Signal sending method and device based on semantic features, and signal processing method and device

Abstract

This disclosure provides a signal transmission method, signal processing method, and apparatus based on semantic features, relating to the field of communication technology, specifically semantic communication technology. The transmitting end acquires information to be transmitted, extracts a first semantic feature, and converts the first semantic feature into first bit data including first important bit data and first non-important bit data according to a conversion rule. The first bit data is mapped to corresponding constellation points, and a transmission signal is obtained based on the constellation points and transmitted to the receiving end. The constellation points corresponding to the first non-important bit data are centered on the constellation points corresponding to the first important bit data. The receiving end performs the reverse operation of the signal transmission method based on semantic features, acquires the transmission signal sent by the transmitting end, and processes the transmission signal to obtain recovered information. This allows communication resources to be preferentially allocated to the first important bit data, effectively improving the fidelity of semantic information recovery and meeting transmission requirements.

Description

Technical Field

[0001] This disclosure relates to the field of communication technology, specifically to the field of semantic communication technology, and in particular to signal transmission methods, signal processing methods, apparatus, and electronic devices based on semantic features. Background Technology

[0002] In 6th Generation Mobile Communication Technology (6G), semantic communication is a core technology. It upgrades the communication paradigm from the traditional "bit-accurate transmission" to "understanding and conveying the meaning and intent of information." By deeply integrating artificial intelligence technology, semantic communication systems can extract more compact semantic features needed to complete tasks from raw data (such as images and text) for transmission, thereby significantly improving communication efficiency in bandwidth-constrained scenarios.

[0003] Currently, under poor channel conditions, errors in some critical bit data can directly lead to severe distortion of semantic information, thereby significantly reducing the overall performance of the communication system. Summary of the Invention

[0004] This disclosure provides a signal transmission method, signal processing method, apparatus, and electronic device based on semantic features.

[0005] According to the first aspect, a signal transmission method based on semantic features is provided, the method being applied at a transmitting end, the method comprising: Obtain the information to be sent, and extract the first semantic feature of the information to be sent; According to the conversion rule, the first semantic feature is converted into first bit data including first important bit data and first unimportant bit data. The conversion rule is used to indicate the way the first semantic feature is converted into the first bit data. The first important bit data and the first unimportant bit data are determined based on the semantic importance of the first semantic feature. The first bit data is mapped to the corresponding constellation point, and the transmission signal is obtained based on the constellation point. The constellation point corresponding to the first non-important bit data is centered on the constellation point corresponding to the first important bit data. The transmission signal is sent to the receiving end.

[0006] According to a second aspect, a signal processing method is provided, the method being applied at a receiving end, the method comprising: Acquire a transmission signal, wherein the transmission signal is generated and transmitted using the signal transmission method based on semantic features as described in any one of the first aspects; Processing the transmitted signal yields second bit data, which includes second important bit data and second unimportant bit data; According to the conversion rule, the second bit data is converted into a second semantic feature. The conversion rule is used to indicate the method of converting the first semantic feature into the first bit data during the generation of the transmission signal. Based on the second semantic feature, the recovery information is obtained.

[0007] According to a third aspect, a signal transmission device based on semantic features is provided, the device being applied at a transmitting end, the device comprising: An acquisition unit is used to acquire information to be sent and extract the first semantic feature of the information to be sent; A conversion unit is configured to convert the first semantic feature into first bit data including first important bit data and first unimportant bit data according to a conversion rule, wherein the conversion rule is used to indicate the method of converting the first semantic feature into the first bit data, and the first important bit data and the first unimportant bit data are determined based on the semantic importance of the first semantic feature. A modulation unit is used to map the first bit data to a corresponding constellation point and obtain a transmission signal based on the constellation point, wherein the constellation point corresponding to the first non-important bit data is centered on the constellation point corresponding to the first important bit data. The transmitting unit is used to transmit the transmitting signal to the receiving end.

[0008] According to a fourth aspect, a signal processing apparatus is provided, characterized in that the apparatus is applied at a receiving end, the apparatus comprising: An acquisition unit is used to acquire a transmission signal, wherein the transmission signal is generated and transmitted using the signal transmission method based on semantic features as described in any one of the first aspects; The demodulation unit is used to process the transmitted signal to obtain second bit data including second important bit data and second non-important bit data; A conversion unit is used to convert the second bit data into a second semantic feature according to a conversion rule, wherein the conversion rule is used to indicate the method of converting the first semantic feature into the first bit data during the generation of the transmission signal; The generation unit is used to obtain the recovery information based on the second semantic feature.

[0009] According to the fifth aspect, an electronic device is provided, comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enable the at least one processor to perform the method described in either the first aspect or the second aspect.

[0010] According to a sixth aspect, a non-transitory computer-readable storage medium is provided that stores computer instructions for causing the computer to perform the method described in any implementation of the first aspect or the method described in any implementation of the second aspect.

[0011] According to a seventh aspect, a computer program product is provided, comprising a computer program that, when executed by a processor, implements the method described in any implementation of the first aspect, or the method described in any implementation of the second aspect.

[0012] This disclosure provides a signal transmission method, signal processing method, apparatus, and related equipment based on semantic features. In the semantic feature-based signal transmission method, the transmitting end acquires information to be transmitted and extracts a first semantic feature from the information. According to a conversion rule, the first semantic feature is converted into first bit data, and the first bit data includes first important bit data and first unimportant bit data. The first bit data is mapped to corresponding constellation points, and a transmission signal is obtained based on the constellation points and transmitted to the receiving end. The first important bit data and the first unimportant bit data are determined based on the semantic importance of the first semantic feature, and the constellation point corresponding to the first unimportant bit data is centered on the constellation point corresponding to the first important bit data. This allows for the determination of the first important bit data based on semantic importance, enabling communication resources to be preferentially allocated to the first important bit data. This prioritizes the allocation of communication resources to high-value semantic information, more effectively improving the quality of the final task completion and the fidelity of semantic information recovery, thus meeting transmission requirements. Correspondingly, in the signal processing method, the reverse operation of the semantic feature-based signal transmission method is adopted. The transmitted signal is acquired, and processed to obtain second bit data including second important bit data and second non-important bit data. According to the conversion rule, the second bit data is converted into a second semantic feature. Finally, based on the second semantic feature, the recovered information is obtained. This allows for the reverse processing of the transmitted signal to obtain the recovered information. Furthermore, the semantic feature-based signal transmission and signal processing methods provided in this disclosure maintain high compatibility with traditional modulation methods, while the introduced additional processing complexity is significantly lower than the performance gain, exhibiting low complexity and high compatibility.

[0013] It should be understood that the description in this section is not intended to identify key or important features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0014] The accompanying drawings are provided to better understand this solution and do not constitute a limitation of this disclosure. Wherein: Figure 1 This is a schematic diagram illustrating an exemplary application scenario provided by an embodiment of the present disclosure; Figure 2 A flowchart illustrating a signal transmission method based on semantic features provided in this disclosure embodiment; Figure 3 A schematic diagram illustrating the information transmission process provided in this embodiment of the disclosure; Figure 4 A schematic diagram illustrating a two-dimensional assignment vector mapped to distance, provided as an embodiment of this disclosure; Figure 5 A schematic diagram of a constellation distribution provided in an embodiment of this disclosure; Figure 6 A schematic flowchart of a signal processing method provided in an embodiment of this disclosure; Figure 7 A schematic diagram of a signal transmission device based on semantic features provided in this disclosure embodiment; Figure 8 This is a schematic diagram of the structure of a signal processing device provided in an embodiment of the present disclosure; Figure 9 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this disclosure. Detailed Implementation

[0015] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0016] To facilitate understanding and explanation of the technical solutions provided in the embodiments of this disclosure, the background technology of this disclosure will be described first.

[0017] Semantic communication technology can extract semantic features from raw data for transmission. However, different components of semantic features are not equally important to the final task, i.e., there is a difference in "semantic importance." This difference directly leads to bit inequivalence after quantizing continuous semantic features into a bit stream—errors in some bits can have a fatal impact on semantic understanding and task performance, while errors in other bits have a smaller impact.

[0018] Currently, mainstream wireless communication systems generally employ classic modulation schemes such as M-ary Quadrature Amplitude Modulation (MQAM) at the physical layer. These schemes treat all bits as equally important during design. In other words, "important bits" carrying crucial semantic information face the same probability of error as ordinary bits. Under poor channel conditions, errors in important bits can directly lead to severe distortion of semantic information, thereby significantly reducing the overall performance of the communication system.

[0019] Based on this, this disclosure provides a signal transmission method, signal processing method, apparatus, and related equipment based on semantic features. In the signal transmission method based on semantic features, the transmitting end acquires information to be transmitted and extracts a first semantic feature from the information. According to a conversion rule, the first semantic feature is converted into first bit data, and the first bit data includes first important bit data and first unimportant bit data. The first bit data is mapped to corresponding constellation points, and a transmission signal is obtained based on the constellation points and transmitted to the receiving end. The first important bit data and the first unimportant bit data are determined based on the semantic importance of the first semantic feature, and the constellation point corresponding to the first unimportant bit data is centered on the constellation point corresponding to the first important bit data. This allows for the determination of the first important bit data based on semantic importance, enabling communication resources to be preferentially allocated to the first important bit data. This prioritizes the allocation of communication resources to high-value semantic information, more effectively improving the quality of the final task completion and the fidelity of semantic information recovery, thus meeting transmission requirements.

[0020] Correspondingly, in the signal processing method, the reverse operation of the semantic feature-based signal transmission method is adopted. The transmitted signal is acquired, and processed to obtain second bit data including second important bit data and second non-important bit data. According to the conversion rule, the second bit data is converted into a second semantic feature. Finally, based on the second semantic feature, the recovered information is obtained. This allows for the reverse processing of the transmitted signal to obtain the recovered information. Furthermore, the semantic feature-based signal transmission and signal processing methods provided in this disclosure maintain high compatibility with traditional modulation methods, while the introduced additional processing complexity is significantly lower than the performance gain, exhibiting low complexity and high compatibility.

[0021] To facilitate understanding of the technical solutions provided in the embodiments of this disclosure, the following description is in conjunction with the accompanying drawings. Figure 1 The application scenarios of the signal transmission method and signal processing method based on semantic features provided in the embodiments of this disclosure are described.

[0022] Figure 1 This diagram illustrates exemplary application scenarios in which the various methods and apparatuses described herein can be implemented according to embodiments of this disclosure. A transmitting end 101 acquires information to be transmitted and extracts a first semantic feature from the information. The transmitting end 101 converts the first semantic feature into first bit data, including first important bit data and first unimportant bit data, according to a conversion rule. The first important bit data and the first unimportant bit data are determined based on the semantic importance of the first semantic feature. The transmitting end 101 maps the first bit data to corresponding constellation points. The constellation point corresponding to the first unimportant bit data is centered on the constellation point corresponding to the first important bit data. The transmitting end 101 obtains a transmission signal based on the constellation points and transmits the transmission signal to the receiving end.

[0023] Receiver 102 acquires the transmission signal sent by transmitter 101. Receiver 102 processes the transmission signal to obtain second bit data, which includes second important bit data and second unimportant bit data. Receiver 102 converts the second bit data into a second semantic feature according to a conversion rule. Receiver 102 obtains the recovered information based on the second semantic feature.

[0024] By employing a bit allocation strategy based on semantic importance awareness, targeted protection is applied to important bits during modulation and demodulation. This allows communication resources to be prioritized for high-value semantic information, thereby more effectively improving the final task completion quality and semantic information recovery fidelity with the same resource consumption. Furthermore, while maintaining high compatibility with traditional modulation methods, the introduced additional processing complexity is significantly lower than the performance gain, exhibiting low complexity and high compatibility. It can be efficiently integrated into communication technology frameworks with minimal modification costs, demonstrating strong practicality and deployability.

[0025] In terms of application scenarios, the signal transmission method and signal processing method based on semantic features provided in this disclosure can be widely used in communication systems including the transmitter 101 and the receiver 102. This disclosure does not limit the application scenarios.

[0026] It is understood that the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0027] The various embodiments disclosed herein are not isolated entities, but rather complementary. The solutions involved in each embodiment are intended to exemplify the overall solution from a specific aspect or dimension. The content of each embodiment in this disclosure can be integrated with other embodiments to constitute one of the many feasible solutions disclosed herein. This design makes this disclosure an organic whole, providing comprehensive and systematic technical support for achieving the inventive objectives, meeting the needs of different application scenarios, and ensuring the integrity and effectiveness of the technical solutions of this disclosure in multiple dimensions and levels.

[0028] To facilitate understanding of the technical solutions provided in the embodiments of this disclosure, the signal transmission method and signal processing method based on semantic features provided in the embodiments of this disclosure will be described below with reference to the accompanying drawings.

[0029] The implementation of the signal transmission method based on semantic features will be explained below.

[0030] See Figure 2 As shown, this figure is a flowchart illustrating a signal transmission method based on semantic features provided in an embodiment of this disclosure. Figure 2 As shown, the signal transmission method based on semantic features may include steps S201-S204. The signal transmission method based on semantic features provided in this disclosure can be applied to a transmitting end or can also be used as a transmitting device.

[0031] Step S201: Obtain the information to be sent and extract the first semantic feature of the information to be sent.

[0032] The information to be sent is the information that the sending end needs to transmit to the receiving end. The information to be sent can be generated by the sending end itself, or it can be information received by the sending end from other devices.

[0033] This disclosure does not limit the type or content of the information to be sent. The information to be sent may be represented in one or more forms, including text, images, audio, and video.

[0034] After obtaining the information to be sent, the first semantic feature is extracted from it. As an example, a semantic feature extraction model can be used to process the information to be sent and obtain the first semantic feature. The semantic feature extraction model can be a pre-trained artificial intelligence model, trained using sample information and the corresponding semantic features. For example, Figure 3This is a schematic diagram illustrating the information transmission process provided in this embodiment of the disclosure. The semantic feature extraction model is a semantic encoder. The sending end can extract the first semantic feature from the information to be sent based on the semantic encoder. For different types of information to be sent, corresponding semantic feature extraction models can be used. For example, for text, a semantic feature extraction model capable of processing text can be used to extract the first semantic feature. For images, a semantic feature extraction model capable of processing images can be used to extract the first semantic feature. As an example, the semantic feature extraction model can be a ResNet-18 semantic decoder. For audio, a semantic feature extraction model capable of processing audio can be used to extract the first semantic feature. In addition, this disclosure does not limit the dimension of the first semantic feature and can be flexibly set according to needs. For example, the first semantic feature may include a 512-dimensional feature.

[0035] Step S202: According to the conversion rules, the first semantic feature is converted into first bit data including first important bit data and first unimportant bit data.

[0036] The conversion rules indicate how to convert a first semantic feature into first bit data. According to the conversion rules, the first semantic feature can be converted into first bit data. The first bit data includes first important bit data and first unimportant bit data. First important bit data is the bit data obtained by converting semantic features with higher semantic importance. First unimportant bit data is the bit data obtained by converting semantic features with lower semantic importance.

[0037] As an example, see Figure 3 As shown, the first semantic feature can be quantized first to obtain the first bit data.

[0038] This disclosure does not limit the specific content of the conversion rules or the corresponding method for obtaining the first bit data. As an example, two possible implementation methods are described below for converting the first semantic feature into first bit data including first important bit data and first unimportant bit data according to the conversion rules.

[0039] The first type: the conversion rules include preset quantification standards.

[0040] Quantization standards are used to indicate how the first bit of data is converted, how the semantic importance of the first bit of data is determined, and the proportion of the first important bit of data to the total first bit of data.

[0041] According to the quantization standard, the first semantic feature is converted into the first bit data. This allows features with high semantic importance to be converted into first important bit data, and determines the first important bit data and the first unimportant bit data included in the first bit data.

[0042] As an example, following the IEEE 754 unified quantization standard, the first semantic feature is converted into the first bit data. IEEE 754 includes a sign bit (S), an exponent bit (E), and a mantissa bit (M). The exponent bit (E) has the highest semantic importance, followed by the sign bit (S), and the mantissa bit (M) has the lowest semantic importance. The semantic importance of different types of data bits is pre-set and does not need to be recalculated. After obtaining the bit data corresponding to the feature in each dimension, they are sorted in the order of exponent bit (E), sign bit (S), and mantissa bit (M). The first important bit data is determined according to the pre-defined proportion of the first important bit data to the total first important bit data. For example, the proportion of the first important bit data to the total first important bit data can be 25% or 30%. Thus, the exponent bit, or the exponent bit and all or part of the sign bit, from the bit data converted from the feature in each dimension can be used as the first important bit data, and the remaining bit data can be used as the first unimportant bit data.

[0043] The second type: the transformation rules include the way to measure semantic importance and the bit allocation model.

[0044] Specifically, the semantic importance of each dimension of the features included in the first semantic feature is first determined. As an example, a semantic analysis model or a knowledge base can be used to determine the semantic importance of each dimension of the features included in the first semantic feature. The semantic analysis model can be an artificial intelligence model, pre-trained using training features and corresponding labels (used to indicate semantic importance). The knowledge base can pre-store the semantic importance between tasks and features, as well as between features, so that the semantic importance of features can be retrieved through queries.

[0045] Using a semantic analysis model, we can determine the semantic importance of the first semantic feature, which is consistent with the semantics to be expressed by the information to be sent.

[0046] As an example, the semantic importance of a feature in each dimension can be represented by an importance vector. The semantic importance of a feature in each dimension is determined based on the relevance between semantics and the task, as well as the relevance between semantics. For example, the semantic importance of a feature in each dimension (e.g., the importance vector) is the product of the relevance vector between semantics and the task and the relevance vector between semantics.

[0047] After determining the semantic importance of each feature, a corresponding number of bits are assigned to each feature based on the bit allocation model and the semantic importance of each feature. The bit allocation model is used to determine the number of bits allocated to a feature based on its semantic importance. As an example, the bit allocation model can be an artificial intelligence model. For instance, the bit allocation model can be the Dynamic Proximal Policy Optimization (DPPO) model. The DPPO model is based on a Markov Decision Process (MDP) construction task, using an Actor (policy network)-Critic (value network) dual-network architecture, and combines trajectory sampling and policy updates with pruning to achieve iterative optimization, ultimately learning the optimal bit allocation policy based on semantic importance.

[0048] The number of bits allocated to a feature output by the bit allocation model can be represented by the distance between a two-dimensional allocation vector and the origin. The two dimensions of the two-dimensional allocation vector are the number of significant bits allocated to the feature in that dimension and the total number of bits allocated to the feature in that dimension. For example, the two-dimensional allocation vector can be represented as (x, y), where x is the number of significant bits allocated to the feature in that dimension, and y is the total number of bits allocated to the feature in that dimension. The value of x is greater than or equal to the value of y. The distance between the two-dimensional allocation vector and the origin is the distance between (x, y) and (0, 0). See also Figure 4 As shown in the figure, this is a schematic diagram of mapping a two-dimensional allocation vector to a distance. This allows the two-dimensional allocation vector to be converted into a one-dimensional vector, which reduces the decision complexity and training difficulty of the bit allocation model. It eliminates the need to directly learn the probability distribution of high-dimensional actions; bit allocation decisions can be completed by simply outputting the probability of a one-dimensional index. This simplifies the network structure design and parameter optimization process, and ensures the effectiveness and standardization of bit allocation.

[0049] Based on the number of bits allocated to each feature, the first semantic feature is converted into first bit data. Specifically, each feature included in the first semantic feature can be quantized according to the number of bits allocated to each feature, resulting in bit data for each feature, and thus the first bit data. As an example, a non-subtractive uniform jitter quantization method can be used to convert the floating-point feature values ​​of each feature included in the first semantic feature into discrete binary bit data. In the converted feature bit data, important bits are in the high-order bits, i.e., the first n bits, and unimportant bits are in the low-order bits, i.e., the remaining bits. Here, n is determined based on the number of important bits allocated to that feature.

[0050] Step S203: Map the first bit of data to the corresponding constellation point, and obtain the transmission signal based on the constellation point.

[0051] See Figure 3 As shown, after obtaining the first bit of data, a stream splitter can be used to split the first bit of data, extracting and combining all the first important bits included in the first bit of data to obtain the first important bit data stream. Similarly, all the first unimportant bits included in the first bit of data can be extracted and combined to obtain the first unimportant bit data stream. This facilitates the separate processing of the first important bits in the first important bit data stream and the first unimportant bits in the first unimportant bit data stream.

[0052] After obtaining the first bit of data, channel coding is performed to obtain the transmitted signal.

[0053] During the mapping process, refer to the constellation diagram distribution rules. Map the most important bits to a more robust (error-resistant) position in the constellation diagram, and map less important bits to more error-prone positions. For example, the constellation point corresponding to the first less important bit is centered on the constellation point corresponding to the most important bit. This ensures the constellation point corresponding to the most important bit is more reliable and reduces semantic distortion. See, for example... Figure 5 As shown, this figure is a schematic diagram of a constellation distribution provided in this disclosure.

[0054] Taking the first important bit data stream and the first unimportant bit data stream obtained above as an example, the two bit data streams are mapped and transmitted through important bit pre-amplitude modulation. For example, the first p bits are taken from the first important bit data stream as the prefix of the modulation symbol, and the first s bits are taken from the first unimportant bit data stream as the suffix of the modulation symbol, forming an n (n=p+s) bit modulation symbol. Then, the first p+1 to the first 2p bits are taken from the first important bit data stream as the prefix of the modulation symbol, and the first s+1 to the first 2s bits are taken from the first unimportant bit data stream as the suffix of the modulation symbol, forming a second n (n=p+s) bit modulation symbol. This process continues until all the first important bit data streams and the first unimportant bit data streams are converted into modulation symbols. It should be noted that if the first important bit data stream or the first unimportant bit data stream contains insufficient bits, zeros can be padded to form an n-bit modulation symbol. During constellation mapping, the p bits of prefix data are mapped to a 2^p-QAM constellation point as the center point, and the s bits of suffix data generate a set of constellation points distributed around the center point. The distance between the suffix points and the center point is controlled by the scaling factor α, thereby achieving a flexible trade-off between the first important bit data and the first unimportant bit data in terms of transmission reliability.

[0055] The scaling factor α can range from [0, 1] and is used to reflect the diffusion degree between the first non-essential bit data and the first essential bit data, thereby determining the bit error rate distribution of the first essential bit data / the first non-essential bit data. It also affects the symbol average energy, achieving a dynamic balance in the transmission reliability of the first essential bit data and the first non-essential bit data to adapt to different channel states, semantic feature importance distributions, and task requirements. The scaling factor can be preset manually or combined with the theoretical bit error rate derivation and exponential moving average (EMA) optimization during the DPPO model training process to achieve adaptive updates that dynamically converge with the system state.

[0056] The transmitted signal is obtained by orthogonal amplitude adjustment based on constellation points. The transmitting end then transmits the transmitted signal to the receiving end.

[0057] Step S204: Send a transmission signal to the receiving end.

[0058] See Figure 3 As shown, the transmitting end sends the signal to the receiving end through a physical channel. This disclosure does not limit the transmitting end to determining the receiving end or the method of sending the signal; these can be flexibly configured as needed.

[0059] Based on the above steps S201-S204, it can be seen that while maintaining compatibility with traditional Multiple Quadrature Amplitude Modulation (MQAM) schemes, the present invention can significantly improve information transmission performance and enhance the protection of important information based on semantic content.

[0060] Correspondingly, this disclosure also provides a signal processing method. See [link to relevant documentation]. Figure 6 As shown, this figure is a schematic flowchart of a signal processing method provided in an embodiment of this disclosure. Figure 6 As shown, the signal processing method may include steps S601-S604. The signal processing method provided in this embodiment can be applied to a receiving end.

[0061] Step S601: Obtain the transmission signal.

[0062] The transmitted signal is generated and sent using the semantic feature-based signal transmission method described above. See also... Figure 3 As shown, the receiving end obtains the transmission signal sent by the sending end through the physical channel.

[0063] Step S602: Process the transmitted signal to obtain the second bit data, which includes the second important bit data and the second non-important bit data.

[0064] Step S602 is the reverse operation of step S203. The transmitted signal is demodulated to obtain the second important bit data and the second unimportant bit data.

[0065] For example, see Figure 3 As shown, channel decoding is performed on the transmitted signal to obtain the second important bit data stream and the second unimportant bit data stream.

[0066] As an example, based on the scaling factor at the transmitting end, the constellation point at the center is identified. The second important bit data included in the modulation symbol is demodulated first, followed by the second unimportant bit data included in the modulation symbol, resulting in the second important bit data stream and the second unimportant bit data stream. The second important bit data stream and the second unimportant bit data stream correspond to the first important bit data stream and the first unimportant bit data stream generated by the transmitting end.

[0067] It should be noted that the scaling factor can be sent from the sender to the receiver to achieve synchronization between the sender and receiver.

[0068] like Figure 3 As shown, the second important bit data stream and the second unimportant bit data stream can be split based on the inverse sorter to obtain the bit data corresponding to the feature of each dimension, which is the second bit data.

[0069] This allows the receiving end to easily split the second important bit data stream into important bit data corresponding to the features of each dimension, and the second unimportant bit data stream into unimportant bit data corresponding to the features of each dimension. Then, the important and unimportant bit data corresponding to the features of each dimension can be combined to obtain the bit data corresponding to the features of each dimension. Step S603: According to the conversion rules, convert the second bit data into the second semantic feature.

[0070] Step S603 is the reverse operation corresponding to step S202. The receiving end, based on the conversion rules, reversely converts the second bit of data into the second semantic feature.

[0071] In one possible implementation, when the conversion rule is fixed, for example, when the sender converts the first semantic feature into the first bit of data according to the IEEE 754 unified quantization standard, the sender and receiver can agree on it in advance.

[0072] In another possible implementation, when the conversion rule is not fixed—for example, when the sender converts the first semantic feature into first bit data based on a method for measuring semantic importance and a bit allocation model—the sender and receiver synchronize the conversion rules. Specifically, the sender can send the method for measuring semantic importance and the number of bits allocated to each dimension of the feature generated using the bit allocation model to the receiver, and then perform inverse quantization on the bit data corresponding to each dimension of the feature to obtain the second semantic feature.

[0073] Step S604: Obtain the recovery information based on the second semantic feature.

[0074] Step S604 is the reverse operation of part of step S201. Based on the obtained second semantic features, the second semantic features are converted into information to obtain the recovered information. For example... Figure 3 As shown, the receiving end can use the semantic decoder to process the second semantic feature, obtain the recovered information, and complete the semantic communication between the sending end and the receiving end.

[0075] The signal transmission and signal processing methods based on semantic features disclosed herein can innovate at the channel level. By dividing the first important bit data and the first non-important bit data and mapping them separately on the constellation diagram, communication resources can be preferentially tilted towards the first important bit data, ensuring that the constellation points corresponding to the first important bit data are more reliable. In physical layer transmission, semantic importance is fully utilized, so that the bits with higher importance can obtain more targeted reliability guarantees in transmission and reduce semantic distortion.

[0076] As an example, the signal transmission and signal processing methods based on semantic features provided in this disclosure can be applied to intelligent visual perception and recognition scenarios in a 6G environment, such as image classification tasks based on Dynamic Proximal Policy Optimization (DPPO). Specifically, the transmitter uses ResNet-18 as a semantic encoder to extract image features, and then allocates bit resources for each feature dimension through a DPPO-based bit allocator, i.e., determines the number of bits. The state space of the DPPO-based bit allocator includes the current feature value, the global semantic importance vector, and the proportion of remaining bits. The action space of the DPPO-based bit allocator consists of all two-dimensional vectors that satisfy the constraints (number of important bits, total number of bits), and is mapped to a one-dimensional index by calculating the Euclidean distance from the origin (0,0) to reduce decision complexity. After allocation, each feature is non-subtractive uniform jitter quantization according to the total number of allocated bits, and the high-order bits are separated from the low-order bits according to the first important bit data and the first non-important bit data, and sent to the first important bit data stream and the first non-important bit data stream respectively. During the modulation phase, an Important Bits Prefix Major Quadrature Amplitude Modulation (IBP-MQAM) scheme is used, taking 2 bits from the first important bit data stream as a prefix and 4 bits from the first non-important bit data stream as a suffix to form a 6-bit modulation symbol. The prefix in the modulation symbol maps to the center point of the QPSK constellation, and the suffix generates 16 distribution points around the center. The density of the distribution is controlled by a scaling factor to generate the transmitted signal. The receiver acquires the transmitted signal, recovers the second important bit data stream and the second non-important bit data stream (i.e., dual-bit data stream) through nearest neighbor demodulation, and then obtains the bit data corresponding to the features of each dimension based on the second important bit data stream and the second non-important bit data stream. The semantic features are reconstructed according to the allocation scheme of the bit allocator of the DPPO at the transmitter, and finally, the image classification is completed by the semantic decoder.

[0077] As another example, the signal transmission and signal processing methods based on semantic features provided in this disclosure can be applied to resource-constrained text semantic communication applications, such as text transmission based on Latent Dirichlet Allocation (LDA) topic modeling. Specifically, the transmitter uses an LDA model to extract topic distribution features from the text, and through sparsification, retains only the three topics with the highest probabilities, along with their indices and probability values. For the quantized features, the transmitter statically divides the first important bit data stream and the first unimportant bit data stream according to the IEEE 754 single-precision floating-point standard. For example, the 5-bit code of the topic index and the 8-bit exponent of the probability value are used as important bit data, while the 23-bit mantissa of the probability value is used as unimportant bit data. This division does not require complex online calculations but effectively distinguishes the importance levels of information. The divided dual-bit stream is input into an IBP-MQAM modulator and transmitted using a modulation structure with a 2-bit prefix and a 4-bit suffix. At the receiver, the second bit data recovered through demodulation can accurately reconstruct the topic index and probability value, thus rebuilding the sparse topic distribution vector.

[0078] Based on the semantic feature-based signal transmission method provided in the above-described method embodiments, this disclosure also provides a semantic feature-based signal transmission device, which is applied to the transmitting end. The semantic feature-based signal transmission device will be described below with reference to the accompanying drawings.

[0079] like Figure 7 As shown, the semantic feature-based signal transmission device 700 provided in this embodiment includes: The acquisition unit 701 is used to acquire the information to be sent and extract the first semantic feature of the information to be sent; The conversion unit 702 is configured to convert the first semantic feature into first bit data including first important bit data and first unimportant bit data according to a conversion rule. The conversion rule is used to indicate the method of converting the first semantic feature into the first bit data. The first important bit data and the first unimportant bit data are determined based on the semantic importance of the first semantic feature. The modulation unit 703 is used to map the first bit data to the corresponding constellation point and obtain the transmission signal based on the constellation point, wherein the constellation point corresponding to the first non-important bit data is centered on the constellation point corresponding to the first important bit data. The transmitting unit 704 is used to transmit the transmitting signal to the receiving end.

[0080] In one possible implementation, the conversion unit 702 is specifically used to convert the first semantic feature into first bit data according to a preset quantization standard, and to determine that the first bit data includes first important bit data and first unimportant bit data. The quantization standard is used to indicate the method of converting the first bit data, the method of determining the semantic importance of the first bit data, and the proportion of the first important bit data to the first bit data.

[0081] In one possible implementation, the conversion unit 702 is specifically used to determine the semantic importance of the features in each dimension included in the first semantic feature; based on the bit allocation model and the semantic importance of the features in each dimension, allocate a corresponding number of bits to the features in each dimension; convert the first semantic feature into first bit data according to the number of bits allocated to the features in each dimension, and determine the first bit data including first important bit data and first unimportant bit data.

[0082] In one possible implementation, the number of bits is represented by the distance between a two-dimensional allocation vector and the origin, wherein the two dimensions of the two-dimensional allocation vector are the number of bits of the first important bit data allocated to the feature of the dimension and the total number of bits allocated to the feature of the dimension.

[0083] In one possible implementation, the distance between the constellation point corresponding to the first non-significant bit data and the constellation point corresponding to the first significant bit data is determined based on a scaling factor, which indicates the degree of diffusion of the constellation point corresponding to the first non-significant bit data relative to the constellation point corresponding to the first significant bit data.

[0084] In the embodiments of this disclosure, the specific processing of each unit and the resulting technical effects in the signal transmission device 700 based on semantic features can be referred to the relevant descriptions of each step in the embodiments corresponding to the aforementioned signal transmission method based on semantic features, and will not be repeated here.

[0085] Based on the signal processing method provided in the above-described embodiments, this disclosure also provides a signal processing apparatus applied at a receiving end. The signal processing apparatus will be described below with reference to the accompanying drawings.

[0086] like Figure 8 As shown, the signal processing apparatus 800 provided in this embodiment includes: The acquisition unit 801 is used to acquire the transmission signal, which is generated and transmitted using the signal transmission method based on semantic features described in the above embodiments; Demodulation unit 802 is used to process the transmitted signal to obtain second bit data including second important bit data and second non-important bit data; The conversion unit 803 is used to convert the second bit data into a second semantic feature according to a conversion rule, wherein the conversion rule is used to indicate the method of converting the first semantic feature into the first bit data during the generation of the transmission signal; The generation unit 804 is used to obtain recovery information based on the second semantic feature.

[0087] In this embodiment of the disclosure, the specific processing of each unit in the signal processing apparatus 800 and the resulting technical effects can be referred to the relevant descriptions of each step in the embodiments corresponding to the aforementioned signal processing method, and will not be repeated here.

[0088] Figure 9 A schematic block diagram of an example electronic device 900 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0089] like Figure 9 As shown, device 900 includes a computing unit 901, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 902 or a computer program loaded into random access memory (RAM) 903 from storage unit 908. RAM 903 may also store various programs and data required for the operation of device 900. The computing unit 901, ROM 902, and RAM 903 are interconnected via bus 904. Input / output (I / O) interface 905 is also connected to bus 904.

[0090] Multiple components in device 900 are connected to I / O interface 905, including: input unit 906, such as keyboard, mouse, etc.; output unit 907, such as various types of monitors, speakers, etc.; storage unit 908, such as disk, optical disk, etc.; and communication unit 909, such as network card, modem, wireless transceiver, etc. Communication unit 909 allows device 900 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0091] The computing unit 901 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 901 performs the various methods and processes described above, such as semantic feature-based signal transmission methods or signal processing methods. For example, in some embodiments, semantic feature-based signal transmission methods and signal processing methods can be implemented as computer software programs tangibly contained in a machine-readable medium, such as storage unit 908. In some embodiments, part or all of the computer program can be loaded and / or installed on device 900 via ROM 902 and / or communication unit 909. When the computer program is loaded into RAM 903 and executed by the computing unit 901, one or more steps of the semantic feature-based signal transmission methods or signal processing methods described above can be performed. Alternatively, in other embodiments, the computing unit 901 may be configured by any other suitable means (e.g., by means of firmware) to perform a signal transmission method or a signal processing method based on semantic features.

[0092] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0093] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable information processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0094] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0095] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; 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 sound input, voice input, or tactile input).

[0096] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an information server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0097] Computer systems can include client devices and servers. Client devices and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is created by computer programs running on the respective computers and establishing a client-server relationship between them. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0098] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0099] The specific embodiments described above do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure.

Claims

1. A signal transmission method based on semantic features, characterized in that, The method is applied at the sending end, and the method includes: Obtain the information to be sent, and extract the first semantic feature of the information to be sent; According to the conversion rule, the first semantic feature is converted into first bit data including first important bit data and first unimportant bit data. The conversion rule is used to indicate the way the first semantic feature is converted into the first bit data. The first important bit data and the first unimportant bit data are determined based on the semantic importance of the first semantic feature. The first bit data is mapped to the corresponding constellation point, and a transmission signal is generated based on the constellation point. The constellation point corresponding to the first non-important bit data is centered on the constellation point corresponding to the first important bit data. The transmission signal is sent to the receiving end.

2. The method according to claim 1, characterized in that, The step of converting the first semantic feature into first bit data, which includes first important bit data and first unimportant bit data, according to the conversion rules includes: According to a preset quantization standard, the first semantic feature is converted into first bit data, and the first bit data includes first important bit data and first unimportant bit data. The quantization standard is used to indicate the method of converting the first bit data, the method of determining the semantic importance of the first bit data, and the proportion of the first important bit data to the first bit data.

3. The method according to claim 1, characterized in that, The step of converting the first semantic feature into first bit data, which includes first important bit data and first unimportant bit data, according to the conversion rules includes: Determine the semantic importance of each dimension of the features included in the first semantic feature; Based on the bit allocation model and the semantic importance of the features in each dimension, the number of bits corresponding to the features in each dimension is determined. Based on the number of bits corresponding to the features of each dimension, the first semantic feature is converted into first bit data, and the first bit data includes first important bit data and first unimportant bit data.

4. The method according to claim 3, characterized in that, The number of bits is represented by the distance between a two-dimensional allocation vector and the origin. The two dimensions of the two-dimensional allocation vector are the number of bits of the first important bit data allocated to the feature of the dimension and the total number of bits allocated to the feature of the dimension.

5. The method according to any one of claims 1-4, characterized in that, The distance between the constellation point corresponding to the first non-significant bit data and the constellation point corresponding to the first significant bit data is determined based on a scaling factor, which is used to indicate the degree of diffusion of the constellation point corresponding to the first non-significant bit data relative to the constellation point corresponding to the first significant bit data.

6. A signal processing method, characterized in that, The method is applied at the receiving end, and the method includes: Acquire a transmission signal, wherein the transmission signal is generated and transmitted using the signal transmission method based on semantic features as described in any one of claims 1-5; Processing the transmitted signal yields second bit data, which includes second important bit data and second unimportant bit data; According to the conversion rule, the second bit data is converted into a second semantic feature. The conversion rule is used to indicate the method of converting the first semantic feature into the first bit data during the generation of the transmission signal. Based on the second semantic feature, the recovery information is obtained.

7. A signal transmission device based on semantic features, characterized in that, The device is used at the transmitting end, and the device includes: An acquisition unit is used to acquire information to be sent and extract the first semantic feature of the information to be sent; A conversion unit is configured to convert the first semantic feature into first bit data including first important bit data and first unimportant bit data according to a conversion rule, wherein the conversion rule is used to indicate the method of converting the first semantic feature into the first bit data, and the first important bit data and the first unimportant bit data are determined based on the semantic importance of the first semantic feature. A modulation unit is used to map the first bit data to a corresponding constellation point and obtain a transmission signal based on the constellation point, wherein the constellation point corresponding to the first non-important bit data is centered on the constellation point corresponding to the first important bit data. The transmitting unit is used to transmit the transmitting signal to the receiving end.

8. A signal processing apparatus, characterized in that, The device is used at a receiving end, and the device includes: An acquisition unit is used to acquire a transmission signal, wherein the transmission signal is generated and transmitted using the signal transmission method based on semantic features as described in any one of claims 1-5; The demodulation unit is used to process the transmitted signal to obtain second bit data including second important bit data and second non-important bit data; A conversion unit is used to convert the second bit data into a second semantic feature according to a conversion rule, wherein the conversion rule is used to indicate the method of converting the first semantic feature into the first bit data during the generation of the transmission signal; The generation unit is used to obtain recovery information based on the second semantic feature.

9. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the signal transmission method based on semantic features as described in any one of claims 1-5, or to perform the signal processing method as described in claim 6.

10. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to execute the signal transmission method based on semantic features according to any one of claims 1-5, or to execute the signal processing method according to claim 6.

11. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the signal transmission method based on semantic features according to any one of claims 1-5, or performs the signal processing method according to claim 6.

Images