Real-time Broadcast Message Processing System and Method Based on Distributed Edge Computing

Abstract

This invention proposes a real-time broadcast message processing system and method based on distributed edge computing. The method includes: collecting raw broadcast data; constructing broadcast text based on the raw broadcast data; converting the broadcast text into a phoneme encoding sequence and inputting the phoneme encoding sequence into a speech synthesis model to obtain audio data; constructing a priority scheduling score, which is calculated based on urgency, network availability, device idle time, and regional adaptation; executing a preset logical judgment and outputting a scheduling instruction packet based on the priority scheduling score and a preset scheduling threshold; executing corresponding actions based on the scheduling instruction packet and generating a unique identifier for completed broadcast actions. This achieves consistency in task response, adaptability of the broadcast link, and verifiability of execution results in a distributed edge environment.

Description

Technical Field

[0001] This invention belongs to the field of data processing, and in particular relates to a real-time broadcast message processing system and method based on distributed edge computing. Background Technology

[0002] With the large-scale deployment of networked and domestically produced broadcasting equipment, the traditional broadcast scheduling method centered on a central server is facing the problem of insufficient adaptability. Current public address systems mainly rely on a central platform to uniformly generate audio content and issue control commands, with each terminal node often only passively broadcasting and lacking autonomous processing capabilities. This architecture can meet the requirements for small-scale, low-frequency broadcasting tasks, but in complex scenarios such as smart campuses and smart transportation, due to frequent broadcasting tasks, significant regional differences, and high timeliness requirements, existing solutions expose problems such as excessive central load, fixed scheduling paths, and slow node response. Especially in distributed edge computing architectures, the quality of network connections between nodes is uncertain, and some terminals may experience high latency, low bandwidth, or even intermittent disconnection with the central system. This makes the traditional method relying on central scheduling and voice generation unable to meet the real-time, continuous, and regional autonomy requirements of broadcasting tasks. At the same time, existing systems lack semantic layer processing capabilities during message parsing, and can only achieve template matching based on field mapping. They cannot support dynamic judgment of urgency, regional adaptability, or scheduling priority, leading to broadcasting delays, disordered order, or playback failures during execution. In real-world scenarios, the heterogeneity of regional terminals makes it difficult to adapt to a unified content format and broadcasting method, thereby exacerbating the problem of system instability. Summary of the Invention

[0003] The purpose of this invention is to design a real-time broadcast message processing system and method based on distributed edge computing, which can autonomously complete operations such as broadcast content construction, broadcast path selection and scheduling instruction generation in complex environments such as when the central system is unreachable and the network status is unstable, thereby realizing closed-loop landing processing of broadcast tasks.

[0004] To achieve the above objectives, a method for real-time processing of broadcast messages based on distributed edge computing is provided in a first aspect of the present invention, the method comprising:

[0005] Collect raw broadcast data, which includes template number, template interpolation parameters, urgency level, and broadcast area number; construct broadcast text based on the raw broadcast data;

[0006] The broadcast text is converted into a phoneme encoding sequence, and the phoneme encoding sequence is input into a speech synthesis model to obtain audio data;

[0007] A priority scheduling score is constructed, which is calculated based on urgency, network availability, device idle time, and region adaptation. The region adaptation is obtained by comparing the broadcast region number specified by the task with the broadcast region number of the current node. Based on the priority scheduling score and a preset scheduling threshold, a preset logical judgment is performed to output a scheduling instruction packet.

[0008] Execute the corresponding action based on the scheduling instruction packet, and generate a unique identifier for the completed broadcast action.

[0009] Furthermore, after the broadcast text is constructed, field consistency is checked. If the number of template interpolation parameters is insufficient or the format is incorrect, empty placeholders are filled in. If the urgency exceeds the preset range, it is limited to the boundary value within the closed interval. If the broadcast area number is missing, the area number bound by the current device by default is used.

[0010] Furthermore, the phoneme encoding sequence is obtained by decomposing the broadcast text character by character using rule matching and static pinyin mapping, matching phonemes against the local static pinyin dictionary, and introducing short pause symbols for sentence segmentation.

[0011] Furthermore, the speech synthesis model adopts a compressed version of the Transformer-based nested structure, which includes: a three-layer self-attention network, each layer nesting local convolutional channels; a 128-dimensional representation in each layer, with the activation function fixed as a modified linear unit; and the last layer outputs a frame-level spectrogram with a shape of (m×80), where m is the number of time frames.

[0012] Furthermore, a synthesis speed control term is introduced into the speech synthesis model to reduce latency when the urgency is high.

[0013] Furthermore, the presupposed logic is as follows:

[0014] When the priority scheduling score is higher than the preset scheduling threshold and the network availability is higher than the network availability threshold, it is immediately broadcast via IP.

[0015] When the priority scheduling score is higher than the preset scheduling threshold and the network availability is lower than the network availability threshold, 100V simulated backup broadcast is immediately used.

[0016] When the priority scheduling score is lower than the preset scheduling threshold and the urgency is higher than the minimum urgency threshold, the queue is delayed.

[0017] The task is discarded when the urgency level falls below the minimum urgency threshold.

[0018] Furthermore, the scheduling instruction package includes: the broadcast path selection result, whether to broadcast immediately, the audio resource bound to the broadcast, and the task execution priority identifier.

[0019] Furthermore, when broadcasting via IP, the audio interface module is invoked to send audio data to the IP speaker device group bound to the current node; when broadcasting via 100V analog backup, the audio driver interface is invoked to output audio data to the analog channel controlled by the power amplifier module through the sound card driver.

[0020] Furthermore, the unique identifier contains a quick summary of the audio data, its urgency, and a timestamp, represented by a hash function.

[0021] A second aspect of the invention provides a real-time broadcast message processing system based on distributed edge computing, the system comprising:

[0022] The field extraction unit is used to collect raw broadcast data, which includes template number, template interpolation parameters, urgency level, and broadcast area number; and to construct broadcast text based on the raw broadcast data.

[0023] An audio generation unit is used to convert the broadcast text into a phoneme encoding sequence and input the phoneme encoding sequence into a speech synthesis model to obtain audio data.

[0024] The scheduling judgment unit is used to construct a priority scheduling score, which is calculated by urgency, network availability, device idle time and region adaptation. The region adaptation is obtained by comparing the broadcast region number specified by the task with the broadcast region number of the current node. Based on the priority scheduling score and a preset scheduling threshold, a preset logical judgment is performed to output a scheduling instruction packet.

[0025] The broadcast execution unit is used to execute corresponding actions according to the scheduling instruction package and generate a unique identifier for the completed broadcast actions.

[0026] The beneficial technical effects of the present invention are at least as follows:

[0027] To address the aforementioned issues, this invention provides a real-time broadcast message processing system and method based on distributed edge computing. This system can autonomously complete operations such as broadcast content construction, broadcast path selection, and scheduling instruction generation in complex environments such as when the central system is unreachable or the network is unstable, achieving closed-loop processing of broadcast tasks. The system first parses broadcast message fields locally to construct the semantic structure of the broadcast task on the terminal. Then, it combines a template library and parameter mapping to generate the final broadcast text, which is then converted into playable audio data using a locally deployed lightweight speech synthesis model. The synthesized audio task is given a broadcast priority score based on multi-dimensional status parameters such as urgency, regional adaptability, network availability, and device idle time, ultimately determining whether to execute broadcast immediately, use an IP channel or an analog channel, and whether to enter a delay queue. After the scheduling module outputs structured control instructions, the system can drive the corresponding broadcast channel to complete audio playback and generate a unique identifier based on the task region, urgency, and timestamp, completing the task execution loop. The entire process does not rely on central node control; all key logic is autonomously completed at the edge, exhibiting strong anti-interference capabilities and flexible scheduling.

[0028] Through the above mechanism, this invention, guided by task semantics and with executable audio tasks as the core processing unit, achieves a high degree of integration of content generation and broadcasting behavior in broadcast scenarios, and constructs an intelligent broadcasting system architecture for complex deployment environments. Attached Figure Description

[0029] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0030] Figure 1 This is a flowchart of the real-time broadcast message processing method based on distributed edge computing of the present invention.

[0031] Figure 2 This is a framework diagram of the real-time broadcast message processing system based on distributed edge computing of the present invention. Detailed Implementation

[0032] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0033] In one or more embodiments, such as Figure 1 As shown, a real-time processing method for broadcast messages based on distributed edge computing is disclosed, the method comprising the following:

[0034] S1: Collect raw broadcast data, which includes template number, template interpolation parameters, urgency level, and broadcast area number; construct broadcast text based on the raw broadcast data;

[0035] Specifically, this step is performed at the edge nodes of the broadcast system. Its purpose is to convert broadcast messages sent by the central control system into a field structure with explicit broadcast semantics. Since broadcast messages are typically compressed, using template numbers and parameter padding to transmit content, and also include a small amount of control information, structural parsing and semantic recovery are necessary at the edge nodes. This step is characterized by minimizing the number of fields and transmission redundancy to achieve a localized understanding of the three core elements: "what to broadcast, who to broadcast to, and whether it is urgent," and outputting them in a standard structure for use by the speech synthesis and scheduling decision-making modules. This processing method ensures that the edge nodes retain full autonomous broadcasting capabilities even when the central system is offline or delayed.

[0036] The input is a structured message, denoted as M, sent from the central system to the edge nodes via a network broadcast protocol. It contains the following four data fields:

[0037] θ: Template number, corresponding to the index of the broadcast text template stored locally, specified by the central task generator;

[0038] α: Template interpolation parameter, which is a string or a list of numbers used to fill the placeholder positions in the template;

[0039] δ: Urgency, a real number in the interval [0,1], indicating the scheduling priority of the current task;

[0040] φ: Broadcast area number, which can be an integer or a predefined code, used to identify the physical area to be broadcast.

[0041] The above data is generated by the control system during the task issuance phase, received by the communication module, and cached in the memory of the edge terminal.

[0042] Specifically, after receiving M, the edge system calls the locally stored template set. Find the corresponding template using index θ. Pre-installed during the deployment phase, the content format is like "Please evacuate the area [ ] as soon as possible", where [ ] represents the position to be filled.

[0043] The template obtained after the search is denoted as

[0044] Further, a template filling operation is performed, embedding the parameter α sequentially into the corresponding slots of the template to construct the complete broadcast text s:

[0045]

[0046] Among them, symbols This indicates the sequential insertion operation between the string template and the parameter. For example, if... If α is "Please go to the assembly point in [ ] immediately", and α is "Library Area A", then s is "Please go to the assembly point in Library Area A immediately".

[0047] After the template is populated, the system performs field consistency checks. If the number of parameters for α is insufficient or the format is incorrect, an empty placeholder "—" is automatically filled in. If δ exceeds the allowed range, it is restricted to boundary values ​​within a closed interval. If φ is missing, the default region number bound to the current device is used. Finally, a set of structured semantic fields is constructed.

[0048] E = {s, δ, φ};

[0049] In this set, s is the text input for the subsequent speech synthesis module, δ will be used for scheduling and sorting, and φ determines the distribution range of the broadcast terminal.

[0050] S2: Convert the broadcast text into a phoneme encoding sequence, and input the phoneme encoding sequence into a speech synthesis model to obtain audio data;

[0051] Specifically, in this step, the system processes the text 's' to be broadcast, converting it into a phoneme encoding sequence 'x'. This process uses rule matching and static pinyin mapping, eliminating the need for online word segmentation or dynamic parsing. The input 's' is broken down character by character, and phoneme matching is performed against a local static pinyin dictionary (memory usage less than 500KB), incorporating short pause symbols '#'. pau Used for sentence segmentation, outputting a phoneme sequence x = {x1, x2, ..., x} n The length should not exceed 60.

[0052] Subsequently, the phoneme sequence x is fed into the locally deployed speech synthesis model. The model adopts a compressed version of the Transformer-based nested structure, which includes: a three-layer self-attention network, each layer nested with local convolutional channels; each layer has a 128-dimensional representation, with the activation function fixed as the rectified linear unit (ReLU); the last layer outputs a frame-level spectrogram y, which has a shape of (m×80), where m is the number of time frames.

[0053] The model is defined as follows:

[0054]

[0055] Where, λ reg ρ is a regularization factor used for compression control to prevent the spread of acoustic spectrum redundancy. δThe synthesis speed control term is designed based on the urgency level δ, indicating that the delay should be reduced when the urgency level is high, so that the synthesis is more compact and smooth.

[0056] The specific construction is as follows:

[0057] ρ δ =exp(-κ·δ);

[0058] κ is the control constant set by the system (recommended value κ = 1.5). When δ is close to 1, ρ δ Approaching a smaller value guides the model to reduce internal parameters related to speech rate, generating faster-paced, shorter-duration speech spectrograms.

[0059] After the spectrogram is generated, the edge nodes convert y into an audio stream a using a local fast vocoder module v (based on a deconvolutional network, deployed with fixed parameters), with a sampling rate of 16000Hz and an output in 16-bit PCM format. The generated audio is named using a hash index and stored in the local task directory according to the following caching rules:

[0060] a←Cache(y,φ,δ);

[0061] The caching module uses the region number φ as the path index and the urgency δ controls the priority loading weight, ensuring that the scheduling module can quickly retrieve the target task audio during real-time scheduling. The filename uses a hash value of 's' combined with a timestamp for encoding, ensuring that the content and task are bound together and cannot be duplicated.

[0062] The processing latency of the entire process is limited by the model structure and cached I / O, with the total time controlled between 400 and 700 milliseconds, which is suitable for typical configurations in domestic main control environments (1.2GHz ARM + 1GB memory).

[0063] S3: Construct a priority scheduling score, which is calculated based on urgency, network availability, device idle time, and region adaptation. The region adaptation is obtained by comparing the broadcast region number specified by the task with the broadcast region number of the current node. Based on the priority scheduling score and a preset scheduling threshold, execute a preset logical judgment and output a scheduling instruction packet.

[0064] Specifically, this step is executed on the edge node. Its purpose is to transform the audio broadcasting task synthesized in the previous step into an "executable" scheduling instruction package, which determines whether to broadcast it immediately, through which path, and under what conditions to switch to delayed processing.

[0065] In this step, the system also needs to dynamically obtain the following edge node running status parameters:

[0066] η: Network availability score, ranging from [0,1]. A higher value indicates that the current IP broadcast path is unobstructed.

[0067] σ: Node broadcast idle time, range [0,1]. The larger the value, the smaller the current task load of the device.

[0068] φ i : The region number to which the current node belongs. It is preset by the node's local configuration file and is not involved in the formula calculation but is used as a region comparison item.

[0069] First, the priority scheduling score is calculated and denoted as μ, with the following basic form:

[0070] μ=δ+λ1·η+λ2·σ+λ3·Ω(φ,φ i );

[0071] Where: the first term δ represents the urgency of the broadcast task itself; the second term η represents the current network availability; the third term σ represents the current idle level of the device; and the fourth term Ω(φ,φ) represents the current network availability. i ) represents the region adaptation item, defined as:

[0072]

[0073] The broadcast system terminals are highly geographically specific, with different areas typically corresponding to different floors, buildings, or functional zones. Tasks are most efficiently executed by the regional master node. Therefore, introducing an asymmetric regional penalty coefficient γ>0 in the scoring can effectively prevent low-fitness tasks from being mistakenly executed on non-target devices, ensuring closed-loop scheduling of broadcast tasks within the region.

[0074] In the scoring formula above, λ1+λ2+λ3<1, and this value is set through the deployment configuration file and does not support dynamic adjustment, in order to reduce algorithm complexity. Recommended values ​​are such as λ1=0.3, λ2=0.3, λ3=0.3, reserving a margin for δ to dominate.

[0075] Secondly, based on the scoring result μ, the system proceeds to the path selection and strategy generation stage, where a scheduling threshold θ and a minimum emergency threshold δ are set. min Perform the following logical judgment:

[0076]

[0077] In the above judgment, ∈ represents the network availability threshold (recommended to be set to 0.5). When the network is congested, it automatically degrades to analog channel broadcasting to ensure that the voice content is reachable; δ min To minimize urgency and prevent the system task queue from being occupied by a large number of low-priority broadcast tasks.

[0078] Finally, the system generates a broadcast scheduling instruction packet P, which includes: the broadcast path selection result (IP / analog); whether to broadcast immediately; the audio resource a bound to the broadcast; and the task execution priority identifier.

[0079] The instruction packet is written to the local execution task queue and executed by the system scheduling thread.

[0080] S4: Execute the corresponding action based on the scheduling instruction packet, and generate a unique identifier for the completed broadcast action.

[0081] Specifically, the goal of this step is to control the actual physical broadcasting device to complete the audio content broadcasting task on the edge node according to the scheduling instruction package generated in the previous step, and generate a unique identifier for the completed broadcasting behavior, focusing on two actions: (1) driving the broadcast channel to output audio; (2) generating a traceable and non-repeatable task fingerprint.

[0082] The local system determines the required control channel based on r:

[0083] If r = IP, the system calls the audio interface module to send audio data 'a' to the IP speaker device group bound to the current node. This multicast channel is determined by matching the region number 'φ'.

[0084] If r = analog, the system calls the audio driver interface to output the audio data a through the sound card driver to the analog channel controlled by the power amplifier module, and the constant voltage circuit completes the audio broadcast.

[0085] After audio playback starts, the system enters a synchronization wait state, listening to whether the playback device has successfully played the audio. Immediately after playback ends, a unique identifier χ for the task is generated. This identifier must satisfy the following conditions: the generated χ must always be different when the same task is executed at different nodes and at different times; however, the χ must be consistent when the same task is not played repeatedly at the same node.

[0086] The formula is constructed as follows:

[0087]

[0088] Where: H(·) is a locally deployed hash function, such as the domestic SM3; hash(a) is a fast digest of the audio data (such as the first 16 bits of SHA1); The urgency level is mapped to an integer label (e.g., 0.83 → 83); τ is the current system timestamp (in seconds); the symbol || represents the concatenation operation.

[0089] The generated χ is then written to the device's local execution log file, and the task status is marked as "executed". The log is formatted as a structured record with fields including χ, τ, φ, and r, and is recorded by a local lightweight database for subsequent task status queries or distributed comparisons.

[0090] In one or more embodiments, such as Figure 2 As shown, a real-time broadcast message processing system based on distributed edge computing is disclosed, the system comprising:

[0091] The field extraction unit is used to collect raw broadcast data, which includes template number, template interpolation parameters, urgency level, and broadcast area number; and to construct broadcast text based on the raw broadcast data.

[0092] An audio generation unit is used to convert the broadcast text into a phoneme encoding sequence and input the phoneme encoding sequence into a speech synthesis model to obtain audio data.

[0093] The scheduling judgment unit is used to construct a priority scheduling score, which is calculated by urgency, network availability, device idle time and region adaptation. The region adaptation is obtained by comparing the broadcast region number specified by the task with the broadcast region number of the current node. Based on the priority scheduling score and a preset scheduling threshold, a preset logical judgment is performed to output a scheduling instruction packet.

[0094] The broadcast execution unit is used to execute corresponding actions according to the scheduling instruction package and generate a unique identifier for the completed broadcast actions.

[0095] It is worth noting that the specific workflow of the real-time broadcast message processing system based on distributed edge computing provided in this embodiment of the invention is the same as that of the real-time broadcast message processing method based on distributed edge computing described in the above embodiments, and will not be repeated here.

[0096] This invention also provides a real-time broadcast message processing device based on distributed edge computing, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the steps described in the above embodiments of the real-time broadcast message processing method based on distributed edge computing. Figure 1 The steps S1 to S4 described above; or, when the processor executes the computer program, it implements the functions of each module in the above system embodiments.

[0097] For example, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the distributed edge computing-based real-time broadcast message processing device.

[0098] The real-time broadcast message processing device based on distributed edge computing can be a desktop computer, laptop, handheld computer, or cloud server, etc. This device may include, but is not limited to, processors and memory. Those skilled in the art will understand that the real-time broadcast message processing device based on distributed edge computing may also include input / output devices, network access devices, buses, etc.

[0099] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. This processor is the control center of the distributed edge computing-based real-time broadcast message processing device, connecting all parts of the device via various interfaces and lines.

[0100] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the distributed edge computing-based real-time broadcast message processing device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function, etc.; the data storage area may store data created based on the operation of the air conditioner controller, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart memory card (SMC), secure digital card (SD), flash memory card, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0101] The modules integrated into the distributed edge computing-based real-time broadcast message processing device, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0102] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0103] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A method for real-time processing of broadcast messages based on distributed edge computing, characterized in that, The method includes: S1: Collect raw broadcast data, which includes template number, template interpolation parameters, urgency level, and broadcast area number; construct broadcast text based on the raw broadcast data; S2: The broadcast text is converted into a phoneme encoding sequence, and the phoneme encoding sequence is input into the speech synthesis model to obtain audio data; the speech synthesis model adopts a compressed version of the Transformer-based nested structure, including: a three-layer self-attention network, each layer nesting local convolutional channels; each layer has a 128-dimensional representation, with the activation function fixed as the modified linear unit; the last layer outputs a frame-level spectrogram, which has a shape of m×80, where m is the number of time frames and 80 is the preset number of frequency dimensions; S3: Construct a priority scheduling score, which is calculated based on urgency, network availability, device idle time, and region adaptation. The region adaptation is obtained by comparing the broadcast region number specified by the task with the broadcast region number of the current node. Based on the priority scheduling score and a preset scheduling threshold, execute a preset logic judgment and output a scheduling instruction packet. The preset logic is as follows: When the priority scheduling score is higher than the preset scheduling threshold and the network availability is higher than the network availability threshold, it is immediately broadcast via IP. When the priority scheduling score is higher than the preset scheduling threshold and the network availability is lower than the network availability threshold, 100V simulated backup broadcast is immediately used. When the priority scheduling score is lower than the preset scheduling threshold and the urgency is higher than the minimum urgency threshold, the queue is delayed. Discard the task if the urgency level falls below the minimum urgency threshold. S4: Execute the corresponding action based on the scheduling instruction packet, and generate a unique identifier for the completed broadcast action.

2. The real-time broadcast message processing method based on distributed edge computing according to claim 1, characterized in that, After the broadcast text is constructed, field consistency is checked. If the number of template interpolation parameters is insufficient or the format is incorrect, an empty placeholder is filled in. If the urgency exceeds the preset range, it is limited to the boundary value within the closed interval. If the broadcast area number is missing, the area number bound by the current device by default is used.

3. The real-time broadcast message processing method based on distributed edge computing according to claim 1, characterized in that, The phoneme encoding sequence is obtained by decomposing the broadcast text character by character using rule matching and static pinyin mapping, matching phonemes against a local static pinyin dictionary, and introducing short pause symbols for sentence segmentation.

4. The real-time broadcast message processing method based on distributed edge computing according to claim 1, characterized in that, The speech synthesis model also incorporates a synthesis speed control term to reduce latency when urgency is high.

5. The real-time broadcast message processing method based on distributed edge computing according to claim 1, characterized in that, The scheduling instruction package includes: the playback path selection result, whether to play immediately, the audio resource bound to the playback, and the task execution priority identifier.

6. The real-time broadcast message processing method based on distributed edge computing according to claim 1, characterized in that, When broadcasting via IP, the audio interface module is invoked to send audio data to the IP speaker device group bound to the current node; when broadcasting via 100V analog backup, the audio driver interface is invoked to output audio data to the analog channel controlled by the power amplifier module through the sound card driver.

7. The real-time broadcast message processing method based on distributed edge computing according to claim 1, characterized in that, The unique identifier contains a quick summary, urgency, and timestamp of the audio data, which are calculated using a hash function.

8. A real-time broadcast message processing system based on distributed edge computing, characterized in that, The system includes: The field extraction unit is used to collect raw broadcast data, which includes template number, template interpolation parameters, urgency level, and broadcast area number; and to construct broadcast text based on the raw broadcast data. An audio generation unit is used to convert the broadcast text into a phoneme encoding sequence and input the phoneme encoding sequence into a speech synthesis model to obtain audio data. The speech synthesis model adopts a compressed version of the Transformer-based nested structure, which includes: a three-layer self-attention network, each layer nesting local convolutional channels; each layer has a 128-dimensional representation, with the activation function fixed as a modified linear unit; the last layer outputs a frame-level spectrogram with a shape of m×80, where m is the number of time frames and 80 is the preset number of frequency dimensions. The scheduling judgment unit is used to construct a priority scheduling score, which is calculated based on urgency, network availability, device idle time, and region adaptation. The region adaptation is obtained by comparing the broadcast region number specified by the task with the broadcast region number of the current node. Based on the priority scheduling score and a preset scheduling threshold, a preset logic judgment is executed to output a scheduling instruction packet. The preset logic is as follows: When the priority scheduling score is higher than the preset scheduling threshold and the network availability is higher than the network availability threshold, it is immediately broadcast via IP. When the priority scheduling score is higher than the preset scheduling threshold and the network availability is lower than the network availability threshold, 100V simulated backup broadcast is immediately used. When the priority scheduling score is lower than the preset scheduling threshold and the urgency is higher than the minimum urgency threshold, the queue is delayed. Discard the task if the urgency level falls below the minimum urgency threshold. The broadcast execution unit is used to execute corresponding actions according to the scheduling instruction package and generate a unique identifier for the completed broadcast actions.

Images