Channel state aware based dynamic adaptive transmission method and system for OTT video streaming

Abstract

The application relates to an OTT video flow dynamic self-adaptive transmission method and system based on channel state sensing. The method comprises the following steps: collecting multi-dimensional channel parameters in real time through a physical layer driver, performing protocol analysis, noise filtering and structured processing, constructing a four-dimensional phase vector of a channel state by using polar coordinate transformation and high-dimensional space mapping, and generating a continuous evolution track by sliding window integration. Further, the angular velocity, curvature and residence time length characteristics of the phase track are extracted, a discrete anchor table of the channel phase and the video frame priority is constructed, and the sub-priority binding of the channel mutation area and the key video frame is realized. When the channel state is monitored to be significantly changed and stability verification is passed, a scheduling queue instruction is automatically triggered, the key video frame is written into a highest priority slot, and the video transmission reliability and the anti-mutation capability are improved.

Description

Technical Field

[0001] This invention relates to the field of wireless communication and multimedia transmission technology, and in particular to a dynamic adaptive transmission method and system for OTT video streams based on channel state awareness. Background Technology

[0002] The mainstream technical solutions in the field of dynamic adaptive transmission of over-the-top (OTT) video streams generally rely on the periodic sampling and reporting of channel state information, combined with video content-aware models to achieve adaptive adjustment of scheduling strategies. The scheduling module employs methods such as multi-level weighted scoring, dynamic priority scoring based on queuing models, or reinforcement learning to make real-time decisions based on channel indicators reported by the physical layer. These models primarily use global feature engineering or machine learning back-inference methods, attempting to achieve adaptive matching between scheduling priority and channel state through online iteration of model parameters, dynamic reordering of policy tables, and sliding window normalization of historical indicators. Some systems have begun to explore the integration of multi-dimensional channel awareness and GOP structures, but still rely on complex feedback loops and large-scale model inference. These technologies have been applied in carrier-grade OTT platforms and media distribution networks.

[0003] The limitations of the aforementioned existing technologies include: discretely sampled channel parameters are difficult to fully capture the continuous dynamic evolution of the physical channel, resulting in a lag in the identification of sudden channel transitions; priority reordering and scheduling decisions have high computational overhead and large response delays, and the feedback loop is constrained by multi-protocol layer data transmission and model inference, with response time often exceeding the actual rate of change of channel state; when the channel changes rapidly, the system struggles to complete priority switching within milliseconds, affecting video experience and quality of service assurance; scheduling strategies frequently rely on feature normalization, parameter optimization, and online training, increasing system resource consumption; and traditional scheduling methods are difficult to strictly align with the GOP structure, limiting the space for strategy innovation.

[0004] There is a widespread need in this field for: accurate characterization of the continuous evolution of channel physical states and real-time capture of dynamic transitions; rapid mapping of channel states to key frame priority adjustments in the GOP structure at the scheduling layer with extremely low computational overhead, eliminating the fundamental problem of feedback loop lag; and leveraging the periodicity of frame structures to encode channel dynamics into scheduling geometric attributes, achieving hardware-level instruction-based priority locking without delay. There is an urgent need for an OTT video stream priority scheduling scheme that balances the continuity of channel state evolution, real-time scheduling decisions, and lightweight system execution, overcoming obstacles such as discrete sampling, high resource consumption, or slow response in existing technologies, and providing a technical foundation for channel-aware scheduling for video streaming services. Summary of the Invention

[0005] This application provides a channel state-aware OTT video stream dynamic adaptive transmission method and system, aiming to solve how to achieve real-time and accurate scheduling of video keyframes with extremely low computational overhead when the channel state changes rapidly.

[0006] The channel state-aware OTT video stream dynamic adaptive transmission method and system provided in this application specifically include:

[0007] Obtain a multi-dimensional channel parameter stream, which includes at least four of the following: MCS index, channel occupancy rate, interference level, and RSSI fluctuation entropy value. The multidimensional channel parameter stream is subjected to polar coordinate transformation to map the multidimensional channel parameter stream to the coordinate axes of a four-dimensional unit hypersphere, thereby generating a phase vector; Geometric features are extracted from the time series of the phase vector on a four-dimensional unit hypersphere to calculate a phase evolution feature set including angular velocity, curvature, and local dwell time. Cluster analysis is performed on the trajectory on the four-dimensional unit hypersphere using the phase evolution feature set to extract typical transition phase regions. The fixed offset position of each frame is obtained by parsing the GOP structure of the video stream. The typical transition phase regions are associated and matched with the fixed offset positions, and a phase-frame anchoring table is generated based on the matching results. Based on the phase evolution feature set, it is determined whether the phase vector has entered the transition phase region. If a sudden increase in curvature is detected and the angular velocity exceeds a preset threshold, the phase stability filter is triggered to perform continuous residence time verification on the transition phase region. If the phase stability filter confirms that the phase vector has continuously resided in the transition phase region for more than three sampling periods, it queries the corresponding anchor frame position identifier according to the phase-frame anchoring table and generates a scheduling queue instruction. According to the scheduling queue instruction, the video frame marked as the anchor frame position identifier is written into the highest priority slot of the scheduling queue to complete the priority anchoring operation.

[0008] The channel-state-aware OTT video stream dynamic adaptive transmission method and system provided in this application have the following advantages: By upgrading channel state evolution from discrete index sampling to continuous phase space geometric modeling, and mapping multidimensional parameters to a four-dimensional unit hypersphere to construct a phase vector, the trajectory is characterized by differential features such as angular velocity, curvature, and dwell time, thereby improving the sensitivity to channel anomalies and the ability to make pre-decision decisions. Simultaneously, by using a pre-set phase-frame anchoring table and a phase stability filter, after activation is confirmed in the transition phase region, the corresponding keyframe position in the GOP structure is located directly based on the offline mapping relationship and marked and written to the highest priority slot in the scheduling queue. The entire operation involves only table lookup and bitmasking, and with dwell time verification to suppress the risk of false triggering, deterministic scheduling guarantees without real-time computation or model dependence are achieved. These methods collaboratively construct a closed-loop adaptive scheduling system, transforming response delay control into a deterministic anchoring task in the phase space that can be predefined, statically bound, and quickly matched. This avoids the uncertainty of algorithm convergence and runtime resource contention problems in traditional scheduling. The system does not require the deployment of deep learning models, the maintenance of complex intermediate states, or frequent access to storage resources, exhibiting good deployability and scalability. Attached Figure Description

[0009] Figure 1 This is the main flowchart of a channel state-aware OTT video stream dynamic adaptive transmission method and system.

[0010] Figure 2 This is a sub-flowchart of a channel state-aware OTT video stream dynamic adaptive transmission method and system.

[0011] Figure 3 This is another sub-flowchart of the OTT video stream dynamic adaptive transmission method and system based on channel state awareness. Detailed Implementation

[0012] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0013] The following disclosure provides many different embodiments or examples for implementing different structures of the invention. To simplify the disclosure, specific examples of components and arrangements are described below. Of course, these are merely examples and are not intended to limit the invention. Furthermore, reference numerals and / or letters may be repeated in different examples; such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.

[0014] like Figure 1As shown, this application provides a channel state-aware OTT video stream dynamic adaptive transmission method and system, specifically including: S1: Obtain the multi-dimensional channel parameter stream output by the physical layer driver. The multi-dimensional channel parameter stream includes the MCS index, channel occupancy rate, interference level, and RSSI fluctuation entropy value. The multi-dimensional channel parameter stream is used as the input data source for constructing the channel phase space trajectory.

[0015] S2: Based on the multidimensional channel parameter stream, polar coordinate transformation processing is performed to map the MCS index, channel occupancy rate, interference level and RSSI fluctuation entropy value to the horizontal axis, vertical axis, third axis and fourth axis of a four-dimensional unit hypersphere, respectively, to generate a phase vector characterizing the continuous evolution of the channel state.

[0016] S3: Geometric features are extracted from the time series of the phase vector on the four-dimensional unit hypersphere, and a phase evolution feature set including angular velocity, curvature and local dwell time is calculated to quantify the stability and abrupt change trend of the channel state.

[0017] S4: Construct a phase-frame anchoring table using the phase evolution feature set. The phase-frame anchoring table establishes a deterministic mapping relationship between channel phase attributes and video frame priority by offline binding of typical transition phase regions on the four-dimensional unit hypersphere with fixed offset positions in the video stream GOP structure.

[0018] S5: Based on the phase evolution feature set, determine whether the phase vector has entered the transition phase region. If a sudden increase in curvature is detected and the angular velocity exceeds the preset threshold, trigger the phase stability filter to check the continuous dwell time of the transition phase region.

[0019] S6: If the phase stability filter confirms that the phase vector has continuously resided in the transition phase region for more than three sampling periods, then it queries the corresponding anchor frame position identifier according to the phase-frame anchor table and generates a scheduling queue instruction containing bit mask markers.

[0020] S7: Write the video frame marked as the anchor frame position identifier into the highest priority slot of the hardware scheduling queue according to the scheduling queue instruction, and complete the frame-level priority anchoring operation based on channel state evolution phase locking.

[0021] Step S1: Obtain the multi-dimensional channel parameter stream output by the physical layer driver. The multi-dimensional channel parameter stream includes the MCS index, channel occupancy rate, interference level, and RSSI fluctuation entropy value. This multi-dimensional channel parameter stream is used as the input data source for constructing the channel phase space trajectory. Specifically, it includes: S1.1: Real-time monitoring and capture processing of the raw signal state data output by the physical layer driver interface to generate an initial channel state sampling sequence containing timestamps. This initial channel state sampling sequence serves as the basic input for subsequent parameter extraction.

[0022] Sub-step S1.1 is located at the data acquisition front end of the transmission strategy decision module. It is responsible for capturing raw signal state data in real time from the physical layer driver interface, constructing an initial channel state sampling sequence with high-precision timestamps, and providing a timing reference for subsequent multi-dimensional parameter analysis.

[0023] Configure the interrupt callback function of the kernel-mode network driver, listen for raw Channel State Information (CSI) packets reported by the physical layer Media Access Control (MAC) sublayer, and establish a low-latency data capture channel.

[0024] Perform byte alignment verification on the received raw binary data stream, extract the hardware counter value in the frame header as the base time reference, and ensure that the data acquisition is synchronized with the underlying hardware clock.

[0025] The system uses a high-precision system clock source to obtain the current absolute timestamp, with precision controlled at the microsecond level. The absolute timestamp is then linearly mapped and calibrated with the hardware counter value to eliminate time errors introduced by software scheduling jitter.

[0026] The calibrated timestamp is appended to the header of each CSI data packet to form a raw signal state record unit containing precise timing information.

[0027] A circular buffer mechanism is used to temporarily store timestamped record units. A buffer overflow protection strategy is set up so that when the buffer occupancy rate exceeds a preset threshold, the oldest non-critical sampling point is automatically discarded to ensure the real-time performance of the latest channel data.

[0028] Perform batch read operations on the data in the circular buffer, reassemble the data stream in ascending order of timestamps, and generate a continuous and ordered initial channel state sampling sequence.

[0029] Through the aforementioned real-time monitoring, clock calibration, and ordered reassembly processing, the discretely reported signal states from the physical layer are transformed into an initial channel state sampling sequence with a unified timing reference, achieving strict alignment of channel data in the time dimension and providing a high-fidelity input object for subsequent multi-protocol parsing.

[0030] For example, in a 5G NR band 3.5GHz scenario, the physical layer driver is configured to report CSI data at a period of 2ms. The system initializes the circular buffer size to 1024 entries, each containing 64 bytes of raw data and an 8-byte timestamp. When a change in the MCS index is detected, the driver triggers an interrupt, and the kernel callback function completes data acquisition within 5 microseconds. The TSC register is called to obtain the nanosecond-level timestamp, which is mapped with the system frame number issued by the base station, and the absolute time deviation is calculated to be less than 100 nanoseconds. If the buffer is full, the earliest 32 sampling points are discarded. In the final output initial channel state sampling sequence, the standard deviation of the time interval between adjacent sampling points is less than 0.1ms, ensuring the temporal continuity of phase space trajectory construction and significantly improving the time resolution of subsequent transition phase region determination.

[0031] S1.2: Perform multi-protocol parsing operation based on the initial channel state sampling sequence, and use the preset physical layer signaling mapping table to separate the original value of modulation and coding strategy index, spectrum resource occupancy count value, co-channel interference intensity measurement value and received signal strength indication fluctuation sample set to generate a multi-dimensional channel parameter set to be cleaned.

[0032] Receive an initial channel state sampling sequence containing timestamps, which serves as the base input for multiprotocol parsing.

[0033] The pre-defined physical layer signaling mapping table is invoked. This mapping table defines the signaling field offset and bit width parameters of the MAC sublayer control frame under different communication protocols.

[0034] A byte-level scan is performed on the binary data stream in the initial channel state sampling sequence, and the bit range where the modulation and coding strategy (MCS) index is located is located according to the mapping table.

[0035] Extract the original value of the MCS index, which directly reflects the theoretical maximum transmission rate and noise immunity level of the current channel.

[0036] Continue parsing the spectrum resource allocation field in the data stream, read the occupancy count value of resource blocks (RBs) or subcarriers, and quantify the load pressure of the current frequency band.

[0037] Retrieve the co-channel interference measurement report field to extract the received signal-to-interference-noise ratio (SINR) or dedicated interference intensity measurement value to characterize the degree of deterioration of the external electromagnetic environment.

[0038] Extract a historical fluctuation sample set of Received Signal Strength Indication (RSSI), which contains signal amplitude records for the most recent N sampling periods.

[0039] The extracted original MCS index values, spectrum resource occupancy count values ​​(i.e., channel occupancy rate), co-channel interference intensity measurement values ​​(i.e., interference level), and RSSI fluctuation sample sets are encapsulated into structured data units.

[0040] Through multi-protocol parsing operations, unstructured underlying driver data is separated into a set of multi-dimensional channel parameters to be cleaned with clear physical meaning, realizing the accurate transformation from the original bit stream to semantic channel features, and providing standardized input for subsequent entropy calculation and timing alignment.

[0041] For example, in 5G NR TDD mode, the physical layer signaling mapping table defines the MCS index as a conventional bit in the downlink control information. The system performs a bitmasking operation on the initial sampling sequence to extract the MCS index value of 24, corresponding to 64QAM modulation and a code rate of 0.92. Parsing the spectrum resource field reveals that the number of occupied resource blocks is 48, accounting for 60% of the available bandwidth. Reading the interference measurement field yields a SINR value of 12dB. The RSSI sample set within the last 10ms is extracted, containing 20 amplitude values ​​ranging from -75dBm to -68dBm. These parameters are encapsulated into a multi-dimensional channel parameter set to be cleaned, where MCS, occupancy rate, and interference value are scalars, and RSSI is a vector. This process takes less than 2 microseconds, ensuring real-time parameter extraction, eliminating data heterogeneity caused by protocol differences, and enabling subsequent modules to perform phase modeling based on a unified semantic space.

[0042] S1.3: Perform sliding window entropy calculation on the received signal strength indication fluctuation sample set in the multidimensional channel parameter set to be cleaned, and use the Shannon entropy algorithm to quantify the randomness of the signal amplitude in order to generate RSSI fluctuation entropy parameters that characterize channel stability.

[0043] Receive an RSSI fluctuation sample set from the multidimensional channel parameter set to be cleaned. This sample set contains continuously received signal strength indication sample values ​​within a preset sliding time window.

[0044] Perform probability distribution statistics on the RSSI fluctuation sample set, construct a discrete amplitude interval histogram, calculate the frequency of sampling points in each amplitude interval, and generate a normalized probability distribution sequence.

[0045] Based on the Shannon entropy algorithm, information entropy quantization is performed on the normalized probability distribution sequence to calculate the RSSI fluctuation entropy value.

[0046] The calculated raw entropy value is subjected to linear mapping to transform it into a standard numerical range of 0 to 1, eliminating the influence of dimensions and generating a standardized channel stability metric.

[0047] By using a sliding window entropy calculation and standardization process, the RSSI fluctuation sample set from the previous step is transformed into RSSI fluctuation entropy parameters that characterize the strength of channel randomness, thereby achieving a quantitative description of channel stability characteristics and providing a fourth dimension of input data for subsequent phase space mapping.

[0048] For example, in a Wi-Fi 6 band 5GHz scenario, a sliding window length of 20ms is configured, containing 40 RSSI sampling points. The RSSI amplitude range of -90dBm to -30dBm is divided into 12 equally wide intervals. When the channel is in a static, interference-free state, the RSSI sampling points are concentrated in a single interval near -45dBm with a probability close to 1, while the probability of occurrence in other amplitude intervals is close to 0. The calculated entropy value H is approximately 0.05 bits, and the normalized entropy value after mapping is 0.025, characterizing a highly stable channel. When there is co-channel interference causing rapid signal fluctuations, the sampling points are evenly distributed across multiple intervals, with the probability of occurrence in each amplitude interval being approximately 0.083. The calculated entropy value is approximately 3.58 bits, and the normalized entropy value after mapping is 0.89, characterizing severe channel fluctuations. The system outputs this normalized entropy value in real time to step S1.4 as an independent stability criterion beyond interference levels, significantly improving the ability of the phase space to distinguish between small amplitude fluctuations and macroscopic jumps.

[0049] S1.4: Based on the original value of the modulation and coding strategy index, the spectrum resource occupancy count value, the co-channel interference intensity measurement value, and the generated RSSI fluctuation entropy value parameter, perform timing alignment and outlier removal processing, use linear interpolation algorithm to fill in missing data points and filter noise data exceeding the physical threshold to generate a standardized multidimensional channel parameter stream.

[0050] The system receives a set of multidimensional channel parameters to be cleaned, including the original value of the modulation and coding strategy index, the spectrum resource occupancy count value, the co-channel interference intensity measurement value, and the RSSI fluctuation entropy value. Each parameter has an independent timestamp.

[0051] Extract the timestamp sequence of each dimension parameter, calculate the time interval difference between adjacent sampling points, and identify missing data segments with time intervals exceeding a preset threshold.

[0052] For the identified missing data segments, locate the nearest valid sampling points before and after the missing point, and obtain their corresponding parameter values ​​as interpolation boundary conditions.

[0053] By applying a linear interpolation algorithm, the parameter estimates for intermediate missing moments are calculated based on the relative position weights of the missing points on the time axis.

[0054] The calculated interpolation results are filled into the corresponding missing positions in the original data stream to restore the continuity of the time series.

[0055] Load the threshold table of legal value ranges for each physical layer parameter, including the maximum and minimum values ​​of the MCS index, the 0-100% range of channel occupancy, the dynamic range of interference intensity, and the non-negative constraint of entropy value.

[0056] Traverse the time-aligned multidimensional parameter stream and compare the current sampled value with the physical threshold boundary of the corresponding dimension point by point.

[0057] All abnormal data points that exceed the physical threshold range are marked and determined to be invalid data caused by sensor noise or transmission errors.

[0058] A forward-filling strategy is adopted, which replaces the current outlier with the valid and valid value of the previous time step of the outlier to maintain the stability of the data flow.

[0059] Perform a dimensional consistency check on the data stream after anomaly removal to ensure that the four dimensional parameters meet the input requirements for subsequent polar coordinate mapping in terms of numerical magnitude.

[0060] By employing the aforementioned time alignment, linear interpolation filling, and threshold filtering methods, discrete and noisy multidimensional channel parameters are transformed into a time-continuous and numerically valid standardized multidimensional channel parameter stream, thereby achieving the high-fidelity data foundation required for phase space trajectory construction.

[0061] S1.5: Perform structured encapsulation processing on the standardized multidimensional channel parameter stream, and package the MCS index, channel occupancy rate, interference level and RSSI fluctuation entropy value into continuous data blocks according to the preset four-dimensional vector data format to generate a multidimensional channel parameter stream that can be directly called by the phase space mapping unit.

[0062] Step S2: Perform polar coordinate transformation based on the multidimensional channel parameter stream, mapping the MCS index, channel occupancy rate, interference level, and RSSI fluctuation entropy value to the horizontal, vertical, third, and fourth axes of a four-dimensional unit hypersphere, respectively, to generate a phase vector characterizing the continuous evolution of the channel state. Specifically, this includes: S2.1: Obtain the MCS index data and channel occupancy data from the multidimensional channel parameter stream. Based on the two-dimensional plane mapping algorithm from Cartesian coordinate system to polar coordinate system, perform joint vector synthesis processing on the MCS index data and channel occupancy data to generate the first quadrant basic phase component that characterizes the channel transmission capability and load status.

[0063] MCS index and channel occupancy data are extracted from the standardized multidimensional channel parameter stream and used as input variables for constructing the basic phase components in the first quadrant.

[0064] A linear mapping transformation is performed on the MCS index to convert its discrete integer values ​​into normalized amplitude components, thus eliminating the dimensional differences between different modulation orders.

[0065] A linear mapping transformation is performed on the MCS index to convert its discrete integer values ​​into normalized amplitude components, eliminating the dimensional differences between different modulation orders; at the same time, the channel occupancy data is scaled to directly convert its percentage values ​​into normalized load components in the range of 0 to 1.

[0066] The normalized amplitude of the MCS is set as the horizontal axis coordinate of a two-dimensional Cartesian coordinate system, and the normalized load component is set as the vertical axis coordinate.

[0067] Based on the transformation logic from Cartesian coordinates to polar coordinates, the phase angle of the coordinate point on the two-dimensional plane is calculated to characterize the coupling relationship between channel transmission capability and load status.

[0068] The fundamental phase angle is calculated using the following formula:

[0069] in, Based on the phase angle, The normalized amplitude of MCS. To normalize the load components, the range of the arctan function is limited to the first quadrant to match the physical meaning.

[0070] Calculate the Euclidean distance from the coordinate point to the origin, and generate the radial modulus characterizing the overall channel performance.

[0071] By combining the phase angle and the radial magnitude, a first-quadrant basic phase component vector containing both directional and intensity information is constructed.

[0072] By using the above-mentioned joint vector synthesis processing method, the discrete MCS index and channel occupancy rate are transformed into continuous geometric phase attributes, realizing the structured representation of the channel state in the two-dimensional subspace, and providing basic components with clear physical meaning for subsequent four-dimensional hyperspherical mapping.

[0073] For example, in 5GN RTD mode, the MCS index ranges from 0 to 27. When the acquired MCS index is 18 and the channel occupancy rate is 60%, the minimum MCS index value supported by the system is 0, and the maximum MCS index value supported by the system is 27. The calculated normalized MCS amplitude is (18-0) / (27-0), which is approximately 0.667. load We directly take the value as 0.6. Substituting this into the phase angle formula, φ equals arctan(0.6 / 0.667), and the calculated result is approximately 0.732 radians. The radial modulus is calculated as sqrt(0.667). 2 +0.6 2The result is approximately 0.897. The generated basic phase component vector in the first quadrant is (0.732, 0.897). This vector accurately reflects the current channel state as being in a medium-to-high modulation order and with a medium load. Compared to the original discrete values, this continuous phase component can more smoothly reflect the small fluctuations in channel quality, significantly improving the resolution of subsequent phase trajectory tracking.

[0074] S2.2: Receive interference level data and RSSI fluctuation entropy value data from the multidimensional channel parameter stream, and use high-dimensional space orthogonal extension transformation technology to map the interference level data into a third-dimensional axial vector and the RSSI fluctuation entropy value data into a fourth-dimensional axial vector to construct orthogonal disturbance phase components that characterize the channel environmental noise characteristics and signal stability.

[0075] The interference level data and RSSI fluctuation entropy value data received from the multidimensional channel parameter stream are used as the input source for constructing the orthogonal disturbance phase component.

[0076] Perform a linear mapping transformation on the interference level data to compress its numerical range to the [-1,1] interval, generating a third-dimensional axial scalar that characterizes the intensity of channel environmental noise.

[0077] Using a three-dimensional spatial rotation matrix algorithm, the third-dimensional axial scalar is projected onto the Z-axis direction of the four-dimensional hypersphere to construct a noise interference vector component perpendicular to the MCS-occupancy plane.

[0078] Logarithmic normalization is performed on the RSSI fluctuation entropy data to eliminate the dimensional differences in the randomness of signal amplitude and generate a fourth-dimensional axial scalar that characterizes channel stability.

[0079] By applying the Gram-Schmidt orthogonalization process, we ensure that the W-axis vector generated by the fourth-dimensional axial scalar maintains a strict orthogonal relationship with the vectors of the first three dimensions, thus eliminating coupling interference between multidimensional parameters.

[0080] The noise interference vector component and the signal stability vector component are superimposed to construct an orthogonal disturbance phase component that characterizes the noise characteristics of the channel environment and the signal stability.

[0081] By using a high-dimensional space orthogonal extension transformation, the discrete interference and entropy parameters from the previous step are transformed into a four-dimensional disturbance vector with geometric orthogonal properties, thereby achieving the decoupled representation of the channel's non-ideal state in the phase space.

[0082] S2.3: Based on the basic phase component and the orthogonal perturbation phase component in the first quadrant, perform a four-dimensional hyperspherical radius normalization constraint operation to forcibly project the axial vectors of the four dimensions onto the unit hyperspherical surface to eliminate dimensional differences and generate a normalized four-dimensional channel state vector with a unified geometric metric standard.

[0083] Sub-step S2.3 is located in the core phase space mapping link of the transmission strategy decision module. It is responsible for projecting the basic phase components and orthogonal disturbance phase components with different physical dimensions generated in the previous steps onto the four-dimensional unit hyperspherical surface, eliminating the dimensional differences between MCS index, channel occupancy rate, interference level and RSSI fluctuation entropy value, and generating a normalized four-dimensional channel state vector with a unified geometric metric standard.

[0084] Receive the fundamental phase component vector V from the first quadrant of S2.1 base With the orthogonal perturbation phase component vector V from S2.2 pert The two are then superimposed in four-dimensional Euclidean space to construct the original four-dimensional composite vector V containing all channel state information. raw .

[0085] Calculate the original four-dimensional composite vector V raw The Euclidean norm, i.e. the vector magnitude, is used as a scaling factor for normalization. This magnitude reflects the comprehensive energy intensity of the channel state at the current moment in the multidimensional parameter space.

[0086] The L2 norm normalization algorithm is used to transform the original four-dimensional composite vector V. raw Divide each dimension component by the calculated vector magnitude and perform scalar division to ensure that the transformed vector length strictly converges to 1.

[0087] The following formula is used to perform the normalization constraint calculation for the radius of a four-dimensional hypersphere:

[0088] in, Normalized four-dimensional channel state vector, This is the original four-dimensional composite vector. for The L2 norm.

[0089] The normalized vector components are truncated using floating-point precision, retaining six significant digits after the decimal point. This eliminates the small cumulative errors introduced by hardware floating-point operations, ensuring that the vector endpoint falls precisely on the unit hypersphere surface.

[0090] The processed, normalized four-dimensional channel state vector is encapsulated into a fixed-length data structure, marked with the current timestamp, and used as the standard input for subsequent calculations of geodesic distance and angular velocity.

[0091] By using the above-mentioned vector superposition, modulus calculation and L2 norm normalization methods, the multidimensional channel parameters with different dimensions in the previous step are transformed into a normalized four-dimensional channel state vector with a unified geometric metric standard, realizing the unbiased mapping of the channel state on the four-dimensional unit hypersphere, laying the data foundation for subsequent phase evolution analysis based on geometric features.

[0092] For example, in a Wi-Fi 6 high-frequency band communication scenario, the first quadrant fundamental phase component vector output by S2.1 is [0.8, 0.6, 0, 0], and the orthogonal perturbation phase component vector output by S2.2 is [0, 0, 0.5, 0.5]. The system performs vector superposition to obtain the original four-dimensional composite vector V. raw =[0.8,0.6,0.5,0.5]. Calculate its L2 norm ||V. raw ||=sqrt(0.8 2 +0.6 2 +0.5 2 +0.5 2 = sqrt(0.64 + 0.36 + 0.25 + 0.25) = sqrt(1.5) ≈ 1.2247. Perform normalization, V norm =[0.8 / 1.2247,0.6 / 1.2247,0.5 / 1.2247,0.5 / 1.2247]≈[0.6532,0.4899,0.4082,0.4082]. Verify the modulus as sqrt(0.6532). 2 +0.4899 2 +0.4082 2 +0.4082 2 )≈1.0000. The final output normalized vector eliminates the dimensional difference between MCS and interference intensity, so that subsequent curvature calculation is no longer dominated by the numerical range of a single parameter, which significantly improves the geometric consistency of phase trajectory analysis.

[0093] S2.4: Apply a time-series sliding window integration mechanism to the normalized four-dimensional channel state vector, and combine it with the hyperspherical geodesic distance calculation logic to connect the discrete normalized four-dimensional channel state vectors into a continuous time evolution path, so as to generate a phase vector that can reflect the smooth transition and abrupt change trend of the channel state in real time.

[0094] The system receives a normalized four-dimensional channel state vector sequence after S2.3 normalization. This sequence contains discrete sampling points that are strictly ordered by timestamps, which serve as the basic input data for constructing a continuous phase trajectory.

[0095] Initialize a fixed-length time sliding window, with the window size set to the number of sampling points covering at least one complete GOP period, to cache historical phase vectors to support local integration operations.

[0096] The hyperspherical geodesic distance is calculated for adjacent normalized four-dimensional channel state vectors within the sliding window. The shortest arc length between two points is solved using the inverse cosine function, and the geometric displacement of the channel state on a unit hypersphere is quantified.

[0097] The geodesic distance between the vectors at time t and time t-1 is calculated using the following formula. :

[0098] in, The normalized four-dimensional vector at the current moment. The normalized four-dimensional vector from the previous time step. This represents the vector dot product operation, where arccos is the inverse cosine function, and the result ranges from [0, π].

[0099] Based on the calculated geodesic distance sequence, cumulative path integration is performed to accumulate the discrete distance scalars and generate the total evolution path length L from the window start point to the current point. t .

[0100] For the total evolution path length L t By performing time differential approximation and combining it with the sampling time interval Δt, the phase evolution rate per unit time is calculated to form preliminary phase vector components.

[0101] An exponentially weighted moving average filter is introduced to smooth the evolution rate, suppress instantaneous jitter caused by physical layer measurement noise, and preserve the macroscopic trend characteristics of channel state changes.

[0102] The smoothed evolution rate vector is spatiotemporally concatenated with the original normalized vector to construct a composite phase vector containing both positional and motion trend information.

[0103] By using the above geodesic distance integration and smoothing filtering methods, the discrete normalized vector from the previous step is transformed into a phase vector with continuity and directionality, realizing the morphological transformation of the channel state from a static point to a dynamic trajectory, and providing high-fidelity input for subsequent curvature and angular velocity feature extraction.

[0104] For example, in a 5GNRTDD configuration, the sliding window length is set to 20 sampling points, corresponding to a time span of 40ms. Assume the normalized vector v at time t... t Given the vector v at time t-1, with a range of [0.6, 0.4, 0.3, 0.5]. t-1 The range is [0.58, 0.42, 0.31, 0.49]. The calculated dot product is 0.9982. Substituting this into the formula, we obtain the geodesic distance d. tThe instantaneous evolution rate is approximately 0.06 radians. If the sampling interval Δt is 2 ms, the instantaneous evolution rate is 30 radians / second. After passing through an exponentially weighted moving average filter with a weighting coefficient of 0.9, the output is a smoothed evolution rate. The final generated phase vector not only contains the current channel coordinates on the hypersphere but also embeds the cumulative evolution path information over the past 40 ms, significantly enhancing the ability of the phase trajectory to characterize the channel's abrupt change trend, thus greatly reducing the false alarm rate of subsequent transition detection.

[0105] S2.5: Output the phase vector to the phase evolution feature extraction unit, and use the phase vector as the sole input source for subsequent calculation of angular velocity, curvature and local dwell time, so as to complete the final morphological transformation from discrete physical layer indicators to continuous geometric phase space trajectory.

[0106] Step S3: Extract geometric features from the time series of the phase vector on the four-dimensional unit hypersphere, and calculate the phase evolution feature set including angular velocity, curvature, and local dwell time to quantify the stability and abrupt change trends of the channel state. Specifically, this includes: S3.1: Perform differential operation on the continuous sampling points of the phase vector within the preset sliding time window to obtain the phase displacement vector sequence between adjacent sampling points, and use the phase displacement vector sequence as the input basic data for calculating the instantaneous motion state.

[0107] Receive the phase vector sequence output from step S2, which consists of four-dimensional unit hypersphere coordinate points marked with consecutive timestamps, as the raw data source for phase displacement calculation.

[0108] The size of the preset sliding time window is set to N sampling periods, where the value of N is determined according to the channel coherence time, and is usually set to 5 to 10 sampling points to balance noise suppression and real-time response.

[0109] Extract the phase vector V(t) at the current time t and the phase vector V(t-1) at the previous time t-1 from the phase vector sequence, ensuring that the two vectors have the same geometric metric reference on the four-dimensional hypersphere.

[0110] Perform four-dimensional spatial vector difference operation to calculate the phase displacement vector ΔV between adjacent sampling points. Perform modulus verification on the calculated phase displacement vector ΔV to remove abnormal displacements of zero or minimum values ​​caused by quantization error or transmission jitter, so as to ensure the data validity of subsequent feature extraction.

[0111] The verified phase displacement vector sequence is stored in a circular buffer to form a displacement trajectory data stream with time continuity, which serves as the direct input basis for calculating the instantaneous angular velocity in step S3.2.

[0112] Through the above differential operation, the static phase coordinate sequence generated in the previous step is transformed into a dynamic phase displacement vector sequence, realizing the preliminary quantization of the channel state evolution speed and direction, and providing high-precision kinematic feature data for subsequent identification of phase transition boundaries.

[0113] For example, suppose the sliding window size N is 6 and the sampling period is 2ms. At t=100ms, the phase vector V(100)=[0.8,0.4,0.3,0.2] is obtained, and at t=98ms, V(98)=[0.75,0.45,0.35,0.25] is obtained. The phase displacement vector is obtained by performing a difference operation as [0.05,-0.05,-0.05,-0.05]. The Euclidean norm of this displacement vector is calculated as the displacement amplitude. If the amplitude is greater than the preset noise threshold of 0.01, it is determined to be a valid displacement and stored in the buffer. This process is implemented in the Field-Programmable Gate Array (FPGA) hardware logic through a parallel subtractor array. The single operation latency is less than 10ns, which significantly improves the real-time performance of phase trajectory tracking and ensures the accuracy of displacement feature capture in the case of rapid channel fading.

[0114] S3.2: Based on the phase displacement vector sequence, perform vector magnitude normalization and time interval ratio calculation to obtain the instantaneous angular velocity scalar that characterizes the rate of change of the channel state on the four-dimensional unit hypersphere, and use the instantaneous angular velocity scalar as the first feature parameter for evaluating the channel change trend.

[0115] Receive the phase displacement vector sequence output from step S3.1, which contains the geometric position difference vectors of adjacent sampling points on a four-dimensional unit hypersphere. Perform Euclidean norm calculation on each displacement vector in the phase displacement vector sequence to obtain the geometric length modulus of the vector in four-dimensional space, which serves as a fundamental physical quantity characterizing the amplitude of channel state changes.

[0116] Obtain the timestamp difference between the current sampling point and the previous sampling point, construct a discrete time interval scalar sequence, and use it as a time scale reference for quantizing the phase evolution process.

[0117] By employing the ratio logic of vector magnitude to time interval, the geometric length magnitude is divided by the corresponding time interval scalar to calculate the rate of change of phase arc length per unit time, i.e., the instantaneous angular velocity scalar.

[0118] The instantaneous angular velocity scalar is calculated using the following formula:

[0119] in, For instantaneous angular velocity scalar, for The Euclidean norm of the phase displacement vector at time step This represents the time interval between adjacent sampling points.

[0120] The calculated instantaneous angular velocity scalar is subjected to moving average filtering to eliminate minor jitter caused by physical layer measurement noise and retain the effective component that reflects the macroscopic trend changes of the channel.

[0121] The filtered instantaneous angular velocity scalar is labeled as the first-dimensional feature parameter. This parameter quantifies the speed of the channel state's movement in the phase space and is directly related to the severity of sudden changes in the channel environment.

[0122] By using the above-mentioned vector magnitude normalization and time interval ratio calculation processing method, the phase displacement vector sequence of the previous step is transformed into an instantaneous angular velocity scalar that characterizes the rate of change of the channel state, realizing the first dimension of quantitative evaluation of the channel change trend and providing a basic input for subsequent curvature calculation.

[0123] For example, the sampling time interval Δt is set to 2ms. At time t, the phase shift vector ΔV t The four-dimensional components are (0.05, -0.03, 0.02, 0.01). Calculate its Euclidean norm ||ΔV||. t || is sqrt(0.05) 2 +(-0.03) 2 +0.02 2 +0.01 2 The instantaneous angular velocity ω is approximately 0.062. Substituting this into the formula, we calculate ω = 0.062 / 0.002 = 31.2 rad / s. If the angular velocity suddenly increases to 150 rad / s in the next instant, the channel is determined to have entered a period of rapid change. This calculation process is implemented in an FPGA using a pipelined architecture, with a single operation latency of less than 50 ns, significantly improving the response speed to dynamic changes in the channel.

[0124] S3.3: Utilize the instantaneous angular velocity scalar and its time-varying derivative information to perform second-order differential geometric calculations to obtain the spatial curvature value describing the degree of curvature of the trajectory of the phase vector on the four-dimensional unit hypersphere, and use the spatial curvature value as the second-dimensional feature parameter for identifying the phase transition boundary.

[0125] The instantaneous angular velocity scalar sequence output from S3.2 is received and used as the dynamic input basis for calculating the spatial curvature. The phase displacement vector sequence obtained from S3.1 is divided by the sampling time interval to obtain the instantaneous displacement rate of change vector sequence V. t Then, perform a first-order difference on the instantaneous displacement rate of change vector sequence to obtain the rate of change vector a. t .

[0126] Based on the Riemannian geometric properties of a four-dimensional unit hypersphere, the geodesic curvature calculation formula is used to perform nonlinear coupling operations between the instantaneous displacement rate of change vector and the rate of change vector. The spatial curvature value is calculated using the following formula:

[0127] in, This represents the spatial curvature value. It is the instantaneous displacement rate vector of the dynamic phase vector on the four-dimensional unit hypersphere (i.e., the result of time normalization of the phase displacement vector of adjacent sampling points). , This represents the generalized cross product operation in four-dimensional space. This represents the magnitude of the vector.

[0128] The calculated raw curvature values ​​are subjected to moving average filtering to eliminate high-frequency noise interference caused by physical layer sampling jitter, generating smoothed spatial curvature time series data.

[0129] The smoothed spatial curvature value is compared with the preset transition boundary threshold in real time to mark the time point when the curvature increases significantly, and to establish the geometric boundary for the transition of the phase trajectory from the stable state to the abrupt state.

[0130] By using second-order differential geometric processing, the instantaneous angular velocity result from the previous step is transformed into a spatial curvature value describing the degree of trajectory curvature, thus achieving the expected technical effect of identifying phase transition boundaries.

[0131] For example, the sampling period is set to 2ms, and the sliding window length is 5 sampling points. When the channel rapidly switches from a high MCS state of Wi-Fi 6 to a low MCS state due to interference, the instantaneous angular velocity surges from 0.5 rad / s to 2.0 rad / s within 10ms. After first-order difference calculation, the angular velocity acceleration reaches 150 rad / s². 2 Substituting the values ​​into the curvature formula above, the calculated spatial curvature κ is 0.85. This value far exceeds the preset stability threshold of 0.1. Based on this, the system determines that the channel has entered the transition phase region, accurately capturing the drastic fluctuations in the channel state. This provides precise geometric characteristics for subsequent priority anchoring, significantly improving the accuracy of scheduling response.

[0132] S3.4: Based on the comparison between the instantaneous angular velocity scalar and the preset static threshold, a continuous counting and accumulation operation is performed to obtain the duration of the phase vector remaining relatively stable in a specific phase region, i.e., the local dwell time, and the local dwell time is used as the third-dimensional feature parameter for filtering noise interference.

[0133] The instantaneous angular velocity scalar sequence output by S3.2 is received and used as the reference input data for judging the motion state of the channel phase trajectory.

[0134] A static threshold parameter is set, which is determined based on the maximum permissible angular displacement rate of noise disturbance per unit hypersphere, and is used to distinguish between steady-state dwell and dynamic transition of the channel.

[0135] The instantaneous angular velocity scalar at the current moment is compared with a preset static threshold. If the instantaneous angular velocity is less than the static threshold, the current sampling point is determined to be in the phase stable region.

[0136] When a phase-stable region is identified, the continuous counting accumulator is activated to increment the count value of the current dwell period by one, in order to record the duration of the channel state within the specific phase region.

[0137] If the instantaneous angular velocity is greater than or equal to the stationary threshold, it is determined that a phase transition has occurred in the channel, the continuous counting accumulator is immediately reset to zero, and the accumulation process of the current dwell time is terminated.

[0138] The accumulated continuous count values ​​are multiplied by the sampling time interval to calculate the local dwell time of the phase vector in a specific phase region, which remains relatively stable.

[0139] By using the above comparison and counting accumulation process, the instantaneous angular velocity of the previous step is transformed into a local dwell time index that characterizes channel stability, thus achieving effective filtering of phase noise interference.

[0140] For example, the sampling time interval is set to 2ms, and the preset static threshold is 0.05 rad / ms. When the channel is in the stable transmission phase of Wi-Fi 6, the instantaneous angular velocities of five consecutive sampling points are 0.01, 0.02, 0.01, 0.03, and 0.02 rad / ms, respectively, all less than the threshold. The continuous counter accumulator increments from 0 to 5, calculating a local dwell time of 10ms. This duration indicates that the channel is in the stable phase region, and subsequent scheduling strategies can maintain the current priority. If the angular velocity of the sixth sampling point suddenly increases to 0.08 rad / ms, exceeding the threshold, the counter is reset to 0, the local dwell time is cleared, and the transition detection logic is triggered, significantly improving the accuracy of the response to channel abrupt changes.

[0141] S3.5: Perform multi-dimensional data encapsulation and structured recombination processing on the instantaneous angular velocity scalar, spatial curvature value, and local dwell time to generate a phase evolution feature set containing complete information on channel stability and abrupt change trends, and use the phase evolution feature set as the sole input source for subsequent transition phase region determination logic.

[0142] like Figure 2As shown, step S4: Construct a phase-frame anchoring table using the phase evolution feature set. The phase-frame anchoring table establishes a deterministic mapping relationship between channel phase attributes and video frame priorities by offline binding of typical transition phase regions on the four-dimensional unit hypersphere to fixed offset positions in the video stream GOP structure. Specifically, this includes: S4.1: Obtain the historical channel state dataset as the input data source, perform clustering analysis algorithm based on the phase vector sequence generated by polar coordinate transformation, extract the set of typical transition phase regions that characterize the channel state change mode, and form the basic phase feature library for anchor table construction.

[0143] Obtain the historical channel state dataset, which contains time-series records of MCS index, channel occupancy, interference level, and RSSI fluctuation entropy values ​​collected over a long period of time.

[0144] Polar coordinate transformation preprocessing is performed on the original multidimensional parameters in the historical dataset to map the four-dimensional physical layer index at each time step to a phase vector on a four-dimensional unit hypersphere, thus constructing a high-dimensional phase trajectory database.

[0145] The density-based spatial clustering algorithm DBSCAN was used to perform unsupervised clustering analysis on vector points in the phase trajectory database. The neighborhood radius parameter epsilon was set to 0.15, and the minimum number of samples was set to min. samples The value is 20 to identify high-density clustered regions and sparse transition regions in phase space.

[0146] Calculate the geodesic distance between each phase vector point and other points in its neighborhood, eliminate noise points and outliers, and retain the set of phase vectors belonging to the core cluster to form candidate stable phase regions and candidate transition boundary regions.

[0147] Curvature gradient analysis is performed on the phase vector sequence within the candidate transition boundary region. Continuous segments with a curvature change rate exceeding a preset threshold of 0.8 are extracted and marked as typical transition phase regions. These regions characterize the rapid evolution process of the channel state from stable to abrupt change.

[0148] The geometric center of the extracted typical transition phase region is calculated to generate a central phase vector representing the characteristics of the region. The spatial coverage radius of the region is calculated to form a phase feature descriptor containing the center coordinates and the boundary range.

[0149] The feature descriptors of all typical transition phase regions are structured and encapsulated according to spatial location indexes to build a basic phase feature library, which serves as the sole data source for subsequent spatiotemporal alignment mapping with the video GOP structure.

[0150] By using DBSCAN clustering and curvature gradient analysis, the historical phase trajectory data from the previous step is transformed into a set of discretized and characteristic typical transition phase regions, which realizes the automated extraction and standardized description of channel mutation modes, laying the data foundation for building a low-latency lookup table mechanism.

[0151] For example, 5G channel log data from the past 24 hours was collected at a sampling frequency of 1kHz, resulting in 86.4 million sampling points. After polar coordinate transformation, the DBSCAN algorithm was applied, with epsilon=0.15 and min... samples =20, clustering time 120 seconds. Fifteen typical transition phase regions were identified. Region A corresponds to a scenario where the MCS drops sharply from 28 to 15 and the interference level increases from -80dBm to -60dBm, with a center phase vector of [0.45, 0.62, -0.33, 0.51] and a spatial radius of 0.08. Region B corresponds to a scenario where the channel occupancy rate surges from 30% to 85%, with a center phase vector of [-0.21, 0.75, 0.12, -0.44] and a spatial radius of 0.12. The final generated basic phase feature library contains 15 entries, each storing the center vector, radius, and region ID. The data size is only 4KB, significantly reducing the computational overhead of online queries and ensuring sub-millisecond response time.

[0152] S4.2: Perform video stream structure parsing operations based on a set of typical transition phase regions, and perform time axis alignment marking on the fixed offset positions of I-frames, P-frames and B-frames in the standard GOP structure to generate a GOP frame position feature set containing frame type identifiers and relative temporal indexes, so as to establish a time reference system for the video content.

[0153] The set of typical transition phase regions output from step S4.1 is received as input reference data for the video stream structure analysis operation.

[0154] Deeply analyze the Group of Pictures (GOP) structure of a standard video stream to extract the coding dependencies and timestamp information of I-frames, P-frames, and B-frames.

[0155] Establish a timeline reference system with the GOP start time as zero, and calculate the fixed time offset of each video frame relative to the GOP start time.

[0156] Based on the inter-frame prediction dependency topology, I-frames are assigned the highest base weight identifier, P-frames are assigned the medium base weight identifier, and B-frames are assigned the low base weight identifier.

[0157] The frame type identifier is structurally bound to the calculated relative temporal index to generate a GOP frame location feature set containing frame ID, type label, and time offset.

[0158] Through the above processing method, the phase region data from the previous step is transformed into a video frame position feature set with a time reference, achieving precise alignment between the channel state and the time dimension of the video content, and providing a unified spatiotemporal reference for subsequently constructing a deterministic mapping relationship. For example, for a video stream using the H.264 encoding standard, the GOP length is set to 16 frames, with a structure of IPBBPBBPBBPBBPBB. The parser identifies the first frame as an I-frame with a time offset of 0ms and assigns it a type identifier of 0x01; the second frame is a P-frame with a time offset of 33ms and is assigned a type identifier of 0x02; the third and fourth frames are B-frames with time offsets of 66ms and 99ms respectively and are assigned a type identifier of 0x03. This process continues, generating a GOP frame position feature set containing 16 entries, where each entry explicitly records the frame's logical sequence number, physical time offset, and decoding dependency level within the GOP, ensuring accurate location of key frames requiring priority protection in subsequent mapping.

[0159] S4.3: Utilize the basic phase feature library and GOP frame position feature set to perform multi-dimensional correlation mapping operation, logically bind the spatial geometric attributes of typical transition phase regions with the key frame offsets in the GOP frame position feature set, generate preliminary channel phase to video frame priority mapping relationship pairs, and establish a causal connection between phase change and frame importance.

[0160] The set of typical transition phase regions and the set of GOP frame location features from the basic phase feature library are received as the input data source for multidimensional correlation mapping operations. Spatial geometric attribute extraction is performed on the set of typical transition phase regions, and the centroid coordinates and coverage radius of each region on the four-dimensional unit hypersphere are calculated to generate phase region feature vectors characterizing the intensity of channel abrupt changes.

[0161] Frame dependency parsing is performed on the GOP frame location feature set. Based on the video coding standard, the reference levels of I-frames, P-frames, and B-frames in the decoding sequence are determined, and initial importance weight coefficients are assigned to different frame types. A causal mapping model from channel phase to video frame priority is constructed, and the rate of curvature change in the phase region feature vector is nonlinearly coupled with the frame importance weight coefficients.

[0162] The weighted Euclidean distance algorithm is used to calculate the matching degree between the current transition phase region and the fixed offset position of each GOP, and the key frame sequence most affected by channel mutation is selected.

[0163] Candidate frames are sorted according to the matching degree index, and the top N frames with the highest matching degree are selected as the strong guarantee candidate set. The strong guarantee candidate set is logically bound to typical transition phase regions to generate a preliminary mapping tuple containing phase region IDs and frame offset index pairs.

[0164] Collision detection is performed on the initial mapping tuples. If the matching degree of multiple GOP positions corresponding to the same phase region is similar, the mapping relationship of the highest priority frame is retained according to the frame type priority rule.

[0165] By using a multi-dimensional correlation mapping operation, the basic phase features and GOP structure information from the previous step are transformed into preliminary channel phase-to-video frame priority mapping pairs, realizing a deterministic causal connection between phase change and frame importance, and providing a logical basis for subsequent hardware instruction encoding.

[0166] For example, consider a typical transition phase region A where the MCS index drops sharply from 15 to 5 and the interference level jumps from 2dBm to 15dBm. Its centroid coordinates correspond to a specific geometric sector on a four-dimensional hypersphere. The GOP structure is analyzed, with I-frame weights set to 1.0, P-frame weights to 0.6, and B-frame weights to 0.3. The curvature C of this phase region is calculated to be 0.85, yielding f(C) = 0.92. For the first frame (I-frame) in the GOP, the matching degree M = 1.0 × 0.92 = 0.92; for the third frame (P-frame), the matching degree M = 0.6 × 0.92 = 0.552. The system determines that the first frame has the highest matching degree and binds its offset 1 to phase region A, generating a mapping tuple {Region_A,Offset_1}. This process ensures that the most critical I-frames are prioritized during periods of severe channel degradation, significantly improving the anti-interference capability of the video stream.

[0167] S4.4: Discretize the channel phase to video frame priority mapping relationship and use a bit mask generation algorithm to convert the mapping relationship into a binary instruction format to generate standardized scheduling entries containing anchor frame position identifiers, so as to realize the instruction data structure that can be directly recognized by the hardware scheduling queue.

[0168] Receive the channel phase to video frame priority mapping pair generated by S4.3. This data contains the spatial geometric attribute index of typical transition phase regions and the logical binding information of GOP keyframe offset.

[0169] Bit-plane encoding conversion is performed on the GOP frame position feature set in the mapping relationship pair. Based on the fixed temporal index of I-frame, P-frame and B-frame in the standard GOP structure of the video stream, an initial frame type bitmap vector with a length of 16 bits is constructed, where the most significant bit corresponds to the starting I-frame of the GOP, and the subsequent bits correspond to the P-frame and B-frame in sequence.

[0170] A bitmask generation algorithm is used to perform a bitwise AND operation between the initial frame type bitmap vector and the preset priority weight template. Positions marked as strong protection frames are set to 1, and other non-critical frame positions are set to 0, thus generating a preliminary binary priority instruction word.

[0171] Hardware protocol adaptation encapsulation is performed on the initial binary priority instruction word. A 4-bit phase region index prefix is ​​added to the instruction word header to identify the specific transition phase region type that triggers the scheduling strategy, forming a composite data structure containing region identifier and frame-level mask.

[0172] The composite data structure is subjected to checksum appending processing. The cyclic redundancy check algorithm is used to calculate an 8-bit checksum and append it to the end of the instruction word to ensure data integrity during offline storage and online reading, and generate the final standardized scheduling entry.

[0173] By using discretization encoding and bitmask generation, the logical mapping relationship from the previous step is transformed into a binary instruction format that can be directly parsed by the hardware scheduling queue, achieving sub-millisecond priority decision response with zero computational overhead.

[0174] For example, for a video stream with a GOP length of 16, when a typical transition phase region A with a sudden drop in MCS and a surge in interference is detected, the system extracts the corresponding GOP frame position features and determines that the first frame (I-frame) and the fourth frame (P-frame) are strong guarantee frames. A 16-bit initial bitmap vector is constructed, with the first and fourth bits set to 1 and the rest set to 0, resulting in the binary string 10010000000000000. This is then bitwise ANDed with the priority weight template, keeping the mask unchanged. A 4-bit phase region index 0011 (representing region A) is added to the header, forming 20 bits of intermediate data. The CRC-8 checksum is calculated to be 0xA5 and appended to the tail, finally generating a 28-bit standardized scheduling entry 0011100100000000000000010100101. This entry is directly written to read-only memory, requiring no floating-point operations during online retrieval; a single memory read is sufficient to obtain the complete hardware scheduling instructions, significantly reducing scheduling latency.

[0175] S4.5: Based on standardized scheduling entries, perform static lookup table construction operations, write all standardized scheduling entries into read-only storage units in the spatial index order of the four-dimensional unit hypersphere, and generate the final phase-frame anchor table to complete the deterministic mapping and solidification from continuous channel phase evolution characteristics to discrete hardware scheduling instructions.

[0176] The system receives a set of standardized scheduling entries generated through discretization encoding. This set contains transition phase region indices and corresponding GOP frame bitmask instructions. It then performs spatial hash mapping calculations on the standardized scheduling entries, converting continuous phase region coordinates into unique linear memory address indices based on the latitude and longitude division rules of the four-dimensional unit hypersphere.

[0177] A Z-order curve filling algorithm is used to mesh the four-dimensional hyperspherical surface, ensuring that geometrically adjacent phase regions maintain locality at memory addresses to optimize cache hit rate. An address mapping table header for the read-only memory (ROM) is constructed, defining a fixed word length for each memory cell. The high 16 bits store the phase region identifier, and the low 16 bits store the control word for the hardware scheduling queue. All standardized scheduling entries are traversed, and the bitmask instruction for each entry is written to the low 16-bit data area of ​​the corresponding linear memory address. Simultaneously, the phase region identifier is written to the high 16-bit index area, completing the physical hardening of the key-value pairs.

[0178] Parity check code generation is performed on the written data, and an 8-bit error correction code is added to each storage unit to ensure data integrity during offline table lookup. All storage units are arranged in ascending address order to form a continuous static lookup table structure, which is then burned into the cache or firmware storage area of ​​the transmission strategy decision module.

[0179] Through the above static lookup table construction operation, the logical mapping relationship generated in the previous step is transformed into a read-only data structure that can be directly addressed by the physical layer, realizing the zero-latency deterministic mapping solidification from continuous channel phase evolution characteristics to discrete hardware scheduling instructions, and eliminating runtime computation overhead.

[0180] For example, a four-dimensional hypersphere is divided into 4096 uniform grid cells, each corresponding to a memory address. For a typical transition phase region A, its center coordinates are mapped to address 0x0A3F. In the standardized scheduling entry, the bitmask for I-frame priority is 0x8001. The system writes 0x0A3F into the high 16 bits of address 0x0A3F, writes 0x8001 into the low 16 bits, and appends a checksum of 0xA5. When the real-time channel state falls into region A, the hardware directly reads the content of address 0x0A3F and obtains the priority instruction within one clock cycle, without any floating-point operations or memory allocation, significantly reducing scheduling response latency.

[0181] like Figure 3 As shown, step S5: Based on the phase evolution feature set, determine whether the phase vector has entered the transition phase region. If a sudden increase in curvature is detected and the angular velocity exceeds a preset threshold, trigger the phase stability filter to perform continuous residence time verification on the transition phase region. Specifically, this includes: S5.1: Obtain the instantaneous curvature value and instantaneous angular velocity value from the phase evolution feature set as input conditions, use the mutation detection algorithm to perform first-order difference calculation on the instantaneous curvature value and perform threshold comparison processing on the instantaneous angular velocity value to generate a phase transition initial judgment signal containing curvature surge flag and angular velocity over-limit flag.

[0182] The instantaneous curvature and instantaneous angular velocity values ​​received from the phase evolution feature set are used as the real-time input data source for the mutation detection algorithm.

[0183] A first-order forward difference operation is performed on the instantaneous curvature value sequence within a continuous time window to calculate the rate of curvature change between adjacent sampling points, thereby quantifying the geometric curvature acceleration trend of the channel state trajectory on the four-dimensional hypersphere.

[0184] The calculated rate of curvature change is compared with a preset curvature abrupt change threshold. If the rate of curvature change exceeds the threshold, it is determined that the channel trajectory has undergone a dramatic change in geometric shape, and a high-level curvature abrupt increase flag is generated.

[0185] In parallel, instantaneous angular velocity values ​​within the same time window are acquired and compared with an upper threshold for angular velocity determined based on historical channel stability statistics. When the instantaneous angular velocity value exceeds the upper threshold, it indicates an abnormally rapid migration rate of the channel state in the phase space, generating a logic high-level angular velocity over-limit flag. A bitwise combination operation is performed on the generated curvature surge flag and the angular velocity over-limit flag to encapsulate them into a phase transition initial judgment signal containing two-dimensional decision information.

[0186] Through the parallel threshold decision and flag bit encapsulation process described above, continuous geometric feature values ​​are transformed into discrete binary decision signals, enabling rapid preliminary screening of channel phase transition events and providing a low-latency triggering basis for subsequent stability verification.

[0187] For example, the curvature abrupt change threshold is set to 0.15 rad / ms. 2 The upper limit threshold for angular velocity is 0.8 rad / ms. In the current sampling period t, the system acquires an instantaneous curvature value of 0.22 rad / m, and the instantaneous curvature value at the previous moment was 0.05 rad / m. The calculated rate of change of curvature Δ is 0.17 rad / ms. 2 Since 0.17 is greater than 0.15, the curvature surge flag is set to 1. Simultaneously, an instantaneous angular velocity value of 0.92 rad / ms is acquired; since 0.92 is greater than 0.8, the angular velocity exceedance flag is set to 1. The system combines the two flags into a binary sequence "11" to generate a phase transition initial judgment signal. This signal directly triggers the subsequent logical AND operation of S5.2, eliminating the need for floating-point comparison delay and significantly improving response speed.

[0188] S5.2: Based on the curvature surge flag and angular velocity over-limit flag in the initial phase transition judgment signal, perform logical AND operation processing to mark the moment that simultaneously satisfies the curvature surge condition and the angular velocity over-limit condition as the potential transition start point, so as to generate a potential transition event identifier carrying timestamp information.

[0189] The system receives the initial phase transition judgment signal output from S5.1. This signal includes Boolean flags representing curvature abrupt changes and angular velocity exceeding limits. The curvature abrupt increase flag is extracted as the first logical input variable, and the angular velocity exceeding limit flag is extracted as the second logical input variable. A hardware-level bitwise AND operation is performed on the first and second logical input variables. The operation result is true only if both flags are simultaneously high, indicating that the current moment satisfies the double abrupt change condition. The moment deemed true is marked as the potential transition start point, and the corresponding high-precision system timestamp is recorded. A potential transition event identifier carrying timing index information is generated based on this timestamp. Through logical AND operation processing, the discrete flags from the previous step are transformed into a potential transition event identifier with spatiotemporal uniqueness, achieving precise locking of moments of drastic channel state changes and providing an accurate time reference for subsequent stability verification.

[0190] For example, the sampling period is set to 2ms. At t=100ms, the curvature surge flag and the angular velocity exceedance flag output by S5.1 are both 1. A logical AND operation is performed: 1AND1=1. The result is true, and t=100ms is marked as the potential transition start point. A potential transition event identifier "Event_ID_0x9A" is generated and associated with a timestamp of 100ms. At t=102ms, the curvature surge flag is 1, and the angular velocity exceedance flag is 0. A logical AND operation is performed: 1AND0=0. The result is false, and no event identifier is generated. This mechanism ensures that subsequent processes are triggered only when curvature and angular velocity synchronization is abnormal, effectively filtering false alarms caused by fluctuations in a single indicator and significantly improving the accuracy of transition detection.

[0191] S5.3: Obtain the potential transition event identifier and its corresponding timestamp information, call the phase stability filter to perform continuous sampling and tracking processing on the phase vector sequence following the potential transition event identifier, so as to count the continuous residence period count of the phase vector in the transition phase region.

[0192] The potential transition event identifier carrying timestamp information generated by S5.2 is received and used as the trigger input signal for the phase stability filter. The initial time t0 and the corresponding phase vector coordinate P(t0) recorded in the identifier are extracted, and the continuous dwell period counter N is initialized to 0.

[0193] The sliding time window tracking mechanism is started, and the subsequent phase vector sequence P(t0+Δt), P(t0+2Δt)... is continuously obtained from the physical layer driver interface with a preset sampling interval Δt as the step size.

[0194] The phase vector P(t) at the current sampling time ti i Perform the region attribution determination and calculate P(t).i The hyperspherical geodesic distance d between the point Ck, the center point of the typical transition phase region extracted in S4.1, and the point S4.1 is... i .

[0195] The calculated geodesic distance d i Compared with the preset region radius threshold R thresh Compare them. If d i Less than or equal to R thresh If the current phase vector is determined to be within the transition phase region, the counter is incremented, and N = N + 1 is set.

[0196] If d i Greater than R thresh If the phase vector is determined to have escaped the transition region, the tracking process is immediately terminated, the counter N is reset to 0, and the current potential transition event identifier is discarded.

[0197] During continuous tracking, the maximum dwell time record is updated in real time until the counter N reaches the preset minimum dwell period threshold or an overflow interruption occurs.

[0198] Through the above continuous sampling and tracking and region attribution determination processing, the instantaneous change signal in the previous step is transformed into a continuous dwell period count value that characterizes the stability of the channel state. This effectively filters out false transition events caused by phase noise and ensures that only a continuous and stable channel deterioration or improvement trend can trigger subsequent priority scheduling actions.

[0199] S5.4: Perform numerical comparison processing based on the continuous dwell period count value and the preset minimum dwell period threshold to determine whether the continuous dwell period count value is greater than or equal to three sampling periods, so as to generate a phase dwell verification result that characterizes the validity of the transition event.

[0200] The continuous dwell period count obtained from the phase stability filter statistics is used as the core quantitative indicator to determine the validity of the transition event. This count reflects the number of sampling points that remain within the current transition phase region after the phase vector detects a sudden increase in curvature and an excessive angular velocity.

[0201] The system reads the preset minimum dwell period threshold parameter, which is set to 3 sampling periods, corresponding to a physical time of approximately 6 milliseconds. This threshold is designed to filter out false transitions caused by transient noise or signal jitter.

[0202] The numerical comparison logic is executed to perform a greater than or equal to relationship operation between the continuous dwell period count value and the minimum dwell period threshold, generating a Boolean-type phase dwell period verification result flag.

[0203] If the continuous dwell period count is less than 3, the current channel state change is determined to be transient noise interference, the phase dwell period verification result is marked as invalid, the system discards the current potential transition event identifier and resets the counter.

[0204] If the continuous dwell period count is greater than or equal to 3, it is determined that the channel state has stabilized and entered a new phase region. The phase dwell period verification result is marked as valid, and the transition event is confirmed to have scheduling response value.

[0205] By using the above numerical comparison and threshold determination processing method, the continuous sampling statistical results of the previous step are transformed into a binary verification signal that characterizes the authenticity of the transition event, thereby achieving robust filtering of high-frequency channel noise and ensuring that priority adjustment response is only made to persistent channel deterioration or improvement.

[0206] For example, suppose the system sampling frequency is 500Hz, i.e., the sampling period is 2ms. When step S5.3 counts that the phase vector has continuously resided in the transition region for 5 sampling points, the continuous residence period count is 5. The preset minimum residence period threshold is 3. Performing a comparison 5≥3, the result is true. The system generates a "valid" verification result, allowing entry into step S6 to query the anchor table. If the RSSI fluctuates momentarily due to Wi-Fi co-channel interference, and only 2 sampling points are continuously resided, the count is 2. Performing a comparison 2≥3, the result is false. The system generates an "invalid" verification result, discards this trigger, avoids prioritizing errors in non-critical frames, and significantly reduces ineffective scheduling overhead.

[0207] S5.5: Perform status confirmation processing based on the phase dwell verification result. If the phase dwell verification result is confirmed to be valid, output the final verification pass instruction containing the transition phase region confirmation information. If the phase dwell verification result is confirmed to be invalid, discard the potential transition event flag and reset the continuous dwell period count value to complete the noise immunity verification operation of the transition phase region.

[0208] The phase dwell time verification result is received as the input criterion for the state confirmation logic. This result indicates whether the continuous dwell time count value of the phase vector in the transition phase region meets the preset minimum dwell time threshold requirement.

[0209] The phase dwell time verification result is processed by Boolean logic branch judgment. If the verification result indicates that the continuous dwell time count value is greater than or equal to three sampling periods, the current channel state change is determined to be a valid transition event.

[0210] Based on the determination of a valid transition event, a final verification pass instruction containing confirmation information of the transition phase region is generated. This instruction carries the timestamp of the current moment and the corresponding phase region encoding identifier, which is used to trigger subsequent priority mapping operations.

[0211] If the phase dwell time check result indicates that the continuous dwell time count is less than three sampling periods, then the current channel state fluctuation is determined to be instantaneous noise interference or unstable jitter, which is an invalid transition event.

[0212] For invalid transition events, the potential transition event identifier is discarded, and all temporary state data associated with the event is cleared from the memory buffer to prevent the generation of erroneous scheduling instructions.

[0213] The continuous dwell period counter is synchronously reset to its initial zero value, and the internal state machine of the phase stability filter is reset, causing it to re-enter the listening mode waiting for the next potential transition start point.

[0214] Through the above status confirmation and reset process, the numerical comparison result of the previous step is transformed into a definite scheduling enable signal or silent reset action, which realizes effective filtering of channel phase noise and anti-interference verification of scheduling decisions, ensuring that priority adjustment is triggered only when the channel undergoes a continuous and stable transition, thereby ensuring the stability and real-time performance of video stream transmission.

[0215] For example, the minimum dwell period threshold is set to 3 sampling periods, with each sampling period corresponding to a 2ms time interval. When the phase stability filter counts a continuous dwell period of 4, the verification result is true, and the system outputs a verification pass instruction containing the phase region index code 0x1A, triggering step S6 to query the anchor table. When the count of continuous dwell periods is 2, the verification result is false, the system discards the detected curvature surge event, clears the counter, and does not generate any scheduling instructions, thus avoiding false responses to brief signal fluctuations and significantly reducing ineffective scheduling overhead.

[0216] Step S6: If the phase stability filter confirms that the phase vector has continuously resided in the transition phase region for more than three sampling periods, then the corresponding anchor frame position identifier is queried according to the phase-frame anchoring table, and a scheduling queue instruction containing bitmask markers is generated. Specifically, this includes: S6.1: Based on the continuous dwell time verification result output by the phase stability filter, the phase vector in the transition phase region is processed by regional coding to generate a standardized phase region index code that represents the current channel evolution state, which serves as the sole addressing basis for subsequent table lookup operations.

[0217] The phase dwell verification result output by the phase stability filter is received. This result includes a valid confirmation flag of the transition event and the corresponding current coordinate value of the phase vector.

[0218] For confirmed valid phase vectors, a four-dimensional hyperspherical mesh mapping process is performed to discretize the continuous four-dimensional spatial coordinates into finite index units to eliminate lookup jitter caused by floating-point precision differences.

[0219] The uniform quantization algorithm is used to divide the values ​​of MCS, channel occupancy, interference level and RSSI fluctuation entropy into intervals, and the value range of each dimension is divided into 16 equal-width intervals to generate discretized interval indices for the four dimensions.

[0220] The discretized interval indexes of the four dimensions are shifted and combined according to the preset bit weights to construct a unique integer phase region code value, ensuring that different phase states correspond to a unique code identifier, and generating a standardized phase region index code.

[0221] Boundary constraint checks are performed on the generated standardized phase region index codes. If the index value exceeds the predefined range of legal transition phase regions, it is corrected to the nearest legal boundary index to prevent illegal memory access.

[0222] By combining gridded mapping with bit weights, the continuous phase vector from the previous step is transformed into a deterministic, standardized phase region index code, achieving a lossless conversion from continuous geometric space to discrete hardware addressing space, providing a unique and stable addressing basis for subsequent zero-latency table lookup operations.

[0223] For example, when the phase stability filter confirms the validity of the transition event, the four-dimensional coordinates of the current phase vector are obtained as (MCS=18.5, Occupancy=0.72, Interference=0.45, Entropy=1.2). 16-interval uniform quantization is performed on each dimension. The maximum value of the MCS dimension is set to 31, and the interval width is 1.9375. 18.5 / 1.9375 is rounded to obtain index 9; the maximum value of the Occupancy dimension is set to 1.0, and the interval width is 0.0625. 0.72 / 0.0625 is rounded to obtain index 11; the maximum value of the Interference dimension is set to 1.0, and the interval width is 0.0625. 0.45 / 0.0625 is rounded to obtain index 7; the maximum value of the Entropy dimension is set to 2.0, and the interval width is 0.125. 1.2 / 0.125 is rounded to obtain index 9. The index code 39801 is obtained by calculating the bit-weighted shift combination. The index code 39801 is directly used as a hash key to point to a specific storage unit in the phase-frame anchor table, ensuring that the mapping from channel state to scheduling instructions is completed within microseconds, avoiding the uncertainty delay caused by floating-point comparison.

[0224] S6.2: Using the standardized phase region index code as the input key, a deterministic hash retrieval operation is performed on the offline constructed phase-frame anchor table to extract the set of fixed offset positions of the GOP structure that are strictly bound to the current transition phase region, and obtain the key frame logical address sequence to be guaranteed.

[0225] The normalized phase region index code generated in step S6.1 is received. This index code serves as the unique input key for deterministic hash retrieval and is directly mapped to the preset phase-frame anchor table address space in read-only memory.

[0226] Based on the standardized phase region index code, high-bit truncation and low-bit mask separation operations are performed to extract the main index bit used to locate the hash bucket and the sub-index bit used to resolve conflicts, ensuring that memory addressing initialization is completed within a single instruction cycle.

[0227] The primary hash slot of the phase-frame anchor table is accessed using the extracted primary index bit. The pointer offset stored in the slot is read, which points to the starting address of the secondary data block containing the specific GOP structure mapping relationship.

[0228] To address potential hash collisions, a linear probing or chaining method is used to traverse the secondary data blocks, comparing each sub-index bit with the tag field in the data block bit by bit until a completely matching phase region identifier is found.

[0229] Once a tag is successfully matched, the pre-bound GOP structure fixed offset position set is immediately extracted from the corresponding data record. This set is stored in the form of a compressed bitmap, which clearly identifies the relative timing indexes of the I-frames, P-frames and key B-frames that need to be prioritized in the current channel transition state.

[0230] The extracted set of fixed offset positions is decompressed and reconstructed with logical addresses. The relative temporal index is converted into the absolute frame sequence number in the video stream transmission queue, and a sequence of logical addresses of key frames to be protected is generated.

[0231] Through the aforementioned deterministic hash retrieval and address reconstruction process, the phase region index code generated in the previous step is transformed into a specific keyframe logical address sequence, achieving zero-latency mapping from abstract channel state to specific video frame priority. This significantly reduces the computational overhead of scheduling decisions and eliminates the response lag of traditional algorithms. For example, the standardized phase region index code is 0x3A5F, the main index bit 0x3A points to bucket 58 of the hash table, and the read pointer offset is 0x100. The data block is traversed at the offset address, and the sub-index bit 0x5F matches the second entry label. The bitmap 0x00000009 stored in this entry is extracted, and after decompression, the offset {0,3} is obtained, corresponding to frame 1 (I-frame) and frame 4 (P-frame) in the GOP. The system immediately generates the logical address sequence [Frame_0,Frame_3]. The entire process takes less than 10 nanoseconds, ensuring that keyframes can be instantly identified and prioritized during channel degradation.

[0232] S6.3: Based on the key frame logical address sequence, perform bit-plane mapping conversion on the frame sequence number in the video stream transmission queue to generate an initial binary bitmap vector containing the anchor frame position identifier, thereby achieving initial alignment from logical address to physical scheduling slot.

[0233] The system receives a sequence of keyframe logical addresses, containing the relative offset indices of the video frames to be secured within the Group of Pages (GOP) structure. It constructs a zero-initialized binary bitmap vector with the same size as the current transmission window, and a length equal to the maximum concurrent slot number of the hardware scheduling queue. It iterates through each offset index value in the keyframe logical address sequence, calculating its absolute physical slot number within the current sliding time window. Modulo operations are used to map the relative offsets to the physical address space of the circular buffer, ensuring the indices do not go out of bounds. Bit-plane positioning is performed on the calculated absolute physical slot number to determine its specific bit position in the binary bitmap vector. The value of this bit position is flipped from 0 to 1, marking it as a strongly secured state. This mapping and flipping operation is repeated until all keyframe indices in the sequence have been processed. Through bit-plane mapping transformation, the logical address sequence from the previous step is converted into an initial binary bitmap vector containing anchor frame position identifiers, achieving initial alignment from logical addresses to physical scheduling slots and providing a directly usable bitmask data foundation for subsequent hardware instruction encapsulation.

[0234] For example, the hardware scheduling queue depth is set to 64, and the current GOP size is 8. The keyframe logical address sequence output by the phase-frame anchoring table is {0, 3}, corresponding to the I-frame and the 3rd P-frame, respectively. A 64-bit binary bitmap vector is initialized, with all bits set to 0. For index 0, its absolute physical slot number is calculated as the current window start pointer plus 0. Assuming the start pointer is 10, the absolute number is 10. The 10th bit of the bitmap vector is set to 1. For index 3, its absolute physical slot number is calculated as 10 plus 3, which equals 13. The 13th bit of the bitmap vector is set to 1. The final generated initial binary bitmap vector has 1s in the 10th and 13th bits, and the remaining bits are 0. This bitmap vector directly represents the physical slots that need to be scheduled first, eliminating address translation latency and significantly improving scheduling response speed.

[0235] S6.4: Based on the initial binary bitmap vector, perform a bitmask encapsulation operation adapted to the hardware protocol to generate a scheduling queue instruction containing bitmask markers that conforms to the underlying scheduling interface specification, and complete the final conversion of priority decision signals into execution layer control words.

[0236] Step S7: Write the video frame marked with the anchor frame position identifier into the highest priority slot of the hardware scheduling queue according to the scheduling queue instruction, completing the frame-level priority anchoring operation based on channel state evolution phase locking. Specifically, this includes: S7.1: Obtain the scheduling queue instruction containing bitmask markers as the input data source, parse and process the scheduling queue instruction to extract the encoded anchor frame position identifier and the corresponding target priority level parameter.

[0237] Receive the scheduling queue instruction containing bitmask markers generated in step S6. This instruction serves as the sole input data source for this step, and its data structure conforms to the underlying bus transmission protocol specification, including an instruction header, a payload area, and a checksum tail.

[0238] The binary stream parsing operation is performed on the scheduling queue instructions. Based on the predefined instruction set architecture specification, the starting offset address of the instruction load area is located, and the encoded control word sequence is separated out.

[0239] A bit-field extraction algorithm is used to mask and filter the high 16 bits of the control word sequence, remove irrelevant state bits, and accurately extract the anchor frame position identifier field that represents the logical position of the video frame.

[0240] The captured anchor frame position identifier field is converted into an unsigned integer value, which directly corresponds to the frame index offset within the GOP structure, thus establishing the logical identity of the frame to be prioritized.

[0241] The lower 8 bits of the control word sequence are simultaneously shifted and decoded to parse out the target priority level parameter, which quantifies the queue slot weight level that the hardware scheduler needs to allocate.

[0242] The parsed anchor frame location identifier and target priority level parameter are subjected to integrity verification to ensure that the combination of the two conforms to the preset scheduling semantic constraints and to prevent illegal instructions from causing hardware state machine abnormalities.

[0243] Through the bit-field parsing and decoding process described above, the binary scheduling queue instruction generated in the previous step is transformed into a structured anchor frame position identifier and target priority level parameter, realizing a lossless mapping from the underlying control signal to the upper-level scheduling logic parameter, and providing an accurate execution basis for subsequent address index matching.

[0244] For example, in a 5G NR channel environment, after the phase stability filter confirms the validity of the transition event, step S6 generates a 32-bit hardware scheduling instruction 0x00A30005. The system parses this instruction, first extracting the high 16 bits 0x00A3, which, after bit field filtering, yields the decimal value 163, confirming it as the anchor frame position identifier for the 163rd frame (i.e., the next I-frame) in the current GOP. Then, the low 8 bits 0x05 are extracted and decoded to obtain the target priority level parameter 5, representing the highest urgent transmission level. The verification step confirms that identifier 163 is within the legal frame range [0, 200] and that priority 5 is a valid value. Finally, the anchor frame position identifier 163 and the target priority level 5 are output, ensuring that the MAC layer driver can accurately lock the 163rd frame and assign it the highest transmission weight, significantly reducing the probability of critical frames being dropped in congested channels.

[0245] S7.2: Based on the anchor frame position identifier and target priority level parameter, perform address index matching processing on the current video frame sequence to be transmitted to generate a strong guarantee frame address mapping table containing specific memory address pointers.

[0246] The anchor frame location identifier and target priority level parameter obtained from S7.1 are received and used as the input control word for address index matching.

[0247] Traverse the circular buffer descriptor linked list of the current video frame sequence to be transmitted, and extract the globally unique frame sequence number field from each frame control block.

[0248] The anchor frame position identifier is compared bit by bit with the extracted frame sequence number to identify the target frame object that logically needs to be scheduled first. For a successfully matched frame control block, its corresponding physical memory start address pointer and data length metadata are read.

[0249] A dynamic hash mapping structure is constructed, using the target priority level as the key, and the matched physical memory address pointers are encapsulated as mapping table entries. Through Direct Memory Access (DMA) descriptor generation logic, the physical address pointers are converted into a bus address format addressable by the hardware scheduler. Integrity checks are performed on the generated address mapping table to ensure that the physical addresses of all strongly protected frames point to valid video data payload regions.

[0250] Through the above address index matching and mapping conversion process, the logical frame identifier of the previous step is transformed into a strongly guaranteed frame address mapping table containing pointers to specific physical memory addresses, realizing a precise mapping from logical scheduling instructions to physical data location, and providing a zero-latency address reference basis for subsequent hardware queue loading.

[0251] For example, the video frame sequence to be transmitted contains 60 frame control blocks, each occupying 32 bytes of memory. Anchor frame position identifiers indicate that frames 1 and 4 in the GOP must be transmitted with priority. The system traverses the buffer, comparing and finding that the frame number of the 1st control block is 1001 (corresponding to frame 1 of the GOP), and the frame number of the 4th control block is 1004 (corresponding to frame 4 of the GOP). The physical address of the 1st control block is read as 0x80001000, with a length of 150KB; the physical address of the 4th control block is read as 0x80005000, with a length of 40KB. A hash mapping table is constructed, mapping priority level 0xFF to the address list [0x80001000, 0x80005000]. After DMA conversion, bus addresses 0xC0001000 and 0xC0005000 are generated. The addresses are verified to be valid, and an address mapping table containing two physical pointers is output. This process is completed within two CPU clock cycles, significantly improving address resolution efficiency and ensuring the immediate accessibility of key frame data.

[0252] S7.3: Using the strong guarantee frame address mapping table, scan and query the status register of the idle slot in the hardware scheduling queue to locate and lock the physical index number of the highest priority available slot.

[0253] The system receives a strongly guaranteed frame address mapping table as input, which contains pointers to the physical memory addresses of video frames to be prioritized for transmission and their corresponding frame length information. It then traverses the hardware scheduling queue's idle slot status register array, which consists of multiple 32-bit status words, each corresponding to a group of physical transmit descriptor slots. A bitwise AND operation is performed on the status registers of the current scan cycle, using a preset all-idle mask value of 0xFFFFFFFF as the operand to quickly filter out completely unoccupied slot groups. If the bitwise AND result is non-zero, it indicates the existence of available slots. The leading zero counting instruction CLZ is then called to process the result value, calculating the position index of the most significant bit 1. This position index is converted into a linear physical slot number, and combined with the base address offset, the absolute physical index number of the target slot in SRAM is calculated. An atomicity test is performed on the calculated physical index number, and a setting operation is performed. A bus latch mechanism is used to prevent multi-core concurrent access conflicts, ensuring that the slot is exclusively locked by the current process. If the atomic operation succeeds, the lock status flag is written to the local cache, and a slot allocation handle containing the target physical index number is generated. Through the above hardware-level bit operations and atomic latching mechanism, the abstract address mapping relationship is transformed into a deterministic physical storage resource occupation, achieving zero-contention slot locking within microseconds and ensuring the real-time and deterministic nature of high-priority frame scheduling.

[0254] For example, the hardware scheduling queue contains 1024 transmit descriptor slots, divided into 32 status register groups, each managing 32 slots. The strongly guaranteed frame address mapping table points to I-frame data at memory address 0x8000A000. The system scans the 5th status register group and reads the value 0x0000FF00. A bitwise AND operation 0x0000FF00 AND 0xFFFFFFFF is performed, resulting in 0x0000FF00. The Count Leading Zeros (CLZ) instruction is called to process 0x0000FF00, detecting that the 8th bit is the most significant bit (counting from 0, actually the 15th bit, depending on the specific architecture definition; here it is assumed to be the 8th available bit scanned first). Convert to linear index: 5 32 + 8 = 168. An atomic test is performed on index 168 and set, returning a success status. The system locks index 168 exclusively for the current task and generates a slot allocation handle. This process takes less than 50 nanoseconds, significantly better than the millisecond latency of traditional software linked list queries, ensuring resource reservation is completed within the channel transition window.

[0255] S7.4: Based on the physical index number of the highest priority slot, write the memory address pointer in the strong guarantee frame address mapping table into the control register of the hardware scheduling queue to complete the atomic loading operation of the strong guarantee frame data block to the highest priority slot.

[0256] Obtain the memory address pointer and the physical index number of the highest priority slot from the strong guarantee frame address mapping table, and use them as the input data source for the atomic load operation.

[0257] Perform a mutex lock status check on the hardware scheduling queue control register to confirm that no other process is currently using write permissions, and establish an exclusive access channel to prevent pointer overwrite errors caused by data races.

[0258] The starting memory address pointer of the strong-protection frame is loaded into the CPU general-purpose register, and a transfer descriptor structure containing the source address, target slot index, and data length is constructed using the direct memory access controller.

[0259] By using bus locking instructions or atomic swap primitives, the key control fields in the transfer descriptor are written into the control register group of the hardware scheduling queue in one go, ensuring that the address pointer and priority identifier are updated synchronously within the same clock cycle.

[0260] By using a hardware-level atomicity guarantee mechanism, the risk of race conditions in a multi-threaded environment is eliminated, enabling zero-latency and conflict-free loading of strongly guaranteed frame data blocks to the highest priority slot.

[0261] For example, in a 5G NR channel environment, when the MCS index is detected to drop sharply from 15 to 8 and the interference level exceeds -70dBm, the system determines that it has entered the transition phase region. At this time, the strong guarantee frame is the first I-frame in the GOP, with a memory starting address of 0x8A000000. The highest priority slot index of the hardware scheduling queue is 0. The system generates a transmission descriptor containing a source address of 0x8A000000, a destination index of 0, and a length of 4096 bytes. An atomic write operation is performed on the control register address 0x40001000 using the LDREX / STREX instructions of the ARM architecture. This operation is completed within one CPU instruction cycle, taking approximately 0.5 nanoseconds. Compared to the microsecond-level context switching overhead required for traditional software queue insertion, this atomic loading operation significantly reduces scheduling latency, ensuring that critical video frames receive immediate transmission guarantees when the channel deteriorates, effectively avoiding video stuttering caused by scheduling delays.

[0262] S7.5: Based on the highest priority slot that has completed the loading operation, send a forced preemptive transmission trigger signal to the media access control layer driver to perform a frame-level priority anchoring operation based on channel state evolution phase locking and immediately start data transmission.

[0263] This application also provides a system for a channel state-aware dynamic adaptive transmission method for OTT video streams. This system corresponds to the aforementioned method and can implement all the technical processes described in the method section. The system includes: a channel awareness module, a phase mapping module, a feature extraction module, a table building module, a transition detection module, a scheduling instruction generation module, and a priority anchoring module. Specifically: The channel sensing module is used to acquire a multi-dimensional channel parameter stream, which includes at least four of the following: MCS index, channel occupancy rate, interference level, and RSSI fluctuation entropy value. The phase mapping module is used to perform polar coordinate transformation on the multidimensional channel parameter stream, mapping the multidimensional channel parameter stream to the coordinate axes of a four-dimensional unit hypersphere to generate a phase vector; The feature extraction module is used to extract geometric features from the time series of phase vectors on a four-dimensional unit hypersphere, and calculate a phase evolution feature set including angular velocity, curvature and local dwell time. The table building module is used to perform cluster analysis on the trajectory on the four-dimensional unit hypersphere using the phase evolution feature set, extract typical transition phase regions, parse the GOP structure of the video stream to obtain the fixed offset position of each frame, associate and match the typical transition phase regions with the fixed offset positions, and generate a phase-frame anchoring table based on the matching results. The transition detection module is used to determine whether the phase vector has entered the transition phase region based on the phase evolution feature set. If a sudden increase in curvature is detected and the angular velocity exceeds a preset threshold, the phase stability filter is triggered to check the continuous dwell time of the transition phase region. The scheduling instruction generation module is used to generate a scheduling queue instruction if the phase stability filter confirms that the phase vector has continuously resided in the transition phase region for more than three sampling periods, and then queries the corresponding anchor frame position identifier according to the phase-frame anchor table. The priority anchoring module is used to write the video frame marked as the anchor frame position identifier into the highest priority slot of the scheduling queue according to the scheduling queue instruction, thus completing the priority anchoring operation.

[0264] For those skilled in the art, various other corresponding changes and modifications can be made based on the technical solutions and concepts described above, and all such changes and modifications should fall within the protection scope of the claims of this invention.

[0265] Unless otherwise defined, the technical or scientific terms used herein shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains. The terms “first,” “second,” “third,” and similar terms used in this patent application specification and claims do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Similarly, the terms “an” or “a” and similar terms do not indicate a quantity limitation, but rather indicate the presence of at least one. The terms “comprising” or “including” and similar terms mean that the element or object preceding “comprising” or “including” encompasses the element or object listed following “comprising” or “including” and its equivalents, and do not exclude other elements or objects. The “multiple” mentioned in the embodiments of this application refers to two or more. A and / or B indicate three possibilities: A; B; and A and B.

[0266] The above description is merely an exemplary embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and such modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A channel-state-aware dynamic adaptive transmission method for OTT video streams, characterized in that, include: Obtain a multi-dimensional channel parameter stream, which includes at least four of the following: MCS index, channel occupancy rate, interference level, and RSSI fluctuation entropy value. The multidimensional channel parameter stream is subjected to polar coordinate transformation to map the multidimensional channel parameter stream to the coordinate axes of a four-dimensional unit hypersphere, thereby generating a phase vector; Geometric features are extracted from the time series of the phase vector on a four-dimensional unit hypersphere to calculate a phase evolution feature set including angular velocity, curvature, and local dwell time. Cluster analysis is performed on the trajectory on the four-dimensional unit hypersphere using the phase evolution feature set to extract typical transition phase regions. The fixed offset position of each frame is obtained by parsing the GOP structure of the video stream. The typical transition phase regions are associated and matched with the fixed offset positions, and a phase-frame anchoring table is generated based on the matching results. Based on the phase evolution feature set, it is determined whether the phase vector has entered the transition phase region. If a sudden increase in curvature is detected and the angular velocity exceeds a preset threshold, the phase stability filter is triggered to perform continuous residence time verification on the transition phase region. If the phase stability filter confirms that the phase vector has continuously resided in the transition phase region for more than three sampling periods, it queries the corresponding anchor frame position identifier according to the phase-frame anchoring table and generates a scheduling queue instruction. According to the scheduling queue instruction, the video frame marked as the anchor frame position identifier is written into the highest priority slot of the scheduling queue to complete the priority anchoring operation.

2. The channel state-aware OTT video stream dynamic adaptive transmission method according to claim 1, characterized in that, The multidimensional channel parameter stream is subjected to polar coordinate transformation to map the multidimensional channel parameter stream to the coordinate axes of a four-dimensional unit hypersphere, generating a phase vector, including: The MCS index data and channel occupancy data in the multidimensional channel parameter stream are obtained. Based on the two-dimensional plane mapping algorithm from Cartesian coordinate system to polar coordinate system, the MCS index data and channel occupancy data are jointly vector synthesized to generate the first quadrant basic phase component that characterizes the channel transmission capacity and load status. The system receives interference level data and RSSI fluctuation entropy value data from a multidimensional channel parameter stream. Using high-dimensional space orthogonal extension transformation technology, the interference level data is mapped to a third-dimensional axial vector and the RSSI fluctuation entropy value data is mapped to a fourth-dimensional axial vector to construct orthogonal disturbance phase components that characterize the channel environmental noise characteristics and signal stability. Based on the fundamental phase component and the orthogonal perturbation phase component in the first quadrant, a four-dimensional hyperspherical radius normalization constraint operation is performed to forcibly project the axial vectors of the four dimensions onto the surface of the unit hyperspherical surface, so as to eliminate the difference in dimensions and generate a normalized four-dimensional channel state vector with a unified geometric metric standard. By applying a time-series sliding window integration mechanism to the normalized four-dimensional channel state vector and combining it with the hyperspherical geodesic distance calculation logic, the discrete normalized four-dimensional channel state vectors are connected into a continuous time evolution path to generate a phase vector that can reflect the smooth transition and abrupt change trend of the channel state in real time. The phase vector is output to the phase evolution feature extraction unit, and this phase vector is used as the sole input source for subsequent calculation of angular velocity, curvature and local dwell time, so as to complete the final morphological transformation from discrete physical layer indicators to continuous geometric phase space trajectory.

3. The channel state-aware OTT video stream dynamic adaptive transmission method according to claim 1, characterized in that, The process involves using the phase evolution feature set to perform cluster analysis on the trajectory on the four-dimensional unit hypersphere, extracting typical transition phase regions, parsing the GOP structure of the video stream to obtain the fixed offset positions of each frame, associating and matching the typical transition phase regions with the fixed offset positions, and generating a phase-frame anchoring table based on the matching results, including: Historical channel state datasets are used as input data sources. Based on the phase vector sequence generated by polar coordinate transformation, a clustering analysis algorithm is performed to extract a set of typical transition phase regions that characterize abrupt changes in channel state, forming a basic phase feature library for anchor table construction. Based on the set of typical transition phase regions, a video stream structure parsing operation is performed. The fixed offset positions of I-frames, P-frames, and B-frames in the standard GOP structure are time-axis aligned and marked to generate a GOP frame position feature set containing frame type identifiers and relative temporal indexes, so as to establish a time reference system for the video content. By performing multidimensional correlation mapping operations using the basic phase feature library and the GOP frame position feature set, the spatial geometric attributes of typical transition phase regions are logically bound to the key frame offsets in the GOP frame position feature set, generating preliminary channel phase to video frame priority mapping pairs to establish a causal connection between phase change and frame importance. The channel phase to video frame priority mapping relationship is discretized and encoded. A bitmask generation algorithm is used to convert the mapping relationship into a binary instruction format to generate a standardized scheduling entry containing anchor frame position identifiers, so as to realize the instruction data structure that can be directly recognized by the hardware scheduling queue. Based on the standardized scheduling entries, a static lookup table construction operation is performed. All standardized scheduling entries are written into the read-only storage unit in the spatial index order of the four-dimensional unit hypersphere to generate the final phase-frame anchor table, so as to complete the deterministic mapping and solidification from continuous channel phase evolution characteristics to discrete hardware scheduling instructions.

4. The channel state-aware OTT video stream dynamic adaptive transmission method according to claim 1, characterized in that, The acquisition of the multidimensional channel parameter stream includes: The raw signal state data output by the physical layer driver is monitored and parsed to separate the MCS index, spectrum occupancy, interference intensity and RSSI fluctuation sample set; Perform sliding window entropy calculation on the RSSI fluctuation sample set to generate RSSI fluctuation entropy parameters; The obtained parameters are then subjected to timing alignment and anomaly removal, and encapsulated into the multidimensional channel parameter stream.

5. The channel-state-aware OTT video stream dynamic adaptive transmission method according to claim 1, characterized in that, The angular velocity is obtained in the following way: Perform a difference operation on the sampling points of the phase vector within the sliding time window to obtain a phase displacement vector sequence; The instantaneous angular velocity scalar is obtained by performing modulus normalization and time interval ratio calculation on the phase displacement vector sequence.

6. The channel state-aware OTT video stream dynamic adaptive transmission method according to claim 5, characterized in that, The curvature is obtained in the following manner: By performing second-order differential geometric calculations using the instantaneous angular velocity scalar and its derivative information, the spatial curvature value of the trajectory of the phase vector on the four-dimensional unit hypersphere is obtained.

7. The channel state-aware OTT video stream dynamic adaptive transmission method according to claim 5, characterized in that, The local dwell time is obtained in the following way: The instantaneous angular velocity scalar is compared with a preset static threshold, and continuous counting is performed to obtain the duration during which the phase vector remains stable within a specific phase region.

8. The method for dynamic adaptive transmission of OTT video streams based on channel state awareness according to claim 1, characterized in that, The continuous dwell time verification of the triggered phase stability filter includes: Perform differential and threshold comparison on the curvature value and the angular velocity value to generate a phase transition initial judgment signal; Perform a logical AND operation on the flag bits in the initial judgment signal to mark the potential transition start point; The phase stability filter is invoked to track the subsequent phase vector sequence and count the continuous dwell period count. The count is then compared with a preset threshold, and the verification result is output.

9. The channel state-aware OTT video stream dynamic adaptive transmission method according to claim 1, characterized in that, The step of querying the corresponding anchor frame position identifier based on the phase-frame anchoring table includes: performing a gridded mapping on the phase vector to generate an index code, retrieving the phase-frame anchoring table using the index code, extracting the GOP fixed offset position, and converting it into the anchor frame position identifier.

10. The system of the channel state-aware OTT video stream dynamic adaptive transmission method according to any one of claims 1-9, comprising: The channel sensing module is used to acquire a multi-dimensional channel parameter stream, which includes at least four of the following: MCS index, channel occupancy rate, interference level, and RSSI fluctuation entropy value. The phase mapping module is used to perform polar coordinate transformation processing on the multidimensional channel parameter stream, mapping the multidimensional channel parameter stream to the coordinate axes of a four-dimensional unit hypersphere, and generating a phase vector; The feature extraction module is used to extract geometric features from the time series of the phase vector on the four-dimensional unit hypersphere and calculate a phase evolution feature set including angular velocity, curvature and local dwell time. The table building module is used to perform cluster analysis on the trajectory on the four-dimensional unit hypersphere using the phase evolution feature set, extract typical transition phase regions, parse the GOP structure of the video stream to obtain the fixed offset position of each frame, associate and match the typical transition phase regions with the fixed offset positions, and generate a phase-frame anchoring table based on the matching results. The transition detection module is used to determine whether the phase vector has entered the transition phase region based on the phase evolution feature set. If a sudden increase in curvature is detected and the angular velocity exceeds a preset threshold, the phase stability filter is triggered to perform continuous residence time verification on the transition phase region. The scheduling instruction generation module is used to generate a scheduling queue instruction if the phase stability filter confirms that the phase vector has continuously resided in the transition phase region for more than three sampling periods, and then queries the corresponding anchor frame position identifier according to the phase-frame anchoring table. The priority anchoring module is used to write the video frame marked as the anchoring frame position identifier into the highest priority slot of the scheduling queue according to the scheduling queue instruction, thereby completing the priority anchoring operation.

Images