An anti-jamming enhancement method for action camera high-rate wi-fi transmission

Abstract

The application discloses an anti-interference enhancement method for high-speed Wi-Fi transmission of a motion camera, and comprises the following steps: acquiring inertial motion data collected by a built-in inertial measurement unit of the motion camera and direction reference information of a receiving end; extracting multi-dimensional motion features based on the inertial motion data; performing real-time classification of a motion mode based on the multi-dimensional motion features, outputting a current motion mode identifier, and at least including a rotation dominant mode, a periodic swing mode and a shock precursor mode and corresponding mode feature parameters; generating a transmission parameter adjustment instruction according to the mode identifier, combining the mode feature parameters and the direction reference information of the receiving end; and performing corresponding transmission parameter feedforward adjustment based on the instruction. The application adopts a response lag limitation which is different from a post-radio frequency feedback, avoids loss and frequent oscillation of a throughput in a high dynamic scene, and improves the wireless link robustness and the real-time performance of video stream transmission in a limit motion environment.

Description

Technical Field

[0001] This invention relates to the field of wireless communication and sensor data fusion technology, and in particular to an anti-interference enhancement method for high-speed Wi-Fi transmission of action cameras. Background Technology

[0002] With the increasing popularity of portable shooting devices, the demand for real-time transmission of high-definition video streams between action cameras and smart terminals is growing. In extreme sports or high dynamic range recording scenarios, maintaining high throughput and low latency of the wireless link is crucial for ensuring smooth real-time video preview and the integrity of massive amounts of business data.

[0003] Currently, portable devices typically use the standard Wi-Fi protocol stack for data interaction, and their link maintenance mainly relies on rate adaptation algorithms and retransmission mechanisms at the media access control layer. These existing rate adaptation algorithms primarily rely on packet loss statistics or acknowledgment frame feedback to perform post-event responses and adjust modulation and coding schemes. Furthermore, when faced with sudden channel degradation, the system often employs a fixed redundancy mechanism for channel coding or an adaptive error correction strategy that only responds after an error is detected. However, when a device experiences high-speed spatial attitude flips, short-period high-frequency physical obstructions, or severe mechanical vibrations, the wireless propagation path changes drastically within a very short time window. In such cases, traditional algorithms based on post-event statistical feedback often have response delays far exceeding the duration of the physical fading itself. This not only fails to provide protection at the moment of fading but also easily leads to continuous oscillations of transmission rate increases and decreases due to feedback lag, potentially even causing overflow of the underlying transmission buffer queue and dense bursts of packet loss.

[0004] Existing link adaptation mechanisms based on pure radio frequency feedback exhibit significant shortcomings in response lag and adaptability when facing complex operating conditions with high dynamics and strong physical interference, easily leading to a sharp drop in throughput and a surge in transmission latency. Therefore, it is necessary to investigate a method that can overcome the limitations of hysteresis feedback and improve the robustness and real-time performance of wireless data stream transmission under complex motion conditions. Summary of the Invention

[0005] Purpose of the invention: To provide an anti-interference enhancement method for high-speed Wi-Fi transmission of action cameras, in order to solve the above-mentioned problems in the prior art.

[0006] Technical solution: A method for enhancing anti-interference capabilities of high-speed Wi-Fi transmission for action cameras, comprising:

[0007] Acquire inertial motion data collected by the built-in inertial measurement unit of the action camera and orientation reference information from the receiver;

[0008] Extracting multidimensional motion features based on inertial motion data;

[0009] Real-time classification of motion patterns based on multidimensional motion features, outputting the current motion pattern identifier and corresponding pattern feature parameters, wherein the classified motion patterns include at least rotation-dominated patterns, periodic oscillation patterns and impact precursor patterns.

[0010] Based on the current motion mode identifier, combined with the mode characteristic parameters, and based on the current motion mode identifier, determining whether to further combine the receiver direction reference information, a transmission parameter adjustment instruction is generated;

[0011] The corresponding transmission parameter feedforward adjustment is executed based on the transmission parameter adjustment command.

[0012] Beneficial effects: This invention employs a response hysteresis limitation that differs from post-event radio frequency feedback, avoiding throughput loss and frequent oscillations in high dynamic scenarios, and improving the robustness of wireless links and the real-time performance of video stream transmission in extreme sports environments. Attached Figure Description

[0013] Figure 1 This is a flowchart illustrating the basic inertial perception and recognition process of the action camera in complex extreme sports scenarios in this application embodiment.

[0014] Figure 2 This is a flowchart illustrating the steps of utilizing the angular velocity component in the mode feature parameters in the embodiments of this application.

[0015] Figure 3 This is a flowchart illustrating the steps of utilizing motion phase and oscillation period in the mode feature parameters in this application embodiment.

[0016] Figure 4 This is a flowchart illustrating the steps in this application to determine whether the duration of zero gravity in the mode feature parameters exceeds a pre-configured threshold for the fall confirmation duration.

[0017] Figure 5 This is a flowchart illustrating the steps of the anti-interference enhancement method for high-speed Wi-Fi transmission of an action camera in this application embodiment. Detailed Implementation

[0018] Example 1, such as Figure 1 As shown, this embodiment details the basic inertial perception recognition process of an action camera in complex extreme sports scenarios, as well as the real-time classification mechanism based on multi-dimensional motion features using a finite state machine. This provides a motion benchmark and classification basis for subsequent feedforward adjustment strategies for different physical interference scenarios. Optionally, the anti-interference enhancement method for high-speed Wi-Fi transmission of the action camera includes:

[0019] Step 101: Obtain inertial motion data collected by the inertial measurement unit built into the action camera and the direction reference information of the receiving end.

[0020] In practical applications of miniaturized portable shooting devices such as action cameras, the real-time performance and accuracy of data acquisition are fundamental to subsequent prediction algorithms. Action cameras typically have a built-in inertial measurement unit (IMU) containing a three-axis accelerometer and a three-axis gyroscope. In this basic process, the system activates the high-speed sampling channel of the IMU to simultaneously acquire raw sequences of three-axis acceleration and three-axis angular velocity at a preset uniform sampling rate.

[0021] Specifically, the uniform sampling rate is selected based on the attitude change frequency characteristics of the action camera. In typical extreme motion scenarios, the attitude change bandwidth of the device usually does not exceed 50 Hz. To satisfy the Nyquist sampling theorem and reserve sufficient oversampling margin for subsequent digital filtering processing, the uniform sampling rate can be set to no less than 200 Hz. The acquired inertial motion data is a six-dimensional time series composed of three-dimensional acceleration components and three-dimensional angular velocity components with uniform timestamps.

[0022] The receiver orientation reference information represents the real-time estimation of the spatial orientation of the Wi-Fi receiver device by the moving camera in its own body coordinate system. Establishing this information is a prerequisite for building a wireless channel spatial prediction model.

[0023] Step 102: Extract multidimensional motion features based on inertial motion data.

[0024] Specifically, since the original inertial motion data contains a large amount of high-frequency mechanical vibration noise and digital noise, it needs to be preprocessed before feature extraction. The system applies a sliding median filter to the signals of each axis, with a window width set to 25 sampling points to suppress pulse interference, and uses a low-pass filter to extract the slowly varying gravity component. For example, a second-order Butterworth low-pass filter with a cutoff frequency of 0.5 Hz can be used to process the triaxial acceleration denoising sequence to obtain the acceleration gravity component reflecting the device's attitude tilt angle. The acceleration gravity component is then subtracted point by point from the denoising sequence to separate the acceleration dynamic component that purely characterizes the motion dynamics.

[0025] Furthermore, multi-dimensional statistical and physical features are extracted from the separated signals. These multi-dimensional motion features specifically include rotational features, periodic motion features, and pre-impact features. For rotational features, the Euclidean norm of the three-axis angular velocity components is calculated to obtain the composite amplitude, and the mean value within a specified time window is statistically analyzed. For periodic motion features, the autocorrelation function is calculated for the composite amplitude sequence of the dynamic acceleration components, and the time delay and normalized amplitude corresponding to the first non-zero peak are extracted. For pre-impact features, the mean value of the total acceleration amplitude within the most recent specified time window is extracted as an index of gravity deviation. These scalar values ​​from each dimension are concatenated to form the multi-dimensional motion features, which are continuously output at an update rate of no less than 50 Hz.

[0026] Step 103: Perform real-time classification of motion patterns based on multi-dimensional motion features, and output the current motion pattern identifier and corresponding pattern feature parameters. The classified motion patterns include at least rotation-dominated patterns, periodic oscillation patterns, and impact precursor patterns.

[0027] To address the issue of drastic fluctuations in Wi-Fi throughput under complex motion scenarios, this embodiment employs a finite state machine based on preset thresholds to determine motion modes in real time. The finite state machine utilizes the multi-dimensional motion features extracted in previous steps to quickly identify three typical physical interference scenarios currently encountered by the device: a rotation-dominated mode caused by antenna null sweep, a periodic oscillation mode caused by human body obstruction, and an impact precursor mode caused by mechanical vibration.

[0028] Accordingly, the state transition rules of the finite state machine are as follows: If the average magnitude of the composite amplitude of the angular velocity in the multidimensional motion characteristics continuously exceeds the pre-configured rotation threshold and the periodic normalized amplitude is lower than the pre-configured lower limit threshold of the period, the current motion mode identifier representing the rotation-dominant mode is output. In some examples, the factory default value of the rotation threshold can be set to 3 radians / second, the duration constraint can be set to more than 100 milliseconds, and the lower limit threshold of the period can be set to 0.4, excluding non-purely periodic motions containing violent rotational components. In the state transition rules of the finite state machine, the impact precursor mode has the highest judgment priority. When the entry conditions of the impact precursor mode and the entry conditions of other modes are met simultaneously, the system prioritizes outputting the current motion mode identifier representing the impact precursor mode, ensuring that the landing impact warning protection is not obscured by other modes.

[0029] Optionally, if the periodic normalization amplitude exceeds a pre-configured period upper limit threshold and the average value of the synthesized angular velocity amplitude is lower than a rotation threshold, a current motion mode identifier representing the periodic oscillation pattern is output. In some examples, the period upper limit threshold can be set to 0.6, indicating that the oscillation motion has a periodic recurrence characteristic.

[0030] Optionally, if the gravity deviation index in the multidimensional motion features is lower than the pre-configured gravity acceleration ratio threshold and the duration exceeds the pre-configured mode entry time threshold, the current motion mode identifier representing the impact precursor mode is output; if none of the above three conditions are met, the current motion mode identifier of the previous moment is maintained or the current motion mode identifier representing the default steady-state mode is output.

[0031] In some examples, the gravitational acceleration ratio threshold can be set to 0.3 times the standard gravitational acceleration G, and the time threshold can be set to 50 milliseconds to filter out transient slight weightlessness interference in daily use.

[0032] While outputting the current motion mode identifier, the system simultaneously extracts and outputs the mode feature parameters necessary for subsequent prediction algorithms under the corresponding mode. In the rotation-dominated mode, the mode feature parameters include instantaneous triaxial angular velocity components; in the periodic oscillation mode, the mode feature parameters include the motion phase and oscillation period calculated based on the autocorrelation peak time delay; in the impact precursor mode, the mode feature parameters include the zero gravity duration and the acceleration residual index; among them, the acceleration residual index is taken as the instantaneous value of the total acceleration amplitude, and its physical meaning is the deviation of the current accelerometer reading from the zero reading under ideal free fall conditions, which is distinguished from the gravity deviation index used for mode classification by taking the average value of the time window.

[0033] If none of the above three mode determination conditions are met, the system maintains the default steady-state transmission mode and does not apply any feedforward intervention to the transmission parameters. At this time, the Wi-Fi link operates autonomously according to the standard rate adaptive algorithm built into the communication chip.

[0034] Step 104: Based on the current motion mode identifier, combined with the mode feature parameters, and based on the current motion mode identifier, determine whether to further combine with the receiver direction reference information, and generate a transmission parameter adjustment instruction.

[0035] Specifically, after the finite state machine undergoes a state transition and stabilizes in a certain motion mode, the system uses the extracted mode feature parameters that are highly correlated with the specific physical interference scenario, combined with the spatial orientation of the receiver in the device coordinate system, i.e. the receiver orientation reference information, to calculate the spatiotemporal characteristics of the upcoming physical channel fading.

[0036] Accordingly, based on feedforward prediction theory, the system pre-calculates the required adjustment strategy several milliseconds or tens of milliseconds before the actual occurrence of physical layer RF fading and encapsulates it as a transmission parameter adjustment command. For different current motion mode identifiers, the transmission parameter adjustment command has different control objects and action granularities, including but not limited to the degradation of modulation and coding schemes, centralized scheduling and allocation of data frame queues in the medium access control layer, and dynamic injection of erasure code redundancy ratio in the application layer.

[0037] Step 105: Execute the corresponding transmission parameter feedforward adjustment based on the transmission parameter adjustment instruction.

[0038] In this embodiment, the generated transmission parameter adjustment command is sent from the application layer to the firmware layer of the Wi-Fi protocol stack, or executed within the system's transmit buffer. Unlike post-event remediation mechanisms that rely on packet loss rate feedback, the forward adjustment of transmission parameters in this embodiment completes pre-deployment and rapid disarmament operations within a very short time window—before the wireless channel has significantly deteriorated, or between the point of deterioration and recovery. Through this type of proactive adjustment that spans the inertial physical domain and the radio frequency channel domain, physical packet loss can be minimized while maintaining maximum throughput.

[0039] Example 2: This example details how, after the finite state machine outputs the rotation-dominant mode identifier, the system uses a kinematic prediction mechanism to calculate the time window during which the antenna radiation null point sweeps across the receiver direction, and implements a pulse-based transmission parameter adjustment strategy.

[0040] Step 201: If the current motion mode identifier represents a rotation-dominant mode, based on the current motion mode identifier, combined with mode characteristic parameters, and determining whether to further combine the receiver direction reference information based on the current motion mode identifier, a transmission parameter adjustment command is generated, including: such as Figure 2 As shown, by utilizing the angular velocity component in the mode characteristic parameters and combining it with the pre-configured antenna null direction and receiver direction reference information, the instantaneous angle between the antenna null direction and the receiver direction and the rate of change of the angle are calculated.

[0041] Specifically, the pre-configured antenna null direction is extracted by pre-measuring the three-dimensional radiation pattern of the motion camera's built-in antenna in a microwave anechoic chamber. The system identifies directional regions where the gain is lower than a preset decibel level of the omnidirectional average, determines the center direction of this region as the null direction unit vector, and pre-stores it in the camera as a firmware parameter. After acquiring the angular velocity component and the receiver direction reference information, the system uses the inverse cosine function arccos to calculate the spatial angular relationship between them. The specific formula for calculating the included angle is as follows:

[0042] ;

[0043] Where θ(t) is the instantaneous angle, n null d is the unit vector corresponding to the pre-configured antenna null direction. rx (t) represents the body orientation vector represented by the receiving end orientation reference information at the current moment, and * is the vector dot product operator.

[0044] In the body coordinate system, the rate of change of the receiving end direction is driven by the camera's own rotation. By taking the time derivative of the above expression for the included angle, the formula for calculating the rate of change of the included angle can be derived as follows:

[0045] θ dot(t)=(omega(t)*(d rx (t)Xn null )) / sin(θ(t));

[0046] Where, θ dot (t) represents the rate of change of the angle, omega(t) is the angular velocity vector composed of the angular velocity components, X is the vector cross product operator, and sin is the sine function. The physical meaning of this formula is that the projection component of the angular velocity vector onto the plane subtended by the null direction and the receiving direction determines the closing velocity of the angle between the two directions. When the calculated rate of change of the angle is negative, it indicates that the antenna null direction is moving towards the receiving direction.

[0047] Step 202, estimating the remaining time for zero-point crossing and the duration of the crossing window based on the instantaneous included angle and the rate of change of included angle, including: comparing the instantaneous included angle with a pre-configured angle protection threshold.

[0048] As an alternative implementation, since the denominator of the above formula for the rate of change of the included angle includes the sine value of the instantaneous included angle, when the direction of the antenna null point almost coincides with the direction of the receiving end, the instantaneous included angle approaches zero, causing the sine value to approach zero and become unstable. To avoid this computational singularity problem, the system introduces a partitioning processing mechanism.

[0049] Specifically, a pre-configured angle protection threshold is defined, which can be set to 1.5 times the pre-stored antenna null angle half-width. The system compares the calculated instantaneous angle with this angle protection threshold in real time to determine whether to use far-field prediction logic or near-field determination logic. The reason for setting the threshold to 1.5 times instead of 1 times is to provide an overlapping buffer to prevent decision gaps when switching from far-field prediction to near-field determination under different angular velocity conditions.

[0050] Step 202a: If the instantaneous included angle is greater than the angle protection threshold, perform first-order linear extrapolation based on the instantaneous included angle and the rate of change of included angle to obtain the remaining time for zero-point crossing.

[0051] Accordingly, when the instantaneous included angle exceeds the angle protection threshold, the system determines that it is currently in the far-field prediction region. Within this region, the calculated rate of change of the included angle is stable and reliable. The system estimates the remaining time required for the antenna radiation null point to reach the receiving end direction using a first-order linear extrapolation method. The specific formula for calculating the remaining time is as follows:

[0052] Δ t_cross =θ(t) / |θ dot (t)|;

[0053] Where, Δ t_cross The remaining time for crossing from zero, θ(t) is the instantaneous angle, |θdot (t)| represents the absolute value of the rate of change of the included angle.

[0054] Meanwhile, the system uses pre-stored zero-angle half-width to estimate the duration of physical channel fading. The specific duration calculation formula is as follows:

[0055] T window =(2*α) / |θ dot (t)|;

[0056] Among them, T window The duration of the crossing window is α, which is the pre-configured antenna null angle half-width.

[0057] As a numerical example, assuming the absolute value of the angle change rate |θ(t)| at a certain moment is approximately 50 radians / second, and the pre-stored antenna null angle half-width is 0.26 radians, substituting these values ​​into the formula yields a crossing window duration of approximately 0.0104 seconds. It can be seen that this crossing window duration differs by an order of magnitude from the tens of milliseconds response period of existing physical layer adaptive algorithms, further confirming the necessity of employing an advance prediction strategy.

[0058] Step 202b: If the instantaneous included angle is not greater than the angle protection threshold, determine the trend of the antenna null direction moving towards the receiving end based on the angular velocity component, estimate the duration of the crossing window based on the amplitude of the angular velocity component and the pre-configured antenna null angle half-width, and generate a pulse modulation coding adjustment command in combination with the duration of the crossing window.

[0059] Specifically, when the instantaneous included angle is equal to or less than the angle protection threshold, the system determines that the zero-point direction has entered the neighborhood of the receiving end direction, and at this time, it switches to the near-field determination region. In this region, the system disables the included angle change rate formula that includes division operations.

[0060] Furthermore, the system determines the motion trend by calculating the sign of the dot product of the angular velocity vector with the plane normal direction subtended by the zero-point direction and the receiving end direction. If the dot product is negative, it indicates that the included angle is still decreasing. Instead of waiting for the reciprocal of the first-order extrapolation time, the system generates a pulse modulation and coding adjustment command and sets the sending time to the current time plus a preset execution delay advance margin. The execution delay advance margin is the transmission delay required from the application layer command to the actual execution parameter switching of the communication chip firmware, which is set to 2 milliseconds in this embodiment.

[0061] In this state, the calculation of the duration of the crossing window obtains the absolute value of the included angle velocity through an equivalent non-division form, avoiding the risk of overflow. Within this near-field direct determination region, the system no longer calculates the formula for the rate of change of the included angle involving division by sin(θ). Instead, it uses the absolute value of the dot product of the angular velocity vector and the normal direction as the equivalent quantity for measuring the projection of the angular velocity into this plane, i.e., |ω(t)·(d rx (t)×n null The equivalent quantity is numerically equal to |θ(t)|·sin(θ(t)). Based on this equivalent quantity, the division-free estimation formula for the duration of the crossing window is: T window =2·α·sin(θ(t)) / |ω(t·(d rx (t)×n null Since the numerator sin(θ(t)) and the denominator sin(θ(t)) factor implicit in the denominator approach zero when θ(t) approaches zero in the near-field region, their ratio remains numerically stable.

[0062] Step 203: Generate a pulse modulation and coding adjustment instruction based on the remaining time of zero-point crossing and the duration of the crossing window, which serves as the transmission parameter adjustment instruction.

[0063] Accordingly, after obtaining the prediction results in the time dimension, the system generates a pulse-based modulation and coding scheme adjustment command. This command carries three core parameters: the down-adjustment time, the recovery time, and the target modulation and coding scheme level. The down-adjustment time is the sum of the current time and the remaining time before zero-point crossing, minus the advance margin. The recovery time is the sum of the down-adjustment time and the duration of the crossing window. The target modulation and coding scheme level can be set to an anti-interference level that can maintain demodulation even under conditions of significant antenna gain reduction, such as quadrature phase shift keying combined with a 50% code rate. Through this mechanism, the down-adjustment of the modulation and coding scheme level exists only in the form of narrow pulses on the time axis, strictly compressing the throughput loss to the unavoidable physical fading range.

[0064] Optionally, the corresponding transmission parameter feedforward adjustment is performed based on the transmission parameter adjustment instruction, including one of the following execution methods: calling the preset external rate coverage interface in the communication chip firmware to send data according to the target level specified in the pulse modulation and coding adjustment instruction within the cross-window duration; or, at the application layer, according to the cross-window duration determined by the pulse modulation and coding adjustment instruction, the data frame to be sent is fragmented, and a low rate protection flag is set in the header of the data frame, so that the underlying transmission module sends the fragmented data frame in low rate protection mode within the cross-window duration.

[0065] At the practical implementation level, this embodiment provides two alternative methods adapted to different levels of hardware openness. The first method is the preferred implementation, suitable for communication chips that allow firmware customization. In this method, an external rate coverage interface is added inside the chip firmware. This interface accepts timestamps and modulation / coding levels. During the coverage period, the chip's internal rate adaptation algorithm suspends autonomous decision-making and obeys the externally specified level for data transmission. After the coverage ends, autonomous operation automatically resumes.

[0066] The second method is a compatibility approach, suitable for scenarios where access to modify the underlying firmware is unavailable. In this method, the system controls the transmission behavior at the application layer. Specifically, during the predicted traversal window, the system proactively reduces the length of the data frame to be transmitted to less than a quarter of its original length, ensuring transmission within the channel coherence time. Simultaneously, a low-rate protection flag is set in the protocol header of the data frame, forcing the underlying chip's built-in rate adaptive algorithm to be more conservative in selecting the transmission level. Although this method cannot specify a particular level, it can still reduce the bit error rate within a very short time window.

[0067] Example 3: This example details how, after the finite state machine outputs a periodic swing mode identifier, the system utilizes the swing period and motion phase obtained from autocorrelation analysis to cross-border fuse the periodic fluctuations of the radio frequency channel quality and implement centralized scheduling of burst transmission based on motion phase locking.

[0068] Step 301: If the current motion mode identifier represents a periodic oscillation mode, based on the current motion mode identifier, combined with mode characteristic parameters, and determining whether to further combine with receiver direction reference information based on the current motion mode identifier, a transmission parameter adjustment command is generated, including: such as Figure 3 As shown, by utilizing the motion phase and oscillation period in the mode feature parameters, combined with the pre-extracted optimal motion phase, the burst transmission window where the current channel quality is higher than the pre-configured channel quality threshold is determined.

[0069] Specifically, in communication scenarios with periodic body occlusion, channel quality fluctuates rhythmically with periodic movements. The system establishes a unified time reference by utilizing the extracted oscillation period and the real-time updated motion phase. Simultaneously, the system extracts the state point corresponding to the minimum channel fading and the smoothest propagation path as the optimal motion phase. Based on this optimal motion phase, extending to both sides, the system defines a continuous phase interval as a burst transmission window. Within the physical time period corresponding to this window, the expected channel quality is higher than a specific threshold setting, providing the physical conditions for high-throughput data transmission.

[0070] Step 302: Based on the phase difference between the current motion phase and the start and end boundaries of the burst transmission window, as well as the swing period, calculate the remaining time until the next burst transmission window opens and the duration of the window; generate a burst transmission scheduling instruction based on the remaining time until the next burst transmission window opens and the duration of the window, as a transmission parameter adjustment instruction.

[0071] Accordingly, the system calculates the phase difference between the current motion phase and the starting boundary of the next burst transmission window, and divides it by the angular frequency calculated based on the oscillation period to obtain the remaining time required to open the window. Subsequently, the system encapsulates the remaining time, the duration of the window after it is opened, and the required target transmission rate together to generate a burst transmission scheduling instruction.

[0072] Step 303: Execute the corresponding transmission parameter feedforward adjustment based on the transmission parameter adjustment instruction, specifically including: centrally sending the data to be sent in the buffer queue within the burst transmission window, and pausing transmission outside the burst transmission window.

[0073] Optionally, upon receiving a burst transmission scheduling instruction, the media access control layer executes specific buffer queue scheduling actions. During the physical period when the burst transmission window is open, the system calls a higher modulation and coding level to centrally clear the data packets accumulated in the transmit buffer queue; during the physical period when the burst transmission window is closed, the system freezes the transmit queue and only performs data buffer accumulation. This mechanism avoids packet loss and retransmission caused by invalid transmissions during poor channel phases.

[0074] Optionally, the optimal motion phase is extracted through an online calibration phase, which includes: synchronously recording the motion phase over multiple oscillation cycles and the reported received signal strength indication value.

[0075] In this embodiment, to accurately establish the mapping relationship between the motion domain and the radio frequency domain, the system performs online calibration in the initial stage of entering the periodic oscillation mode. Within a preset continuous oscillation period, the system maintains normal constant-rate transmission without intervention or scheduling, but simultaneously acquires two sets of time series with aligned timestamps. One set is the continuously output motion phase, and the other set is the received signal strength indication value reported by the underlying communication chip frame by frame.

[0076] Optionally, the motion phase and the received signal strength indication value are averaged over an interval, and a single-frequency sine model is fitted to establish a phase-channel quality mapping table that includes the full-cycle average channel quality, half-peak fluctuation amplitude, and optimal motion phase.

[0077] Furthermore, to suppress random interference from multipath fast fading and reduce the computational load on embedded devices, the system divides the complete oscillation period into multiple equally wide phase intervals, and calculates the average of the received signal strength indicators falling within the same phase interval. The least squares method is then used to fit a single-frequency sinusoidal model to the processed discrete data. The analytical formula used in this fitting process is as follows:

[0078] Q(φ)=Q bar +Δ Q *cos(φ-φ opt );

[0079] Where Q(φ) represents the expected channel quality corresponding to the motion phase φ, Q bar For the average channel quality over the entire period, Δ Q φ is the half-peak fluctuation amplitude of the channel quality. opt This represents the optimal motion phase corresponding to when the channel quality reaches its peak. The phase-channel quality mapping table established using this formula completes the parameterization and solidification of the physical motion laws.

[0080] Optionally, the start and end boundaries of the burst transmission window can be determined by combining the pre-configured threshold coefficients with the phase-channel quality mapping table.

[0081] Specifically, the system defines a channel quality threshold and selects a continuous phase interval greater than this threshold as the burst transmission window. The specific formula for calculating the burst transmission window phase half-width is as follows:

[0082] Δ = arccos(β);

[0083] Where Δ is the phase half-width of the burst transmission window, β is the pre-configured threshold coefficient, which is set to 0.5 in this embodiment, and arccos is the inverse cosine function. Further, the system uses the optimal motion phase as the center point, subtracting and adding the phase half-width respectively to obtain the start and end boundaries of the burst transmission window.

[0084] Optionally, before generating the burst transmission scheduling instruction, the method may also include: obtaining the highest supported transmission rate reported by the communication chip and the average output bit rate of the current video encoder.

[0085] In this embodiment, the constraints of physical layer hardware limits are considered. The system reads the highest supportable transmission rate that the communication chip can maintain under the current signal-to-noise ratio conditions through the low-level interface, and at the same time reads the average output bitrate of the current video encoder from the application layer, providing basic input data for subsequent rate feasibility assessment.

[0086] Optionally, the target transmission rate required for concentrated transmission within the burst transmission window can be calculated.

[0087] Specifically, because the scheduling mechanism pauses transmission outside the window, to avoid buffer queue overflow and meet the real-time constraints of the video stream, the transmission gap outside the window must be compensated for with a higher rate within the window. The specific formula for calculating the target transmission rate is as follows:

[0088] R target =R video / eta; where R target For the target transmission rate, R video eta represents the average output bitrate of the current video encoder, and eta represents the time percentage corresponding to the currently set burst transmission window.

[0089] Optionally, if the target transmission rate is greater than the maximum supported transmission rate, the burst transmission window can be adaptively expanded by calculating the minimum window percentage required to satisfy the physical layer rate constraint.

[0090] Accordingly, when the required target transmission rate exceeds the hardware's physical limits, the original burst transmission window is insufficient to handle the entire backlog of data. The system employs a reverse derivation mechanism to prioritize the transmission needs of the data volume. The formula for calculating the minimum window size is as follows:

[0091] eta min =R video / R max ;

[0092] Among them, eta min R is the minimum window size. max This represents the maximum supported transmission rate.

[0093] The system updates the pre-configured threshold coefficient based on the minimum window size, using the following formula:

[0094] β new =cos(π*eta min );

[0095] Where, β new Here, π is the updated threshold coefficient, π is the constant pi, and cos is the cosine function. For example, suppose the calculated eta... min The value is 0.4. Substituting this into the formula, we can obtain β. new The threshold coefficient is approximately 0.31. The system lowers the threshold coefficient to this value, adaptively expanding the width of the physical transmission window while maintaining the core mechanism of burst scheduling.

[0096] Optionally, if the minimum window percentage corresponding to the expanded burst transmission window reaches the pre-configured upper limit, the burst transmission scheduling in the periodic swing mode is exited, and a suggestion to reduce the bitrate is sent to the current video encoder.

[0097] As an optional implementation, to prevent excessive expansion of the burst transmission window from negating scheduling benefits, the system sets a lower limit constraint on the threshold coefficient. If the window proportion corresponding to the calculated updated threshold coefficient exceeds the pre-configured upper limit proportion, in this embodiment, the pre-configured upper limit proportion is set to 0.8. That is, when the required minimum window proportion exceeds 80% of the entire swing cycle, the scheduling benefits are insufficient to cover the buffering overhead of centralized transmission. This indicates that the current channel has reduced the global physical transmission capacity to below the minimum service quality threshold required by the existing service. At this time, the system stops the centralized scheduling logic, resumes the normal rate transmission mode, and outputs a code rate reduction suggestion containing specific values, such as setting the code rate upper limit to 0.8*R. max This ensures the basic connectivity of data links.

[0098] Optionally, before establishing the phase-channel quality mapping table, a conservative pre-scheduling state is performed, including: initializing the optimal motion phase using the zero-crossing phase of the prior acceleration dynamic component in the multidimensional motion features.

[0099] Specifically, during the transition period when the system has just switched to the periodic oscillation mode and the online calibration has not yet been completed, the system initiates a conservative pre-scheduling mechanism without definite RF feedback. Since the moment when the device is facing forward often has the best line-of-sight propagation conditions in typical oscillation scenarios, the system extracts the phase of the dynamic acceleration component that has undergone a zero-point transition in the pre-extraction feature step, assigns it to the optimal motion phase, and completes the prior estimation of the channel peak.

[0100] Optionally, centralized transmission can be performed by setting an initial burst transmission window based on a pre-configured conservative threshold coefficient.

[0101] Accordingly, to absorb potential phase deviations in prior estimates, the system does not use conventional threshold parameters, but instead employs a conservative threshold coefficient with a larger value. The burst transmission window determined using this conservative threshold coefficient is narrower, for example, the window percentage is limited to within 30%, ensuring that the physical time period of centralized data transmission falls within the range of optimal channel quality with a very high probability, thus avoiding transmission failures caused by misjudgments.

[0102] Optionally, as data accumulates during the online calibration phase, the optimal motion phase and conservative threshold coefficients are linearly transitioned to the fitted steady-state values.

[0103] Specifically, as calibration data is acquired cycle by cycle, the system performs a weighted fusion of parameters based on prior estimates and parameters based on data fitting. By adjusting the corresponding weight ratios, the optimal motion phase smoothly evolves from the initial pure prior value to the final sinusoidal fitted analytical value, while controlling the conservative threshold coefficient to linearly decrease to the target threshold coefficient set in steady state. This transition process fills the strategy vacuum in the initial stage of mode switching and establishes full-time transmission anti-interference protection.

[0104] Example 4: This example details the complete process of how, after the finite state machine outputs the impact precursor mode identifier, the system uses the purely physical prior signal of free fall to inject an outer layer of error correction coding to the data stream during the airborne phase before the landing impact, and performs cooperative decoding at the receiving end.

[0105] Step 401: If the current motion mode identifier represents an impending impact mode, based on the current motion mode identifier, combined with mode characteristic parameters, and determining whether to further combine it with receiver direction reference information, a transmission parameter adjustment instruction is generated, including: such as... Figure 4 As shown, it is determined whether the zero gravity duration in the mode feature parameters exceeds the pre-configured fall confirmation duration threshold, and whether the acceleration residual index in the mode feature parameters is lower than the pre-configured acceleration ratio threshold.

[0106] Specifically, in scenarios such as motocross jumps or extreme skateboarding stunts, the device inevitably undergoes a period of freefall before landing impact. The system executes a dual-judgment logic based on the zero-gravity duration and acceleration residual indicators extracted by the front-end module. The pre-configured fall confirmation duration threshold can be set to 50 milliseconds to eliminate false judgments caused by transient hand disturbances during daily use. The pre-configured acceleration ratio threshold can be set to 0.15 times the gravitational acceleration to confirm that the current state is pure freefall rather than low-dynamic conventional motion.

[0107] Step 402: If all conditions are met simultaneously, confirm that the system is in free fall, open the impact warning window, and generate an instruction to inject outer layer error correction code into the data to be transmitted as a transmission parameter adjustment instruction; if the conditions are not met simultaneously, maintain the current transmission parameters unchanged and continue to monitor the mode characteristic parameters.

[0108] When both of the above conditions are met simultaneously, the system determines that a landing impact is imminent, even before any signs of deterioration appear in the physical radio frequency link, and immediately opens an impact warning window at the application layer. Simultaneously, it generates a command stream to control the coding behavior of the application layer. By extracting the free-fall state as a precursor signal, the system obtains a sufficient warning lead time. Taking a jump of 1 meter in height as an example, the total free-fall time is approximately 450 milliseconds. Subtracting the 50 milliseconds required for determination and confirmation, the system can obtain a warning lead time of approximately 400 milliseconds, far exceeding the processing latency required for subsequent parameter switching.

[0109] Step 403: Execute the corresponding transmission parameter feedforward adjustment based on the transmission parameter adjustment instruction. Specifically, this includes: applying outer error correction processing to the data to be transmitted in the transmission buffer queue within the impact warning window.

[0110] Accordingly, upon receiving the warning command, the system immediately initiates an outer error correction mechanism for newly entered service data in the transmission buffer queue. This protection state continuously covers the period of severe oscillation during the landing impact. After the actual landing, when the system detects that the total acceleration amplitude has fallen back to within 20% of the standard gravitational acceleration and has remained there for more than 200 milliseconds, it confirms that the mechanical oscillation has subsided, withdraws the outer error correction processing, and restores the standard data transmission mode.

[0111] Optionally, within the impact warning window, an outer error correction process is applied to the data to be sent in the transmission buffer queue, including: dividing the data to be sent into equal-length source data groups.

[0112] Specifically, to implement packet-level channel coding at the application layer, the system extracts continuous data to be transmitted from the video encoder or other service modules and divides it into data blocks of uniform length. Furthermore, the payload length of each source data packet is aligned according to the underlying communication protocol, such as matching the maximum transmission unit size of the underlying data frame, to avoid error propagation caused by secondary segmentation of single-frame data packets by the physical layer.

[0113] Optionally, multiple consecutive source data can be grouped into a single encoding group according to a pre-configured encoding dimension.

[0114] Specifically, the system sets the encoding dimension based on a preset safety margin. This encoding dimension specifies the number of original data blocks required to form a single logical recovery unit. Following the temporal sequence of the data flow, the system groups and aggregates a specified number of source data into independent encoding groups, which serve as the input matrix for subsequent mathematical encoding operations.

[0115] Optionally, outer erasure coding is performed within the coding group to generate redundant groups with a number that conforms to the pre-configured redundancy dimension.

[0116] In this embodiment, the outer erasure coding can employ the Reed-Solomon algorithm. In other embodiments, the outer erasure coding can also employ well-known system erasure codes such as fountain codes or low-density parity-check codes, which can be selected by those skilled in the art based on coding complexity and latency requirements. The system applies algebraic operations byte-by-byte in the column direction to all source data packets within a single coding group, calculating and generating a specified number of redundant packets. The number of redundant packets is determined by a pre-configured redundancy dimension. The generated redundant packets and the source data packets constitute the complete transmission sequence of the coding group, enabling the receiver to recover all the original data through computation even if it loses packets not exceeding the number of redundancy dimensions.

[0117] Optionally, an interleaved block consisting of multiple coding groups is maintained simultaneously. The source data packets and redundant data packets in each group are extracted and sent in a round-robin manner across coding groups, thereby distributing continuous burst packet loss to different coding groups.

[0118] Specifically, the ability of a single coding group to withstand continuous packet loss has a physical upper limit. To mitigate extremely long burst fading caused by landing impacts, the system introduces a cross-coding group interleaving scheduling strategy. The system establishes multiple coding group transmission buffers in parallel in memory; this number of parallel buffers is the interleaving depth. During the transmission phase, the system does not use a sequential transmission mode, but instead sequentially retrieves one data packet from each parallel coding group and submits it to the underlying transmission module.

[0119] Under this interleaving logic, the interleaving parameters need to satisfy specific physical constraint formulas. Specifically, the product of the interleaving depth and the redundancy dimension must be greater than or equal to the maximum number of consecutive packet losses within the expected burst window. In actual video preview services, to avoid the buffer depth caused by interleaving transmission violating real-time constraints, the system needs to perform parameter trade-offs. Assuming the encoding dimension is set to 4 and the redundancy dimension to 2, and the interleaving depth is set to 150, the total number of packets in the interleaved block, including source data and redundant data, is 900. At a typical transmission rate of 70 Mbps, the total transmission latency is approximately 155 milliseconds, lower than the preset latency limit of 200 milliseconds. This configuration can tolerate 300 consecutive frames of loss, covering the initial 50-millisecond period of the impact process when packet loss is most concentrated.

[0120] Optionally, applying outer error correction processing to the data to be sent in the transmission buffer queue further includes: embedding an outer encoding header at the beginning of each source data packet and redundant packet, the outer encoding header carrying an encoding enable flag, encoding group number, intra-group packet number, encoding dimension, and redundancy dimension.

[0121] Accordingly, to achieve stateless collaborative operation between the sender and receiver, the sender reserves a fixed space at the front end of the payload of each application layer packet for writing a 4-byte outer encoding header. The encoding enable flag occupies 1 bit and indicates whether the current packet belongs to the warning protection flow. The encoding group sequence number occupies 15 bits and uniquely identifies the logical block to which the current packet belongs in the global timing. The intra-group packet sequence number occupies 8 bits and marks the relative position of the current packet within the group. The encoding dimension and redundancy dimension each occupy 4 bits and carry real-time parameter configurations. Since a data frame typically contains 1500 bytes, the protocol overhead caused by this 4-byte extra header structure is less than 0.27%, and the impact on throughput is kept to a very low level.

[0122] Optionally, when the receiving end parses that the encoding enable flag is enabled, it independently stores the received packets into the corresponding buffer slots using the encoding group sequence number, and when the number of received packets in any encoding group reaches the encoding dimension, it triggers erasure decoding based on the outer encoding header to restore all source data packets in that group.

[0123] Specifically, at the communication receiving device, the application layer software parses the outer coding header frame by frame. Since each frame of data carries a complete group number and intra-group index characteristics, the interleaving structure is transparent at the receiving end. The receiving module only needs to assemble the packets into the corresponding memory slots based on the parsed coding group number. The system maintains an arrival counter for each slot. As soon as the counter accumulates to the value of the coding dimension indicated by the outer coding header, regardless of whether the received data is source data or redundant data, the system immediately calls the corresponding erasure matrix algorithm to recover the packet content erased by the burst fading, removes the coding header, and submits it to the upper-layer decoder, thus achieving continuous service flow under severe physical impacts.

[0124] Example 5: This example details the active and passive initialization process in the direction of the receiver, and how to distinguish between physical space movement and multipath environmental fluctuations through a variance discrimination mechanism, and perform source-weighted closed-loop correction.

[0125] Optionally, during the initial establishment phase of the communication connection, a test frame sequence is sent to the receiving end, and the attitude quaternion change trajectory output by the inertial measurement unit is recorded simultaneously.

[0126] Specifically, in the initial stage of establishing a wireless connection between the action camera and the receiving device, the system lacks accurate relative spatial orientation awareness. To establish this benchmark, the system prioritizes an active calibration strategy. A sequence of test frames is continuously sent to the receiving end via application-layer signaling; the length of this transmission period can be set to no more than 2 seconds. Within this test time window, the system uses the three-axis angular velocity data output by the inertial measurement unit to integrate in real time, tracking and recording the continuous change trajectory of the body attitude quaternion in the world coordinate system.

[0127] Optionally, based on the received signal strength indication value reported by the receiver for the test frame sequence, the forward axis direction of the camera corresponding to the optimal sampling point of the signal is selected as the initial receiver direction.

[0128] Accordingly, when the receiving end sends back an acknowledgment frame, the system synchronously extracts the received signal strength indication value corresponding to each frame. Under short-range line-of-sight propagation physical conditions, the captured signal strength reaches its maximum when the antenna main lobe is facing the receiving end. After the test, the system filters the recorded data and extracts the top 10% of the sampling points with the highest received signal strength indication values. The unit vector of the camera's forward axis direction at the corresponding time of these sampling points is extracted, and then subjected to arithmetic averaging and normalization. The resulting composite vector is used as the optimal estimate of the initial receiving end direction in the world coordinate system.

[0129] Furthermore, as an alternative, if the total rotation angle spanned by the attitude quaternion is less than 30 degrees within a 2-second test period, it indicates that the amplitude of the body attitude change is too small to distinguish signal differences in spatial direction. In this case, the system extracts the camera forward axis direction at the moment of connection establishment as the initial estimate, and triggers a passive fast convergence mechanism during the subsequent 30-second run. In this accelerated convergence mode, the forgetting factor is temporarily reduced and the correction update frequency is increased.

[0130] Optionally, in subsequent operation, the real-time attitude quaternion is continuously updated based on the inertial motion data through continuous integration, and the initial receiver orientation is transformed to the current body coordinate system to obtain continuously updated receiver orientation reference information.

[0131] Specifically, after initialization, within each sampling period, the system constructs the currently acquired three-axis angular velocity as a pure quaternion angular increment, multiplies it with the attitude quaternion from the previous moment, and maintains the real-time attitude quaternion update. Using the conjugate rotation operator of this real-time attitude quaternion, the initial receiver orientation, fixed in the world coordinate system, is dynamically transformed to the current body coordinate system. The resulting vector is the receiver orientation reference information, representing the real-time spatial orientation of the receiver from the camera's own perspective. The update delay is only the inertial sampling period, not exceeding 5 milliseconds.

[0132] Optionally, the method further includes performing source-weighted online correction on the receiver direction reference information.

[0133] Specifically, during long-term operation, the user's physical movement and the integral drift of the inertial sensor will gradually cause an accumulated deviation in the established receiver orientation reference information. Simultaneously, the complex multipath propagation environment will also cause random fluctuations in the radio frequency signal. If the spatial orientation is corrected indiscriminately using signal fluctuation feedback, it will lead to system miscalibration. Therefore, the system introduces a weighted correction mechanism based on the identification of the deviation source.

[0134] Optionally, the actual packet loss rate and received signal strength indication value during transmission are compared with the corresponding predicted values ​​calculated based on the transmission parameter adjustment command for each motion mode, and the prediction deviation for each motion mode is calculated.

[0135] Accordingly, after generating and executing the transmission parameter adjustment instructions for each mode, the system collects the actual packet loss rate and received signal strength indication value at the specified frequency. In the rotation-dominated mode, the system checks the actual packet loss rate characteristics within the zero-point crossing window. If no increase is found, the system calculates the prediction deviation in the form of an angle based on the time difference between the peak value of the actual packet loss rate and the prediction center time.

[0136] The conversion method is as follows: multiply the time difference between the actual peak packet loss rate and the predicted zero-point crossing center time by the angular velocity scalar at that moment to obtain the prediction error in angular form. In periodic oscillation mode, the system checks the average received signal strength indication value within the burst window. If the system is systematically low, a prediction error in phase form is generated. The conversion method is as follows: substitute the average received signal strength indication value within the burst window into the phase-channel quality mapping table for reverse lookup, calculate the offset based on the peak phase, and obtain the prediction error in phase form.

[0137] Optionally, the variances of the trend component and the fluctuation residuals of the prediction bias can be calculated separately to identify whether the prediction bias is dominated by multipath environmental fluctuations.

[0138] In this implementation, for the prediction bias extracted from the rotation mode, since it originates from deterministic measurements of the antenna null physical characteristics and has minimal multipath interference, the system assumes it reflects the drift trend. For the prediction bias extracted from the periodic mode, the system statistically analyzes the linear regression slope over the most recent 10 swing cycles as the trend component and calculates the residual variance. If the residual standard deviation exceeds twice the absolute value of the trend component, the system determines that environmental fluctuations dominate the prediction bias and do not reflect physical spatial movement; conversely, if the residual standard deviation is small, the system determines that the bias mainly originates from receiver movement or integral drift.

[0139] Optionally, adaptive weights are configured based on the prediction bias of different mode sources based on the identification results, and the initial receiver direction is corrected by an exponentially weighted moving average.

[0140] As an optional implementation, the system maintains a weighted cumulative value for the receiver orientation correction based on the aforementioned identification results. Specifically, the prediction bias from the rotation mode is configured with a fixed weight of 0.8; the prediction bias from the periodic mode is configured with a dynamic weight, set to 0.5 when the trend component is dominant and to 0 when environmental fluctuations are dominant. The system uses this weighted cumulative value to perform an update operation on the initial receiver orientation, with the correction formula as follows:

[0141] d rx_world_new =Rot(Δ φ *e corr )*d rx_world_old ;

[0142] Where, d rx_world_new The updated initial receiver direction vector; Rot is the rotation operator that rotates the receiver by a specified angle around a specified axis; Δ φ e is the angle correction amount for weighted cumulative calculation; corrThe corrected rotation axis direction is a unit vector inferred from the deviation. Specifically, for prediction deviations from the dominant rotation mode, the corrected rotation axis direction is taken as the direction of the normalized angular velocity at the moment the prediction deviation occurs; for prediction deviations from the periodic oscillation mode, the corrected rotation axis direction is taken as the normal direction of the oscillation motion plane. Those skilled in the art can determine the appropriate corrected axis direction based on the physical motion geometry corresponding to the source of the deviation; d rx_world_old This is the initial receiver direction vector before the update.

[0143] Furthermore, to prevent drastic changes in direction estimation caused by sudden multipath mutations, the system sets a reasonable truncation constraint on the angle correction amount in a single operation, forcibly limiting the absolute value of the angle correction amount in a single operation to no more than 2 degrees.

[0144] Example 6: This example details how a smooth transition control is implemented through a linear ramp mechanism and a coding group boundary alignment mechanism when the system state machine jumps.

[0145] Optionally, when a change in the current motion mode identifier is detected, a transition time window of a pre-configured duration is initiated.

[0146] Specifically, during continuous system operation, the state classification module outputs the current motion mode identifier in each frame and compares it with the identifier of the previous frame. If they are inconsistent, the system confirms a change in the physical interference scenario and generates a mode switching trigger signal. Upon receiving this trigger signal, the system temporarily suspends the full execution of the new mode and instead initiates a transition mechanism. The length of the transition time window can be set to 200 milliseconds. Within this time window, the system allocates computing resources to perform fusion and smoothing processing of the two sets of transmission parameter adjustment strategies.

[0147] Optionally, within the transition time window, the adjustment amplitude of the transmission parameter adjustment command corresponding to the previous motion mode is controlled to decay to zero according to a linear ramp, and the adjustment amplitude of the transmission parameter adjustment command corresponding to the new motion mode is controlled to rise from zero to full amplitude.

[0148] In this embodiment, to prevent the underlying communication protocol stack from falling into scheduling chaos due to parameter jumps, the system implements gradual intervention within the transition time window. The specific manifestation of the intervention intensity differs for different motion modes. In transitions involving pulsed modulation and coding scheme switching, the intervention intensity is manifested as the modulation and coding scheme downsampling amplitude. In this case, the downsampling amplitude during the transition gradually decreases from full amplitude until it stops downsampling, or gradually increases from no downsampling to full amplitude downsampling.

[0149] Furthermore, in transitions involving burst transmission scheduling, the intensity of the effect is reflected in the degree to which transmission is paused outside the burst transmission window. In this case, the transmission control logic during the transition gradually transitions from a paused transmission state to a continuous transmission state, or performs a reverse gradual transition. In transitions involving outer layer error correction coding, the intensity of the effect is reflected in the redundancy dimension. In this case, the number of additional redundant packets during the transition gradually decreases from the target setpoint to 0, or gradually increases from 0 to the target setpoint. By implementing linear ramp control on the above-mentioned intensity levels, the system maintains the continuity of the transmission link during mode switching and avoids sudden data congestion.

[0150] Optionally, when it involves the transition of applying an outer layer of error correction processing to the data to be sent, the gradual change in the strength of the action is limited to performing a group-by-group increase or decrease of the redundancy dimension with the complete coding group as the boundary.

[0151] As an alternative implementation, during the transition of error correction coding strategies, since the decoding at the receiving end depends on a fixed matrix dimension, changing the redundancy dimension within a single coding group will cause the encoding and decoding parameters at both ends to lose synchronization. Therefore, the system introduces strict boundary alignment constraints, keeping the coding dimension and redundancy dimension within each coding group fixed to ensure that the receiving end can correctly decode the coding group based on the parameters carried in the data frame header.

[0152] Accordingly, the system restricts any changes to encoding parameters to occur only at the boundaries of adjacent encoding groups. That is, after the current encoding group has completed the encoding and distribution of all packets, the system can adopt different encoding dimensions and redundancy dimensions for the next encoding group. During the transition time window, the redundancy dimension of each newly opened encoding group is reduced by 1 or increased by 1 relative to the previous encoding group, until the redundancy dimension reaches the target set value.

[0153] Furthermore, taking a coding dimension of 4 and a target redundancy dimension of 2 as an example, during the transition process of the system exiting the impact precursor mode, the parameters of the first transition coding group are configured to include 4 source data groups and 2 redundant groups. The parameters of the second transition coding group are changed to include 4 source data groups and 1 redundant group. The parameters of the third transition coding group are further changed to include 4 source data groups and 0 redundant groups, that is, only pure source data. After that, the system removes the coding instructions and enters normal transmission.

[0154] Optionally, at a typical transmission rate of 70 megabits per second, the physical transmission time for each coding group is approximately 0.7 to 1 millisecond. Therefore, the stepwise transition of parameters for 2 to 3 coding groups can be completed within 2 to 3 milliseconds, which is much shorter than the preset transition time window of 200 milliseconds. By performing gradual changes in action strength at the granularity of complete coding groups, the receiver can decode independently group by group, unaffected by the negative impact of the overall parameter change trend during the transition.

[0155] Numerical analysis of the above embodiments shows that the feedforward adjustment mechanism can compress the response delay of transmission parameter changes from tens of milliseconds in traditional post-feedback algorithms to within a few milliseconds. In the rotation-dominated mode, the down-adjustment window of the modulation coding level is precisely limited to the physical fading duration, significantly reducing the effective throughput loss compared to the low-rate operation caused by feedback lag of traditional algorithms, which lasts for several times or even tens of times longer. In the periodic oscillation mode, centralized scheduling of burst transmissions avoids the waste of resources in invalid transmissions and probe retransmissions within obstructed blind zones. In the impulsive precursor mode, the application layer outer-layer error correction coding transparently resolves long burst packet losses that the underlying layer cannot recover autonomously at the receiver. Those skilled in the art will understand that the specific performance improvement may vary depending on antenna radiation pattern characteristics, movement speed, multipath environment complexity, and hardware platform differences.

[0156] Example 7: According to one aspect of this application, the anti-interference enhancement method for high-speed Wi-Fi transmission of an action camera further includes, as follows: Figure 5 As shown:

[0157] S1. Inertial motion data acquisition and multidimensional motion feature extraction.

[0158] S11, synchronous acquisition of six-axis inertial data.

[0159] The high-speed sampling channel of the action camera's built-in six-axis inertial measurement unit is activated, and raw three-axis acceleration and angular velocity sequences are simultaneously acquired at a uniform sampling rate of no less than 200 Hz. Both sets of sequences share the same timestamp reference, and each sampling point includes a timestamp, three acceleration component values, and three angular velocity component values, forming a six-dimensional synchronized time series. The sampling rate is selected based on the following criteria: the attitude change bandwidth of the action camera in extreme motion scenarios typically does not exceed 50 Hz, and a 200 Hz sampling rate satisfies the Nyquist criterion while reserving sufficient oversampling margin for subsequent filtering.

[0160] S12, Sliding window noise reduction filtering and gravity-dynamic component separation.

[0161] The original triaxial acceleration and triaxial angular velocity sequences were read, and a sliding median filter with a window width of 25 sampling points was applied to each axis signal to suppress high-frequency quantization noise from the inertial sensor and pulse interference introduced by mechanical vibration, thus obtaining the triaxial acceleration denoising sequence and the triaxial angular velocity denoising sequence.

[0162] Based on this, the gravity component and dynamic component are separated from the triaxial acceleration denoising sequence. A second-order Butterworth low-pass filter with a cutoff frequency of 0.5 Hz is used to extract the slowly varying acceleration gravity component, which reflects the device's attitude tilt angle relative to the direction of gravity. The acceleration gravity component is subtracted from the triaxial acceleration denoising sequence point by point to obtain the acceleration dynamic component, which characterizes the pure motion acceleration of the device after eliminating the influence of gravity. The cutoff frequency of 0.5 Hz is chosen because the frequency of the device's attitude tilt angle change caused by human movement is usually below 0.1 Hz, while the main energy distribution of motion acceleration is above 1 Hz; 0.5 Hz can effectively separate the two.

[0163] S13, Multi-source motion feature extraction and feature vector construction.

[0164] Three sets of features were extracted from the triaxial angular velocity denoising sequence, the acceleration dynamic component, and the triaxial acceleration denoising sequence, respectively.

[0165] The first group is rotational features: calculate the Euclidean norm of the three components of the triaxial angular velocity denoising sequence to obtain the instantaneous angular velocity composite amplitude, and statistically analyze the mean and variance of the composite amplitude within the sliding window of the most recent 0.5 seconds to form three scalars.

[0166] The second group is the periodic motion characteristics: the autocorrelation function of the synthetic amplitude sequence of the acceleration dynamic components is calculated within the most recent 2-second window, and the time delay corresponding to the first non-zero peak of the autocorrelation function is extracted to reflect the length of the main motion period and the normalized amplitude of the peak, reflecting the significance of the periodicity. The value ranges from 0 to 1, forming two scalars.

[0167] The third group consists of pre-impact characteristics: the Euclidean norms of the three components of the triaxial acceleration denoised sequence (including the gravity component) are calculated to obtain the total acceleration amplitude. Under normal stationary conditions, this value is approximately one times the gravitational acceleration, and approaches zero during free fall. The mean of the total acceleration amplitude within the most recent 50-millisecond window is extracted as a gravity deviation index, along with the cumulative duration for which this index remains below 0.3 times the gravitational acceleration, forming two scalars.

[0168] The above three sets of seven scalars are concatenated to form a multidimensional motion feature vector, which is continuously updated and output at a rate of no less than 50 Hz. To address the issue of a sharp drop in throughput caused by antenna null crossing the receiver due to high-speed rotation in high-dynamic scenarios, this solution pre-calculates the null crossing window through kinematic tracking and precisely issues pulse modulation and coding adjustment commands. This strictly compresses the throughput loss to a very small time span when physical fading occurs, avoiding the response lag and rate recovery period of tens of milliseconds inherent in traditional post-feedback algorithms. To address the problem of transmission rate oscillations caused by periodic physical obstructions, this solution utilizes a motion phase-locking mechanism to predict the channel's optimal window and implements centralized scheduling of burst transmissions. This transforms inefficient and ineffective transmissions previously constrained by obstruction blind zones into full-load bursts within the optimal phase window, eliminating link fluctuations caused by blind rate probing and improving the average throughput under periodic interference.

[0169] For the challenge of dense burst packet loss caused by severe mechanical vibrations such as extreme landings, this invention uses the free fall state as a purely physical prior signal. During the airborne phase before the physical impact arrives, cross-group interleaved outer layer error correction coding is injected into the service data. This smoothly resolves long burst packet loss that could otherwise lead to a serious deterioration in transmission quality at the application layer, avoiding overflow of the underlying transmission queue and a surge in retransmission delay.

[0170] Correspondingly, to address the issue that multipath fast fading can easily lead to miscalibration of the space reference, a variance discrimination mechanism is introduced to extract the interference of environmental radio frequency fluctuations on physical space prediction and implement weighted correction, thereby ensuring the long-term convergence accuracy and robustness of the space prediction reference under complex electromagnetic and high dynamic environments.

[0171] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A method for enhancing anti-interference capabilities of high-speed Wi-Fi transmission for action cameras, characterized in that, include: Acquire inertial motion data collected by the built-in inertial measurement unit of the action camera and orientation reference information from the receiver; Extracting multidimensional motion features based on inertial motion data; Real-time classification of motion patterns based on multidimensional motion features, outputting the current motion pattern identifier and corresponding pattern feature parameters, wherein the classified motion patterns include at least rotation-dominated patterns, periodic oscillation patterns and impact precursor patterns. Based on the current motion mode identifier, combined with the mode characteristic parameters, and based on the current motion mode identifier, determining whether to further combine the receiver direction reference information, a transmission parameter adjustment instruction is generated; Execute the corresponding transmission parameter feedforward adjustment based on the transmission parameter adjustment command; If the current motion mode identifier indicates a rotation-dominant mode, based on the current motion mode identifier, combined with mode characteristic parameters, and determining whether to further incorporate receiver orientation reference information based on the current motion mode identifier, a transmission parameter adjustment command is generated, including: Using the angular velocity component in the mode characteristic parameters, combined with the pre-configured antenna null direction and receiver direction reference information, the instantaneous angle between the antenna null direction and the receiver direction and the rate of change of the angle are calculated. Estimate the remaining time for zero-point crossing and the duration of the crossing window based on the instantaneous angle and the rate of change of the angle; Based on the remaining time of zero-point crossing and the duration of the crossing window, a pulse-type modulation and coding adjustment instruction is generated as a transmission parameter adjustment instruction. If the current motion mode identifier represents a periodic oscillation mode, based on the current motion mode identifier, combined with mode characteristic parameters, and determining whether to further incorporate receiver direction reference information based on the current motion mode identifier, a transmission parameter adjustment command is generated, including: By utilizing the motion phase and oscillation period in the mode feature parameters, and combining them with the pre-extracted optimal motion phase, a burst transmission window is determined when the current channel quality is higher than the pre-configured channel quality threshold. Based on the phase difference between the current motion phase and the start and end boundaries of the burst transmission window, as well as the swing period, calculate the remaining time until the next burst transmission window opens and the duration of the window. Based on the remaining time until the next burst transmission window opens and the duration of the window, a burst transmission scheduling instruction is generated as a transmission parameter adjustment instruction. Based on the transmission parameter adjustment command, the corresponding transmission parameter feedforward adjustment is executed, specifically including: sending the data to be sent in the buffer queue in a concentrated manner within the burst transmission window, and pausing transmission outside the burst transmission window; If the current motion mode identifier indicates an impending impact pattern, based on the current motion mode identifier, combined with mode characteristic parameters, and determining whether to further incorporate receiver direction reference information based on the current motion mode identifier, a transmission parameter adjustment instruction is generated, including: Determine whether the zero gravity duration in the mode feature parameters exceeds the pre-configured fall confirmation duration threshold, and whether the acceleration residual index in the mode feature parameters is lower than the pre-configured acceleration ratio threshold. If all conditions are met, it is confirmed that the object is in free fall, the impact warning window is opened, and an instruction is generated to inject the outer layer error correction code into the data to be sent, as an instruction to adjust the transmission parameters. If the conditions are not met simultaneously, the current transmission parameters will remain unchanged and the mode characteristic parameters will continue to be monitored. Based on the transmission parameter adjustment command, the corresponding transmission parameter feedforward adjustment is executed, specifically including: within the impact warning window, applying outer layer error correction processing to the data to be sent in the transmission buffer queue.

2. The method according to claim 1, characterized in that, Real-time classification of motion patterns based on multi-dimensional motion features, outputting the current motion pattern identifier and corresponding pattern feature parameters, including: If the mean amplitude of the composite angular velocity in the multidimensional motion features continuously exceeds the pre-configured rotation threshold and the periodic normalization amplitude is lower than the pre-configured lower limit of the period threshold, output the current motion mode identifier representing the rotation-dominant mode. If the periodic normalization amplitude exceeds the pre-configured period upper limit threshold and the average value of the synthesized angular velocity amplitude is lower than the rotation threshold, output the current motion mode identifier that represents the periodic oscillation mode. If the gravity deviation index in the multidimensional motion features is lower than the pre-configured gravity acceleration ratio threshold and the duration exceeds the pre-configured mode entry time threshold, output the current motion mode identifier representing the impact precursor mode. If none of the above three conditions are met, the current motion mode identifier from the previous moment will remain unchanged or the current motion mode identifier representing the default steady-state mode will be output.

3. The method according to claim 1, characterized in that, Estimate the remaining time for zero-point crossing and the duration of the crossing window based on the instantaneous angle and the rate of change of the angle, including: The instantaneous included angle is compared with the pre-configured angle protection threshold; If the instantaneous included angle is greater than the angle protection threshold, a first-order linear extrapolation is performed based on the instantaneous included angle and the rate of change of the included angle to obtain the remaining time for zero-point crossing; If the instantaneous included angle is not greater than the angle protection threshold, the trend of the antenna null direction moving towards the receiving end is determined based on the angular velocity component. The duration of the crossing window is estimated based on the amplitude of the angular velocity component and the pre-configured antenna null angle half-width, and a pulse modulation and coding adjustment command is generated in combination with the duration of the crossing window.

4. The method according to claim 1, characterized in that, Based on the transmission parameter adjustment command, the corresponding transmission parameter feedforward adjustment is performed, including one of the following execution methods: The external rate coverage interface preset in the communication chip firmware is invoked to send data according to the target level specified in the pulse modulation coding adjustment instruction within the duration of the crossing window; Alternatively, at the application layer, the data frame to be sent can be fragmented according to the duration of the crossing window determined by the pulse modulation and coding adjustment instruction, and a low-rate protection flag can be set in the header of the data frame, so that the underlying transmission module can send the fragmented data frame in low-rate protection mode within the duration of the crossing window.

5. The method according to claim 1, characterized in that, The optimal motion phase is extracted through an online calibration phase, which includes: Simultaneously record the motion phase and the received signal strength indication value within multiple swing cycles; The motion phase and the received signal strength indication value are averaged over an interval, and a single-frequency sine model is fitted to establish a phase-channel quality mapping table that includes the full-cycle average channel quality, half-peak fluctuation amplitude, and optimal motion phase. By combining the pre-configured threshold coefficients with the phase-channel quality mapping table, the start and end boundaries of the burst transmission window are determined.

6. The method according to claim 1, characterized in that, Before generating burst transmission scheduling instructions, the following is also included: Obtain the maximum supported transmission rate from the communication chip report and the average output bitrate of the current video encoder; Calculate the target transmission rate required for concentrated transmission within the burst transmission window; If the target transmission rate is greater than the maximum supported transmission rate, the burst transmission window is adaptively expanded by calculating the minimum window percentage required to satisfy the physical layer rate constraint. If the minimum window percentage corresponding to the expanded burst transmission window reaches the pre-configured upper limit, exit the burst transmission scheduling of the periodic swing mode and suggest reducing the bitrate output to the current video encoder.

7. The method according to claim 5, characterized in that, Before establishing the phase-channel quality mapping table, a conservative pre-scheduling state is performed, including: The optimal motion phase is initialized using the zero-crossing phase of the prior dynamic acceleration component in the multidimensional motion features; Centralized transmission is performed by setting an initial burst transmission window based on a pre-configured conservative threshold coefficient. As data accumulates during the online calibration phase, the optimal motion phase and conservative threshold coefficients are linearly transitioned to the fitted steady-state value.

Images