuit can be understood as the process of the processor loading instructions to implement the functions of some or all of the above units or modules. Furthermore, it can also be a hardware circuit designed for artificial intelligence, which can be understood as an ASIC, such as a Neural Network Processing Unit (NPU), a Tensor Processing Unit (TPU), or a Deep Learning Processing Unit (DPU).

[0248] Figure 7A is a schematic diagram of the structure of the communication device 7100 proposed in an embodiment of this disclosure. The communication device 7100 can be a network device (e.g., access network device, core network device, etc.), a terminal (e.g., user equipment, etc.), a chip, chip system, or processor that supports the network device in implementing any of the above methods, or a chip, chip system, or processor that supports the terminal in implementing any of the above methods. The communication device 7100 can be used to implement the methods described in the above method embodiments; for details, please refer to the descriptions in the above method embodiments.

[0249] As shown in Figure 7A, the communication device 7100 includes one or more processors 7101. The processor 7101 can be a general-purpose processor or a dedicated processor, such as a baseband processor or a central processing unit (CPU). The baseband processor can be used to process communication protocols and communication data, while the CPU can be used to control communication devices (e.g., base stations, baseband chips, terminal devices, terminal device chips, DUs or CUs, etc.), execute programs, and process program data. Optionally, the communication device 7100 can be used to execute any of the above methods. Optionally, one or more processors 7101 can be used to invoke instructions to cause the communication device 7100 to execute any of the above methods.

[0250] In some embodiments, the communication device 7100 further includes one or more transceivers 7102. When the communication device 7100 includes one or more transceivers 7102, the transceiver 7102 performs at least one of the communication steps (e.g., steps S201, S202, but not limited thereto) in the above method, such as sending and / or receiving, while the processor 7101 performs at least one of other steps (e.g., steps S201, S202, but not limited thereto). In optional embodiments, the transceiver may include a receiver and / or a transmitter, which may be separate or integrated. Optionally, the terms transceiver, transceiver unit, transceiver, transceiver circuit, interface circuit, interface, etc., can be used interchangeably; the terms transmitter, sending unit, transmitter, sending circuit, etc., can be used interchangeably; and the terms receiver, receiving unit, receiver, receiving circuit, etc., can be used interchangeably.

[0251] In some embodiments, the communication device 7100 further includes one or more memories 7103 for storing data. Optionally, all or part of the memories 7103 may be located outside the communication device 7100. In optional embodiments, the communication device 7100 may include one or more interface circuits 7104. Optionally, the interface circuits 7104 are connected to the memories 7102 and can be used to receive data from the memories 7102 or other devices, and to send data to the memories 7102 or other devices. For example, the interface circuits 7104 can read data stored in the memories 7102 and send the data to the processor 7101.

[0252] The communication device 7100 described in the above embodiments may be a network device or a terminal, but the scope of the communication device 7100 described in this disclosure is not limited thereto, and the structure of the communication device 7100 may not be limited by FIG. 7A. The communication device may be a standalone device or a part of a larger device. For example, the communication device may be: (1) a standalone integrated circuit IC, or chip, or chip system or subsystem; (2) a collection of one or more ICs, optionally, the IC collection may also include storage components for storing data and programs; (3) an ASIC, such as a modem; (4) a module that can be embedded in other devices; (5) a receiver, terminal device, smart terminal device, cellular phone, wireless device, handheld device, mobile unit, vehicle device, network device, cloud device, artificial intelligence device, etc.; (6) others, etc.

[0253] Figure 7B is a schematic diagram of the structure of the chip 7200 according to an embodiment of this disclosure. For cases where the communication device 7100 can be a chip or a chip system, the schematic diagram of the chip 7200 shown in Figure 7B can be referred to, but is not limited thereto.

[0254] Chip 7200 includes one or more processors 7201. Chip 7200 is used to perform any of the above methods.

[0255] In some embodiments, chip 7200 further includes one or more interface circuits 7202. Optionally, terms such as interface circuit, interface, and transceiver pin can be used interchangeably. In some embodiments, chip 7200 further includes one or more memories 7203 for storing data. Optionally, all or part of the memories 7203 may be located outside chip 7200. Optionally, interface circuit 7202 is connected to memory 7203, and interface circuit 7202 can be used to receive data from memory 7203 or other devices, and interface circuit 7202 can be used to send data to memory 7203 or other devices. For example, interface circuit 7202 can read data stored in memory 7203 and send the data to processor 7201.

[0256] In some embodiments, the interface circuit 7202 performs at least one of the communication steps (e.g., steps S201, S202, but not limited thereto) in the above-described method, such as sending and / or receiving. For example, the interface circuit 7202 performing the communication steps (e.g., steps S201, S202, but not limited thereto) refers to the interface circuit 7202 performing data interaction between the processor 7201, the chip 7200, the memory 7203, or the transceiver device. In some embodiments, the processor 7201 performs at least one of other steps (e.g., steps S201, S202, but not limited thereto).

[0257] The modules and / or devices described in the various embodiments, such as virtual devices, physical devices, and chips, can be combined or separated arbitrarily as needed. Optionally, some or all steps can also be performed collaboratively by multiple modules and / or devices, which is not limited here.

[0258] This disclosure also proposes a storage medium storing instructions that, when executed on the communication device 7100, cause the communication device 7100 to perform any of the above methods. Optionally, the storage medium is an electronic storage medium. Optionally, the storage medium is a computer-readable storage medium, but not limited thereto; it may also be a storage medium readable by other devices. Optionally, the storage medium may be a non-transitory storage medium, but not limited thereto; it may also be a temporary storage medium.

[0259] This disclosure also provides a program product that, when executed by the communication device 7100, causes the communication device 7100 to perform any of the above methods. Optionally, the program product is a computer program product.

[0260] This disclosure also proposes a computer program that, when run on a computer, causes the computer to perform any of the above methods.

Claims

1. A model determination method, characterized in that, Performed by a training device, the method includes: A training sample set is determined, wherein the samples in the training sample set include channel measurement data corresponding to the sensing reference signal, and the labels of the samples include the sensing results corresponding to the channel measurement data; The first initial model and the second initial model are trained based on the training sample set. The first initial model at the end of training is used as the semantic quantization model, and the second initial model at the end of training is used as the perceptual model. The semantic quantization model's input includes channel measurement data, and its output includes predicted semantic quantization data corresponding to the channel measurement data. The perception model's input includes the predicted semantic quantization data, and its output includes the predicted perception result corresponding to the channel measurement data.

2. The method according to claim 1, characterized in that, During the training of the first initial model and the second initial model based on the training sample set, the difference between the predicted perception result and the actual perception result is used as the loss function.

3. The method according to claim 2, characterized in that, Training may be stopped if at least one of the following conditions is met: The result of the loss function is within the threshold range; The number of training cycles is greater than or equal to the number threshold.

4. The method according to any one of claims 1 to 3, characterized in that, The predicted perception result includes at least one of the following: Distance perception results; Speed ​​perception results; Angle perception results.

5. The method according to any one of claims 1 to 4, characterized in that, The channel measurement data includes at least one of the following: Relevant information about the sensing reference signal; The channel state information matrix corresponding to the sensing reference signal; The complex number corresponding to the channel state information matrix; The spectral information of the channel corresponding to the sensing reference signal; The sensing reference signal corresponds to the point cloud information of the channel.

6. The method according to claim 5, characterized in that, The relevant information of the sensing reference signal includes at least one of the following: The amplitude of the sensing reference signal; The phase of the sensing reference signal; The I-channel information of the sensing reference signal; The Q-path information of the sensing reference signal.

7. The method according to claim 5, characterized in that, The spectral information includes at least one of the following: Time delay spread spectrum information; Doppler spectral information; Micro-Doppler spectral information; Angular spectrum information; Signal intensity spectrum information.

8. A model training method, characterized in that, Performed by a first training device, the method includes: A training sample set is determined, wherein the samples in the training sample set include channel measurement data corresponding to the sensing reference signal, and the labels of the samples include the sensing results corresponding to the channel measurement data; The first initial model is trained based on the training sample set, and the first initial model at the end of training is used as the semantic quantization model. The input to the semantic quantization model includes channel measurement data, and the output of the semantic quantization model includes... The predicted semantic quantization data corresponding to the channel measurement data.

9. The method according to claim 8, characterized in that, The method further includes: The predicted semantic quantization data output by the first initial model is sent to the second training device for the second device to train the second initial model; Wherein, the second initial model at the time of training stop serves as the perception model, the input of the perception model includes the predicted semantic quantization data, and the output of the perception model includes the predicted perception result corresponding to the channel measurement data.

10. A model training method, characterized in that, Performed by a second training device, the method includes: Receive the predicted semantic quantization data output by the first initial model trained by the first training device; The second initial model is trained based on the predicted semantic quantization data; The input to the second initial model includes the predicted semantic quantization data.

11. A semantic quantization method, characterized in that, The method, executed by a first device in which the semantic quantization model is deployed, includes: The first channel measurement data is determined by measuring the sensing reference signal; The first channel measurement data is input into the semantic quantization model to determine the first predicted semantic quantization data corresponding to the first channel measurement data.

12. The method according to claim 11, characterized in that, The semantic quantization model is determined based on the following method: A training sample set is determined, wherein the samples in the training sample set include channel measurement data corresponding to the sensing reference signal, and the labels of the samples include the sensing results corresponding to the channel measurement data; The first initial model and the second initial model are trained based on the training sample set. The first initial model at the end of training is used as the semantic quantization model, and the second initial model at the end of training is used as the perceptual model. The semantic quantization model's input includes channel measurement data, and its output includes predicted semantic quantization data corresponding to the channel measurement data. The perception model's input includes the predicted semantic quantization data, and its output includes the predicted perception result corresponding to the channel measurement data.

13. The method according to claim 11, characterized in that, The semantic quantization model is determined based on the following method: A training sample set is determined, wherein the samples in the training sample set include channel measurement data corresponding to the sensing reference signal, and the labels of the samples include the sensing results corresponding to the channel measurement data; The first initial model is trained based on the training sample set, and the first initial model at the end of training is used as the semantic quantization model. The input to the semantic quantization model includes channel measurement data, and the output of the semantic quantization model includes predicted semantic quantization data corresponding to the channel measurement data.

14. The method according to any one of claims 11 to 13, characterized in that, The method further includes: The first predicted semantic quantization data is sent to the second device, wherein the second device is equipped with the perception model, and the perception model determines the first predicted perception result corresponding to the first channel measurement data based on the first predicted semantic quantization data.

15. A model determining device, characterized in that, The device includes: The processing module is configured to determine a training sample set, wherein the samples in the training sample set include channel measurement data corresponding to a sensing reference signal, and the labels of the samples include sensing results corresponding to the channel measurement data; and to train a first initial model and a second initial model based on the training sample set, wherein the first initial model at the end of training is used as a semantic quantization model, and the second initial model at the end of training is used as a sensing model. The semantic quantization model's input includes channel measurement data, and its output includes predicted semantic quantization data corresponding to the channel measurement data. The perception model's input includes the predicted semantic quantization data, and its output includes the predicted perception result corresponding to the channel measurement data.

16. A model training device, characterized in that, The device includes: The processing module is configured to determine a training sample set, wherein the samples in the training sample set include channel measurement data corresponding to a sensing reference signal, and the labels of the samples include sensing results corresponding to the channel measurement data; and to train a first initial model based on the training sample set, wherein the first initial model at the end of training is used as a semantic quantization model. The input to the semantic quantization model includes channel measurement data, and the output of the semantic quantization model includes predicted semantic quantization data corresponding to the channel measurement data.

17. A model training device, characterized in that, The device includes: The receiving module is configured to receive the predicted semantic quantization data output by the first initial model trained by the first training device; The processing module is configured to train the second initial model based on the predicted semantic quantization data; The input to the second initial model includes the predicted semantic quantization data.

18. A semantic quantization device, characterized in that, The device includes: The processing module is configured to determine first channel measurement data by measuring a sensing reference signal; and input the first channel measurement data into a semantic quantization model to determine first predicted semantic quantization data corresponding to the first channel measurement data.

19. A communication device, characterized in that, include: One or more processors; The communication device is used to execute the model training method of any one of claims 1 to 10, and / or the semantic quantization method of any one of claims 11 to 14.

20. A storage medium storing instructions, characterized in that, When the instruction is executed on the communication device, the communication device performs the model training method of any one of claims 1 to 10, and / or the semantic quantization method of any one of claims 11 to 14.

21. A program product, characterized in that, When the above-mentioned program product is executed by a communication device, the communication device performs the model training method of any one of claims 1 to 10, and / or the semantic quantization method of any one of claims 11 to 14.

Images