Model-based information service providing method, system, device and medium

Abstract

The present application relates to the field of communication technology, especially to a kind of information service providing method, system and equipment based on model, applied to the downlink scene of the same category service information being issued by base station to the first quantity of user equipment, and specific technical scheme includes: the base station converts the service information into semantic encoding;The base station divides the semantic encoding, obtains common part and individual part;The base station transmits the common part and the individual part to the user equipment;The user equipment combines the common part and the individual part, obtains second semantic encoding;The user equipment decodes the second semantic encoding into second service information.The present application can carry out semantic encoding and semantic splitting to service information and carry out multiple access transmission, can significantly reduce the data transmission amount in communication service, greatly improves information transmission efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of communication technology, in particular to a model-based information service providing method, system, device and medium. BACKGROUND

[0002] In the future Internet of Everything network, network nodes tend to be intelligent, and the intelligence of network nodes leads to rapid expansion of information space and even dimension disaster, which exacerbates the difficulty of representing information bearing space, causes the traditional network service capability to be difficult to match the high-dimensional information space, and results in too large data volume of communication transmission, so that the information service system cannot continuously meet the needs of people for complex, diverse and intelligent information transmission.

[0003] In order to transmit more information, the existing technology generally compresses the service information to be transmitted at the signal source end, for example, using a high compression rate image coding format when transmitting an image. At present, various types of service information (such as text, image, voice, video, etc.) have mature compression technology. The above technology belongs to signal source end coding technology. In order to transmit more information on fewer channels, the characteristics of the signal source and the channel can be considered comprehensively, that is, signal source and channel joint coding.

[0004] There is existing technology that connects multiple user equipment to multiple base stations to establish multiple channels, so that the base station can transmit multiple information to multiple user equipment at the same time. However, in the case of keeping the number of channels unchanged, there is existing technology that allows multiple user equipment to share channels by time division, frequency division and other methods, that is, multiple access technology. However, the utilization rate of the channel and the transmission efficiency of the service information in the existing technology are difficult to further improve. SUMMARY

[0005] In view of the above problems existing in the prior art, a model-based information service providing method, system, device and medium are provided.

[0006] The present application includes a model-based information service providing method, comprising:

[0007] The base station converts the service information into semantic coding;

[0008] The base station divides the semantic coding to obtain a common part and an individual part;

[0009] The base station transmits the common part and the individual part to the user equipment;

[0010] The user equipment combines the common part and the individual part to obtain second semantic coding;

[0011] The user equipment decodes the second semantic coding into second service information.

[0012] Optionally, the common part has the same value, opposite value, positive similarity or negative similarity with the value of the corresponding position of the common part of other semantic encoding; and the individual part is the part of the semantic encoding after removing the common part.

[0013] Optionally, before the base station transmits the common part and the individual part to the user equipment, the base station notifies the user equipment to transmit.

[0014] Optionally, the base station transmits the common part and the individual part to the user equipment, including: the base station transmits a first number of the individual parts to a first number of the user equipment in different time-frequency resources.

[0015] Optionally, the base station transmits the common part and the individual part to the user equipment, including: the base station selects one of the first number of the common information to transmit to the first number of the user equipment, or the base station selects two or more of the common information to fuse and transmits to the first number of the user equipment, and the values of the corresponding positions of the common parts participating in the fusion are adjusted to be the same or positively similar.

[0016] Optionally, the base station converts the service information into semantic encoding, including: the base station uses a pre-configured semantic encoding model to convert the service information to be transmitted into semantic encoding; the base station generates the semantic encoding model locally, or the base station obtains the semantic encoding model from other devices.

[0017] Optionally, the user equipment decodes the second semantic encoding into second service information, including: the user equipment uses a pre-configured semantic decoding model to decode the second semantic encoding into second service information; the user equipment generates the semantic decoding model locally, or the user equipment obtains the semantic decoding model from other devices.

[0018] Optionally, the semantic encoding model or the semantic decoding model is divided into a plurality of model components according to a division rule for storage or transmission; the division rule includes at least one of the following: horizontal layer division; vertical cross-layer division; random division; division based on data volume; division based on function; division based on use.

[0019] Optionally, the base station divides the semantic encoding to obtain the common part and the individual part, including: based on a position mask representing common semantics, obtaining the common part from the semantic encoding; and based on a position mask representing individual semantics, obtaining the individual part from the semantic encoding.

[0020] Optionally, the user equipment combines the common part and the individual part to obtain a second semantic code, comprising: the user equipment combines the common part and the individual part into a second semantic code based on the position mask representing the common semantic and the position mask representing the individual semantic.

[0021] Optionally, the service information comprises at least one of the following: text, audio, image, video, point cloud, one-dimensional waveform data, radar signal.

[0022] Optionally, the semantic code adopts source channel joint coding or source coding.

[0023] Optionally, the transmission adopts any one of the following: time division multiple access (TDMA), frequency division multiple access (FDMA), interleaved multiple access (IDMA), space division multiple access (SDMA), orthogonal frequency division multiple access (OFDMA), non-orthogonal multiple access (NOMA).

[0024] The application also includes a model-based information service providing method, applied to a base station in a downlink scenario, characterized in that:

[0025] The base station converts the service information to be transmitted into a semantic code;

[0026] The base station splits the semantic code to obtain a common part and an individual part;

[0027] The base station transmits the common part and the individual part to the user equipment.

[0028] Optionally, the method further comprises: the common part and the common part of other semantic codes have the same, opposite, positive similar or negative similar values at the corresponding positions; and the individual part is the part after removing the common part from the semantic code.

[0029] Optionally, the method further comprises: the base station sends a transmission request for transmitting service information to the user equipment; and the base station receives a confirmation message for transmission sent by the user equipment.

[0030] Optionally, the base station transmits the common part and the individual part to the user equipment, comprising: the base station selects one of the first number of common information to transmit to the first number of user equipment, or the base station selects two or more common information from the first number of common information to fuse and transmit to the first number of user equipment, and the values of the corresponding positions of the common part participating in the fusion are adjusted to be the same or positively similar.

[0031] Optionally, the base station transmits the common part and the individual part to the user equipment, comprising: the base station transmits a first number of the individual parts to a first number of the user equipment at different time-frequency resources.

[0032] Optionally, the base station converts the service information into semantic encoding, comprising: the base station converts the service information to be transmitted into semantic encoding using a pre-configured semantic encoding model; the base station generates the semantic encoding model locally, or the base station obtains the semantic encoding model from other devices.

[0033] Optionally, the semantic encoding model is divided into several model components according to a division rule for storage or transmission; the division rule comprises at least one of the following: horizontal layer division; vertical layer division; random division; data volume-based division; function-based division; purpose-based division.

[0034] Optionally, the base station divides the semantic encoding to obtain the common part and the individual part, comprising: obtaining the common part from the semantic encoding based on a position mask representing common semantics; obtaining the individual part from the semantic encoding based on a position mask representing individual semantics.

[0035] Optionally, the service information comprises at least one of the following: text, audio, image, video, point cloud, one-dimensional waveform data, radar signal.

[0036] Optionally, the semantic encoding adopts source channel joint encoding or source encoding.

[0037] Optionally, the transmission adopts any one of the following: time division multiple access (TDMA), frequency division multiple access (FDMA), interleaved multiple access (IDMA), space division multiple access (SDMA), orthogonal frequency division multiple access (OFDMA), non-orthogonal multiple access (NOMA).

[0038] The application also includes a model-based information service providing method, applied to a user equipment in a downlink scenario, characterized in that:

[0039] The user equipment receives a transmission request for transmitting service information from the base station;

[0040] The user equipment receives the common part and the individual part from the base station;

[0041] The user equipment combines the common part and the individual part to obtain a second semantic encoding;

[0042] The user equipment decodes the second semantic encoding into a second service information.

[0043] Optionally, the method further comprises: the base station receiving transmission requests of service information from a second number of user devices; and based on the second number of transmission requests, the base station instructing the first number of user devices to transmit.

[0044] Optionally, the method comprises: the common part and the individual part being obtained by the user device dividing semantic encoding; and the semantic encoding being obtained by the user device encoding service information to be transmitted.

[0045] Optionally, the user device decoding the second semantic encoding into second service information comprises: the user device decoding the second semantic encoding into second service information using a pre-configured semantic decoding model; and the user device generating the semantic decoding model locally or obtaining the semantic decoding model from other devices.

[0046] Optionally, the semantic decoding model is divided into a plurality of model components according to a division rule for storage or transmission; and the division rule comprises at least one of the following: horizontal layer division; vertical layer division; random division; division based on data volume; division based on function; and division based on use.

[0047] Optionally, the common part has the same value, opposite value, positive similarity or negative similarity with the common part of other semantic encoding at a corresponding position; and the individual part is a part of the semantic encoding after removing the common part.

[0048] Optionally, the user device receiving the common part and the individual part from the base station comprises:

[0049] The base station receives one copy of the common part and the first number of individual parts in different time-frequency resources.

[0050] Optionally, the service information comprises at least one of the following: text, audio, image, video, point cloud, one-dimensional waveform data, and radar signal.

[0051] Optionally, the semantic encoding adopts source channel joint encoding or source encoding.

[0052] Optionally, the transmission adopts any one of the following: time division multiple access (TDMA), frequency division multiple access (FDMA), interleaved multiple access (IDMA), space division multiple access (SDMA), orthogonal frequency division multiple access (OFDMA), and non-orthogonal multiple access (NOMA).

[0053] The application further comprises a model-based information service providing system applied to an uplink scenario of a first number of user devices to a base station, comprising:

[0054] An encoding module, the base station converts the service information into semantic encoding;

[0055] A code splitting module, the base station splits the semantic encoding to get common part and individual part;

[0056] A transmission module, the base station transmits the common part and the individual part to the user equipment;

[0057] A code combining module, the user equipment combines the common part and the individual part to get second semantic encoding;

[0058] A decoding module, the user equipment decodes the second semantic encoding into second service information.

[0059] Optionally, the system corresponds to the method applied to the uplink scenario of the first number of user equipment to the base station one by one. The system comprises modules corresponding to the subdivision steps of the foregoing method.

[0060] The application also includes a model-based information service providing system, applied to the base station in the downlink scenario, characterized in that:

[0061] An encoding module, the base station converts the service information into semantic encoding;

[0062] A code splitting module, the base station splits the semantic encoding to get common part and individual part;

[0063] A transmission module, the base station transmits the common part and the individual part to the user equipment.

[0064] Optionally, the system corresponds to the method applied to the base station in the downlink scenario one by one. The system comprises modules corresponding to the subdivision steps of the foregoing method.

[0065] The application also includes a model-based information service providing system, applied to the user equipment in the downlink scenario, characterized in that:

[0066] A grouping module, the user equipment receives the transmission request of the transmission service information from the base station;

[0067] A receiving module, the user equipment receives the common part and the individual part from the base station;

[0068] A code combining module, the user equipment combines the common part and the individual part to get second semantic encoding;

[0069] A decoding module, the user equipment decodes the second semantic encoding into second service information.

[0070] Optionally, the system corresponds to the aforementioned method applied to the user equipment in the downlink scenario. The system comprises modules corresponding to the sub-steps of the aforementioned method.

[0071] The application further includes an electronic device comprising a processor, a communication interface, a memory and a communication bus, wherein the communication bus is used to complete mutual communication between the processor, the communication interface and the memory.

[0072] The memory is used to store a computer program.

[0073] The processor is used to execute the computer program stored in the memory, so as to realize the information service providing method in any one of the above technical solutions.

[0074] The application further includes a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program realizes the information service providing method in any one of the above technical solutions when executed by a processor.

[0075] The application further includes a computer program product, which, when executed on a computer, causes the computer to execute the information service providing method in any one of the above technical solutions.

[0076] The application provides a model-based information service providing method, system, device and medium, which can utilize a pre-trained model to perform semantic encoding and multiple access transmission on service information of multiple user equipment, can significantly reduce data transmission amount in communication service, and greatly improves information transmission efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0077] Reference will now be made to the drawings to describe the embodiments of the application in more detail. However, the accompanying drawings are only used for illustration and explanation, and do not constitute a limitation on the scope of the application.

[0078] Figure 1 A schematic diagram of an application scenario provided for the embodiments of the application;

[0079] Figure 2 A step flowchart of an information service providing method provided for the embodiments of the application;

[0080] Figure 3 A principle block diagram of an information service providing system provided for the embodiments of the application;

[0081] Figure 4 A schematic diagram of one of semantic encoding provided for the embodiments of the application;

[0082] Figure 5A schematic diagram of the second semantic encoding provided for the embodiment of the present application;

[0083] Figure 6 A schematic diagram of semantic encoding splitting provided for the embodiment of the present application;

[0084] Figure 7 A schematic diagram of the extraction matrix of the common part provided for the embodiment of the present application;

[0085] Figure 8 A schematic diagram of the extraction vector of the common part provided for the embodiment of the present application;

[0086] Figure 9 A schematic diagram of the extraction matrix of the individual part provided for the embodiment of the present application;

[0087] Figure 10 A block diagram of an electronic device for implementing the method of the embodiment of the present application. DETAILED DESCRIPTION

[0088] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of protection of the present application.

[0089] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict.

[0090] Multiple access: namely multiple access. When multiple users access a common transmission medium to realize communication with each other, it is necessary to distinguish the users in a certain domain. Commonly used domains for distinguishing include a time domain, a frequency domain, and a space domain.

[0091] Time-frequency resource: when transmitting data between a sending end and a receiving end, multiple devices participating in transmission are divided into respective transmission intervals according to time division and frequency division. When a device is allocated to a specific time domain interval and / or a frequency domain interval, the time domain interval and / or the frequency domain interval become a time-frequency resource available to the device. That is, the device performs multiple access transmission in the time-frequency resource.

[0092] Model division multiple access: the present case is distinguished by model-extracted semantics, which can be referred to as a semantic domain. This distinguishing manner can be called model division multiple access (MDMA) for short. Specific semantic multiple access is described in detail below.

[0093] Service information: information involved in information service. According to the different types of information, it can be divided into text service information, image service information, voice service information, video service information, point cloud service information, etc. It can also be divided into structured information and unstructured information. It can also be all possible types of service information for communication transmission.

[0094] Semantic encoding: semantic information contained in service information. According to the different types of semantics, it can be divided into text semantic encoding, image semantic encoding, audio semantic encoding, video semantic encoding, point cloud semantic encoding, etc. The possible categories of semantic encoding are related to the possible categories of service information, and can be all possible types of semantic encoding for communication transmission. The generation of semantic encoding is also related to the semantic encoding model, which can use all possible models of artificial intelligence, deep learning, pattern recognition, etc.

[0095] Semantic multiple access: after converting the service information to be transmitted into semantic encoding, the semantic encoding is divided into common part and individual part, and the common part and the individual part corresponding to each semantic encoding are transmitted for multiple access transmission.

[0096] Common part: after multiple service information is converted into multiple semantic encodings, the common part obtained by superimposing the multiple semantic encodings in the semantic domain.

[0097] Individual part: after the service information is converted into semantic encoding, the semantic encoding can be divided into common part and individual part. The semantic encoding obtained by converting the service information is the individual part that is different from the common part in the semantic domain.

[0098] Transmitting end and receiving end: there is a channel connection between the transmitting end and the receiving end, the transmitting end (transmitter) has an encoder, and the receiving end (receiver) has a decoder. The present application is based on semantic encoding and semantic decoding, so the transmitting end of the present application has a semantic encoder; the receiving end of the present application has a semantic decoder. The transmitting end and the receiving end of the present application work in the semantic layer, and the bottom layer is still the Shannon physical layer.

[0099] Artificial intelligence model: including artificial intelligence semantic encoding model and artificial intelligence semantic decoding model. The artificial intelligence semantic encoding model is used to encode the service information from the signal source into semantic information. The semantic information is transmitted on the channel. The artificial intelligence semantic decoder is used to decode the semantic information transmitted on the channel into service information. According to the different types of semantics, it can be divided into text artificial intelligence model, audio artificial intelligence model, image artificial intelligence model, video artificial intelligence model, point cloud artificial intelligence model, etc. The possible categories of artificial intelligence model are related to the possible categories of semantic encoding and the possible categories of service information, and can be all possible types of model for communication transmission.

[0100] The general application scenario of this invention is a downlink scenario with a base station and multiple user devices. Without loss of generality, the general application scenario of this invention can have one transmitter and M receivers, where M is a positive integer.

[0101] As attached Figure 1 In the exemplary scenario shown, the transmitting device includes a base station 402. The receiving device includes a first user equipment 403 and a second user equipment 404. A channel 401 connects the two user equipments to the base station.

[0102] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0103] Example 1

[0104] This invention provides a model-based method for providing information services, which utilizes models to provide information services. The method steps are as follows: Figure 2 As shown, it includes:

[0105] Step S101: The base station converts the service information into semantic code;

[0106] Step S102: The base station divides the semantic code to obtain a common part and a unique part;

[0107] Step S103: The base station transmits the common part and the individual part to the user equipment;

[0108] Step S104: The user equipment combines the common part and the individual part to obtain the second semantic code;

[0109] In step S105, the user equipment encodes and decodes the second semantic information into second service information.

[0110] (i) Steps prior to step S101, the base station prepares to transmit service information.

[0111] by Figure 1 For example, base station 402 is preparing to transmit service information 406 (not shown in the figure) to first user equipment 403, and base station 402 is preparing to transmit service information 407 (not shown in the figure) to second user equipment 404.

[0112] Therefore, base station 402 will send a transmission request for service information 406 to the first user equipment 403; base station 402 will send a transmission request for service information 407 to the second user equipment 404.

[0113] Optionally, the base station 402 obtains the service information 406 and the service information 407 at different time. The source of the service information 406 and the source of the service information 407 can be the same or different. The source of the service information 406 and the destination of the service information 407 can be the same or different. For example, the base station 402 transmits both the service information 406 and the service information 407 to the first user equipment 403, and still can achieve the effect of semantic multiple access transmission by transmitting the common information only once.

[0114] Optionally, the base station 402 can decide to start multiple access transmission when the number of service information received is insufficient, and at this time the base station 402 needs to find other service information to group for multiple access transmission. For example, after obtaining the transmission requirement of the service information 406, the base station 402 detects whether there is other service information of the same category to be transmitted, which can be grouped for multiple access transmission. For example, after obtaining the transmission requirement of the service information 406 to be transmitted to the first user equipment 403, the base station 402 detects whether there is other service information that can be transmitted together for multiple access transmission. The base station 402 detects that there is a transmission requirement of the service information 407 to be transmitted to the second user equipment 404, and then the base station 402 can process the service information 406 and the service information 407 together, i.e. multiple access transmission.

[0115] Optionally, the base station 402 can also decide to start multiple access transmission after receiving service information sufficient to be grouped for multiple access transmission. For example, assuming that multiple access transmission requires 2 service information, the base station 402 has already received the transmission requirements of the service information 406 and 407, and the base station 402 starts to decide to start multiple access transmission.

[0116] Optionally, the transmission of the base station 402 can be passive or active. For example, the base station passively receives the transmission requirement of the source or the user equipment; or the base station 402 actively decides to transmit.

[0117] Optionally, after the base station obtains the service information, it needs to judge whether the service information meets the condition of multiple access transmission. For example, the base station 402 needs to judge whether the service information 406 and the service information 407 meet the condition of transmission on the channel 401.

[0118] For another example, the condition includes whether the service information is service information of the same category. For example, two or more service information are images, and then common semantic encoding can be extracted from the two images to be transmitted.

[0119] For another example, the condition also includes whether the transmission time window of the service information 406 and 407 matches. Only when the service information that matches in time is transmitted for multiple access transmission, time delay will not be caused.

[0120] For example, the condition further includes whether the base station has a communication connection with two user equipments and whether there are suitable time-frequency resources to form a common channel. If the signal of one user equipment is too weak or the time-frequency resources are too different from another, it is difficult to perform multiple access transmission.

[0121] The above-mentioned various measures of obtaining a service information transmission request can effectively screen service information that can be subjected to multiple access transmission, so that it is possible to transmit information to multiple user equipments using the same channel. The effect of more user equipments being served while the number of channels remains unchanged is achieved, and the effect of more information being transmitted while the number of channels remains unchanged is also achieved.

[0122] (II) Referring to step S101, the base station converts the service information into semantic encoding. In this embodiment, to achieve multiple access transmission of semantic encoding, a plurality of service information is obtained and converted into semantic encoding.

[0123] Optionally, the base station converts the service information into semantic encoding, including: the base station uses a pre-configured semantic encoding model to convert the service information to be transmitted into semantic encoding; the base station generates the semantic encoding model locally, or the base station obtains the semantic encoding model from other equipment.

[0124] Optionally, the pre-configured semantic encoding model can be classified according to the type of service information (i.e. the type of semantic encoding after encoding). Taking the type of service information as an example, the semantic encoding model is an image semantic encoding model, and the corresponding semantic decoding model is also an image semantic decoding model.

[0125] Optionally, the semantic encoder is configured in pairs with a semantic decoder, and the semantic decoder can decode the semantic encoding generated according to the original service information to obtain the restored service information. The network equipment at both ends (the sending end and the receiving end) of the channel for transmitting one service information should be configured with a semantic encoding model and a semantic decoding model that are functionally corresponding (i.e. encoding and decoding are a reversible process).

[0126] Optionally, the semantic encoding model and the semantic decoding model can be common artificial intelligence models.

[0127] Specifically, the semantic encoding model in this embodiment can adopt an image semantic encoding model based on a convolutional neural network algorithm; the semantic decoding model can adopt an image semantic decoding model based on an adversarial generative network technology or other deep learning network technology. It should be noted that the second service information obtained by using the semantic decoding model to recover the semantic encoding is highly similar to the service information of the sending end. For example, the service information is a cat image, the semantic information (semantic encoding) is obtained by using the semantic encoding model to extract the semantics of the cat image, the intermediate is decomposed into common information and individual information, and the semantic information (second semantic encoding) is obtained by recombining the common information and the individual information after channel transmission to the receiving end; and the cat image (second service information) is obtained by using the semantic decoding model to reconstruct the semantic information. The cat image recovered by the semantic decoding model has some differences in image quality compared with the original cat image, but the contents of the two are consistent, and the visual experience of the user is almost the same.

[0128] Optionally, the semantic encoding model is preconfigured in the plurality of sending ends respectively; and the semantic decoding model corresponding to the semantic encoding model is preconfigured in the plurality of receiving ends respectively. For network devices that frequently transmit common categories of service information (such as text, image, audio, video, etc.), it is necessary to preconfigure the semantic encoders and the semantic decoders of these categories.

[0129] Optionally, the semantic encoding model or the semantic decoding model is divided into a plurality of model components according to a division rule for storage or transmission; the division rule includes at least one of the following: horizontal layer division; vertical layer division; random division; data volume-based division; function-based division; and purpose-based division.

[0130] Optionally, the image semantic encoding model encodes the input image service data into a fixed length of semantic encoding. The fixed length can be determined according to the range of the transmitted image service data (which can be determined from the richness of image semantics and the scale of the training data set). If the image transmitted on the channel is limited to a certain subdivision field, the fixed length can be relatively small; if the image transmitted on the channel involves more categories and more semantics, the fixed length can be relatively large.

[0131] For example, based on the OpenImage data set, the semantic encoding model encodes a 512*512*3 (RGB three image channels) color input image to obtain 4096-bit semantic encoding. The compression ratio of the semantic encoding is about 4096 / (512*512*3)=0.52%.

[0132] The semantic encoding model and the semantic decoding model are trained based on a domain general data set, and thus can be compatible with all business information in the same domain. For example, the OpenImage data set has complete data labels and a large amount of data, and can cover the encoding requirements of all business information in the image domain.

[0133] Similarly, for different types of business information, including but not limited to images, voices, texts, videos, and point clouds, semantic encoding models and semantic decoding models of each type can be trained respectively, so that the models deployed on the communication device can normally process all business information that needs to be transmitted in the communication.

[0134] As shown in FIG. 4, it is a schematic diagram of semantic encoding extracted by the semantic encoding model. FIG. 4 does not draw all 4096-bit semantic encodings, but uses a 12-bit semantic encoding segment as an illustration. Figures 4-5 Figures 4-5 FIG. 4 does not draw all 4096-bit semantic encodings, but uses a 12-bit semantic encoding segment as an illustration.

[0135] Referring to step S102, the base station divides the semantic encoding to obtain a common part and an individual part.

[0136] Optionally, the common part has the same value, opposite value, positive similarity or negative similarity with the common part of other semantic encodings at a corresponding position; and the individual part is a part of the semantic encoding after removing the common part.

[0137] The base station converts each business information into semantic encoding based on the semantic encoding model. Experiments prove that the semantic encoding includes a common part and an individual part, and can be distinguished according to the position.

[0138] As shown in FIG. 5, taking an application scenario as an example, it is assumed that x is business information 406 (not shown in the figure) prepared to be transmitted to the first user equipment 403, and semantic encoding Sx is obtained after the business information 406 is encoded semantically; it is assumed that y is business information 407 (not shown in the figure) prepared to be transmitted to the second sending end 404, and Sy is the semantic encoding of the business information 407 after semantic encoding. Figure 1 It is assumed that a segment of the semantic encoding Sx is as shown in FIG. 6; it is assumed that a segment of the semantic encoding Sy is as shown in FIG. 7; each semantic encoding bit has a numerical position sequence 1-12.

[0139] Figure 4 It is assumed that a segment of the semantic encoding Sx is as shown in FIG. 6; it is assumed that a segment of the semantic encoding Sy is as shown in FIG. 7; each semantic encoding bit has a numerical position sequence 1-12. Figure 5

[0140] ​​​For commonalities, certain bits in the semantic codes obtained by specific types of semantic coding models for that type of business information are "common" (or "invariant"). Taking image semantic coding models based on convolutional neural networks as an example, convolutional layers extract basic image elements at various scales, and some basic elements remain unchanged across different images. These basic image elements are reflected in the image semantic coding; that is, in the image semantic codes obtained from different images, certain coded bits are "common": identical or positively similar, or opposite or negatively similar.

[0141] Experiments have shown that the commonalities include four cases: the values ​​of the commonalities at corresponding positions are the same, opposite, positively similar, or negatively similar. "Same" means that the encoded values ​​of the two commonality encoded segments are exactly the same; "positively similar" means that the encoded values ​​of the two commonality encoded segments are highly similar, or their similarity is higher than a predetermined threshold. "Opposite" means that after bitwise inversion, one commonality encoded segment is exactly the same as the other; "negatively similar" means that after bitwise inversion, one commonality encoded segment is highly similar to the other, or their similarity is higher than a predetermined threshold.

[0142] Optionally, when the semantic coding model encodes the input business data into a fixed-length semantic code, the positions of the common and unique parts of the semantic code remain relatively constant. That is, the common and unique parts can be obtained from relatively fixed positions. For example, if the semantic coding model obtains a 4096-bit semantic code, and for multiple different input images, assume the output of the first bit remains unchanged, then the code of the first bit belongs to the common part. Assuming the output of the 101st bit changes with the input image, then the semantic code of the 101st bit belongs to the unique part.

[0143] Optionally, each position of the semantic code generated by a specific semantic coding model has an attribute ("C" or "P"), where C represents commonality, that is, the position belongs to the commonality part, and P represents individuality, that is, the position belongs to the individuality part.

[0144] by Figure 1 Taking the scenario of 1 base station and 2 user equipment as an example, the transmission of service information 406 and service information 407 meets the conditions for multiple access transmission, meaning that they are of the same type and meet other multiple access transmission conditions. The encoding yields semantic codes Sx and Sy, both of which include common and unique components.

[0145] That is, semantic encoding Sx can be decomposed into a common part Scx and a unique part Spx; while semantic encoding Sy can be decomposed into a common part Scx and a unique part Spy. Among them, Scx and Sxc meet the conditions of being the same, opposite, positively similar, and negatively similar.

[0146] The following addresses the appendix Figures 4-9 The semantic code segment is 12 bits long for illustrative purposes. It is assumed that in the 12-bit semantic code obtained by the semantic coding model, the 1st, 2nd, and 3rd bits and the 7th, 8th, and 9th bits are known to be common parts of the semantic code.

[0147] With attachment Figure 4 For example, the voice code Sx obtained by the first user equipment 403 includes 12 coding bits. Its common part Scx consists of 1, 2, 3, 7, 8, and 9 bits. Its unique part Spx consists of 4, 5, 6, 10, 11, and 12 bits.

[0148] With attachment Figure 5 For example, the voice code Sy obtained by the second user equipment 404 includes 12 coding bits. Its common part Scy is bits 1, 2, 3, 7, 8, and 9. Its unique part Spy is bits 4, 5, 6, 10, 11, and 12.

[0149] The common part Scx is as follows: the 1st, 2nd and 3rd positions are [6, 3, 9] respectively; the 7th, 8th and 9th positions are [3, 7, 8] respectively.

[0150] The common part Scy is as follows: the 1st, 2nd and 3rd positions are [6, 3, 9] respectively; the 7th, 8th and 9th positions are [-3, -7, -8] respectively.

[0151] The common parts Scx and Sxc of the two codes satisfy the conditions of being the same, opposite, positively similar, or negatively similar. Specifically, each bit of the common part satisfies any one of the following conditions: same, opposite, positively similar, or negatively similar. (See attached...) Figure 4 and attached Figure 5 In specific examples, the 1st, 2nd, and 3rd positions belong to the same common part; the 7th, 8th, and 9th positions belong to opposite common parts.

[0152] Optionally, the base station transmits the common part and the individual part to the user equipment, including: the base station selects one from a first number of common information and transmits it to a first number of user equipments, or the base station selects two or more common information from the first number of common information, fuses them, and transmits them to the first number of user equipments, wherein the values ​​of the corresponding positions of the common parts participating in the fusion are adjusted to be the same or positively similar.

[0153] To adjust the value of the corresponding position of the common part to the same or positive similarity, it must be adjusted to a certain common standard, called standard common part.

[0154] For example, taking the semantic encoding common part Scx of the first user equipment as the standard common part, the semantic encoding common part Scy of the second user equipment is adjusted to the same or positive similarity with Scx. That is, taking the 1st-3rd bit [6, 3, 9] and the 7th-9th bit [3, 7, 8] of Scx as the standard, the 1st-3rd bit [6, 3, 9] of Scy is kept unchanged, and the 7th-9th bit [-3, -7, -8] of Scy is adjusted to [3, 7, 8] by bit inversion.

[0155] The above example is the opposite case of the common part. If the common part is negative similarity, the negative similar bit is also adjusted to positive similarity by bit inversion.

[0156] As shown in the accompanying Figure 6 , it is a structural schematic diagram of marking the semantic encoding Sy common part and the individual part. The accompanying Figure 5 uses 12-bit semantic encoding as an example. Each position has a letter, and the letter C indicates that the bit belongs to the common semantic encoding, and the letter P indicates that the bit belongs to the individual semantic encoding. For the common part, the letter C without a minus sign indicates that the bit is the same or positively similar to the standard common part, and the letter C with a minus sign indicates that the bit is opposite or negatively similar to the standard common part. Figure 6 The process of obtaining the common part from the semantic encoding is to extract all the common encoding (all bits with C or -C) according to the position attribute of the semantic encoding to form the common part. The process of obtaining the individual part from the semantic encoding is to extract all the individual encoding (all bits with P) according to the position attribute of the semantic encoding to form the individual part.

[0157] Optionally, a position extraction matrix is used to extract the individual part or the common part.

[0158] As shown in the accompanying

[0159] , the total length of the semantic encoding is 12, of which the position sequence numbers of the common part are 1, 2, 3, 7, 8, and 9. A position extraction matrix U of the common part is constructed as shown in the accompanying Figure 5 , which is a 6-row 12-column matrix. Each row is a one-hot vector. In the first row, the 1st bit is 1 and the rest are 0; in the second row, the 2nd bit is 1 and the rest are 0; in the third row, the 3rd bit is 1 and the rest are 0. In the fourth row, the 7th bit is -1 and the rest are 0; in the fifth row, the 8th bit is -1 and the rest are 0; in the sixth row, the 9th bit is 1 and the rest are 0. Figure 7

[0160] ​Appendix Figure 7 The commonality extraction matrix shown is just one method. The extraction of individual characteristics is done in a similar way.

[0161] Let U be the matrix representing the location extraction of the common components, and S be the semantic encoding vector. Then, the formula for calculating the common component vector C is: C = S * U'. The superscript comma indicates the transpose of the matrix. For example, matrix U is shown below. Figure 7 As shown.

[0162] Let T be the matrix representing the location extraction of the personality component, and S be the semantic encoding vector. Then, the formula for calculating the personality component vector P is: P = S * T'. The superscript comma indicates the transpose of the matrix. For example, matrix T is shown below. Figure 9 As shown.

[0163] Optionally, a location mask can be used to extract the individual characteristics or the common characteristics.

[0164] Optionally, the step of semantically encoding each piece of business information to obtain a common part and a unique part corresponding to each piece of business information includes:

[0165] Based on the semantic coding model, each piece of business information is converted into semantic code;

[0166] Based on the location mask representing common semantics, the common part is obtained from the semantic encoding;

[0167] The personality component is obtained from the semantic encoding based on the location encoding that represents the personality semantics.

[0168] As attached Figure 8 As shown, it is known Figure 5 The semantic encoding Sy shown has a total length of 12, with the common part numbered 1, 2, 3, 7, 8, and 9. Then, a... Figure 8 The positional mask vector shown has bits 1, 2, and 3 as 1, bits 7, 8, and 9 as -1, and the rest as 0. By using the positions with values ​​of 1 or -1 in this positional mask vector, two operations can be performed simultaneously: 1) obtaining the common semantic mask; 2) adjusting the opposite or negatively similar parts (bits 7, 8, and 9) of the common semantic encoding with the standard common part (…). Figure 4 The 1st, 2nd, 3rd, 7th, 8th, and 9th positions are the same or very similar.

[0169] The personalized semantic mask can be obtained by using the positions with a value of 0 in the positional mask vector. The values ​​1, -1, and 0 used as mask values ​​are merely examples; other methods can be used to achieve the same result, and will not be elaborated further.

[0170] The above is an example of the extraction process of the common part and the individual part of the semantic encoding. The combination process of the common part and the individual part has similar principles.

[0171] (IV) Before step S103, the base station informs the user equipment to perform multiple access transmission

[0172] Optionally, before the base station transmits the common part and the individual part to the user equipment, the base station informs the user equipment to perform transmission.

[0173] The content of the multiple access transmission by the base station can include the following content:

[0174] 1) Which user equipment performs this multiple access transmission;

[0175] 2) All user equipment need to receive the common part;

[0176] 3) Each device needs to receive its own individual part;

[0177] 4) Which time-frequency resources are used for the transmission of the common part and the individual part.

[0178] Optionally, for a single user equipment, the content it receives can only be the content it should know, for example: the first user equipment 403 only receives how it will receive the common part and the individual part related to itself. Sometimes the first user equipment 403 needs to decode all the individual information of the receiving end user equipment, and then take out its own. Sometimes the first user equipment 403 does not need to decode all of them, and it can decode its own.

[0179] Optionally, the base station can split the common part and the individual part before deciding to perform multiple access transmission. If it subsequently decides to perform multiple access transmission, the base station can transmit the common part and the individual part in different time-frequency resources according to the needs of multiple access transmission. Even if the base station does not subsequently perform multiple access transmission, the split common part and individual part can still be transmitted normally: the common part will be sent to each receiving end, and the individual part will also be sent to each receiving end.

[0180] (V) Referring to step S103, the base station transmits the common part and the individual part to the user equipment.

[0181] Optionally, the semantic encoding adopts source-channel joint encoding or source encoding.

[0182] Optionally, the transmission adopts any of the following: time division multiple access (TDMA), frequency division multiple access (FDMA), interleaved multiple access (IDMA), space division multiple access (SDMA), orthogonal frequency division multiple access (OFDMA), non-orthogonal multiple access (NOMA).

[0183] Take the scenario of 1 base station 2 user equipments as shown in FIG. 1. There are three transmission modes. Figure 1

[0184] 1) transmitting one common part Scx to the first user equipment 403 in separate time-frequency resources;

[0185] 2) transmitting one common part Scy to the second user equipment 404 in separate time-frequency resources;

[0186] 3) transmitting the common part Sc to the first user equipment 403 and the second user equipment 404 simultaneously in the same time-frequency resources. The Sc represents Scx or Scy or the weighted average of Scx and Scy. Because the base station can obtain all the common information, sometimes the weighted average of multiple common information can better describe the information of all common information.

[0187] If transmitting one common part Scx takes one resource, then transmitting one common part Scy only takes one time-frequency resource, and transmitting a weighted average of Scx and Scy also only takes one time-frequency resource.

[0188] Optionally, the base station transmits the common part and the individual part to the user equipment, including that the base station transmits a first number of the individual parts to a first number of the user equipment in different time-frequency resources.

[0189] Optionally, the first number of the user equipment respectively transmits the first number of the individual parts in different time-frequency resources of the same channel.

[0190] Take the scenario of 1 base station 2 user equipments as shown in FIG. 1. There are three transmission modes. Figure 1 Optionally, there is a transmission mode. First, transmit one individual part Spx to the first user equipment 403 and the second user equipment 404 (actually only useful for the first user equipment 403); then transmit one individual part Spy to the first user equipment 403 and the second user equipment 404 (actually only useful for the second user equipment 404); finally, transmit the common part Sc to the first user equipment 403 and the second user equipment 404. The above transmission order can be adjusted arbitrarily as long as the user equipment can receive in order. Therefore, time-frequency resources are finally saved.

[0191]

[0192] ​​Optionally, for three service information x, y, z, assume that their semantic encodings are Sx, Sy, Sz respectively. Then the common part is Sc.

[0193] For the semantic encoding Sx, its common part is Scx and its individual part is Spx.

[0194] For the semantic encoding Sy, its common part is Scy and its individual part is Spy.

[0195] For the semantic encoding Sz, its common part is Scz and its individual part is Spz.

[0196] The transmission of the common parts Scx, Scy, Scz has seven combinations: 1) only Scx is transmitted; 2) only Scy is transmitted; 3) only Scz is transmitted; 4) the weighted average of Scx and Scy is transmitted; 5) the weighted average of Scx and Scz is transmitted; 6) the weighted average of Scy and Scz is transmitted; 7) the weighted average of Scx, Scy and Scz is transmitted. Therefore, it is expressed as Sc.

[0197] Optionally, there is a transmission mode that transmits Spx, Spy, Spz and Sc in sequence. The above transmission order can be adjusted arbitrarily as long as the base station can receive them in sequence.

[0198] Optionally, for n service information x1, x2, …, xn, assume that their semantic encodings are S1, S2, …, Sn respectively. Then the common part is Sc. The individual part corresponding to S1 is Sp1, the individual part corresponding to S2 is Sp2, …, and the individual part corresponding to Sn is Spn. The transmission order of Sc, Sp1, Sp2, …, Spn can be arbitrarily specified.

[0199] Therefore, for the multiple access transmission of n service information, one common part and n-1 individual parts are finally sent through the channel. The transmission resources of n-1 common parts are finally saved.

[0200] (Six) Referring to step S104, the user equipment combines the common part and the individual part to obtain a second semantic encoding.

[0201] Optionally, the user equipment decodes the second semantic encoding into second service information, including: the user equipment uses a pre-configured semantic decoding model to decode the second semantic encoding into second service information; the user equipment generates the semantic decoding model locally, or the user equipment obtains the semantic decoding model from other equipment.

[0202] Optionally, the user equipment combines the common part and the individual part to obtain the second semantic code, including: the user equipment combines the common part and the individual part into the second semantic code based on the position mask representing the common semantic and the position mask representing the individual semantic.

[0203] According to the extraction process of the common part and the individual part of the semantic encoding part described above, the combination process of the common part and the individual part of the semantic decoding process is similar in principle.

[0204] Optionally, the semantic decoding of the common part and the individual part corresponding to each service information is performed to obtain the second service information corresponding to each service information, including:

[0205] The common part and the individual part are combined into the second semantic code based on the position mask representing the common semantic and the position mask representing the individual semantic.

[0206] The second semantic code is converted into the second service information corresponding to the service information based on the semantic decoding model.

[0207] In the application, the common part and the individual part are combined into the second semantic code based on the position mask representing the common semantic and the position mask representing the individual semantic. Figure 1 Taking the application scenario of 2-service information multiple access transmission as an example, it is assumed that the communication channel transmission is a common part Sc and two individual parts Spx and Spy.

[0208] The first user equipment can restore Sx through the common part Sc and the individual part Spx.

[0209] The second user equipment can restore Sy through the common part Sc and the individual part Spy.

[0210] For n-service information multiple access transmission, how to restore n semantic codes according to the common part and n individual parts is similar.

[0211] Optionally, the process of restoring the second semantic code by using the common part and the individual part includes:

[0212] 1) Initialize the second semantic code;

[0213] 2) According to the position of the common part, the common part is sequentially filled into the second semantic code; for the positions of opposite or negative correlation, the corresponding bits are inverted and filled in;

[0214] 3) According to the position of the individual part, the individual part is sequentially filled into the second semantic code.

[0215] As shown in the position of the common part and the individual part in the application, the specific process can be: Figure 6 As shown in the position of the common part and the individual part in the application, the specific process can be:

[0216] 1) Initialize the 12-bit second semantic code;

[0217] 2) According to the appendix Figure 6 The common part positions (C and -C), namely positions 1, 2, 3, 7, 8, and 9, are filled into the corresponding positions of the second semantic code in sequence; for the position of -C (positions 7, 8, and 9), an inversion operation is required;

[0218] 3) According to the appendix Figure 6 The positions (P) of the personality part, namely the 4th, 5th, 6th, 10th, 11th and 12th bits, are used to fill the corresponding positions of the second semantic code with the 6-bit personality part in sequence.

[0219] Optionally, a location mapping matrix can be used to combine the individual parts and the common parts into a second semantic code.

[0220] As attached Figure 7 The common parts shown are extracted using the mapping matrix U, as shown in the attached figure. Figure 9 The location extraction matrix T of the individual part shown can be used as a location mapping matrix.

[0221] For a given semantic code, let the individual part be vector P and the common part be vector C. Then the second semantic code S = C*U + P*T.

[0222] With attachment Figure 5 The commonalities in the example are numbered 1, 2, 3, 7, 8, and 9. Therefore, the commonality C is a 6-bit vector. Similarly, the individuality P is also a 6-bit vector.

[0223] Multiplying C by U maps the 6-bit C to the corresponding positions in the 12-bit matrix (i.e., bits 1, 2, 3, 7, 8, and 9). Multiplying P by T maps the 6-bit P to the corresponding positions in the 12-bit matrix (i.e., bits 4, 5, 6, 10, 11, and 12). Meanwhile, the positions in the U matrix corresponding to opposite or negative correlations are -1.

[0224] (vii) Referring to step S105, the user equipment encodes and decodes the second semantic information into second service information.

[0225] The user equipment decodes the second semantic code into second service information by: the user equipment using a pre-configured semantic decoding model to decode the second semantic code into second service information; the user equipment generating the semantic decoding model locally, or the user equipment obtaining the semantic decoding model from another device.

[0226] Optionally, the pre-configured semantic decoding model can be classified according to the type of service information (i.e. the type of encoded semantics). Taking the type of service information as an example, the semantic encoding model is an image semantic encoding model, and the corresponding semantic decoding model is also an image semantic decoding model.

[0227] Optionally, the semantic encoder is configured in pairs with a semantic decoder, which can decode the semantic code generated by the semantic encoder from the original service information to obtain the restored service information. The network devices at both ends (sending end and receiving end) of the channel for transmitting a service information should be configured with semantic encoding models and semantic decoding models corresponding in function (i.e. encoding and decoding is a reversible process).

[0228] Optionally, the semantic encoding model and the semantic decoding model can be common artificial intelligence models.

[0229] Specifically, the semantic decoding model in the embodiment can adopt a picture semantic decoding model based on a convolutional neural network algorithm; the semantic decoding model can also adopt a picture semantic decoding model based on an adversarial generative network technology or other deep learning network technology.

[0230] Optionally, the semantic decoding model is pre-configured in the receiving end respectively. For network devices that frequently transmit common types of service information (such as text, image, audio, video, etc.), it is necessary to pre-configure the semantic encoders and semantic decoders of these types.

[0231] Optionally, the semantic encoding model or the semantic decoding model is divided into several model components according to a division rule for storage or transmission; the division rule includes at least one of the following: horizontal layer division; vertical layer division; random division; division based on data volume; division based on function; division based on use.

[0232] Similarly, for different types of service information, including but not limited to image, voice, text, video, and point cloud, the semantic encoding model and the semantic decoding model of each type can be trained respectively, so that the model deployed on the communication device can normally process all service information that needs to be transmitted in communication.

[0233] (Eight) Information transmission efficiency

[0234] Finally, the efficiency of the service information transmission of the present embodiment is calculated. It is assumed that the semantic encoding model and the semantic decoding model are obtained based on the OpenImage image dataset. The size of the input image service information is 512*512*3, and the size of the semantic encoding obtained by the semantic encoding model is 4096. It is statistically obtained by experiments that the number of bits of the individual semantics is 1000 bits, and thus the number of bits of the common semantics is 4096-1000=3096 bits.

[0235] The efficiency (compression rate) of the semantic multiple-address transmission of the present embodiment on two images is calculated. The data of the two images is 512*512*3*2. The semantic encoding is 4096*2, wherein the common part is 3096 bits. The individual part of the two images is 1000 bits. Therefore, the data finally transmitted on the channel is 3096+1000*2=5096 bits. Therefore, the semantic encoding is compressed from 512*512*3*2 to 4096*2, and then multiple-address encoded to 5096, and the compression rate is 5096 / (512*512*3*2)=0.32%.

[0236] The efficiency (compression rate) of the semantic multiple-address transmission of the present embodiment on three images is calculated. The data of the three images is 512*512*3*3. The semantic encoding is 4096*3, wherein the common part is 3096 bits. The individual part of the three images is 1000 bits. Therefore, the data finally transmitted on the channel is 3096+1000*3=6096 bits. Therefore, the semantic encoding is compressed from 512*512*3*3 to 4096*3, and then multiple-address encoded to 5096, and the compression rate is 6096 / (512*512*3*3)=0.26%.

[0237] The efficiency (compression rate) of the semantic multiple-address transmission of the present embodiment on four images is calculated. The data of the four images is 512*512*3*4. The semantic encoding is 4096*4, wherein the common part is 3096 bits. The individual part of the four images is 1000 bits. Therefore, the data finally transmitted on the channel is 3096+1000*4=7096 bits. Therefore, the semantic encoding is compressed from 512*512*3*4 to 4096*4, and then multiple-address encoded to 7096, and the compression rate is 7096 / (512*512*3*4)=0.23%.

[0238] The efficiency (compression rate) of the embodiment when transmitting n images by semantic multiple access is calculated. The data of n images is 512*512*3*n. The semantic code is 4096*n, of which the common part is 3096 bits. The individual part of n images is 1000 bits. Therefore, the data transmitted on the channel is 3096+1000*n bits. Therefore, the semantic code is encoded from 512*512*3*n to 4096*n, and then multiple access encoded to 3096+1000*n, and the compression rate is (3096+1000*n) / (512*512*3*n).

[0239] Alternatively, considering that the semantic encoding model and the semantic decoding model may also need to be transmitted on the channel. At this time, the worst case is that all the sending ends and the receiving end need to receive the model from the channel: for example, the sending end does not have a semantic encoding model, only a semantic decoding model, and the receiving end does not have a semantic decoding model, only a semantic encoding model. Therefore, the model itself (in the form of model parameters, i.e., numerical features) needs to be propagated on the channel. Since the transmission of model parameters is the same, the semantic encoding model parameters can be multiplexed when sent to two sending ends, and the semantic decoding model parameters can also be multiplexed when sent to two receiving ends. Assuming that the size of the semantic encoding model is M1 and the size of the semantic decoding model is M2. Then in the worst case of model loss, the compression rate is (3096+1000*n+M1+M2) / (512*512*3*n)

[0240] Alternatively, when n is large, the number of users supported by multiple access transmission is limited. Assuming that the maximum number of users supported by multiple access transmission on the channel is 8, then n service information needs to be split into ceiling(n / 8) times of transmission, where ceiling is the ceiling function.

[0241] For example, if the sending end device now needs to transmit 1000 3M pictures, i.e., service information, to the receiving end device, the transmission data amount using the traditional communication method is 3M*1000, about 3G. Through the above technical solution, before the sending end device transmits the 1000 pictures, the transmission method provided by the application can be used to store the trained artificial intelligence model as a semantic encoding model in the sending end device before providing information service, and in the process of providing information service, first use the semantic encoding model stored in the sending end device to extract the semantics of each 3M picture and extract the semantic features, and after analyzing the importance of the semantic features, generate a description statement, which can compress the original 3M picture into a semantic information of about 30K, i.e., a semantic code. Then the size of the 1000 pictures after compression by the semantic encoding model is about 30K*1000=30M.

[0242] At this time, the semantic multiple access transmission is performed again on the 1000 semantic encodings, assuming that the multiple access transmission can process 8 semantic encodings each time, a total of 125 times.

[0243] For the 8 semantic encodings of each multiple access transmission, it is assumed that the common part accounts for 80% and the individual part accounts for 20% when multiple access transmission. The common encoding is 30K*0.8=24K, and the individual part that needs to be transmitted is 30K*0.2*8=48K, so the data amount of each multiple access transmission is 24K+48K=72K.

[0244] A total of 72K*125=9M is required for 125 transmissions. Then the total size of 1000 pictures after multiple access processing needs to be transmitted is about 9M.

[0245] After receiving the semantic information, the receiving end device can decode it using the pre-stored semantic decoding model, and reconstruct the image using the received semantic information, that is, the third service information. It should be noted that the semantic decoding model in the embodiment is a pre-stored artificial intelligence model in the receiving end device. As can be seen, the method provided by the present application can significantly reduce the communication transmission data amount, and greatly improve the transmission efficiency. It should be noted that the service information of the present application includes but is not limited to pictures, and can also be video, audio, text and other forms of information. The core of the present application is to use the model to complete the information extraction and information recovery of the service information, and to reduce the communication transmission data amount.

[0246] The application takes the user experience quality as a standard evaluation model to provide the ability of obtaining, storing, transmitting and processing services for information services, for example, when the service information is an image, the user experience quality evaluation standards for the image include but are not limited to the peak signal to noise ratio (PSNR) and the structural similarity (SSIM), the PSNR is an objective standard for evaluating images, and the SSIM is an index for measuring the similarity of two images, which is used to compare the original image (i.e. service information) with the image (i.e. third service information) obtained after reconstruction of the second service information; when the service information is audio, the subjective method or the objective method or the combination of the two can be used to evaluate the audio quality, the user experience quality evaluation standard can be a subjective evaluation standard, which is scored by human for audio, such as the mean opinion score (MOS), or the method of calculating the MCD (Mel Cepstral Distortion) as an objective evaluation; when the service information is video or other forms of information, there are corresponding user experience quality standards, and these evaluation standards are commonly used in the prior art, therefore, the following will not be repeated. Through the above technical scheme, the transceiver device compresses the picture pixels into semantic information for transmission based on the shared model, greatly reduces the amount of information represented, reduces the required transmission bandwidth, and improves the network efficiency.

[0247] The embodiment can form a technical solution in combination with the contents of other embodiments of the application, and the contents of the embodiment can also be combined into a technical solution in other embodiments of the application.

[0248] Embodiment two

[0249] The application also includes a model-based information service providing method, which is applied to a base station in a downlink scenario, and has the characteristics that:

[0250] The base station converts the service information to be transmitted into semantic encoding;

[0251] The base station splits the semantic encoding to obtain a common part and an individual part;

[0252] The base station transmits the common part and the individual part to the user equipment.

[0253] Optionally, the method further includes that the common part has the same value, opposite value, positive similarity or negative similarity with the common part of other semantic encoding at the corresponding position; and the individual part is the part after removing the common part from the semantic encoding.

[0254] Optionally, the method further comprises: the base station sending a transmission request of the transmission of the service information to the user equipment; and the base station receiving a confirmation message of the transmission sent by the user equipment.

[0255] Optionally, the base station transmitting the common part and the individual part to the user equipment comprises: the base station selecting one of the first number of common information to transmit to the first number of user equipment, or the base station selecting two or more of the first number of common information to transmit to the first number of user equipment after fusion, and the values of the corresponding positions of the common part participating in the fusion are adjusted to be the same or positively similar.

[0256] Optionally, the base station transmitting the common part and the individual part to the user equipment comprises: the base station transmitting the first number of individual parts to the first number of user equipment in different time-frequency resources.

[0257] Optionally, the base station converts the service information into semantic encoding comprises: the base station converts the service information to be transmitted into semantic encoding using a pre-configured semantic encoding model; the base station generates the semantic encoding model locally, or the base station obtains the semantic encoding model from other devices.

[0258] Optionally, the semantic encoding model is divided into several model components according to a division rule for storage or transmission; the division rule comprises at least one of the following: horizontal layer division; vertical layer division; random division; division based on data volume; division based on function; division based on use.

[0259] Optionally, the base station divides the semantic encoding to obtain the common part and the individual part comprises: obtaining the common part from the semantic encoding based on a position mask representing common semantics; and obtaining the individual part from the semantic encoding based on a position mask representing individual semantics.

[0260] Optionally, the service information comprises at least one of the following: text, audio, image, video, point cloud, one-dimensional waveform data, and radar signal.

[0261] Optionally, the semantic encoding adopts source channel joint encoding or source encoding.

[0262] Optionally, the transmission adopts any one of the following: time division multiple access (TDMA), frequency division multiple access (FDMA), interleaved multiple access (IDMA), space division multiple access (SDMA), orthogonal frequency division multiple access (OFDMA), and non-orthogonal multiple access (NOMA).

[0263] Embodiment three

[0264] The application also discloses a model-based information service providing method applied to a user equipment in a downlink scene.

[0265] The user equipment receives a transmission request of transmission service information from the base station.

[0266] The user equipment receives the common part and the individual part from the base station.

[0267] The user equipment combines the common part and the individual part to obtain a second semantic code.

[0268] The user equipment decodes the second semantic code into second service information.

[0269] Optionally, the method further comprises that the user equipment receives a transmission request of service information from the base station; and based on the transmission request, the user equipment agrees to the transmission of the base station.

[0270] Optionally, the method comprises that the common part and the individual part are obtained by the base station dividing a semantic code; and the semantic code is obtained by the base station encoding service information to be transmitted.

[0271] Optionally, the user equipment decoding the second semantic code into second service information comprises that the user equipment uses a pre-configured semantic decoding model to decode the second semantic code into second service information; the user equipment generates the semantic decoding model locally, or the user equipment obtains the semantic decoding model from other equipment.

[0272] Optionally, the semantic decoding model is divided into a plurality of model components according to a division rule for storage or transmission; the division rule comprises at least one of the following: horizontal layer division; vertical layer division; random division; division based on data volume; division based on function; division based on use.

[0273] Optionally, the common part has the same value, opposite value, positive similarity or negative similarity with a common part of other semantic codes at a corresponding position; and the individual part is a part of the semantic code after removing the common part.

[0274] Optionally, the user equipment receiving the common part and the individual part from the base station comprises:

[0275] The base station receives one copy of the common part and the first number of individual parts in different time-frequency resources.

[0276] Optionally, the service information comprises at least one of the following: text, audio, image, video, point cloud, one-dimensional waveform data and radar signal.

[0277] Optionally, the semantic encoding employs source-channel joint encoding or source encoding.

[0278] Optionally, the transmission employs any of the following: time division multiple access (TDMA), frequency division multiple access (FDMA), interlaced multiple access (IDMA), space division multiple access (SDMA), orthogonal frequency division multiple access (OFDMA), non-orthogonal multiple access (NOMA).

[0279] Embodiment four

[0280] As shown in the accompanying drawings Figure 3 The application also includes a model-based information service providing system applied to a base station in a downlink scenario, characterized in that:

[0281] An encoding module 101, the base station converts the service information into semantic encoding;

[0282] A code division module 102, the base station divides the semantic encoding to obtain a common part and an individual part;

[0283] A transmission module 103, the base station transmits the common part and the individual part to the user equipment;

[0284] A code combination module 104, the user equipment combines the common part and the individual part to obtain second semantic encoding;

[0285] A decoding module 105, the user equipment decodes the second semantic encoding into second service information.

[0286] Optionally, the system corresponds one-to-one to the aforementioned method applied to the base station in the downlink scenario. The system includes modules corresponding to the sub-steps of the aforementioned method.

[0287] Embodiment five

[0288] The application also includes a model-based information service providing system applied to a base station in a downlink scenario, characterized in that:

[0289] An encoding module, the base station converts the service information to be transmitted into semantic encoding;

[0290] A code division module, the base station divides the semantic encoding to obtain a common part and an individual part;

[0291] A transmission module, the base station transmits the common part and the individual part to the user equipment.

[0292] Optionally, the system corresponds to the aforementioned method applied to the base station in the downlink scenario. The system comprises modules corresponding to the sub-steps of the aforementioned method.

[0293] Embodiment six

[0294] The application further comprises a model-based information service providing system, a user equipment applied to the downlink scenario, characterized in that:

[0295] a grouping module, the user equipment receives a transmission request of transmission service information from the base station;

[0296] a receiving module, the user equipment receives the common part and the individual part from the base station;

[0297] a code grouping module, the user equipment groups the common part and the individual part to obtain a second semantic code;

[0298] a decoding module, the user equipment decodes the second semantic code into a second service information.

[0299] Optionally, the system corresponds to the aforementioned method applied to the user equipment in the downlink scenario. The system comprises modules corresponding to the sub-steps of the aforementioned method.

[0300] Embodiment seven

[0301] According to the embodiments of the application, the application further provides an electronic device, a readable storage medium and a computer program product.

[0302] The application further comprises an electronic device comprising a processor, a communication interface, a memory and a communication bus, wherein the communication bus is used to complete the mutual communication among the processor, the communication interface and the memory.

[0303] The memory is used to store a computer program.

[0304] The processor is used to execute the computer program stored in the memory to realize the information service providing method in any one of the above technical solutions.

[0305] The application further comprises a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program is executed by a processor to realize the information service providing method in any one of the above technical solutions.

[0306] The application further comprises a computer program product, when the computer program product is executed on a computer, the computer program product makes the computer execute the information service providing method in any one of the above technical solutions.

[0307] Figure 10 A schematic block diagram of an example electronic device 500 that can be used to implement embodiments of the present application is shown. The electronic device is intended to represent various forms of digital computers, such as laptops, desktops, tablets, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown here, their connections and relationships, and their functions, are meant to be examples only, and are not intended to limit the implementations of the present application described and / or claimed in this document.

[0308] As shown in Figure 10 , the device 500 includes a computing unit 501 that can perform various appropriate actions and processes in accordance with a computer program stored in a read-only memory (ROM) 502 or a computer program loaded into a random access memory (RAM) 503 from a storage unit 508. Various programs and data required for the operation of the device 500 can also be stored in the RAM 503. The computing unit 501, the ROM 502, and the RAM 803 are connected to each other through a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0309] Various components in the device 500 are connected to the I / O interface 505, including an input unit 506, such as a keyboard, a mouse, etc.; an output unit 507, such as various types of displays, speakers, etc.; a storage unit 508, such as a magnetic disk, an optical disk, etc.; and a communication unit 509, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 509 allows the device 500 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0310] The computing unit 501 can be various general and / or special purpose processing components with processing and computing capabilities. Some examples of the computing unit 501 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 501 performs various methods and processes described above, such as the testing method. For example, in some embodiments, the testing method can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 508. In some embodiments, part or all of the computer program can be loaded and / or installed onto the apparatus 500 via the ROM 502 and / or the communication unit 509. When the computer program is loaded onto the RAM 503 and executed by the computing unit 501, one or more steps of the testing method described above can be performed. Alternatively, in other embodiments, the computing unit 501 can be configured to perform the testing method by any other appropriate means, such as by means of firmware.

[0311] Various implementations of the systems and techniques described above can be realized in digital electronic circuitry, integrated circuitry, a field programmable gate array (FPGA), an application specific integrated circuit (ASIC), a system on a chip (SOC), a programmable logic device (PLD), a computer hardware, firmware, software, and / or combinations thereof. These various implementations can include implementation in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be special or general purpose, coupled to receive data and instructions from, and to transmit data and instructions to, a storage system, at least one input device, and at least one output device.

[0312] Program code for carrying out methods of the present application can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the program code, when executed by the processor or controller, produces a means for implementing the functions / acts specified in the flowcharts and / or block diagrams. The program code can be executed entirely on a machine, partially on a machine, partially on a machine and partially on a remote machine or entirely on a remote machine or server.

[0313] In the context of the present application, a machine-readable medium can be a tangible medium that can contain or store program for use by or in connection with an instruction execution system, apparatus, or device. Machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable medium can include, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of machine-readable storage medium would include one or more lines of electrical wire, portable computer diskette, hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or Flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination of the foregoing.

[0314] To provide for interaction with a user, the systems and techniques described here can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can be used to provide for interaction with a user as well; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0315] The systems and techniques described here can be implemented in a computing system that includes a back end component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front end component (e.g., a user computer having a graphical user interface or a Web browser through which a user can interact with an implementation of the systems and techniques described here), or any combination of such back end, middleware, or front end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0316] The computer system can include clients and servers. A client and server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server can be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0317] It should be understood that the various forms of flow shown above can be used to reorder, add, or remove steps. For example, the steps described in the present disclosure can be performed in parallel, in series, or in a different order, without limitation herein, so long as the desired results of the technology disclosed herein are achieved.

[0318] The specific embodiments discussed herein should not be construed as limiting the scope of the disclosure. Those skilled in the art will recognize that modifications, combinations, sub-combinations, and alternatives can be made to the specific embodiments without departing from the spirit and scope of the disclosure. Any further modifications, equivalents, and / or alterations of the specific embodiments are intended to fall within the scope of the present disclosure.

Claims

1. A model-based information service providing method applied to a downlink scenario in which a base station transmits a same type of service information to a first number of user equipments, characterized in that: the base station converts the service information into a semantic code; the base station divides the semantic code to obtain a common part and a unique part; the base station transmits the common part and the unique part to the user equipments; the user equipments combine the common part and the unique part to obtain a second semantic code; and the user equipments decode the second semantic code into a second service information. the common part has the same, opposite, positive similar or negative similar values as the common part of other semantic codes at corresponding positions; and the unique part is the part of the semantic code after removing the common part. Before the base station transmits the common part and the unique part to the user equipments, the base station notifies the user equipments of the transmission. The base station transmits the common part and the unique part to the user equipments by: transmitting a first number of the unique parts to a first number of the user equipments at different time-frequency resources; or selecting one of a first number of the common information to transmit to a first number of the user equipments, or selecting two or more of the common information to transmit to a first number of the user equipments after fusion, and adjusting the values of the common parts at corresponding positions to be the same or positive similar. The base station converts the service information into a semantic code by: using a pre-configured semantic code model to convert the service information to be transmitted into a semantic code; or generating the semantic code model locally, or obtaining the semantic code model from other devices. The user equipments decode the second semantic code into a second service information by: using a pre-configured semantic decoding model to decode the second semantic code into a second service information; or generating the semantic decoding model locally, or obtaining the semantic decoding model from other devices. The semantic code model or the semantic decoding model is divided into a plurality of model components according to a division rule for storage or transmission.

2. The method of claim 1, wherein, The division rule includes at least one of the following: horizontal layer division; vertical layer division; random division; division based on data volume; division based on function; and division based on use.

3. The method of claim 1, wherein, The base station divides the semantic code to obtain a common part and a unique part by: obtaining the common part from the semantic code based on a position mask representing common semantics; and obtaining the unique part from the semantic code based on a position mask representing unique semantics. The user equipments combine the common part and the unique part to obtain a second semantic code by:

4. The method of claim 1, wherein, ​ ​ 5. The method of claim 1, wherein, ​ ​ ​ 6. The method of claim 1, wherein, ​ ​ ​ 7. The method of claim 6, wherein, ​ ​ 8. The method of claim 1, wherein, ​ ​ ​ 9. The method of claim 8, wherein, ​ The user equipment combines the common part and the individual part into a second semantic code based on the position mask representing common semantics and the position mask representing individual semantics.

10. The method of claim 1, wherein, The service information includes at least one of the following: text, audio, image, video, point cloud, one-dimensional waveform data, and radar signal.

11. The method of claim 1, wherein, The semantic code adopts source channel joint coding or source coding.

12. The method of claim 1, wherein, The transmission adopts any one of the following: time division multiple access (TDMA), frequency division multiple access (FDMA), interlaced multiple access (IDMA), space division multiple access (SDMA), orthogonal frequency division multiple access (OFDMA), and non-orthogonal multiple access (NOMA).

13. A model-based information service providing method applied to a base station in a downlink scenario, characterized in that: The base station converts the service information to be transmitted into a semantic code; The base station splits the semantic code to obtain a common part and an individual part; The base station transmits the common part and the individual part to the user equipment.

14. The method of claim 13, wherein, Further comprising: The common part has the same, opposite, positive similarity, or negative similarity value as the common part of other semantic codes at the corresponding position; The individual part is the part of the semantic code after removing the common part.

15. The method of claim 13, wherein, Further comprising: The base station sends a transmission request for transmitting service information to the user equipment; The base station receives a confirmation message for transmission sent by the user equipment.

16. The method of claim 13, wherein, The base station transmits the common part and the individual part to the user equipment, including: The base station selects one of the first number of common information to transmit to the first number of user equipment, or The base station selects two or more common information from the first number of common information to fuse and transmit to the first number of user equipment, and the values of the corresponding positions of the common part participating in the fusion are adjusted to be the same or positively similar.

17. The method of claim 13, wherein, The base station transmits the common part and the individual part to the user equipment, including: The base station transmits the first number of individual parts to the first number of user equipment in different time-frequency resources.

18. The method of claim 13, wherein, The base station divides the semantic code to obtain a common part and an individual part, including: Based on the position mask representing common semantics, the common part is obtained from the semantic code; Based on the position mask representing individual semantics, the individual part is obtained from the semantic code.

19. A model-based information service providing method applied to a user equipment in a downlink scenario, characterized in that: The user equipment receives a transmission request for transmitting service information sent by the base station; The user equipment receives the common part and the individual part sent by the base station; The user equipment combines the common part and the individual part to obtain a second semantic code; The user equipment decodes the second semantic code into a second service information.

20. The method of claim 19, wherein, Further comprising: The base station receives a transmission request for transmitting service information sent by the second number of user equipment; Based on the second number of transmission requests, the base station notifies the first number of user equipment to transmit.

21. The method of claim 19, wherein, Including: The common part and the individual part are obtained by the user equipment dividing the semantic code; The semantic code is obtained by encoding service information to be transmitted by the user equipment.

22. The method of claim 21, wherein, The common part has the same value, opposite value, positive similarity or negative similarity with the common part of other semantic codes at the corresponding position; and the individual part is the part after the common part is removed from the semantic code.

23. The method of claim 19, wherein, The user equipment receives the common part and the individual part transmitted by the base station, including: The base station receives one copy of the common part and the first number of individual parts in different time-frequency resources.

24. A model-based information service providing system applied to an uplink scenario of a first number of user equipment to a base station, characterized in that: An encoding module, the base station converts service information to be transmitted into a semantic code; A code splitting module, the base station splits the semantic code to obtain a common part and an individual part; A transmission module, the base station transmits the common part and the individual part to the user equipment; A code combining module, the user equipment combines the common part and the individual part to obtain a second semantic code; A decoding module, the user equipment decodes the second semantic code into second service information.

25. A model-based information service providing system applied to a base station in a downlink scenario, characterized in that: An encoding module, the base station converts service information to be transmitted into a semantic code; A code splitting module, the base station splits the semantic code to obtain a common part and an individual part; A transmission module, the base station transmits the common part and the individual part to the user equipment.

26. A model-based information service providing system applied to a user equipment in a downlink scenario, characterized in that: An encoding module, the user equipment receives a transmission request of transmission service information from a base station; A receiving module, the user equipment receives the common part and the individual part transmitted by the base station; A code combining module, the user equipment combines the common part and the individual part to obtain a second semantic code; A decoding module, the user equipment decodes the second semantic code into second service information.

27. An electronic device, comprising: A processor, a communication interface, a memory and a communication bus, wherein the communication bus is used to complete mutual communication among the processor, the communication interface and the memory; The memory is used to store a computer program; The processor is used to execute the computer program stored in the memory to implement the information service providing method in any one of claims 1-12, 13-18 and 19-23.

28. A computer-readable storage medium, characterized in that, The computer program is stored in the computer readable storage medium, and when the computer program is executed by the processor, the information service providing method in any one of claims 1-12, 13-18 and 19-23 is implemented.

29. A computer program product, characterised in that, When the computer program product is executed on the computer, the computer is caused to execute the information service providing method in any one of claims 1-12, 13-18 and 19-23.

Images