5G broadcast television image communication system fused with deep learning and method thereof

Abstract

The invention relates to a 5G broadcast television image communication system fused with deep learning and a method thereof. The system comprises an image processing module, a channel processing module, a data transmission module, a channel decoding module, an image reconstruction and enhancement module and a system optimization and self-adaption module. Through an image compression technology based on Vision Transform, an image is divided into image blocks with a fixed size, and a multi-head attention mechanism is utilized to extract global context and local features, so that efficient compression is realized. Meanwhile, the channel coding and decoding unit based on Transform enhances the anti-noise capability of the system by generating redundant bits and correcting cross-layer error codes, an image reconstruction and enhancement module is combined with a generative adversarial network (GAN) and a super-resolution model, detail compensation and enhancement can be carried out on the decompressed image, and through a system optimization and self-adaptive adjustment module, the image reconstruction and enhancement module can carry out self-adaptive adjustment on the image. A transmission strategy is dynamically adjusted according to a channel state and transmission feedback, and efficient and stable radio and television image communication is realized in a complex 5G network environment.

Description

Technical Field

[0001] The present invention relates to a 5G broadcast television image communication system and method integrating deep learning, and in particular to a 5G broadcast television image communication system and method integrating deep learning applied in the field of image communication technology. Background Art

[0002] Although the existing 5G broadcast television image communication technology can achieve real-time transmission to a certain extent by taking advantage of high bandwidth and low latency, it still faces many technical challenges in terms of network environment fluctuations, increased image quality requirements, and system anti-interference capabilities. For example, compression and reconstruction technology in image transmission, channel anti-interference mechanism, and how to adaptively adjust transmission strategies according to network status are all issues that need to be urgently addressed in existing technologies.

[0003] Chinese invention patent CN118138802B discloses a 5G broadcast television image communication method integrating deep learning. This method constructs a link interference prediction model and dynamically adjusts the compression and encoding parameters according to the real-time signal-to-noise ratio in the communication link to ensure the transmission stability of television images. However, this technology has the following shortcomings: Limited dynamic adjustment capability: Although this patent can adjust the modulation parameters and bandwidth based on the signal-to-noise ratio, the signal-to-noise ratio prediction model it relies on may respond slowly to sudden network fluctuations, especially in a complex 5G network environment. This static parameter set adjustment mechanism is difficult to adapt to real-time changes. Limitations of compression technology: The JPEG compression algorithm and LDPC coding scheme used in this patent can reduce the amount of data, but the loss of details of the compressed image is relatively serious, making it difficult to meet the transmission requirements of high-quality images while ensuring real-time performance.

[0004] Chinese invention patent CN114615499B discloses a semantic optical communication system for image transmission, which transmits encoded image data through optical signals and uses a transformation module to process the image to ensure the transmission and restoration of the image. This technology is mainly aimed at the optical communication transmission of images, and has certain innovations, but has the following limitations: Restriction of transmission medium: This patent relies on optical signals to transmit image data. Although optical communication has the advantages of high bandwidth and low latency, in the broadcasting and television scenario, the coverage and convenience of 5G networks are more suitable for application, and this technology does not have sufficient 5G network adaptation capabilities. The image reconstruction quality is low: Although the encoding and decoding module of this patent can restore the optical signal to image data, it fails to solve the problem of detail loss and blurring that may occur during the transmission of the image, especially in the case of high-resolution images, the image quality will decline.

[0005] The above patented design has made certain contributions to improving the efficiency and stability of image transmission, but it still has limitations, such as insufficient dynamic adjustment capabilities, large quality loss after image compression, and difficulty in coping with sudden network fluctuations. Therefore, the existing technical solutions still need to be further optimized and improved when facing high-resolution, real-time image transmission under 5G networks. Summary of the invention

[0006] In view of the above-mentioned prior art, the technical problem to be solved by the present invention is how to improve the efficiency of image compression and transmission, enhance the anti-interference ability, and enhance the resolution and clarity of the reconstructed image while achieving adaptive adjustment.

[0007] To solve the above problems, the present invention provides a 5G broadcast television image communication system and method integrating deep learning, including an image processing module, a channel processing module, a data transmission module, a channel decoding module, an image reconstruction and enhancement module, and a system optimization and adaptation module;

[0008] An image processing module, used for preprocessing and compressing the original image, the image processing module includes an image preprocessing unit and an image compression unit;

[0009] The image preprocessing unit is used to perform format conversion and normalization processing on the input image;

[0010] The image compression unit divides the image into fixed-size image blocks based on Vision Transformer, and extracts global and local features through a multi-head attention mechanism to achieve low-dimensional compressed representation of the image;

[0011] A channel processing module, used for performing channel coding and state estimation on the compressed image data, the channel processing module comprises a channel coding unit and a channel state estimation unit;

[0012] The channel coding unit generates redundant bits through a Transformer-based encoder to enhance the data's ability to resist noise;

[0013] The channel state estimation unit estimates the state of the 5G channel in real time based on the Transformer timing prediction model and dynamically adjusts the coding scheme according to the channel state;

[0014] A data transmission module, used for transmitting data through a 5G network, the data transmission module comprising a channel transmission unit, the channel transmission unit transmitting the encoded image data through the 5G network, and monitoring errors in data transmission through an error monitoring and feedback unit;

[0015] A channel decoding module is used to receive and decode image data. The channel decoding module includes a channel decoding unit. The channel decoding unit decodes the received data based on the multi-head attention mechanism of Transformer, and detects and corrects data errors through an error correction unit.

[0016] An image reconstruction and enhancement module, used for reconstructing and enhancing the decoded image data, the image reconstruction and enhancement module comprises an image decompression unit and an image enhancement unit;

[0017] The image decompression unit uses the Transformer model to decompress the decoded low-dimensional data and restore it to a high-dimensional image representation;

[0018] The image enhancement unit enhances the details and improves the resolution of the reconstructed image based on the generative adversarial network (GAN) and super-resolution technology;

[0019] System optimization and adaptive module, used to optimize and dynamically adjust the entire system. The system optimization and adaptive module includes a joint training optimization unit and an adaptive adjustment unit;

[0020] The joint training optimization unit optimizes the models of image processing, channel coding, channel decoding and image reconstruction modules through end-to-end joint training;

[0021] The adaptive adjustment unit dynamically adjusts the coding and transmission strategies according to the channel state estimation result and data transmission feedback.

[0022] In the above-mentioned 5G broadcasting and television image communication system and method integrating deep learning, the image compression unit based on VisionTransformer can improve the compression rate while maintaining the details and global features of high-quality images, thereby realizing efficient transmission of images under the 5G network. Through Transformer-based channel coding and multi-head attention mechanism, redundant bits are generated to enhance the data's anti-noise ability, thereby improving the channel's anti-interference ability. Through the generative adversarial network (GAN) and super-resolution model, the details of the compression loss can be effectively compensated during the reconstruction process after image transmission, so that the reconstructed image can be improved in resolution and clarity. The adaptive adjustment module can dynamically adjust the transmission strategy according to changes in the channel state, ensuring stable and efficient image transmission even in complex and frequently fluctuating network environments.

[0023] As a further improvement of the present application, the image compression unit divides the image into multiple image blocks of fixed sizes based on the Vision Transformer model, and extracts the global context and local feature information of each image block through a multi-head attention mechanism to perform image compression processing.

[0024] As a further improvement of the present application, the channel coding unit generates redundant bits through a Transformer-based multi-head attention mechanism and performs parallel processing on channel transmission data to enhance the channel's ability to resist noise.

[0025] As a further improvement of the present application, the channel state estimation unit predicts the channel state based on the time series Transformer model through historical channel data collected in real time, and dynamically adjusts the coding and transmission strategies according to the prediction results.

[0026] As another improvement of the present application, the channel decoding unit identifies erroneous bits in the received data through a Transformer model based on a multi-head attention mechanism, and performs cross-layer error correction through an error correction unit.

[0027] As another improved supplement of the present application, the image enhancement unit in the image reconstruction and enhancement module adopts a Transformer-based generative adversarial network (GAN) and a super-resolution model to identify blurred areas of the image through a multi-head attention mechanism, perform detail compensation and enhancement, and improve the resolution of the image.

[0028] As another improved supplement of the present application, the system optimization and adaptation module jointly optimizes the model parameters of the image processing module, channel processing module, channel decoding module and image reconstruction module through an end-to-end joint training optimization unit to improve the overall communication quality and efficiency of the system.

[0029] As another improvement of the present application, the adaptive adjustment unit dynamically adjusts the channel coding and data transmission strategy according to the channel state estimation results and feedback information during the data transmission process to adapt to the ever-changing 5G network communication environment.

[0030] A 5G broadcast television image communication system and method integrating deep learning, characterized in that it includes the following steps:

[0031] S1: preprocessing the original image, including format conversion and normalization, and dividing the image into multiple image blocks of fixed sizes;

[0032] S2: Through the Vision Transformer-based image compression unit, the multi-head attention mechanism is used to extract the global context and local features of the image block and compress the image into a low-dimensional representation;

[0033] S3: Channel coding is performed on the compressed image data through a Transformer-based channel coding unit to generate redundant bits and enhance the data's ability to resist noise;

[0034] S4: Transmit the encoded image data through the 5G network, monitor the errors in the transmission process in real time, and feedback to the sender;

[0035] S5: The receiving end decodes the received coded data through the Transformer-based channel decoding unit and uses the multi-head attention mechanism to detect and correct bit errors;

[0036] S6: The decoded low-dimensional data is restored to a high-dimensional image representation through a Transformer-based image decompression unit;

[0037] S7: Through the Transformer-based image enhancement unit, use the generative adversarial network (GAN) or super-resolution model to compensate for the details and improve the resolution of the reconstructed image;

[0038] S8: Through the system optimization and adaptive adjustment module, the channel coding and data transmission strategies are dynamically adjusted according to the channel state estimation results and transmission feedback information to adapt to changes in the 5G network environment.

[0039] In summary, this application has the following beneficial effects:

[0040] 1. Efficient image compression and transmission. In terms of image processing and compression, this application uses an image compression unit based on VisionTransformer (ViT). By dividing the image into fixed-size image blocks and using the multi-head attention mechanism in ViT to extract global context and local features, efficient image compression is achieved. ViT can accurately extract the core information in the image and reduce unnecessary redundant data, thereby achieving a higher compression rate. This not only reduces the amount of data, but also reduces bandwidth usage. During the compression process, the system retains important global and local feature information, avoiding the common problem of detail loss in traditional compression methods, and ensuring the quality of the image after decompression.

[0041] 2. Powerful anti-interference and error correction capabilities. During data transmission, the complex channel environment of the 5G network may cause various noise and interference problems. In order to meet these challenges, this application enhances the system's anti-noise and anti-interference capabilities through a Transformer-based channel coding and decoding unit. Through a multi-head attention mechanism to process different feature dimensions of data in parallel, the system can generate redundant bits and enhance the robustness of the data in the channel, enabling it to resist the common noise and interference in 5G networks. During channel decoding, cross-layer error correction technology is used to detect and repair erroneous bits in the data layer by layer to ensure data integrity. Compared with traditional error detection methods, the layer-by-layer correction method can improve the accuracy and efficiency of error correction.

[0042] 3. Superior performance of image reconstruction and enhancement. In terms of image reconstruction and enhancement, this application improves the quality of decoded images by introducing generative adversarial networks (GANs) and super-resolution technology. GANs can compensate for the detail information lost during transmission by generating high-quality images, especially when the image is blurred or low-resolution. Through super-resolution technology, the system can identify blurred areas in the image and restore the details to improve the visual effect of the image. Combined with the multi-head attention mechanism, the system can restore the image details more accurately, so that the image after transmission and decompression not only maintains the original visual effect, but also improves the resolution and details.

[0043] 4. Dynamic adaptive adjustment and system optimization. This application introduces a system optimization and adaptive adjustment module, which can adjust the system parameters in real time according to the changes in channel status and transmission feedback to adapt to the changing network environment. The system can dynamically adjust the redundant bits and data transmission strategies in the channel coding according to the real-time collected channel status. For example, when it is detected that the channel state has deteriorated, the system will increase the redundant bits to ensure the stability of data transmission; when the channel state is good, the system will reduce the redundant bits to improve the transmission efficiency. Through end-to-end joint training, the system can continuously optimize the collaborative effect between image processing, channel coding, decoding and reconstruction modules to ensure that the performance of each module is optimized, thereby improving the overall communication quality of the system.

[0044] 5. Efficient radio and television communication system in 5G environment. This application makes full use of the high bandwidth and low latency characteristics of the 5G network, combined with a deep learning model, so that the system shows great advantages in the field of radio and television communications. Due to the improvement of image compression technology and the optimization of channel coding, the system of this application can transmit high-resolution images with low latency, which is suitable for radio and television applications that require real-time image transmission. Through the combination of redundant bit generation, multi-head attention mechanism and error correction mechanism, the system can maintain high reliability of data transmission in the 5G environment. Whether in high-speed mobile scenarios or signal unstable environments, the system can maintain data integrity. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 It is the overall interactive diagram of this application;

[0046] Figure 2 Local interaction for this application Figure 1 ;

[0047] Figure 3 Local interaction for this application Figure 2 ;

[0048] Figure 4 Local interaction for this application Figure 3 ;

[0049] Figure 5 This is an interactive flow chart of this application. DETAILED DESCRIPTION

[0050] Three embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0051] Example 1

[0052] like Figures 1 to 4 As shown, this embodiment provides a 5G broadcast television image communication system integrating deep learning, aiming to improve image transmission efficiency, noise resistance and image quality. The system includes an image processing module, a channel processing module, a data transmission module, a channel decoding module, an image reconstruction and enhancement module, and a system optimization and adaptation module. The specific implementation and working process of each module are described in detail below.

[0053] The function of the image processing module is to preprocess and compress the input original image.

[0054] Image preprocessing unit: receives the original image and performs format conversion and normalization on it. Format conversion ensures that the format of the input image is unified for subsequent processing; normalization is used to standardize the pixel value range of the image and improve the robustness of the image processing model.

[0055] Image compression unit: Based on the Vision Transformer (ViT) architecture, the input image is divided into fixed-size image blocks (such as 16x16 pixels) and converted into vector form. Through the multi-head attention mechanism, ViT extracts the global context and local features of each image block to achieve low-dimensional compressed representation of the image. The multi-head attention mechanism can process different areas of the image in parallel, improve compression efficiency, and retain important image features.

[0056] The main function of the channel processing module is to encode the image data and dynamically adjust it according to the channel status.

[0057] Channel coding unit: Based on the Transformer model, redundant bits are generated to resist noise and interference in 5G channels. Through the multi-head attention mechanism, the encoder can process multiple feature dimensions in the image data stream in parallel and generate redundant bits to enhance the data's noise resistance. In this process, important features in the data stream are weighted multiple times to ensure the robustness of the data in harsh communication environments.

[0058] Channel state estimation unit: Uses the time series transformer model to collect historical channel data (such as signal strength, noise interference, etc.) to perform real-time estimation of channel state. The time series transformer can effectively capture the features in time series data and predict future channel states. Based on the prediction results, the channel state estimation unit will dynamically adjust the channel coding scheme to ensure that data transmission can maintain optimal performance under different channel conditions.

[0059] The data transmission module is responsible for the actual transmission of data through the 5G network.

[0060] Channel transmission unit: Transmits the encoded image data through the 5G network. During the transmission process, the unit monitors the errors in the data transmission and provides real-time feedback on the status of the data transmission through the error monitoring and feedback unit. If an error in the transmission is detected, the system can adjust the channel coding or transmission strategy in time to further improve the stability of data transmission.

[0061] Working process,

[0062] The image processing module receives the original image, first converts and normalizes the image through the image preprocessing unit, then divides the preprocessed image into multiple fixed-size image blocks, and compresses it through the image compression unit based on VisionTransformer. ViT uses a multi-head attention mechanism to extract the global contextual relationship between image blocks and compress the image into a low-dimensional representation.

[0063] The compressed image data enters the channel processing module, and the channel coding unit generates redundant bits based on the Transformer model to improve the anti-noise performance. At the same time, the channel state estimation unit collects historical channel data, makes real-time predictions on the channel state based on the time series Transformer model, and dynamically adjusts the coding scheme to adapt to changes in the channel state.

[0064] The encoded image data is transmitted through the 5G network, and the channel transmission unit is responsible for monitoring possible errors during the transmission process. Once an error is found, the feedback unit will transmit the information back to the channel processing module to adjust the encoding scheme or transmission strategy to ensure the complete transmission of the data.

[0065] Vision Transformer is used for image compression. The global context and local detail features of image blocks are extracted through a multi-head attention mechanism, which greatly improves the compression efficiency. The main features of the original image are retained during the compression process, while the amount of data is reduced, thereby reducing the bandwidth requirements for transmission.

[0066] By generating redundant bits through the Transformer encoder in the channel processing module, the system can effectively resist noise and interference in complex 5G channel environments. The application of the multi-head attention mechanism ensures the parallel processing of features of different dimensions, further improving the robustness of data transmission.

[0067] The time series Transformer model performs well in channel state estimation, can accurately predict the channel state based on historical data, and adjust the coding scheme according to the real-time channel conditions. This dynamic adjustment mechanism enables the system to maintain efficient data transmission in a changing 5G network environment, reducing data packet loss and delay.

[0068] During the data transmission process, the system can monitor errors in real time and adjust the transmission strategy in time through the feedback mechanism to ensure data integrity and transmission reliability.

[0069] Example 1 integrates the Vision Transformer and Transformer models to achieve efficient image compression, noise-resistant channel coding, dynamic channel state estimation, and real-time error monitoring and feedback in the 5G broadcast television image communication system. This example effectively solves the problems of noise interference and data loss in the 5G channel, while greatly improving the transmission efficiency and reliability of image data.

[0070] Example 2

[0071] like Figures 1 to 4 As shown, this embodiment describes the channel decoding, image reconstruction and enhancement, and system optimization and adaptive adjustment functions of the 5G broadcast television image communication system. This embodiment focuses on how to use the multi-head attention mechanism in the Transformer model for channel decoding, how to improve image quality through generative adversarial networks (GANs) and super-resolution technology, and how to optimize the overall performance of the system through joint training and adaptive adjustment.

[0072] The channel decoding module is used to receive and decode image data transmitted through the 5G network, with the focus on using the multi-head attention mechanism to identify erroneous bits in the data and correct them.

[0073] Channel decoding unit: Through the Transformer model based on the multi-head attention mechanism, the received data stream is analyzed to identify the error bits that may occur during the transmission process. The multi-head attention mechanism can process different parts of the received data in parallel and focus on those bits that may contain errors through weight adjustment. Compared with traditional error detection methods, the Transformer model can refine data detection at different levels and improve the accuracy of detection.

[0074] Error correction unit: After identifying the error bits, the system corrects them through the cross-layer error correction unit. Cross-layer error correction adopts a layer-by-layer progressive approach to analyze and repair potential errors in the data layer by layer, thereby ensuring that the data after channel decoding is as close as possible to the original data before transmission.

[0075] After the image data is decoded, the system reconstructs and enhances the decoded image through the image reconstruction and enhancement module to restore the image quality and compensate for the detail information lost during the transmission process.

[0076] Image decompression unit: The decoded low-dimensional image data is decompressed using the Transformer model to restore it to a high-dimensional image representation. Through the self-attention mechanism, the system can effectively restore the global and local details of the image, so that the image remains high quality after decompression.

[0077] Image enhancement unit: Generative adversarial network (GAN) and super-resolution technology are used to improve the resolution and detail of the image. GAN uses the mechanism of two networks confronting each other to make the generated image gradually approach the real image. Combined with the multi-head attention mechanism, the system can accurately identify the blurred or distorted areas in the image, compensate and enhance the details of these areas, and make the clarity and visual effect of the reconstructed image more superior.

[0078] In order to improve the overall performance of the system, the system optimization and adaptation module is responsible for joint optimization and adaptive adjustment of each module in the system.

[0079] Joint training optimization unit: This unit optimizes the model of the image processing module, channel processing module, channel decoding module and image reconstruction module through end-to-end joint training. Specifically, the system optimizes the parameters of the model through multiple rounds of training, so that each module can work together to improve the overall communication quality and image processing effect.

[0080] Adaptive adjustment unit: Based on the channel state estimation results and the information fed back during data transmission, the system dynamically adjusts the channel coding and transmission strategy. By analyzing the real-time transmission data, the system can adjust the number of redundant bits of the channel coding or the transmission strategy to ensure that the system can maintain the best communication performance and image quality under different network conditions.

[0081] Working process

[0082] The system uses the multi-head attention mechanism of the Transformer model through the channel decoding unit to decode the received image data and identify the error bits in the data. The multi-head attention mechanism can process multiple feature dimensions at the same time and analyze the data at different levels, focusing on the bits that may have errors. The identified error bits are corrected by the error correction unit, which uses cross-layer error correction to fine-tune the errors in the data layer by layer, and finally output high-quality decoded data.

[0083] The decoded image data enters the image reconstruction and enhancement module. First, the image decompression unit restores the low-dimensional data to a high-dimensional image representation through the Transformer model, ensuring that the restored image retains the necessary global and local features. Next, the image enhancement unit uses a generative adversarial network (GAN) to identify and repair the blurred areas of the image, and further improves the image details and resolution through super-resolution technology, making the final reconstructed image clearer.

[0084] The system optimization and adaptation module adjusts the channel coding scheme and data transmission strategy in real time according to the results of channel state estimation and feedback information during data transmission. For example, when the channel quality is detected to be degraded, the adaptive adjustment unit will increase the redundant bits in the channel coding to improve the data's anti-noise ability, or adjust the transmission strategy to reduce the data packet loss rate. At the same time, the joint training optimization unit continuously optimizes the model parameters of each module through multiple rounds of joint training to ensure that the overall performance of the system is always in the best state.

[0085] By processing data in parallel through a multi-head attention mechanism, the system can efficiently identify erroneous bits that may occur during transmission, and repair these errors layer by layer through cross-layer error correction units, thereby improving the accuracy and reliability of decoded data.

[0086] By combining the generative adversarial network (GAN) with super-resolution technology, the image reconstruction and enhancement module can identify and repair blurred areas in the image, improve the resolution and details of the image, so that the received image not only maintains the original visual effect, but also has enhanced details.

[0087] The adaptive adjustment unit dynamically adjusts the coding and transmission strategies according to the changes in the channel status, ensuring that the system can maintain efficient communication and image quality under different 5G network conditions. At the same time, the joint training optimization unit continuously improves the collaboration capabilities of various modules in the system through end-to-end training, making the system more stable when transmitting and processing complex image data.

[0088] Through joint training and adaptive adjustment, the system not only improves the compression, transmission and decoding efficiency of images when processing high-bandwidth image communications, but also reduces data loss and transmission delay. This optimization enables stable and high-quality broadcast television image communications in a high-load 5G environment.

[0089] Example 2 improves the overall performance of the 5G broadcast television image communication system through detailed design of channel decoding, image reconstruction and enhancement, and system optimization and adaptive adjustment. The multi-head attention mechanism in the Transformer model not only achieves more accurate error detection and correction, but also improves the visual effect of the image through generative adversarial networks (GANs) and super-resolution technology. The joint training and adaptive adjustment functions of the system further ensure that the system can maintain the best communication quality and image processing effect under changing network conditions.

[0090] Example 3

[0091] like Figures 1 to 5 As shown, this embodiment describes in detail the various steps of the 5G broadcast television image communication method based on integrated deep learning, combines the key technologies in Example 1 and Example 2, further improves the system workflow, and provides a complete technical path, covering the entire process from image preprocessing, compression to transmission, decoding, reconstruction and enhancement.

[0092] Step S1: Image preprocessing

[0093] Preprocess the original input image. First, perform format conversion to convert images of different formats into a format suitable for the system (such as JPEG, PNG, etc.). Then, normalize the image to ensure that the pixel value of the image is within a reasonable range, so that the subsequent deep learning model can extract image features more effectively.

[0094] Step S2: Image compression

[0095] The preprocessed image is compressed by an image compression unit based on Vision Transformer (ViT). The image is first divided into multiple fixed-size image blocks (such as 16x16 pixels), and each image block is converted into a vector form. ViT uses a multi-head attention mechanism to focus on both global context and local features, extract features from these image blocks, and compress them, converting high-dimensional images into low-dimensional representations.

[0096] Step S3: Channel Coding

[0097] The compressed image data is encoded using a Transformer-based channel coding unit. In this step, redundant bits are generated to enhance the data's ability to resist noise. Through the Transformer model's multi-head attention mechanism, the system is able to process different feature dimensions simultaneously and generate redundant bits, ensuring that data can better resist interference and noise in the channel during transmission in the 5G network.

[0098] Step S4: Data transmission and error monitoring

[0099] The compressed and encoded image data is transmitted through the 5G network. During this process, the channel transmission unit monitors the data transmission in real time and identifies possible data loss, interference or noise through the error monitoring unit. If an error occurs, the monitoring unit will send feedback information to the transmitter to adjust the channel coding or transmission strategy.

[0100] Step S5: Channel decoding and error correction

[0101] The receiving end uses a Transformer-based channel decoding unit to decode the received image data. The multi-head attention mechanism can process the received data stream in parallel, focusing on the error bits in the data stream. The system then corrects these error bits through the error correction unit to ensure that the final decoded data matches the original data as closely as possible.

[0102] Step S6: Image decompression

[0103] The decoded low-dimensional image data is restored to a high-dimensional image representation through the Transformer-based image decompression unit. The self-attention mechanism can effectively restore the details and overall structure of the image, so that the image still maintains a high quality after transmission.

[0104] Step S7: Image enhancement and detail compensation

[0105] The decompressed image is enhanced in detail and resolution through the Transformer-based generative adversarial network (GAN) and super-resolution model. The system uses GAN to generate high-quality images, and uses the super-resolution model to identify blurred areas in the image, compensate for and enhance the details, making the final output image clearer and more realistic.

[0106] Step S8: System optimization and adaptive adjustment

[0107] Through the system optimization and adaptive adjustment module, the system dynamically adjusts the channel coding and transmission strategy according to the results of channel state estimation and feedback information during the transmission process. For example, when the channel state deteriorates, the system will increase redundant bits or adjust transmission parameters to ensure the stability of data transmission. At the same time, through end-to-end joint training optimization, the collaborative working efficiency between image processing, channel coding, decoding and reconstruction modules is further improved.

[0108] Image preprocessing and compression (steps S1-S2);

[0109] The system first receives the original image and preprocesses it, ensuring the uniformity of the image format and the efficiency of feature processing through format conversion and normalization. Next, the image is divided into multiple image blocks, and features are extracted through the multi-head attention mechanism in VisionTransformer and compressed into a low-dimensional representation, thereby reducing the bandwidth required for transmission.

[0110] Channel coding and transmission (steps S3-S4);

[0111] The compressed image data generates redundant bits through the Transformer-based channel coding unit to improve noise resistance. Next, the image data is transmitted through the 5G network. The channel transmission unit monitors the errors in data transmission in real time and feeds back information to the sender through error monitoring to adjust the transmission strategy.

[0112] Channel decoding and error correction (step S5);

[0113] The receiver uses a multi-head attention mechanism to decode the received data and repair the detected erroneous bits through the error correction unit. Cross-layer error correction ensures that errors in data are processed and repaired at multiple levels, ensuring the integrity and accuracy of the received data with the original data.

[0114] Image decompression and enhancement (steps S6-S7);

[0115] The decoded low-dimensional image data is restored to a high-dimensional representation through an image decompression unit, and then the image quality is further enhanced through a generative adversarial network (GAN) and a super-resolution model to compensate for the blur and detail loss caused by the transmission process in the image.

[0116] System adaptive adjustment and optimization (step S8);

[0117] During the transmission and processing process, the system dynamically adjusts the channel coding and transmission strategy based on channel state estimation and feedback information, and improves the collaborative efficiency of the entire system through joint training optimization modules to ensure the best communication and processing performance in the 5G environment.

[0118] Through the multi-head attention mechanism of Vision Transformer, the system can reduce the amount of data while maintaining image quality, allowing images to be efficiently transmitted under limited bandwidth.

[0119] Transformer-based channel coding and decoding technology achieves strong resistance to channel noise and interference by generating redundant bits and multi-head attention mechanism, ensuring the stability of data transmission.

[0120] The system can monitor errors in real time during data transmission and adjust transmission strategies through a feedback mechanism. Multi-layer error correction at the receiving end further ensures accurate data decoding.

[0121] Through the generative adversarial network (GAN) and super-resolution model, the system can effectively restore details and clarity after the image is decompressed, compensate for the information lost during the transmission process, and make the final output image clearer and more realistic.

[0122] The system can dynamically adjust the coding and transmission strategies according to changes in channel status and feedback during the transmission process, and continuously optimize the parameter settings of each module through joint training to ensure that the system can maintain optimal performance under different network conditions.

[0123] Example 3 describes in detail the specific method of the 5G broadcast television image communication system based on deep learning, covering the entire process of image preprocessing, compression, transmission, decoding, reconstruction and enhancement. Through the integration of multi-head attention mechanism, generative adversarial network (GAN) and system adaptive adjustment, this method improves the system's transmission efficiency, noise resistance and image quality, making it perform well in the 5G network environment.

[0124] In view of current practical needs, the above-mentioned implementation mode adopted in this application is not limited to the scope of protection. Various changes made within the knowledge scope of technical personnel in this field without departing from the concept of this application still fall within the scope of protection of the present invention.

Claims

1. A 5G broadcasting and television image communication system integrating deep learning, characterized by: It includes image processing module, channel processing module, data transmission module, channel decoding module, image reconstruction and enhancement module, and system optimization and adaptation module; The image processing module is used to preprocess and compress the original image, and the image processing module includes an image preprocessing unit and an image compression unit; The image preprocessing unit is used to perform format conversion and normalization processing on the input image; The image compression unit divides the image into image blocks of fixed size based on Vision Transformer, and extracts global and local features through a multi-head attention mechanism to achieve low-dimensional compressed representation of the image; The channel processing module is used to perform channel coding and state estimation on the compressed image data, and the channel processing module includes a channel coding unit and a channel state estimation unit; The channel coding unit generates redundant bits through a Transformer-based encoder to enhance the noise resistance of the data; The channel state estimation unit estimates the state of the 5G channel in real time based on the Transformer timing prediction model, and dynamically adjusts the coding scheme according to the channel state; The data transmission module is used to perform data transmission through a 5G network. The data transmission module includes a channel transmission unit. The channel transmission unit transmits the encoded image data through the 5G network and monitors errors in data transmission through an error monitoring and feedback unit. The channel decoding module is used to receive and decode image data. The channel decoding module includes a channel decoding unit. The channel decoding unit decodes the received data based on the multi-head attention mechanism of Transformer, and detects and corrects data errors through an error correction unit; The image reconstruction and enhancement module is used to reconstruct and enhance the decoded image data, and the image reconstruction and enhancement module includes an image decompression unit and an image enhancement unit; The image decompression unit decompresses the decoded low-dimensional data using a Transformer model to restore it to a high-dimensional image representation; The image enhancement unit performs detail enhancement and resolution improvement on the reconstructed image based on a generative adversarial network (GAN) and super-resolution technology; The system optimization and adaptation module is used to optimize and dynamically adjust the entire system, and the system optimization and adaptation module includes a joint training optimization unit and an adaptive adjustment unit; The joint training optimization unit performs model optimization on the image processing, channel coding, channel decoding and image reconstruction modules through end-to-end joint training; The adaptive adjustment unit dynamically adjusts the coding and transmission strategies according to the channel state estimation result and the data transmission feedback.

2. The 5G broadcasting and television image communication system integrating deep learning according to claim 1 is characterized in that: The image compression unit is based on the Vision Transformer model, which divides the image into multiple image blocks of fixed sizes, and extracts the global context and local feature information of each image block through a multi-head attention mechanism to perform image compression processing.

3. The 5G broadcasting and television image communication system integrating deep learning according to claim 1 is characterized in that: The channel coding unit generates redundant bits through a Transformer-based multi-head attention mechanism and processes channel transmission data in parallel to enhance the channel's anti-noise capability.

4. The 5G broadcasting and television image communication system integrating deep learning according to claim 1 is characterized in that: The channel state estimation unit predicts the channel state based on the time series Transformer model through historical channel data collected in real time, and dynamically adjusts the coding and transmission strategies according to the prediction results.

5. The 5G broadcasting and television image communication system integrating deep learning according to claim 1, characterized in that: The channel decoding unit identifies error bits in received data through a Transformer model based on a multi-head attention mechanism, and performs cross-layer error correction through an error correction unit.

6. The 5G broadcasting and television image communication system integrating deep learning according to claim 1, characterized in that: The image enhancement unit in the image reconstruction and enhancement module adopts a Transformer-based generative adversarial network (GAN) and a super-resolution model, identifies blurred areas of the image through a multi-head attention mechanism, performs detail compensation and enhancement, and improves the resolution of the image.

7. The 5G broadcasting and television image communication system integrating deep learning according to claim 1, characterized in that: The system optimization and adaptation module jointly optimizes the model parameters of the image processing module, the channel processing module, the channel decoding module and the image reconstruction module through an end-to-end joint training optimization unit to improve the overall communication quality and efficiency of the system.

8. The 5G broadcasting and television image communication system integrating deep learning according to claim 1, characterized in that: The adaptive adjustment unit dynamically adjusts the channel coding and data transmission strategies according to the channel state estimation results and feedback information during the data transmission process to adapt to the ever-changing 5G network communication environment.

9. A communication method applied to a 5G broadcast television image communication system integrating deep learning as described in any one of claims 1 to 8, characterized in that: The communication method includes the following steps: S1: preprocessing the original image, including format conversion and normalization, and dividing the image into multiple image blocks of fixed sizes; S2: Through the Vision Transformer-based image compression unit, the multi-head attention mechanism is used to extract the global context and local features of the image block and compress the image into a low-dimensional representation; S3: Channel coding is performed on the compressed image data through a Transformer-based channel coding unit to generate redundant bits and enhance the data's ability to resist noise; S4: Transmit the encoded image data through the 5G network, monitor the errors in the transmission process in real time, and feedback to the sender; S5: The receiving end decodes the received coded data through the Transformer-based channel decoding unit and uses the multi-head attention mechanism to detect and correct bit errors; S6: The decoded low-dimensional data is restored to a high-dimensional image representation through a Transformer-based image decompression unit; S7: Through the Transformer-based image enhancement unit, use the generative adversarial network (GAN) or super-resolution model to compensate for the details and improve the resolution of the reconstructed image; S8: Through the system optimization and adaptive adjustment module, the channel coding and data transmission strategies are dynamically adjusted according to the channel state estimation results and transmission feedback information to adapt to changes in the 5G network environment.

Images