A method and system for generating a broadcast audio link node timing topology

Abstract

This invention discloses a method and system for generating broadcast audio link node timing topology. Addressing the shortcomings of existing network topology analysis methods, which heavily rely on manual intervention and are highly susceptible to distortion due to clock timing errors and strong electromagnetic interference in equipment rooms, this method injects a low-frequency coded marker signal that avoids power frequency harmonics into the audio stream from the program source; each preset node synchronously acquires audio and binds it with an absolute timestamp, extracting multi-dimensional feature sequences; based on the marker signal, a coarse screening of the correlation of the same link is performed and the nodes are initially sorted by timestamp; a dynamic time warping algorithm is used to calculate the waveform similarity of the multi-dimensional features of adjacent nodes; when the similarity falls below a preset threshold, candidate nodes are re-searched within a preset time window to complete the closed-loop correction of timing; finally, topology data representing the physical connection timing and actual transmission delay is generated. This invention has strong resistance to power frequency interference, can adaptively correct clock errors, and achieves high-precision, automated generation of complex broadcast network topologies.

Description

Technical Field

[0001] This invention relates to the field of broadcast television signal monitoring and network operation and maintenance technology, specifically to a method and system for generating timing topology of broadcast audio link nodes. Background Technology

[0002] Modern broadcast audio transmission systems exhibit extremely high topological complexity. A complete signal link typically involves multiple physical and logical nodes, including program sources, audio preprocessing, digital encoding, radio frequency modulation, power amplification, transmitting antennas, and receiving feedback. With the convergence of multi-frequency and multi-channel broadcast networks at the provincial and municipal levels, accurately grasping the actual transmission sequence and delay status of audio signals between nodes (i.e., node timing topology) is the core foundation for daily system operation and maintenance, equipment status monitoring, and sudden fault location.

[0003] However, existing link topology acquisition and measurement technologies suffer from the following insurmountable technical shortcomings: (1) Manual mapping is extremely inefficient and prone to errors: Traditional topology analysis relies heavily on technicians to check physical cables and equipment interface connections on-site. In complex provincial multi-frequency broadcast networks, drawing a complete link flow diagram often takes several days or even longer. In addition, hidden relay nodes (such as passive signal distributors) in complex systems are easily overlooked, resulting in a high rate of sequential logic errors in multi-node parallel scenarios. Once a link is broken, it is very easy to mislead the direction of troubleshooting.

[0004] (2) Existing automation tools lack underlying audio signal verification: Most existing network topology automatic discovery technologies are based on standard IT protocols (such as SNMP, LLDP) or preset network management connection relationships for logical drawing, without combining the actual physical characteristics of the underlying audio signals for end-to-end verification. When the device management network port is connected but the audio service signal itself is interrupted or severely distorted (e.g., the encoder's internal service process is stuck, or the RF interface impedance is mismatched), the topology shown in the network management diagram deviates greatly from the actual audio flow, and cannot provide real support for business-level operation and maintenance.

[0005] (3) Clock synchronization error leads to inverted timing logic: Some existing monitoring schemes attempt to infer the topology order by comparing the timestamps of the audio collected by each node. However, the node devices in the broadcast room mostly use conventional NTP time synchronization (Network Time Protocol, with synchronization error usually between 10ms and 50ms), while the actual physical transmission delay between adjacent devices is often only 0.1ms to 10ms. This physical contradiction of "time synchronization error being greater than transmission delay" makes it very easy for the order of nodes to be reversed simply by relying on the absolute value of timestamps, resulting in extremely low reliability of topology inference.

[0006] (4) Beacon extraction failure in complex electromagnetic environments: To perform signal tracking and differentiation, some technologies attempt to inject continuous single-frequency long-tone signals at the source end as link beacons. However, in actual high-power broadcast transmitter station equipment rooms, there is extremely strong electromagnetic coupling interference from industrial AC power frequency (50Hz and its harmonics). Conventional low-frequency beacons are easily submerged by environmental noise, or the environmental power frequency noise is mistaken for the injected beacon, resulting in a large number of spurious correlation peaks in subsequent link correlation calculations, and serious false alarms in the system.

[0007] In summary, existing technologies cannot automatically and accurately identify the connection sequence of multiple nodes and calculate the actual transmission delay based on the true physical characteristics of audio signals under strong electromagnetic interference environments. There is an urgent need in this field for a topology generation method that integrates high-precision time synchronization, anti-interference coded beacon extraction, and waveform comparison and correction to solve the problems of slow topology generation, low accuracy, and inability to map the actual service status in existing broadcast network operations and maintenance. Summary of the Invention

[0008] This invention primarily addresses the technical shortcomings of existing technologies, such as the low efficiency of relying on manual methods for broadcast network topology analysis, the lack of underlying signal verification in automated tools, timing inversion caused by conventional timing errors, and the susceptibility of traditional beacons to power frequency electromagnetic interference at transmitting stations. It provides a method and system for generating the timing topology of broadcast audio link nodes.

[0009] The present invention addresses the aforementioned technical problems primarily through the following technical solution: a method for generating timing topology for broadcast audio link nodes, comprising the following steps: S1: At the program source node of the audio transmission link, inject a low-frequency coded marker signal that avoids power frequency harmonics into the audio stream; S2: At each preset node of the audio transmission link, the original audio data containing low-frequency coded marker signals is synchronously collected, and an absolute timestamp is added to each frame of data collected based on the dual-mode time synchronization mechanism; the collected audio data is denoised and multi-dimensional feature sequences are extracted. S3: Based on the low-frequency coded marker signal, perform a coarse screening of the same link correlation on the feature sequences of each node, and sort the coarsely screened nodes according to their absolute timestamps. S4: Calculate the waveform similarity of the multidimensional feature sequences of adjacent nodes using the dynamic time warping algorithm. When the waveform similarity is lower than a preset threshold, re-search the nodes within the preset time window to adjust the sorting, thereby completing the closed-loop correction and verification of the node timing. S5: Based on the verified node timing and the preset node attribute database, generate and output node topology data that characterizes the timing relationship of the physical connection of the link and the actual transmission delay of adjacent nodes.

[0010] This solution introduces spread spectrum digital watermarking technology that avoids power frequency interference, and combines multi-dimensional feature dimensionality reduction and high-precision waveform comparison to construct a self-healing topology generation mechanism of coarse screening and closed-loop correction. This enables high-precision automatic identification of multi-node link order in strong electromagnetic interference environments, and directly calculates the actual microscopic transmission delay of adjacent nodes, providing underlying data support for the automated operation and maintenance and fault location of broadcasting systems.

[0011] In step S2, noise reduction can be performed using wavelet thresholding. For example, the original audio signal can be decomposed into three levels using the Sym8 wavelet basis, and the threshold can be calculated using the general threshold formula λ=σ·sqrt(2lnN). win (where σ is the noise standard deviation, N) win Calculate the noise reduction threshold (for the number of sampling points in the window) and perform soft threshold reconstruction to eliminate the impact of device background noise on subsequent feature comparison.

[0012] Preferably, step S1, the injection step of the low-frequency coded marker signal, specifically includes: Using frequency f m A carrier wave with a prime number and a frequency f m It does not overlap with the industrial standard AC power frequency and its harmonics. The carrier wave is phase-modulated using a pseudo-random binary sequence. The modulated signal is then injected into the digital audio stream through digital weighted mixing. The expression for the generated low-frequency coded marker signal s(t) is: s(t)=0.1A audio ·sin(2πf m t+φ(t)); Among them, A audio Let be the peak amplitude of the original audio signal, t be the time variable, and φ(t) be the additional phase controlled by the pseudo-random binary sequence. The value of the additional phase φ(t) is determined by the logical value of the current symbol corresponding to the pseudo-random binary sequence at time t, and its mapping rule is as follows: When the current symbol is the first logic value, φ(t) = 0; when the current symbol is the second logic value, φ(t) = π; wherein, each symbol of the pseudo-random binary sequence maintains a preset symbol width T. c .

[0013] Broadcast transmission stations often experience extremely strong electromagnetic coupling interference at 50Hz and its harmonics (100Hz, 150Hz, etc.). This invention uses a prime frequency (e.g., 43Hz) as the carrier frequency f. mBased on this, binary phase shift keying (BPSK) is used for digital modulation. The first logic value can correspond to the polarity level "+1" or logic "1" in a pseudo-random binary sequence (such as a 7-bit Barker code), at which point the phase φ(t) = 0; the second logic value corresponds to the polarity level "-1" or logic "0", at which point the phase φ(t) = π. The symbol width T c It can be set to 100ms. This injection action can be completed in the pure digital domain through the digital weighted mixing module built into the program source digital audio processor (DSP), realizing lossless injection of the dark watermark and strong anti-interference.

[0014] Preferably, step S2, the step of extracting the multidimensional feature sequence, includes: dividing the acquired raw audio data into several data frames according to a preset time window; for the w-th frame of data, extracting the short-time energy, zero-crossing rate, and frequency amplitude of at least one preset core frequency; and normalizing the data using the local maximum and minimum values ​​of each frame to obtain the normalized short-time energy E'. w Normalized zero-crossing rate ZCR' w and the normalized frequency amplitude |X|' of each core frequency q,w , where q∈{1,2,…,Q}, and Q is the number of preset core frequencies; The normalized features of the w-th frame are concatenated to construct the i-th node N. i multidimensional feature vector v i (w)=[E' w ,ZCR' w ,|X|' 1,w ,|X|' 2,w ,…,|X|' Q,w The multidimensional feature sequence V is composed of the multidimensional feature vectors of each frame arranged sequentially. i =[v i (1),v i (2),…,v i [(W)], where W is the node N i Total number of frames within the acquisition period.

[0015] In practical broadcast systems (such as 48kHz sampling rate), a 10-second audio file contains up to 480,000 sampling points. Direct comparison not only incurs enormous computational overhead but is also highly susceptible to transient impulse noise. This invention slices the audio into frames according to a preset time window (e.g., a 20ms frame length containing 960 sampling points), compressing each frame into a single frame containing normalized short-time energy E'. w Zero crossing rate ZCR' w and several frequency amplitudes |X|' q,w multidimensional feature vector v i(w). This dimensionality reduction process exponentially reduces the computational load of subsequent dynamic time warping, ensuring that the system can output the entire network topology within 10 minutes.

[0016] In constructing a multidimensional feature vector v i (w) In this invention, a dynamic dimensional expansion design is employed to adapt to detection requirements of varying complexity. Specifically, the basic dimension of the vector includes the normalized short-time energy E'. w With normalized zero-crossing rate ZCR' w Based on this, the system extracts the normalized amplitude |X|' of Q core frequencies according to actual broadcast service requirements. q,w For example, when Q=2, the system extracts the frequency amplitude at 1kHz (core frequency band for broadcast speech) and 3kHz (key frequency band for audio clarity), respectively. The resulting multidimensional feature vector is a four-dimensional vector, i.e., v. i (w)=[E' w ,ZCR' w ,|X|' 1kHz,w ,|X|' 3kHz,w This dynamic stitching of multidimensional features enables the weighted Euclidean distance calculation in the distance matrix D to more accurately capture the microscopic distortions of the signal at different frequency nodes, thereby significantly improving the robustness of the alignment of the Dynamic Time Warping (DTW) algorithm.

[0017] Preferably, in step S3, the specific method for the coarse screening of correlation within the same link is as follows: The collected candidate node signals are processed at a center frequency of f. m After bandpass filtering, a sliding correlation detection is performed with the locally known low-frequency coded marker signal s(t). If the calculated sliding correlation coefficient R≥0.9, the node is determined to belong to the same audio transmission link.

[0018] Due to the discretization characteristics of the signal, the sliding correlation detection here specifically adopts a normalized cross-correlation algorithm based on a discrete time window. Let the local reference sequence be s[n] (length M), the sequence to be tested be x[n], and the sliding step size be m. Then, the formula for discretely calculating the correlation coefficient of the local window is: ; The system extracts the maximum peak value R=max(R[m]) throughout the entire process. This normalized cross-correlation algorithm is insensitive to linear scaling of the signal amplitude. Even if the marked signal is deformed in a multi-stage amplification and attenuation link, it can still maintain an extremely high detection rate and perfectly filter out crosstalk signals that are not in the local frequency.

[0019] Preferably, step S4, which involves calculating waveform similarity using the dynamic time warping algorithm, includes: For the adjacent nodes N after the initial sorting iand N j Let the previous node N be... i The multidimensional feature sequence is V i Subsequent node N j The multidimensional feature sequence is V j Construct a distance matrix D, and the elements d of the distance matrix D. wl Defined as the weighted Euclidean distance between multidimensional feature vectors: d wl =sqrt((v i (w)-v j (l)) T ·Λ·(v i (w)-v j (l))); Where w is node N i The frame number, l is the node N j The frame number, Λ is a preset weight diagonal matrix used to adjust the weight of each feature; Dynamic programming is used to find the cumulative distance D. DTW The minimum optimal path, based on the cumulative distance D. DTW Calculate the waveform similarity S with the length K of the optimal path: S = 1 / (1 + D) DTW / K).

[0020] Λ is a weighted diagonal matrix used to assign different comparison weights to different features (for example, since frequency domain features attenuate less when passing through an FM transmitter, they can be given a higher weight than short-time energy features). The state transition equation (recurrence relation) of the dynamic programming is set as follows: for the non-boundary elements D of the cumulative distance matrix cum (w,l) has a value equal to the distance d from the current point. wl +min{D cum (w-1,l),D cum (w,l-1),D cum (w-1,l-1)};Calculate the endpoint of the matrix, i.e., the minimum cumulative distance D. DTW Then, the optimal alignment path K is obtained through reverse backtracking; a robust similarity formula is used for normalization, which completely eliminates the defect of traditional formulas being easily disturbed by outliers.

[0021] Preferably, step S4, which involves re-searching nodes within a preset time window to adjust the sorting and perform loop closure verification, specifically includes: When adjacent node N i With N j When the waveform similarity S < 0.95, the absolute timestamp difference retrieved in the feature database satisfies |T k -T i |≤T limit The set of candidate nodes, where Ti For node N i absolute timestamp, T k T is the absolute timestamp of the candidate node. limit The preset maximum tolerable transmission delay; Calculate the relationship between each candidate node and node N in the candidate node set. i Waveform similarity; If there is a node in the candidate node set whose waveform similarity satisfies 0.8≤S<0.95, then the node with the highest similarity is selected as the real downstream node, and a link degradation anomaly alarm is output. If the waveform similarity of all nodes in the candidate node set satisfies S < 0.8, then node N is determined to be... i For isolated nodes with broken links, terminate the path search for that branch and output a broken link alarm signal.

[0022] This solution breaks away from the traditional mindset of blindly trusting time stamps. When waveform characteristics do not meet the requirements, the system actively overturns the illusion of physical ordering generated by PTP time synchronization. Furthermore, the two defined gradient thresholds have clear business mapping significance: if the optimal similarity after retrieval falls within the range of [0.8, 0.95), it indicates that the physical link is not broken, but severe signal distortion may occur due to a sharp increase in cable impedance or interface oxidation, triggering a link degradation alarm; if the similarity after a full-domain retrieval is less than 0.8, a fallback logic to prevent infinite loops is triggered, determining that the node is completely disconnected, issuing an island alarm, and directly indicating the link break location to maintenance personnel.

[0023] Preferably, in step S5, the adjacent node N i With N j The actual transmission delay ΔT between true The calculation formula is: ΔT true = (T j -T i )+Δt DTW ; Among them, T j With T i They are nodes N respectively j With node N i The absolute timestamp of the first frame of data recorded; Δt DTW The relative time offset of the dynamic time warping algorithm on the optimal path is calculated using the following formula: ; Where k is the path node number on the optimal path, w k For the k-th alignment point on the optimal path, the corresponding preceding node N i Frame number; l kFor the k-th alignment point of the optimal path, the corresponding true subsequent node N j Frame number; T frame The duration of a single data frame.

[0024] In conventional timestamp ranging schemes, T j -T i This only represents the trigger time difference of packet capture actions between two node acquisition modules and cannot reflect the physical delay of the audio signal flowing in space. This solution combines the macroscopic absolute timestamp with the microscopic DTW optimal path offset Δt. DTW Perfectly integrated. Through alignment point deviation (l) k -w k The calculated frame-level relative slip precisely compensates for the errors caused by asynchronous device acquisition, improving the accuracy of transmission latency measurement to within milliseconds, so that the final output topology data can accurately reflect the small fluctuations in network performance.

[0025] A broadcast audio link node timing topology generation system, running the method described above, includes: Signal injection module: used to inject low-frequency coded marker signals that avoid power frequency harmonics into the program source node; Data synchronization acquisition module: used to synchronously acquire raw audio data containing low-frequency coded marker signals at each preset node of the audio transmission link, and add an absolute timestamp to each frame of data acquired based on a dual-mode time synchronization mechanism; Feature extraction module: used to reduce noise in audio data and extract multidimensional feature sequences; Coarse screening and initial sorting module: used to perform coarse screening of the feature sequences of each node based on the low-frequency coded marker signal, and to initially sort the coarsely screened nodes according to the absolute timestamp; Closed-loop correction module: The dynamic time warping algorithm is used to calculate the waveform similarity of the multidimensional feature sequences of adjacent nodes. When the waveform similarity is lower than a preset threshold, the nodes are re-searched within a preset time window to adjust the sorting, thus completing the closed-loop correction and verification of the node timing. Topology mapping output module: Based on the verified node timing and a preset node attribute database, it generates and outputs node topology data that represents the timing relationship of physical links and the actual transmission delay of adjacent nodes.

[0026] Each module can be implemented either through software program code running on a computer processor or through hardware circuits such as highly integrated field-programmable gate arrays (FPGAs) or dedicated digital signal processors (DSPs) to meet the high requirements of the transmitting station for real-time performance and stability.

[0027] The node time-series topology data generated by this invention can be directly output to the central control network management platform of the broadcast transmission station. It can be used to directly draw transmission node flowcharts and also to support the following specific physical network operation and maintenance services: Precise fault location and hidden node detection: When there are hidden relay nodes in the physical link that are not registered on the drawings (such as privately connected passive distributors or signal delayers), this system can detect abnormal time delay changes (ΔT). true The system automatically visualizes the surge in new feature sequences and their corresponding topologies in the topology. When the system throws a "broken link island" alarm, maintenance personnel can directly go to the physical rack where the faulty node is located, carrying spare cables or boards, based on the topology sequence, eliminating the tedious process of troubleshooting at each level.

[0028] Equipment aging prediction and degradation early warning: By analyzing the actual transmission delay ΔT between adjacent nodes true By conducting long-term dynamic monitoring and comparative analysis, if the latency between two physical devices shows a slow upward trend and the waveform similarity S fluctuates frequently in the range of [0.8, 0.95), the network management platform can determine in advance that there is a serious increase in oxidation resistance or poor interface contact in the physical cable segment, thereby triggering a degradation warning before the service is completely interrupted and guiding maintenance personnel to carry out preventive maintenance.

[0029] Automatic emergency primary / backup link switching: In broadcast systems equipped with intelligent switching control matrices, the topology status output by this system can serve as a high-priority trigger signal. Once a core node is determined to have a "signal link failure," the system not only outputs an alarm but can also directly drive the physical radio frequency switching switch via control commands to seamlessly switch the audio stream to the backup transmission link, ensuring a safe downtime rate for broadcasting.

[0030] The substantial effects of this invention are: 1. Strong anti-interference and precise coarse screening: This solution innovatively adopts a low-frequency dark watermarking technology that combines prime-number frequency carriers with pseudo-random sequence phase modulation, cleverly avoiding the industrial AC power frequency and harmonic interference commonly found in broadcast equipment rooms. Combined with a normalized cross-correlation algorithm, it achieves 100% precise coarse screening of the same link signal under extremely low signal-to-noise ratio, ensuring the reliability of topology sorting from the source.

[0031] 2. The timing correction mechanism significantly improves accuracy: This invention does not blindly rely on the absolute timestamp of network NTP or PTP timing, but instead uses it as an initial candidate condition, innovatively introducing "multi-dimensional feature sequence + DTW dynamic programming" as a closed-loop correction criterion. Even in complex network configurations with extremely short transmission delays between nodes and timing errors, it can still automatically correct timing inversion errors through waveform tracing, resulting in a final node order recognition accuracy of over 99.5%.

[0032] 3. Microscopic Features Directly Empower Business Operations and Maintenance: This solution achieves deep extraction of the underlying physical quantities of signals. On the one hand, by refining similarity thresholds (island and degradation judgment), it successfully prevents algorithm dead loops and directly locates distorted nodes; on the other hand, by fusing absolute timestamps and DTW path offsets, it outputs millisecond-level real physical transmission delays. The final result is no longer a simple connection diagram, but a high-precision digital operation and maintenance testing certificate that can be directly used to diagnose cable aging and internal equipment malfunctions. Attached Figure Description

[0033] Figure 1 This is a flowchart of a method for generating timing topology of broadcast audio link nodes according to the present invention. Detailed Implementation

[0034] The technical solution of the present invention will be further described in detail below through embodiments and in conjunction with the accompanying drawings.

[0035] Example: This example is based on a typical provincial multi-frequency broadcast audio transmission network. The signal passes through multiple nodes in sequence, including the program source production and broadcasting network, digital audio processor, AES3 signal distributor, digital encoder, FM modulator and radio frequency transmitting antenna.

[0036] The embodiments of the present invention address each key preset node (i.e., potential N) in the aforementioned link. i The data acquisition terminals are deployed in a bypass configuration at each node. Each acquisition terminal is equipped with a high-precision dual-mode timing unit consisting of BeiDou (BDS-M8T module) and PTP (IEEE 1588v2 protocol), with a timing accuracy of ≤0.5ms.

[0037] This embodiment provides a method for generating the timing topology of broadcast audio link nodes, based on the aforementioned deployment architecture, such as... Figure 1 As shown, the method specifically includes the following steps: S1: At the program source node of the audio transmission link, inject a low-frequency coded marker signal that avoids power frequency harmonics into the audio stream.

[0038] In broadcast transmission station equipment rooms, extremely strong industrial AC power frequency and harmonic coupling interference at 50Hz, 100Hz, and 150Hz is common. To avoid misjudgment, this embodiment uses frequency f in the digital audio processor (such as a DSP module) of the program source node. m A 43Hz (prime frequency) sine wave is used as the carrier wave, and the carrier wave is modulated by binary phase shift keying (BPSK) using a 7-bit Barker code (pseudo-random binary sequence).

[0039] The mathematical model for the generated low-frequency coded marker signal s(t) is: s(t) = 0.1A audio ·sin(2πfm t+φ(t)); where, A audio The peak amplitude of the original audio signal is used to control the energy ratio of the marker signal, thus achieving the injection of the dark watermark. φ(t) is the additional phase; when the current symbol of the Barker code is logic level "+1" (first logic value), φ(t) = 0; when it is logic level "-1" (second logic value), φ(t) = π. Symbol width T c The setting is 100ms. The modulated signal is incorporated into the service audio stream through a pure digital weighted mixing method and transmitted without loss.

[0040] S2: Synchronously acquire audio data, add timestamps, perform noise reduction processing, and extract multi-dimensional feature sequences.

[0041] Each preset node synchronously acquires an audio PCM data stream containing the aforementioned marker signals, and the dual-mode timing module adds an absolute timestamp T to each frame of data. i .

[0042] To eliminate equipment noise floor, the data is decomposed into three levels using the Sym8 wavelet basis, and the general threshold formula λ=σ·sqrt(2lnN) is applied. win (where σ is the noise standard deviation, N) win Soft thresholding is applied to the number of sampling points in the window.

[0043] Then, dimensionality reduction feature extraction is performed: the audio is sliced ​​into 20ms frames (containing 960 sampling points). For node N... i From the w-th frame of data, extract its short-time energy, zero-crossing rate, and amplitudes at the two core frequencies of 1kHz and 3kHz. After local extremum normalization, construct a four-dimensional feature vector: v i (w)=[E' w ,ZCR' w ,|X|' 1kHz,w ,|X|' 3kHz,w ]; Arrange the feature vectors of each frame sequentially to obtain a multidimensional feature sequence V of length W. i =[v i (1),v i (2),…,v i (W)]. This dimensionality reduction operation compresses the comparison of millions of sampling points to the comparison of thousands of frame sequences, greatly reducing the system's computing power overhead.

[0044] S3: Perform a rough screening of the correlation of the same link and perform preliminary sorting by absolute timestamp.

[0045] This embodiment employs a normalized cross-correlation algorithm based on discrete-time windows for coarse screening. The digital signals of candidate nodes are bandpass filtered at 43Hz to obtain the test sequence x[n], which is then subjected to sliding correlation calculation with the local reference sequence s[n] (length M). The correlation coefficient with a sliding step size of m is: ; The average magnitude of the local reference sequence s[n] over a length M; Let R be the average amplitude of the sequence under test within the current sliding window x[n+m]. The system extracts the peak value R=max(R[m]). Since cross-correlation is not sensitive to amplitude decay and power frequency noise does not have the phase flipping rule of Barker code, when R≥0.9, it can be accurately determined that the node contains the injected marker signal and belongs to the same link. Subsequently, the nodes after coarse screening are arranged in ascending order according to absolute timestamps to obtain the preliminary topology sequence.

[0046] S4: Calculate waveform similarity using the DTW algorithm and perform closed-loop correction and verification.

[0047] For the predecessor node N in the initial sorting i (Total frames W) and subsequent nodes N j (Total number of frames L), construct a W×L distance matrix D. Its element-weighted Euclidean distance is: d wl =sqrt((v i (w)-v j (l)) T ·Λ·(v i (w)-v j (l))); Where Λ is the feature weight diagonal matrix. Next, dynamic programming (DP) is used to find the cumulative minimum distance, defining the cumulative distance matrix D. cum (w,l). Initialize boundary conditions D cum (1,1)=d 11 Then, the following state transition equation is used for recursive calculation: D cum (w,l)=d wl +min{D cum (w-1,l),D cum (w,l-1),D cum (w-1,l-1)}; The above state transition equations are calculated by traversing row by row or column by column until the endpoint element D is calculated. cum (W,L) represents the cumulative minimum distance D between two multidimensional feature sequences under globally optimal alignment. DTW .

[0048] During the above recursive calculation process, each state D is recorded synchronously. cum The smallest preceding node selected at (w,l) (i.e., the coordinate of the previous step corresponding to the min operation). After the calculation is completed, backtracking from the endpoint (W,L) to the starting point (1,1) according to the recorded pointer, the resulting coordinate sequence is the optimal path P=[(w1,l1),(w2,l2),…,(w K ,l K )], where K is the total length of the path.

[0049] Then, the robust waveform similarity is calculated: S = 1 / (1 + D) DTW / K).

[0050] Closed-loop fault-tolerant logic: If S ≥ 0.95, verification passes. If S < 0.95 (indicating timing error causing sorting inversion or link disconnection), the system will fail at |T k -T i Within a time window of ≤100ms, re-traverse other nodes and calculate the relationship with N. i Similarity: (1) If a node with a similarity of 0.8≤S<0.95 is found, the highest value is selected as the real subsequent node, and a link degradation alarm is output to indicate signal distortion caused by cable oxidation or abnormal interface impedance. (2) If all candidate nodes S < 0.8, then determine N. i For an isolated orphan, forcibly terminate the pathfinding for that branch and output a "signal disconnected" alarm location.

[0051] S5: Generate node topology data and map the actual transmission latency.

[0052] For the verified true adjacent node N i With N j This system not only outputs logical topology connections, but also accurately maps the physical transmission delay ΔT. true : ; Among them, T j -T i The macroscopic difference in the timestamps of the two packet capture terminals; the summation part is the deviation of each alignment point on the optimal DTW path (l k -w k The micro-average frame offset of T; frame The frame period is 20ms. This formula perfectly compensates for the absolute error caused by the clock asynchrony of the acquisition modules, outputting high-precision real physical transmission delay at the millisecond level.

[0053] Corresponding to the above method, this embodiment of a broadcast audio link node timing topology generation system specifically includes: Signal injection module: Located at the program source node, it uses an internal DSP chip to generate a 43Hz carrier wave and controls the phase flip according to the Barker code sequence to complete the pure digital domain mixing of the dark watermark.

[0054] Data synchronization acquisition module: distributed at each monitoring node, equipped with ADC unit and BDS / PTP dual-mode clock unit, to complete audio stream sampling and absolute timestamp binding.

[0055] Feature extraction module: Built-in wavelet filtering algorithm and normalization calculation unit to reduce the dimensionality of massive sampling points and convert them into a multi-dimensional feature sequence set.

[0056] Coarse screening and initial sorting module: responsible for performing sliding normalized cross-correlation calculations, filtering background noise nodes that are not in this link, and sorting out the initial sequence according to timestamps.

[0057] Closed-loop correction module: The core computing unit of the system, responsible for performing multi-dimensional DTW dynamic planning and pathfinding, and performing adaptive service routing based on the set two-level similarity threshold network for "link degradation" or "island determination".

[0058] Topology mapping output module: Combined with the pre-built device ledger attribute database, the calculated topology timing and actual latency are converted into standardized interface data (such as JSON format) or a visual topology panel, providing underlying data support for front-end network management monitoring.

[0059] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

[0060] Although this paper makes frequent use of terms such as low-frequency coded marker signals and waveform similarity, the possibility of using other terms is not excluded. These terms are used merely for the convenience of describing and explaining the essence of the invention; interpreting them as any additional limitation would contradict the spirit of the invention.

Claims

1. A method for generating timing topology of broadcast audio link nodes, characterized in that, Includes the following steps: S1: Inject low-frequency coded marker signals into the audio stream at the program source node of the audio transmission link; S2: At each preset node of the audio transmission link, the original audio data containing low-frequency coded marker signals is synchronously collected, and an absolute timestamp is added to each frame of data collected based on the dual-mode time synchronization mechanism; the collected audio data is denoised and multi-dimensional feature sequences are extracted. S3: Based on the low-frequency coded marker signal, perform a coarse screening of the same link correlation on the feature sequences of each node, and sort the coarsely screened nodes according to their absolute timestamps. S4: Calculate the waveform similarity of the multidimensional feature sequences of adjacent nodes using the dynamic time warping algorithm. When the waveform similarity is lower than a preset threshold, re-search the nodes within the preset time window to adjust the sorting, thereby completing the closed-loop correction and verification of the node timing. S5: Based on the verified node timing and the preset node attribute database, generate and output node topology data that characterizes the timing relationship of the physical connection of the link and the actual transmission delay of adjacent nodes.

2. The method for generating a broadcast audio link node timing topology according to claim 1, characterized in that, In step S1, the injection step of the low-frequency coded marker signal specifically includes: Using frequency f m A carrier wave with a prime number is used to perform phase modulation on the carrier wave using a pseudo-random binary sequence, and the modulated signal is injected into a digital audio stream through digital weighted mixing; The expression for the generated low-frequency coded marker signal s(t) is: s(t)=0.1A audio ·sin(2πf m t+φ(t)); Among them, A audio Let be the peak amplitude of the original audio signal, t be the time variable, and φ(t) be the additional phase controlled by the pseudo-random binary sequence. The value of the additional phase φ(t) is determined by the logical value of the current symbol corresponding to the pseudo-random binary sequence at time t, and its mapping rule is as follows: When the current symbol is the first logic value, φ(t) = 0; when the current symbol is the second logic value, φ(t) = π; wherein, each symbol of the pseudo-random binary sequence maintains a preset symbol width T. c .

3. The method for generating a broadcast audio link node timing topology according to claim 2, characterized in that, Step S2, the steps for extracting the multidimensional feature sequence include: The acquired raw audio data is divided into several data frames according to a preset time window. For the w-th frame, short-time energy, zero-crossing rate, and frequency amplitude of at least one preset core frequency are extracted. The local maximum and minimum values ​​of each frame are then used for normalization to obtain the normalized short-time energy E'. w Normalized zero-crossing rate ZCR' w and the normalized frequency amplitude |X|' of each core frequency q,w , where q∈{1,2,…,Q}, and Q is the number of preset core frequencies; The normalized features of the w-th frame are concatenated to construct the i-th node N. i multidimensional feature vector v i (w)=[E' w ,ZCR' w ,|X|' 1,w , |X|' 2,w ,…, |X|' Q,w The multidimensional feature sequence V is composed of the multidimensional feature vectors of each frame arranged sequentially. i =[v i (1),v i (2),…,v i [(W)], where W is the node N i Total number of frames within the acquisition period.

4. The method for generating a broadcast audio link node timing topology according to claim 3, characterized in that, In step S3, the specific method for the coarse screening of correlation within the same link is as follows: The collected candidate node signals are processed at a center frequency of f. m After bandpass filtering, a sliding correlation detection is performed with the locally known low-frequency coded marker signal s(t). If the calculated sliding correlation coefficient R≥0.9, the node is determined to belong to the same audio transmission link.

5. The method for generating a broadcast audio link node timing topology according to claim 4, characterized in that, Step S4, which involves calculating waveform similarity using the dynamic time warping algorithm, includes: For the adjacent nodes N after the initial sorting i and N j Let the previous node N be... i The multidimensional feature sequence is V i Subsequent node N j The multidimensional feature sequence is V j Construct a distance matrix D, and the elements d of the distance matrix D. wl Defined as the weighted Euclidean distance between multidimensional feature vectors: d wl =sqrt((v i (w)-v j (l)) T ·Λ·(v i (w)-v j (l))); Where w is node N i The frame number, l is the node N j The frame number, Λ is a preset weight diagonal matrix used to adjust the weight of each feature; Dynamic programming is used to find the cumulative distance D. DTW The minimum optimal path, based on the cumulative distance D. DTW Calculate the waveform similarity S with respect to the length K of the optimal path: S = 1 / (1 + D) DTW / K).

6. The method for generating a broadcast audio link node timing topology according to claim 5, characterized in that, Step S4, which involves re-searching nodes within a preset time window to adjust the sorting and perform loop closure verification, specifically includes: When adjacent node N i With N j When the waveform similarity S < 0.95, the absolute timestamp difference retrieved in the feature database satisfies |T k -T i |≤T limit The set of candidate nodes, where T i For node N i absolute timestamp, T k T is the absolute timestamp of the candidate node. limit The preset maximum tolerable transmission delay; Calculate the relationship between each candidate node and node N in the candidate node set. i Waveform similarity; If there is a node in the candidate node set whose waveform similarity satisfies 0.8≤S<0.95, then the node with the highest similarity is selected as the real downstream node, and a link degradation anomaly alarm is output. If the waveform similarity of all nodes in the candidate node set satisfies S < 0.8, then node N is determined to be... i For isolated nodes with broken links, terminate the path search for that branch and output a broken link alarm signal.

7. The method for generating a broadcast audio link node timing topology according to claim 6, characterized in that, In step S5, adjacent node N i With N j The actual transmission delay ΔT between true The calculation formula is: ΔT true =(T j -T i )+Δt DTW ; Among them, T j With T i They are nodes N respectively j With node N i The absolute timestamp of the first frame of data recorded; Δt DTW The relative time offset of the dynamic time warping algorithm on the optimal path is calculated using the following formula: ； Where k is the path node number on the optimal path, w k For the k-th alignment point on the optimal path, the corresponding preceding node N i Frame number; l k For the k-th alignment point of the optimal path, the corresponding true subsequent node N j Frame number; T frame The duration of a single data frame.

8. A broadcast audio link node timing topology generation system, characterized in that, The system, which operates the broadcast audio link node timing topology generation method as described in claim 1, comprises: Signal injection module: used to inject low-frequency coded marker signals that avoid power frequency harmonics into the program source node; Data synchronization acquisition module: used to synchronously acquire raw audio data containing low-frequency coded marker signals at each preset node of the audio transmission link, and add an absolute timestamp to each frame of data acquired based on a dual-mode time synchronization mechanism; Feature extraction module: used to reduce noise in audio data and extract multidimensional feature sequences; Coarse screening and initial sorting module: used to perform coarse screening of the feature sequences of each node based on the low-frequency coded marker signal, and to initially sort the coarsely screened nodes according to the absolute timestamp; Closed-loop correction module: The dynamic time warping algorithm is used to calculate the waveform similarity of the multidimensional feature sequences of adjacent nodes. When the waveform similarity is lower than a preset threshold, the nodes are re-searched within a preset time window to adjust the sorting, thus completing the closed-loop correction and verification of the node timing. Topology mapping output module: Based on the verified node timing and the preset node attribute database, it generates and outputs node topology data that represents the timing relationship of the physical connection of the link and the actual transmission delay of adjacent nodes.

Images