Virtual data transmission flow control system for MR head-mounted display

Abstract

The invention provides a virtual data transmission flow control system for an MR head-mounted display. The virtual data transmission flow control system comprises a dynamic key area identification module, a self-adaptive flow scheduling module, a low-delay transmission optimization module and a cross-layer collaborative optimization module, wherein the dynamic key region recognition module is used for recognizing a complex key region in an MR environment in real time based on a user fixation point, a head movement track and scene semantic analysis, and the complex key region comprises a high-dynamic interaction object, a visual focus region and a real-time changing virtual scene element; the dynamic key region identification module obtains user behavior data through multi-modal sensor fusion, predicts the change trend of a user concerned region in combination with a deep learning model, and outputs the priority weight of a key region; the adaptive traffic scheduling module comprises a hierarchical coding unit, a dynamic bandwidth allocation unit and a redundancy control unit; and the self-adaptive flow scheduling module is connected with the dynamic key area identification module.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of flow control, and particularly relates to a virtual data transmission flow control system for an MR head-mounted display. BACKGROUND

[0002] At present, the virtual data transmission flow control system is one of the core technologies of the mixed reality (MR) head-mounted display, and its performance directly determines the fluency and immersion of user experience. However, the existing technical solutions still have many deficiencies in real-time performance, bandwidth efficiency, delay control and resource allocation, which restrict the further popularization and application effect improvement of the MR head-mounted display. The current technology mainly relies on traditional network transmission protocols and static resource allocation strategies. These methods are not up to the task when faced with the high dynamics, low delay requirements and large data volume transmission of MR applications.

[0003] The existing flow control mechanism lacks targeted optimization for the characteristics of the MR scene. Virtual data in the MR environment usually includes high-precision 3D models, real-time dynamic lighting effects and multi-modal interaction information. The transmission requirements of these data are significantly different from traditional video streams or game data. However, existing systems often use general flow scheduling algorithms and do not fully consider the spatial correlation, temporal continuity and user interaction intent of virtual data, resulting in low bandwidth utilization or data transmission redundancy.

[0004] Secondly, the delay problem is particularly prominent. MR applications have very low tolerance for end-to-end delay. A delay of more than 20 milliseconds can cause user dizziness or interaction asynchrony. The current system fails to achieve sufficient fine-grained optimization in congestion control, routing selection and data processing pipeline design. In particular, in a wireless network environment, channel fluctuations and interference further exacerbate delay jitter. The existing technology lacks dynamic adaptive capability and cannot stably maintain low-delay transmission under complex network conditions.

[0005] Furthermore, the resource allocation strategy is rigid. MR head-mounted displays usually need to handle multiple data streams simultaneously, such as environment reconstruction, virtual object rendering, user input feedback, etc. The priority and bandwidth requirements of these data streams dynamically change with user behavior and scene changes. However, existing systems mostly use fixed-weight resource allocation methods that cannot real-time perceive changes in application layer requirements, resulting in insufficient resources for high-priority tasks or excessive bandwidth occupation by low-priority tasks

[0006] In addition, the existing virtual data transmission flow control system has defects in energy efficiency management. MR head-mounted displays are extremely sensitive to power consumption as mobile devices. Traditional flow control algorithms often take performance as the single optimization target, ignoring the impact on device battery life. Frequent data retransmission, invalid data preloading and high-complexity calculation strategies exacerbate device energy consumption, limiting user usage time.

[0007] Current technologies lack cross-layer optimization design concepts. Information isolation between the physical, network, and application layers prevents the system from globally coordinating resources. For example, application layer scenario complexity information is not fed back to the network layer to guide dynamic bitrate adjustment, and physical layer channel state information is not used to optimize application layer rendering strategies. This fragmented design prevents the overall system performance from reaching its optimal level. In summary, current virtual data transmission flow control systems have significant shortcomings in adaptability, real-time performance, energy efficiency, and coordination. A novel intelligent flow control framework is urgently needed to achieve efficient, stable, and low-power virtual data transmission by deeply integrating MR application characteristics, dynamic network environments, and user behavior patterns. Summary of the Invention

[0008] This invention proposes a virtual data transmission flow control system for MR headsets. This technical solution solves the problems of low bandwidth utilization, large latency jitter, and unstable transmission quality in key areas when MR headsets transmit virtual data in complex environments, and realizes intelligent flow control in highly dynamic scenarios.

[0009] The technical solution of the present invention is implemented as follows: a virtual data transmission flow control system for MR headsets, comprising a dynamic key area identification module, an adaptive flow scheduling module, a low-latency transmission optimization module, and a cross-layer collaborative optimization module;

[0010] The dynamic key region recognition module is used to identify complex key regions in the MR environment in real time based on user gaze points, head movement trajectories, and scene semantic analysis. Complex key regions include highly dynamic interactive objects, visual focus areas, and virtual scene elements that change in real time. The dynamic key region recognition module acquires user behavior data through multimodal sensor fusion and combines it with a deep learning model to predict the changing trend of the user's attention area, outputting the priority weight of the key region.

[0011] The adaptive traffic scheduling module includes a hierarchical coding unit, a dynamic bandwidth allocation unit, and a redundancy control unit; the adaptive traffic scheduling module is connected to the dynamic key area identification module and is used to dynamically adjust the data transmission strategy according to the priority weight of the key area.

[0012] The low-latency transmission optimization module dynamically adjusts the FEC redundancy based on network jitter and packet loss rate through the forward error correction strategy optimization unit and the data preloading and cache management unit. This reduces retransmission latency while ensuring data reliability. At the same time, it preloads data in potential critical areas based on user behavior prediction and adopts an intelligent cache replacement strategy to prioritize the retention of high-priority data.

[0013] The cross-layer collaborative optimization module uses the channel awareness and code rate adaptive unit to monitor the wireless channel quality in real time and dynamically adjust the coding code rate to match the current network conditions; through the application layer feedback unit, it feeds back the scenario complexity and user interaction intent information to the network layer to guide the real-time adjustment of traffic scheduling strategies.

[0014] Existing virtual data transmission systems for MR headsets face several technical bottlenecks when dealing with complex interactive environments. Traditional systems lack intelligent recognition capabilities for user-focused visual areas, leading to unreasonable allocation of transmission resources and an inability to prioritize transmission quality for critical interactive areas. Existing flow control mechanisms employ static coding strategies, which struggle to adapt to the dynamically changing visual demands of MR scenarios, resulting in decreased display quality or increased latency for highly dynamic objects. Regarding latency control, conventional forward error correction strategies use fixed redundancy, failing to dynamically adjust based on real-time network conditions, impacting transmission efficiency and making it difficult to balance reliability and real-time performance. Network resource allocation algorithms typically operate independently at a single protocol layer, lacking cross-layer collaborative optimization mechanisms, leading to a disconnect between physical layer channel information and application layer service requirements, hindering globally optimal resource scheduling. Cache management strategies are often based on simple historical access patterns, failing to incorporate intelligent preloading based on user behavior prediction, easily causing content loading delays when users quickly switch between modes. Existing systems lack sufficient accuracy in recognizing complex critical areas, making it difficult to accurately distinguish virtual objects with different dynamic characteristics, resulting in weakly targeted transmission strategies. Layered coding technology does not fully consider human visual characteristics, failing to intelligently allocate coding precision for different areas when bandwidth is limited. The lack of a unified information exchange framework in cross-layer optimization schemes makes it difficult to coordinate and adjust parameters across layers in real time. These technical challenges collectively lead to low transmission efficiency and poor user experience in existing systems under complex MR environments.

[0015] In a preferred embodiment, the hierarchical coding unit performs spatial hierarchical coding on the virtual data stream, divides the key area data into a high-precision layer and a low-precision layer, and adopts a differentiated compression strategy, wherein the high-precision layer uses lossless or near-lossless coding, and the low-precision layer uses lossy coding.

[0016] As a preferred implementation, the dynamic bandwidth allocation unit adjusts the transmission bandwidth of each data stream in real time based on network status monitoring results and key area priority weights, prioritizing data transmission in high-priority key areas and dynamically downgrading the transmission quality of low-priority areas when network congestion occurs.

[0017] As a preferred embodiment, the redundant control unit reduces redundant data transmission in non-critical areas by predicting changes in the user's field of vision and scene dynamics, and adopts an incremental update mechanism to transmit only the data of the scene-changing parts.

[0018] As a preferred embodiment, the dynamic key region identification module employs multimodal sensor fusion technology, specifically including: acquiring gaze point coordinate data through an eye-tracking unit, combining it with head pose information collected by an inertial measurement unit, and constructing a three-dimensional visual attention heatmap; the deep learning model adopts a spatiotemporal convolutional neural network architecture, predicts the distribution probability of key regions within the next N frames by analyzing historical interaction trajectories and scene semantic segmentation results, and outputs dynamic priority coefficients, wherein the priority coefficient of highly dynamic interactive objects increases non-linearly with movement speed.

[0019] As a preferred embodiment, the hierarchical coding unit adopts a region-adaptive coding strategy based on visual saliency, specifically including: using a progressive mesh compression algorithm that preserves geometric features for high-precision layers, removing redundant vertices while maintaining the topological structure; using a texture mapping optimization algorithm based on visual perception for low-precision layers, selectively downsampling in the chroma channel by analyzing the color sensitivity characteristics of the human visual system; and using a differentiated compression strategy to encapsulate data from different precision layers into an extensible coding format through a bitstream encapsulator.

[0020] As a preferred embodiment, the working mechanism of the channel awareness and rate adaptation unit includes: constructing a multi-dimensional channel state information matrix, establishing a channel capacity prediction model by monitoring the changing trends of the received signal strength indication and carrier interference noise ratio; the rate adaptation algorithm adopts a two-layer decision mechanism, with the bottom layer calculating the theoretical transmission rate based on the physical layer measurement results, and the upper layer dynamically adjusting the quantization parameters of video encoding in combination with the QoS requirements of the application layer; when channel degradation is detected, the texture resolution of non-critical areas is reduced first to maintain the transmission quality of geometric data.

[0021] The beneficial effects of this invention after adopting the above technical solutions are as follows: This technical solution achieves accurate prediction of user-focused areas through a dynamic key area identification module, enabling the system to intelligently distinguish virtual objects of different importance, providing a scientific basis for subsequent traffic scheduling. The hierarchical coding strategy adopted by the adaptive traffic scheduling module can implement differentiated processing based on regional importance, prioritizing the transmission quality of key visual elements under limited bandwidth conditions. The dynamic bandwidth allocation mechanism realizes elastic management of network resources, ensuring the transmission stability of high-priority data streams. The redundant control unit significantly reduces unnecessary data transmission through incremental updates, improving bandwidth utilization efficiency. The innovative dynamic FEC strategy of the low-latency transmission optimization module effectively balances the contradiction between transmission reliability and real-time performance, intelligently adjusting the error correction intensity according to network conditions. The intelligent preloading mechanism, combined with user behavior prediction, significantly reduces content loading latency during viewpoint switching. The cross-layer collaborative optimization module constructs a unified information interaction framework, achieving seamless integration between physical layer channel states and application layer service requirements, enabling the system to globally optimize transmission strategies. Through the collaborative work of its various modules, the entire system significantly improves data transmission efficiency of the MR headset in complex environments, ensuring real-time rendering quality in key interactive areas and providing users with a smoother and more stable mixed reality experience. The implementation of this technical solution has significantly improved the display accuracy and motion consistency of virtual objects, resulting in more timely interactive responses and effectively reducing potential dizziness for users. The system possesses excellent environmental adaptability, automatically adjusting and optimizing strategies for different network conditions and application scenarios, demonstrating broad application prospects. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a system flowchart of the present invention. Detailed Implementation

[0024] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0025] Example:

[0026] like Figure 1 As shown, a virtual data transmission flow control system for MR headsets includes a dynamic key area identification module, an adaptive flow scheduling module, a low-latency transmission optimization module, and a cross-layer collaborative optimization module.

[0027] The dynamic key region recognition module is used to identify complex key regions in the MR environment in real time based on user gaze points, head movement trajectories, and scene semantic analysis. Complex key regions include highly dynamic interactive objects, visual focus areas, and virtual scene elements that change in real time. The dynamic key region recognition module acquires user behavior data through multimodal sensor fusion and combines it with a deep learning model to predict the changing trend of the user's attention area, outputting the priority weight of the key region.

[0028] The adaptive traffic scheduling module includes a hierarchical coding unit, a dynamic bandwidth allocation unit, and a redundancy control unit; the adaptive traffic scheduling module is connected to the dynamic key area identification module and is used to dynamically adjust the data transmission strategy according to the priority weight of the key area.

[0029] The low-latency transmission optimization module dynamically adjusts the FEC redundancy based on network jitter and packet loss rate through the forward error correction strategy optimization unit and the data preloading and cache management unit. This reduces retransmission latency while ensuring data reliability. At the same time, it preloads data in potential critical areas based on user behavior prediction and adopts an intelligent cache replacement strategy to prioritize the retention of high-priority data.

[0030] The cross-layer collaborative optimization module uses the channel awareness and code rate adaptive unit to monitor the wireless channel quality in real time and dynamically adjust the coding code rate to match the current network conditions; through the application layer feedback unit, it feeds back the scenario complexity and user interaction intent information to the network layer to guide the real-time adjustment of traffic scheduling strategies.

[0031] In MR applications for remote maintenance of industrial equipment, this system optimizes virtual data transmission through multi-module collaboration. When technicians wear MR headsets to inspect turbine units, the dynamic key area recognition module initiates multimodal sensor fusion: an eye tracker captures the coordinates of the gaze point, an inertial measurement unit collects the six-degree-of-freedom head pose in real time, and a scene semantic segmentation engine identifies highly dynamic interactive objects such as rotating blades and vibration sensor instruments. A spatiotemporal convolutional neural network analyzes the behavioral trajectory within a historical 300 milliseconds, predicts that the visual focus will lock onto the pressure gauge reading area in the next five frames, generates a three-dimensional visual attention heatmap, and outputs priority weight coefficients, with high-speed rotating blades assigned the highest weight due to the risk of motion blur.

[0032] The adaptive traffic scheduling module activates a hierarchical coding mechanism after receiving the weighting coefficients: a progressive mesh compression algorithm is used for the blade geometry model, preserving the curved surface topology with 0.1 mm precision; the pressure gauge texture undergoes visual perception optimization coding, maintaining full-resolution sampling of the blue pointer area in the CIE-Lab color space, while performing 4:2:0 chroma sub-sampling on the gray dial background. The dynamic bandwidth allocation unit monitors channel utilization through the 802.11ax space division multiple access interface. When network congestion reaches a threshold, it automatically increases the transmission bandwidth of the blade model to half of the total bandwidth, while reducing the transmission priority of surrounding static pipes. The redundancy control unit compares inter-frame differences based on an environmental mesh constructed using SLAM, transmitting only data packets showing changes in blade rotation angle.

[0033] The low-latency transmission optimization module synchronously starts with dual-channel protection: the FEC strategy engine dynamically configures Reed-Solomon encoding parameters based on the real-time calculated network jitter index, employing a strong error correction scheme for blade geometry data and a lightweight configuration for texture data. The preloading engine uses an LSTM model to predict the hydraulic system that technicians may inspect later, preloading relevant precision assembly models into the local cache, and uses an improved ARC replacement algorithm to lock the blade data at the top of the LRU list.

[0034] The cross-layer collaborative optimization module constructs an end-to-end control closed loop: When the physical layer detects a 6dB drop in the carrier-to-interference ratio via OFDM pilot signals, the channel awareness unit immediately triggers a rate adaptation mechanism, reducing the background environment rate from 20Mbps to 12Mbps. Simultaneously, it sends a QoS degradation command to the rendering engine through the application layer feedback channel. When semantic analysis detects that a user's finger touches the virtual disassembly tool, the application layer feedback unit injects interaction intent metadata into the MAC layer, triggering an emergency bandwidth reservation mechanism to allocate dedicated transmission time slots for the physical collision data of the tool model.

[0035] The hierarchical encoding unit performs spatial hierarchical encoding on the virtual data stream, dividing key area data into high-precision and low-precision layers, and employs a differentiated compression strategy. The high-precision layer uses lossless or near-lossless encoding, while the low-precision layer uses lossy encoding. In heavy machinery virtual disassembly and assembly training, the hierarchical encoding unit implements region-level encoding based on visual saliency. The high-precision layer uses a geometric feature preservation algorithm for core transmission components: optimizing the mesh topology through surface curvature constraints to accurately maintain the contour accuracy of the gear meshing surface while achieving a high compression ratio. When processing the texture of the equipment casing in the low-precision layer, based on the color perception characteristics of the human visual system, full-resolution sampling is maintained for warning sign areas in CIELAB space, while intelligent downsampling of the chroma channel is implemented for monochrome metal areas. An extensible encoding encapsulator integrates data from different precision layers into a unified bitstream. When the trainee focuses on the gearbox structure, the system automatically allocates bandwidth resources to ensure priority transmission of lossless geometric data of the drive shaft.

[0036] The dynamic bandwidth allocation unit adjusts the transmission bandwidth of each data stream in real time based on network status monitoring results and the priority weight of key areas, prioritizing data transmission in high-priority key areas and dynamically degrading the transmission quality of low-priority areas when network congestion occurs. In the remote surgical guidance system, the dynamic bandwidth allocation unit implements resource scheduling according to the importance of anatomical structures. The network status awareness module analyzes the wireless channel quality fluctuation characteristics in real time, and immediately activates a hierarchical protection mechanism when a decrease in channel capacity is detected: prioritizing the maintenance of lossless transmission quality for augmented reality markers of tumor boundaries; initiating a progressive degradation strategy for 3D models of non-critical tissues, reducing data volume by dynamically adjusting encoding quantization parameters; and reserving dedicated transmission channels for high-priority data streams. This mechanism ensures sub-millimeter-level real-time rendering accuracy of vascular and neural structures during critical surgical stages, effectively maintaining visual consistency in multi-person collaborative operations.

[0037] The redundant control unit reduces redundant data transmission in non-critical areas by predicting changes in user field of view and scene dynamics, and employs an incremental update mechanism to transmit only the data related to scene changes. In the 3D inspection scenario of urban pipeline networks, the redundant control unit implements intelligent updates based on scene dynamic analysis. The visual trajectory prediction engine combines head motion vectors and gaze patterns to predict the operator's field of view migration path; the spatial change detector identifies dynamic elements in the environment and enables a reference frame reuse mechanism for static building structures; when a pose change is detected in a valve device, only the local transformation matrix is ​​transmitted to achieve incremental updates. During viewpoint switching, the system dynamically terminates the data flow in non-visible areas, concentrating resources on the critical facility model in the new field of view, significantly reducing network load while maintaining the continuity of the inspection path.

[0038] The dynamic key area identification module employs multimodal sensor fusion technology, specifically including: acquiring gaze point coordinate data through an eye-tracking unit and combining it with head pose information collected by an inertial measurement unit to construct a 3D visual attention heatmap; the deep learning model uses a spatiotemporal convolutional neural network architecture, predicting the probability of key area distribution within the next N frames by analyzing historical interaction trajectories and scene semantic segmentation results, and outputting dynamic priority coefficients, where the priority coefficient of highly dynamic interactive objects increases non-linearly with movement speed. In the virtual review stage of the automotive clay model, the multimodal fusion system captures the design focus through spatiotemporal analysis. Eye-tracking and inertial measurement data construct a 3D attention distribution model, and the spatiotemporal neural network, combined with historical interaction trajectories, predicts the trend of visual focus migration; the dynamic object analyzer identifies highly active components and dynamically increases their priority coefficients based on motion characteristics. When the designer gazes at the grille structure, the system outputs a heatmap to drive the transmission link to allocate lossless coding resources to that area, completing the seamless migration of attention focus and resource reallocation at the instant the gaze suddenly shifts to the wheel hub.

[0039] The hierarchical coding unit adopts a region-adaptive coding strategy based on visual saliency, specifically including: a progressive mesh compression algorithm that preserves geometric features for high-precision layers, removing redundant vertices while maintaining the topology; a texture mapping optimization algorithm based on visual perception for low-precision layers, selectively downsampling in the chroma channel by analyzing the color sensitivity characteristics of the human visual system; and a differentiated compression strategy that encapsulates data from different precision layers into an scalable coding format through a bitstream encapsulator. In large wind turbine maintenance scenarios, the channel sensing unit achieves cross-layer quality control. The multi-dimensional channel state analysis module constructs a capacity prediction model through signal feature extraction; the rate adaptation mechanism adopts a two-layer decision architecture: physical layer measurement results drive the basic rate calculation, and application layer service requirements guide the dynamic adjustment of coding parameters. When electromagnetic interference is detected causing channel degradation, the system automatically reduces the texture quality of the environmental background, centrally ensures the high-precision point cloud transmission of the gearbox core components, and triggers a multi-band redundant transmission mode under extreme conditions to maintain the reliable delivery of critical data.

[0040] The working mechanism of the channel awareness and rate adaptation unit includes: constructing a multi-dimensional channel state information matrix; establishing a channel capacity prediction model by monitoring the changing trends of received signal strength indication and carrier interference noise ratio; the rate adaptation algorithm adopts a two-layer decision mechanism, with the bottom layer calculating the theoretical transmission rate based on physical layer measurement results, and the upper layer dynamically adjusting the quantization parameters of video encoding in conjunction with the QoS requirements of the application layer; when channel degradation is detected, the texture resolution of non-critical areas is reduced first to maintain the transmission quality of geometric data. In the virtual assembly of spacecraft fuel pipelines, the cross-layer optimization module establishes a joint control mechanism. The physical layer channel monitoring unit perceives multipath fading characteristics in real time and sends quality warnings to the application layer; the service parsing engine identifies the key requirements of the current assembly step and marks the collision detection data of flange docking as the highest priority; the network layer dynamically configures the transmission strategy based on cross-layer parameters: reserving low-latency channels for critical services, triggering multi-path parallel transmission, and coordinating the rendering engine to reduce the display accuracy of non-critical tools. When a wrench tightening operation is detected, the system optimizes the contention window parameters in real time to ensure the immediate transmission of physical feedback data.

[0041] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A virtual data transmission flow control system for MR headsets, characterized in that, It includes a dynamic key area identification module, an adaptive traffic scheduling module, a low-latency transmission optimization module, and a cross-layer collaborative optimization module; The dynamic key region recognition module is used to identify complex key regions in the MR environment in real time based on user gaze points, head movement trajectories, and scene semantic analysis. Complex key regions include highly dynamic interactive objects, visual focus areas, and virtual scene elements that change in real time. The dynamic key region recognition module acquires user behavior data through multimodal sensor fusion and combines it with a deep learning model to predict the changing trend of the user's attention area, outputting the priority weight of the key region. The adaptive traffic scheduling module includes a hierarchical coding unit, a dynamic bandwidth allocation unit, and a redundancy control unit; The adaptive traffic scheduling module is connected to the dynamic key area identification module, and is used to dynamically adjust the data transmission strategy according to the priority weight of key areas. The low-latency transmission optimization module dynamically adjusts the FEC redundancy based on network jitter and packet loss rate through the forward error correction strategy optimization unit and the data preloading and cache management unit. This reduces retransmission latency while ensuring data reliability. At the same time, it preloads data in potential critical areas based on user behavior prediction and adopts an intelligent cache replacement strategy to prioritize the retention of high-priority data. The cross-layer collaborative optimization module utilizes the channel awareness and code rate adaptive unit to monitor the wireless channel quality in real time and dynamically adjust the coding code rate to match the current network conditions. The application layer feedback unit feeds back the scenario complexity and user interaction intent information to the network layer, guiding the real-time adjustment of traffic scheduling strategies.

2. The virtual data transmission flow control system for an MR headset as described in claim 1, characterized in that: The layered coding unit performs spatial layered coding on the virtual data stream, dividing the key area data into a high-precision layer and a low-precision layer, and adopts a differentiated compression strategy, wherein the high-precision layer uses lossless or near-lossless coding, and the low-precision layer uses lossy coding.

3. The virtual data transmission flow control system for an MR headset as described in claim 1, characterized in that: The dynamic bandwidth allocation unit adjusts the transmission bandwidth of each data stream in real time based on network status monitoring results and the priority weight of key areas, prioritizing data transmission in high-priority key areas and dynamically downgrading the transmission quality of low-priority areas when network congestion occurs.

4. A virtual data transmission flow control system for an MR headset as described in claim 1, characterized in that: The redundant control unit reduces redundant data transmission in non-critical areas by predicting changes in the user's field of vision and scene dynamics, and adopts an incremental update mechanism to transmit only the data of the scene changes.

5. A virtual data transmission flow control system for an MR headset as described in claim 1, characterized in that: The dynamic key region identification module adopts multimodal sensor fusion technology, specifically including: acquiring gaze point coordinate data through an eye-tracking unit, and combining it with head pose information collected by an inertial measurement unit to construct a three-dimensional visual attention heatmap; the deep learning model adopts a spatiotemporal convolutional neural network architecture, and predicts the distribution probability of key regions in the next N frames by analyzing historical interaction trajectories and scene semantic segmentation results, and outputs dynamic priority coefficients, wherein the priority coefficient of highly dynamic interactive objects increases nonlinearly with movement speed.

6. A virtual data transmission flow control system for an MR headset as described in claim 1, characterized in that: The hierarchical coding unit adopts a region-adaptive coding strategy based on visual saliency, specifically including: a progressive mesh compression algorithm that preserves geometric features for high-precision layers, removing redundant vertices while maintaining the topological structure; a texture mapping optimization algorithm based on visual perception for low-precision layers, selectively downsampling in the chroma channel by analyzing the color sensitivity characteristics of the human visual system; and a differentiated compression strategy that encapsulates data from different precision layers into an extensible coding format through a bitstream encapsulator.

7. A virtual data transmission flow control system for an MR headset as described in claim 1, characterized in that: The working mechanism of the channel awareness and rate adaptation unit includes: constructing a multi-dimensional channel state information matrix, establishing a channel capacity prediction model by monitoring the changing trends of received signal strength indication and carrier interference noise ratio; the rate adaptation algorithm adopts a two-layer decision mechanism, with the bottom layer calculating the theoretical transmission rate based on the physical layer measurement results, and the upper layer dynamically adjusting the quantization parameters of video coding in combination with the QoS requirements of the application layer; when channel degradation is detected, the texture resolution of non-critical areas is reduced first to maintain the transmission quality of geometric data.

Images