Multi-scene adaptive communication method and system of VOIP product

Abstract

The invention relates to the technical field of VOIP product communication, and discloses a multi-scene adaptive communication method and system of a VOIP product. The method comprises the following steps: carrying out collection and feature extraction on network state parameters to obtain a scene category, and calling reference communication parameters corresponding to the scene category from a scene parameter library; calculating a composite cumulative distribution function based on the reference communication parameters to obtain an optimized communication parameter combination; applying the optimized communication parameter combination to an edge cloud collaborative coding and decoding framework, and carrying out coding and decoding processing on audio and video data to obtain coded communication data; performing block processing on the encoded communication data, and determining a storage position and a communication node of a data block; according to the method, the communication nodes are divided into the master node and the slave node, and the master node and the slave node are subjected to multi-objective optimization to obtain the data transmission path, so that the dynamic optimization of the communication parameters is realized, a VOIP product has scene self-adaptive capability, and the communication service quality of the VOIP product is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of VOIP product communication, and in particular to a multi-scenario adaptive communication method and system for VOIP products. Background Art

[0002] With the rapid development of mobile Internet technology, VOIP products have become an important tool for people's daily communication. However, in actual applications, due to the diversity and complexity of user communication scenarios, traditional VOIP products are difficult to adapt to changes in network environments in different scenarios. For example, in high-speed rail mobile scenarios, rapid changes in network status lead to unstable call quality; in underground communication scenarios, signal attenuation seriously affects communication reliability; in indoor fixed scenarios, network congestion causes call delays and packet loss.

[0003] Existing VOIP products generally use fixed communication parameter configuration schemes and lack the ability to adapt to different scenarios. In terms of codec processing, traditional methods usually concentrate all computing tasks on terminal devices or cloud processing, failing to fully utilize edge computing resources, resulting in uneven system load distribution and low processing efficiency. At the same time, existing technologies do not consider the cooperation mechanism between nodes during data transmission, and cannot effectively deal with abnormal situations such as network failures and link interruptions. Summary of the invention

[0004] The present invention provides a multi-scenario adaptive communication method and system for a VOIP product. The present invention realizes dynamic optimization of communication parameters, enables the VOIP product to have scenario adaptive capability, and improves the communication service quality of the VOIP product.

[0005] In a first aspect, the present invention provides a multi-scenario adaptive communication method for a VOIP product, the multi-scenario adaptive communication method for a VOIP product comprising: Collecting and extracting features of network status parameters to obtain scene categories, and retrieving reference communication parameters corresponding to the scene categories from a scene parameter library; Calculate a composite cumulative distribution function based on the reference communication parameters to obtain an optimized communication parameter combination; Applying the optimized communication parameter combination to the edge-cloud collaborative coding and decoding framework, and performing coding and decoding processing on the audio and video data to obtain coded communication data; Processing the encoded communication data in blocks, and determining storage locations and communication nodes of the data blocks; The communication nodes are divided into master nodes and slave nodes, and multi-objective optimization is performed on the master nodes and slave nodes to obtain a data transmission path.

[0006] In a second aspect, the present invention provides a multi-scenario adaptive communication system for a VOIP product, the multi-scenario adaptive communication system for a VOIP product comprising: A collection module, used to collect and extract features of network status parameters, obtain scene categories, and retrieve reference communication parameters corresponding to the scene categories from a scene parameter library; A calculation module, used for calculating a composite cumulative distribution function based on the reference communication parameters to obtain an optimized communication parameter combination; A coding and decoding module, used to apply the optimized communication parameter combination to the edge-cloud collaborative coding and decoding framework, and perform coding and decoding processing on the audio and video data to obtain coded communication data; A block division module, used for performing block processing on the encoded communication data and determining the storage location and communication node of the data block; The multi-objective optimization module is used to divide the communication nodes into master nodes and slave nodes, and perform multi-objective optimization on the master nodes and slave nodes to obtain a data transmission path.

[0007] In the technical solution provided by the present invention, by constructing a communication parameter adaptive adjustment mechanism of a composite distribution function and a Fox H function optimization algorithm, this solution can accurately evaluate the network status, realize dynamic optimization of communication parameters, enable VOIP products to have scene adaptation capabilities, and ensure call quality in different scenarios. Based on the edge-cloud collaborative coding and decoding framework and the feature extraction method of the pulse neural network SNN, combined with the adaptive redundant coding of the HARQ mechanism, this solution realizes the reasonable allocation of computing resources and the reliability of data transmission, and solves the problem of low processing efficiency in traditional methods. By adopting the distributed caching strategy based on the Lagrange duality principle and the task allocation mechanism based on dynamic programming, this solution realizes efficient storage and rapid recovery of data, and effectively reduces system latency. Through the multi-objective optimization of direct communication technology between devices and the P-SCA algorithm, this solution establishes an efficient multi-node collaboration framework, which can respond in time when a node fails or a link is interrupted, and improves the fault tolerance of the system. Based on a multi-dimensional performance evaluation system, including a comprehensive evaluation of QoS scoring, user experience scoring, and system resource utilization, this solution realizes the precise adjustment of communication parameters and ensures the stable operation of VOIP products in various scenarios. Through the fuzzy comprehensive evaluation algorithm and dynamic parameter update mechanism, this solution has established a complete performance feedback adjustment mechanism, which can continuously optimize the system parameters according to the actual communication effect, thereby improving the communication service quality of VOIP products. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0009] Figure 1 A flowchart of a multi-scenario adaptive communication method for a VOIP product provided in an embodiment of the present application; Figure 2 A schematic block diagram of the structure of a multi-scenario adaptive communication system for a VOIP product provided in an embodiment of the present application. DETAILED DESCRIPTION

[0010] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0011] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change based on actual conditions.

[0012] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0013] It should be further understood that the term “and / or” used in the specification and appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0014] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0015] See also Figure 1 , Figure 1 A flowchart of a multi-scenario adaptive communication method for a VOIP product provided in an embodiment of the present application is shown in FIG. Figure 1As shown, the multi-scenario adaptive communication method of the VOIP product provided in the embodiment of the present application includes steps S100 to S500.

[0016] Step S100: collecting and extracting features of network status parameters to obtain a scene category, and retrieving a reference communication parameter corresponding to the scene category from a scene parameter library; It is understandable that the execution subject of the present invention can be a multi-scenario adaptive communication system of a VOIP product, or a terminal or a server, which is not specifically limited here. The embodiment of the present invention is described by taking a server as the execution subject as an example.

[0017] Specifically, a network status monitoring system is constructed, which collects key network parameters such as bandwidth utilization, delay jitter, packet loss rate, and signal strength in real time to obtain the original data of the network status. The collection of these data is completed through technical means such as network probes, signaling monitoring, and data packet analysis, and a high-precision data synchronization mechanism is used to ensure the time alignment between different parameters. Deep learning is performed on the original network status data to extract the network status feature vector. A deep learning model based on a recurrent neural network, a long short-term memory network, or a variational autoencoder is constructed to capture the time series characteristics of the network status data, and high-dimensional feature vectors are extracted through the hierarchical structure of the model to obtain more detailed network status information. In this process, the time window sliding method is used to enable the deep learning model to effectively perceive the dynamic changes of the network status, while avoiding the impact of short-term abnormal data on the overall analysis results. The network status feature vector is classified by time series features to establish feature time series data. The feature time series data is probabilistically mapped to obtain the probability distribution of scene categories. Based on a Bayesian classifier, a hidden Markov model, or a classifier based on a deep neural network, the probability of the current network status corresponding to different communication scenarios is calculated to obtain the probability distribution of the scene category. The scene categories mainly include high-speed rail mobile scenes, indoor fixed scenes, outdoor roaming scenes, underground communication scenes and elevator scenes. In the classification process, the typical network characteristics of these scenes are combined. For example, the high-speed rail mobile scene is characterized by high-speed switching base station connection, large signal jitter and obvious packet loss rate; the indoor fixed scene has a relatively stable signal strength, small delay jitter and low packet loss rate; the outdoor roaming scene has slight signal fluctuations and high bandwidth availability due to the slow movement of users; the underground communication scene is characterized by high delay, low signal strength and high packet loss rate; and the elevator scene is characterized by sudden signal interruption and extreme network status fluctuations. Based on these characteristics, the classification model is trained to enable it to accurately infer the current scene category under different network conditions and present the possibility of different scenes in the form of probability distribution. After obtaining the scene category probability distribution, the current scene category is determined according to the set threshold or maximum probability criterion. A confidence threshold is set. When the probability of a scene exceeds the threshold, the device is considered to be in the scene; if the probabilities of multiple scenes are close, time smoothing is combined with historical data to reduce the error rate of scene judgment. After determining the specific scenario category, the corresponding benchmark communication parameters are retrieved from the scenario parameter library to adapt to different network environments and optimize the communication experience. The selection of benchmark communication parameters includes coding rate, frame length, sampling rate, quantization bit number and transmission protocol type. In different communication scenarios, the configuration of these parameters is different to meet the requirements of bandwidth adaptability and service quality.

[0018] Step S200, calculating a composite cumulative distribution function based on the reference communication parameters to obtain an optimized communication parameter combination;Specifically, the cumulative probability of bandwidth utilization and delay jitter in the benchmark communication parameters is calculated to obtain the cumulative distribution function of bandwidth utilization and the cumulative distribution function of delay jitter respectively. Bandwidth utilization is an important indicator to measure the occupancy of network resources, while delay jitter reflects the time fluctuation of data packets during transmission. In the calculation process, by statistically modeling the historical communication data and using the empirical distribution function or the probability distribution fitting method based on kernel density estimation, their respective cumulative distribution functions (CDFs), namely the CDF of bandwidth utilization and the CDF of delay jitter, are obtained to characterize the dynamic characteristics of the network under different communication environments. The joint distribution calculation of the bandwidth utilization cumulative distribution function and the delay jitter cumulative distribution function is performed to construct a composite cumulative distribution function. The calculation of the joint distribution is based on the binary joint probability density function and combined with the Copula method to capture the correlation between bandwidth utilization and delay jitter, and obtain a composite cumulative distribution function that is more in line with the actual situation. In this process, the Gaussian mixture model or the probability model based on the Markov process is used to ensure that the calculated composite cumulative distribution function can accurately characterize the time-varying characteristics under different network environments. The composite cumulative distribution function is input into the Fox H function for feature decomposition to extract its key features. The Fox H function is a generalized special function that performs high-order feature analysis on complex probability distributions and uses the Mellin transform for feature extraction. The composite cumulative distribution function is subjected to the Mellin transform, that is, it is mapped to the feature domain of the Fox H function through the integral transform to obtain a mathematical representation that can be used for decomposition. Using the analytical solution of the Fox H function, the distribution function is decomposed into a set of feature sequences, and high-order statistical characteristics such as mean, variance, skewness and kurtosis are extracted. Through the feature decomposition of the Fox H function, a feature sequence that can characterize the dynamic characteristics of the network environment is obtained. The Fox H function feature sequence is recursively iterated to construct an n×m-dimensional optimization mapping matrix, where n represents the number of discrete levels of bandwidth utilization and m represents the number of discrete levels of delay jitter. The construction of the optimization mapping matrix is ​​based on the feature sequence of the Fox H function, and an interpolation algorithm or a deep learning method, such as an autoencoder or a long short-term memory network, is used to ensure that the matrix can be adaptively adjusted under different network conditions. The role of the optimization mapping matrix is ​​to provide a parameter optimization search space, so that the optimal communication parameter combination can be dynamically selected under different network conditions, thereby improving the adaptability of VOIP. The Fox H function optimization algorithm is used to iteratively solve the optimization mapping matrix to obtain the optimal communication parameter value. The Fox H function optimization algorithm is an optimization method based on eigendecomposition and recursive iteration, which can quickly converge to the optimal solution in a high-dimensional parameter space. During the calculation process, the optimization mapping matrix is ​​solved by gradient descent, Newton iteration or random optimization methods (such as genetic algorithms or particle swarm optimization), and the optimal communication parameter combination is determined by minimizing the loss function.During the solution process, the QoE (quality of experience) indicator is considered to ensure that the optimized parameter combination can provide the best communication effect in different scenarios. The optimal communication parameter values ​​are parsed to extract the optimized communication parameter combination, which includes the updated coding rate, frame length, sampling rate, quantization bit number and transmission protocol type. The optimized coding rate should be dynamically adjusted based on the network bandwidth to ensure that basic voice or video quality can be maintained in low-bandwidth environments, while higher sound quality and video clarity can be provided in high-bandwidth environments. The optimization of frame length requires a balance between delay and computational complexity. In an environment with large network jitter, a shorter frame length is used to reduce the delay fluctuation in transmission, while in a stable network environment, the frame length is appropriately extended to reduce the encoding overhead. The adjustment of the sampling rate should be combined with the network conditions and sound quality requirements. When the bandwidth is limited, the sampling rate is reduced to reduce the amount of data, while when the network conditions are good, a higher sampling rate is used to obtain clearer sound quality. The optimization of the number of quantization bits mainly affects the dynamic range and signal-to-noise ratio of the signal. In low-bandwidth scenarios, the number of quantization bits is reduced to reduce the amount of data transmission, while in high-quality voice scenarios, the number of quantization bits is increased to enhance the fidelity of sound quality. The selection of the transmission protocol type is adapted according to the network characteristics. For example, in an environment with a high packet loss rate, the UDP transmission protocol based on FEC (forward error correction) is used, while in an environment that requires data reliability, the transmission mode based on TCP or QUIC protocol is used to ensure data integrity.

[0019] Step S300: Apply the optimized communication parameter combination to the edge-cloud collaborative coding and decoding framework, and perform coding and decoding processing on the audio and video data to obtain coded communication data; Specifically, the optimized communication parameter combination is parsed into different codec parameters suitable for the terminal side and the cloud. The lightweight codec parameters on the terminal side should meet the requirements of low power consumption, low computational complexity and low latency, while the high-performance codec parameters on the cloud side need to have stronger computing power to support higher quality encoding processing. A parameter parsing model is constructed, which, according to the optimized communication parameter combination, splits the parameters into sub-parameter sets applicable to the terminal side and the cloud side respectively according to computing resources, network bandwidth and real-time requirements, so that the terminal device can quickly perform preprocessing of audio and video data, while the cloud side undertakes higher precision encoding and decoding tasks, forming a collaborative processing mechanism. Based on the lightweight codec parameters on the terminal side, the audio and video data is feature extracted and processed by a three-layer pulse neural network SNN, where the number of neurons in the first layer of SNN is 64, the number of neurons in the second layer is 32, and the number of neurons in the third layer is 16, so that the distribution of neurons gradually extracts the key features in the data. The pulse neural network is a computing model that simulates the biological nervous system. It is characterized by using pulse signals to transmit information, and has strong sparse computing capabilities and low power consumption characteristics. It is suitable for efficient feature extraction of audio and video data. Through the membrane potential accumulation mechanism of the pulse neural network, the speech features in the audio signal or the spatial features in the video frame are modeled, thereby reducing the amount of calculation while maintaining the effective information of the signal. After the feature extraction is completed, the audio and video feature data is processed by dynamic time step encoding to optimize the time expression of the data and reduce unnecessary data redundancy. Dynamic time step coding is an adaptive coding technology that dynamically adjusts the time step according to the change rate of the audio and video data, so that a shorter step is used in the high-change area to ensure the accuracy of details, and a longer step is used in the stable area to reduce redundant data. Through this coding strategy, while ensuring the quality of audio and video data, the bandwidth required for data transmission is reduced, and the real-time performance of encoding and decoding is improved. After dynamic time step coding, the data is converted into pulse coded data, which is more suitable for efficient processing in the cloud and can better adapt to communication needs under different network conditions. The pulse coded data and the cloud high-performance encoding and decoding parameters are input into the cloud high-performance encoding and decoding module for data processing to generate initial encoded data. The high-performance codec module in the cloud is based on models with higher computing power, such as deep neural networks or variational autoencoders, which can reconstruct pulse-coded data more accurately and adjust the coding strategy according to different application requirements. For example, in voice communication, a speech enhancement model based on a convolutional neural network is used to reduce noise interference and improve speech clarity; in video communication, a coding strategy based on a spatiotemporal attention mechanism is used to improve video compression efficiency and enhance detail expression capabilities. Through the processing of the high-performance codec module in the cloud, the redundancy of the data is optimized and good communication quality can be ensured under different bandwidth conditions.The HARQ (Hybrid Automatic Repeat Request) mechanism is used to perform redundant coding on the initial coded data to enhance the data's ability to resist packet loss and optimize transmission efficiency. The HARQ mechanism includes an adaptive redundant coding generation unit and a feedback control unit, wherein the adaptive redundant coding generation unit adjusts the redundant coding ratio according to the current network status and adopts forward error correction coding to reduce the degradation of audio and video quality caused by data packet loss. The feedback control unit dynamically adjusts the HARQ parameters based on real-time network feedback information. For example, when a high packet loss rate is detected, the redundant coding ratio is increased, and when the network condition is good, the redundant coding ratio is reduced to reduce additional bandwidth overhead. After the HARQ mechanism generates adaptive redundant coding data, the data is further optimized in combination with the feedback information of the feedback control unit to finally obtain the encoded communication data. An optimization strategy based on deep reinforcement learning is adopted to adaptively adjust the redundancy, grouping strategy and transmission mode of the data under different network conditions. For example, when the network condition is poor, the low-latency transmission mode is preferred, and the redundancy is increased to reduce data loss; when the network condition is good, a more efficient coding strategy is adopted to maximize the data transmission rate. During the data optimization process, error correction coding based on Raptor code or LT code is used to improve the robustness of data transmission and reduce the degradation of audio and video quality caused by packet loss.

[0020] Step S400, dividing the encoded communication data into blocks and determining the storage location and communication node of the data block; Specifically, the encoded communication data is processed in blocks based on the Lagrangian duality principle to form multiple data blocks. Through the Lagrangian duality method in optimization theory, the data block problem is transformed into a constrained optimization problem, in which the goal is to minimize the resource overhead required for data storage and transmission while ensuring data accessibility and system throughput. By constructing a Lagrangian function, the data block strategy is mapped to a convex optimization problem, and the dual variable is used to solve it, so that the data blocks after block division can not only meet the constraints of storage and computing resources, but also improve the overall data processing efficiency. After the data block division is completed, the access frequency statistics and storage cost calculation of each data block are performed to determine the storage strategy and scheduling priority of the data in the system. The access frequency is counted through historical access logs and real-time monitoring traffic data, while the storage cost depends on multiple factors such as the type of storage medium, the occupancy of storage capacity, and the energy consumption overhead required to access data. Priority calculation is performed based on the data block access frequency and storage cost. The calculation of cache priority follows the formula P=f / c, where f represents the access frequency of the data block and c represents the storage cost. The logic of this formula is that when a data block is accessed frequently, its priority should be increased, and when the storage cost is high, its priority should be reduced, thereby improving the availability of high-frequency access data under limited storage resources. The distance and signal strength of adjacent communication nodes are calculated in real time to evaluate the connection quality between each communication node. The node distance value is obtained by GPS location, signal delay measurement or RSSI (received signal strength indication) calculation, while the signal strength value is obtained by real-time wireless channel measurement. After obtaining the node distance and signal strength, the node distance value is compared with the preset distance threshold D to determine whether the physical distance between the communication nodes is within an acceptable range. At the same time, the signal strength value is compared with the preset signal strength threshold S to ensure that the quality of the communication link is stable enough. Through this comparison process, communication nodes suitable for data storage and transmission are screened out, and unstable nodes are excluded. Based on the comprehensive comparison results and cache priority, task allocation calculation is performed through the dynamic programming algorithm to determine the optimal scheduling scheme for the task. The dynamic programming algorithm recursively decomposes the problem and uses the optimal solution of the sub-problem to construct the overall optimal solution. The goal of task allocation calculation is to minimize the total delay of the system. Therefore, an optimization objective function is constructed, which comprehensively considers the cache priority of data blocks, the signal strength of nodes, and the delay of data transmission to minimize the data transmission time of the entire system. The optimization constraints include storage resource constraints, network bandwidth constraints, and node computing power constraints. In the actual solution process, the Bellman recursive equation is used, or combined with heuristic optimization algorithms (such as genetic algorithms, ant colony optimization, or particle swarm optimization) to improve the computing efficiency, so as to quickly obtain the optimal task allocation solution.The data blocks are allocated for storage according to the task allocation scheme to finally determine the storage location and communication node of the data blocks. According to the task allocation scheme, high-priority data blocks are stored on communication nodes with low latency and good signal quality to ensure the stability of data access, while low-priority data blocks are stored on remote or low-frequency access storage nodes to save storage resources. In the storage allocation process, a data redundancy strategy is adopted, that is, multiple copies of data are stored between multiple communication nodes to improve data availability and fault tolerance. The redundancy strategy is implemented based on erasure coding or replication strategy, so that even if a communication node fails, the system can still recover data from other storage nodes, thereby improving overall reliability.

[0021] Step S500: divide the communication nodes into master nodes and slave nodes, and perform multi-objective optimization on the master nodes and the slave nodes to obtain a data transmission path.

[0022] Specifically, the computing power and network connection quality of the communication nodes are comprehensively evaluated to obtain node evaluation parameters. The evaluation process involves multiple key indicators, including computing power, storage resources, bandwidth occupancy, link stability, and delay jitter. The computing power is measured by the node's CPU processing power, memory available space, and current computing load, while the network connection quality is measured by parameters such as bandwidth, signal strength, packet loss rate, and delay. After obtaining these node evaluation parameters, the communication nodes are divided into master nodes and slave nodes according to predefined division criteria. The master node has higher computing power, more stable network connection, and stronger cache and forwarding capabilities, while the slave node is used to assist in data processing and transmission to improve the load balancing and communication efficiency of the overall network. The status information of the master node and the slave node is collected to obtain dynamically changing node status data. These status data include indicators such as computing load, cache capacity, and link quality. The computational load is used to measure whether the processing capacity of the node is close to saturation to prevent overload from affecting data transmission efficiency; the cache capacity reflects the storage capacity of the node and determines its cache capacity in data forwarding, while the link quality directly affects the stability of data transmission, especially in a weak network environment, where the dynamic change of link quality can lead to data loss or transmission delay. Based on the node status data, a connection topology graph G between nodes is constructed. The topology graph is used to describe all communication nodes and their connection relationships, including information such as transmission links, link quality, and bandwidth availability between master and slave nodes. After the topology graph is constructed, a multi-objective optimization calculation is performed on the topology graph G based on the P-SCA algorithm to obtain an optimization parameter group. The P-SCA algorithm is based on the Pareto optimal solution set and weighs different optimization objectives to optimize multiple key performance indicators at the same time. The main optimization objectives include minimizing system delay, maximizing network throughput, and balancing node load. Minimizing system delay means that when selecting a data transmission path, nodes with low latency and high link stability should be given priority for data forwarding, while maximizing network throughput requires that high-bandwidth links be fully utilized during path planning to avoid traffic bottlenecks and maximize the parallel transmission capability of data. Node load balancing is to ensure that communication tasks are reasonably distributed among multiple nodes to prevent some master nodes from affecting the overall communication quality due to overload. In the P-SCA calculation process, reinforcement learning or genetic algorithm is used to search the Pareto optimal solution set to obtain the optimal parameter combination that meets the requirements of multi-objective optimization. The communication path is planned based on the optimized parameter group to obtain the initial transmission path. The master node assumes the responsibility of global communication control and data forwarding, while the slave node is responsible for data processing and relay transmission to optimize data transmission efficiency. In the path planning process, a dynamic weight adjustment method is used to adjust the distribution ratio of data flow according to the load status of the master node and the slave node.For example, in some high-load situations, the master node delegates part of the computing tasks to multiple slave nodes for parallel processing, while in low-load situations, the data flow is centralized to reduce communication overhead. At the same time, considering the dynamic changes of the network, the A* search algorithm or Dijkstra algorithm is used to quickly search the path to ensure the computational efficiency of the initial path. The validity of the initial transmission path is verified to ensure that it can adapt to the real-time changing network environment. An active link monitoring mechanism is used to monitor each node in the path in real time, and the link quality is detected by packet detection or heartbeat signals. By comparing the current link status with historical statistical data, it is determined whether a node has an abnormal situation, and a path verification report is generated based on the detection results. If a node failure or link interruption is detected, the path reconstruction mechanism is immediately triggered to ensure the continuity of data transmission. During the path reconstruction process, the communication path is dynamically adjusted based on the previously optimized parameter group to bypass the faulty node or low-quality link. For example, when a master node cannot process data due to excessive load, the nearest slave node is selected as a replacement, and the data transmission strategy is readjusted to reduce the loss or delay of data packets. At the same time, in the process of path optimization, combined with predictive optimization algorithms, such as LSTM (Long Short-Term Memory Network) or GCN (Graph Neural Network)-based methods, models are trained using historical data to predict in advance which nodes may be overloaded or have reduced link quality, and to adjust the path before the problem occurs, thereby improving system stability. The optimized data transmission path is obtained through the above steps.

[0023] The QoS (Quality of Service) score is calculated based on end-to-end delay, packet loss rate, jitter amplitude, and audio clarity to quantify call quality. End-to-end delay is a key indicator to measure the time required for a data packet to travel from the sender to the receiver, while the packet loss rate determines the stability of voice and video. A high packet loss rate can cause audio and video to freeze or even be interrupted. These two parameters play an important role in QoS score calculation. The jitter amplitude reflects the stability of the network state. When the jitter amplitude is large, it can cause intermittent voice or video to be out of sync. A statistical method is used to perform mean square error analysis on the jitter, and the score is adjusted in combination with the network buffering strategy. At the same time, audio clarity is related to factors such as the encoding and decoding method, sampling rate, and signal-to-noise ratio. It is evaluated using signal distortion or an audio enhancement model based on deep learning. After comprehensively calculating these factors, the QoS score is obtained. The user feedback data is processed through a fuzzy comprehensive evaluation algorithm to obtain a user experience score. The calculation of user experience score needs to comprehensively consider subjective score and objective feedback data. Subjective score is obtained through user survey, satisfaction score of voice interaction system, etc., while objective feedback data is based on call log analysis, such as call duration, frequency of user active call interruption, redial rate and whether user chooses to change network after the call. The core idea of ​​fuzzy comprehensive evaluation algorithm is to quantify multi-dimensional user feedback data by fuzzy mathematics, and unify subjective and objective factors based on fuzzy membership function to obtain the final user experience score. In the calculation process, fuzzy decision matrix is ​​constructed, and hierarchical analysis method is used to perform weighted calculation on different evaluation indicators. At the same time, CPU occupancy, memory usage and bandwidth occupancy are monitored to obtain system resource utilization. By regularly sampling CPU load, memory usage and network traffic data, and combining moving weighted average method to smooth the data, the impact of short-term fluctuations on the evaluation results is reduced, and the calculation of resource utilization can accurately reflect the load of the system. The call quality score, user experience score and system resource utilization are weighted and integrated to obtain the performance evaluation result. A weighted linear regression model is used, with QoS score, user experience score and system resource utilization as input variables, and an optimized set of weighted coefficients is trained based on historical data to ensure that the weight distribution of different indicators meets the actual communication needs. The performance evaluation results are compared with the preset performance threshold to determine whether the current communication status of the system meets expectations. If the performance evaluation result is lower than the set threshold, it indicates that there is room for optimization in the current communication quality, user experience or system resource management, so the parameter adjustment signal is triggered to dynamically optimize the communication parameters of the VOIP system. The generation of the parameter adjustment signal is based on the threshold trigger mechanism, or a reinforcement learning method (such as a Q-Learning-based adaptive strategy) is used to automatically adjust the trigger threshold according to historical data to improve adaptability.When the parameter adjustment signal is triggered, the system updates the benchmark communication parameters in the scenario parameter library based on the current performance evaluation data and the existing scenario classification information, thereby optimizing the communication strategy.

[0024] In the embodiment of the present invention, by constructing a communication parameter adaptive adjustment mechanism of a composite distribution function and a Fox H function optimization algorithm, the scheme can accurately evaluate the network status, realize dynamic optimization of communication parameters, enable VOIP products to have scene adaptation capabilities, and ensure call quality in different scenarios. Based on the edge-cloud collaborative coding and decoding framework and the feature extraction method of the pulse neural network SNN, combined with the adaptive redundant coding of the HARQ mechanism, the scheme realizes the reasonable allocation of computing resources and the reliability of data transmission, and solves the problem of low processing efficiency in traditional methods. By adopting the distributed caching strategy based on the Lagrange duality principle and the task allocation mechanism based on dynamic programming, the scheme realizes efficient storage and rapid recovery of data, and effectively reduces system latency. Through the multi-objective optimization of direct communication technology between devices and the P-SCA algorithm, the scheme establishes an efficient multi-node collaboration framework, which can respond promptly when a node fails or a link is interrupted, and improves the fault tolerance of the system. Based on a multi-dimensional performance evaluation system, including a comprehensive evaluation of QoS score, user experience score and system resource utilization, the scheme realizes the precise adjustment of communication parameters and ensures the stable operation of VOIP products in various scenarios. Through the fuzzy comprehensive evaluation algorithm and dynamic parameter update mechanism, this solution has established a complete performance feedback adjustment mechanism, which can continuously optimize the system parameters according to the actual communication effect, thereby improving the communication service quality of VOIP products.

[0025] In a specific embodiment, the process of executing step S100 may specifically include the following steps: Collect bandwidth utilization, delay jitter, packet loss rate and signal strength to obtain the original data of network status; Perform deep learning on the original data of network status to obtain the network status feature vector, and perform time series feature classification on the network status feature vector to obtain feature time series data; Perform probability mapping on the feature time series data to obtain the probability distribution of scene categories, where the scene categories include high-speed rail mobile scenes, indoor fixed scenes, outdoor roaming scenes, underground communication scenes, and elevator scenes; The corresponding scene category is determined according to the scene category probability distribution, and parameters are retrieved from the scene parameter library based on the scene category to obtain benchmark communication parameters. The benchmark communication parameters include coding rate, frame length, sampling rate, quantization bit number and transmission protocol type.

[0026] Specifically, a network monitoring system is built to collect bandwidth utilization, delay jitter, packet loss rate and signal strength in real time to obtain the original data of network status. It is defined as the ratio of current communication traffic to total available bandwidth, that is: in, Indicates the currently used bandwidth. Indicates the maximum available bandwidth. Bandwidth utilization directly affects the stability and throughput of data transmission. When it is close to 1, the network will be congested, resulting in an increase in packet loss rate. Reflects the variability of the packet arrival interval and is calculated using the mean square error, namely: in, It is The time between arrival of packets, is the mean inter-arrival time, is the number of packets. If If the value is too large, it will cause audio and video synchronization distortion. Defined as the number of lost packets Total number of packets sent The ratio of, that is: when When the signal strength is too high, voice calls may be interrupted or the voice quality may deteriorate. By measuring the received signal power calculate: in, is the received signal power, is the reference power. Lower signal strength leads to packet transmission failure or requires stronger error correction capability. The network state parameters are analyzed using deep learning methods to extract network state feature vectors. Recurrent neural networks or long short-term memory networks are used to process time series data. These models can effectively capture the temporal characteristics of network states. Assume that the input network state data is represented as , then the hidden state of LSTM is calculated as follows: in, is the hidden state at the current time step, and is the weight matrix, is the bias term, Represents the activation function. By training the model, the network state feature vector is extracted, which can highly summarize the characteristics of the current network and be used for subsequent scene classification. The feature vector is classified into time series features to generate feature time series data. This classification task uses a classification model based on Transformer or 1D CNN to learn the time feature patterns of different network scenarios. Assume that the feature time series data is defined as , then the data is represented as a series of network status features: in, Represents the time window length. The goal of the classification model is to map these time series data to different scenario category probability distributions. Probabilistic mapping is performed on the feature time series data to determine which communication scenario the current network status belongs to. The Bayesian classifier is used to calculate the posterior probability of the scenario category: in, Representing feature data Belong to the scene (such as high-speed rail mobile scenes, indoor fixed scenes, etc.) It's in the scene Observed below The probability density of It's a scene The prior probability of It is feature data By calculating these probabilities, the probability distribution of the scenario category is obtained, and the most likely network environment is determined. After determining the scenario category, the benchmark communication parameters are retrieved from the scenario parameter library to optimize the communication quality of VOIP. The benchmark communication parameters include coding rate, frame length, sampling rate, quantization bit number, and transmission protocol type. For example, in the high-speed rail mobile scenario, due to frequent network switching and high packet loss rate, the coding rate is reduced. , using a shorter frame length To reduce the impact of packet loss, adjust the sampling rate and quantization bit To accommodate network jitter: In indoor fixed scenarios, since the network is stable and the bandwidth is sufficient, a higher encoding rate and a larger frame length are used to improve the audio and video quality: Similarly, in underground communication scenarios, due to weak signals, low bit rates and enhanced forward error correction mechanisms are adopted, while in elevator scenarios, brief signal interruptions require the use of ultra-short frame lengths and high-redundancy coding strategies to avoid severe degradation of audio and video quality.

[0027] In a specific embodiment, the process of executing step S200 may specifically include the following steps: The cumulative probability of bandwidth utilization and delay jitter in the reference communication parameters is calculated to obtain the cumulative distribution function of bandwidth utilization and the cumulative distribution function of delay jitter; Perform joint distribution calculation on the bandwidth utilization cumulative distribution function and the delay jitter cumulative distribution function to obtain a composite cumulative distribution function; The composite cumulative distribution function is input into the Fox H function for feature decomposition to obtain the Fox H function feature sequence, where the Fox H function uses Mellin transformation for feature extraction; The Fox H function characteristic sequence is recursively iterated to construct an n×m-dimensional optimization mapping matrix, where n represents the discrete level number of bandwidth utilization and m represents the discrete level number of delay jitter. The optimization mapping matrix is ​​iteratively solved through the Fox H function optimization algorithm to obtain the optimal communication parameter value; The optimal communication parameter values ​​are parsed to obtain an optimized communication parameter combination, which includes an updated coding rate, frame length, sampling rate, quantization bit number, and transmission protocol type.

[0028] Specifically, a large amount of network status data is collected, and the probability distribution of bandwidth utilization and delay jitter is obtained through statistical analysis. It is defined as the ratio of current communication traffic to total available bandwidth, that is: in, Represents the currently used bandwidth. Represents the maximum available bandwidth, The value range is [0,1]. , which is defined as the variance of the packet arrival interval and is calculated using the mean square error: in, Representative The time between arrival of packets, is the average time interval, is the total number of data packets. The cumulative probability of bandwidth utilization and delay jitter in the benchmark communication parameters is calculated, and its cumulative probability distribution function (CDF) is calculated. Cumulative distribution function and They represent the probability that bandwidth utilization and delay jitter are below a certain value: in, and Represent the random variables of bandwidth utilization and delay jitter respectively. Through statistical analysis of a large amount of historical data, we can fit and Mathematical expression of, for example, if bandwidth utilization follows log-normal distribution , then its cumulative distribution function is: in, is the mean bandwidth utilization, is the standard deviation, erf is the error function. Similarly, if the delay jitter follows the gamma distribution , then its cumulative distribution function is: in, and are the shape and scale parameters of the gamma distribution, is the gamma function. and After that, the two functions are jointly distributed to obtain the composite cumulative distribution function The Copula function is used to capture the correlation between bandwidth utilization and delay jitter, and the joint distribution is expressed as: in, is a Copula function, is the correlation parameter. Gumbel Copula is selected to model strong correlation: in, , Controls the strength of the correlation. , and get the complete composite cumulative distribution function. The composite cumulative distribution function is input into the FoxH function for eigendecomposition to extract its main characteristic patterns. The Fox H function is a generalized special function used to analyze complex probability distributions, which is defined as: in, As parameters, is the input variable. The Fox H function uses Mellin transformation for feature extraction, and its transformation form is: Mellin transform can map probability distribution to feature space in order to extract main patterns and reduce dimension. dimensional optimization mapping matrix, where A discrete number of levels representing bandwidth utilization, Represents the discrete level number of delay jitter. Let the mapping matrix be , whose elements represent the optimal communication parameters under different combinations of bandwidth utilization and delay jitter: in, is an optimization function that maps the eigendecomposition results to a set of communication parameters. The Fox H function optimization algorithm is used to iteratively solve the optimization mapping matrix to find the optimal communication parameters. The optimization goal is to minimize the communication delay. and maximize signal quality : in, is the coding rate, is the frame length, is the sampling rate, is the number of quantized bits. Gradient descent method or particle swarm optimization is used to solve and finally the optimal communication parameter value is obtained. The optimized communication parameters are analyzed to adjust the VOIP encoding strategy.

[0029] In a specific embodiment, the process of executing step S300 may specifically include the following steps: Apply the optimized communication parameter combination to the edge-cloud collaborative encoding and decoding framework, perform parameter analysis on the optimized communication parameter combination, and obtain lightweight encoding and decoding parameters on the terminal side and high-performance encoding and decoding parameters on the cloud side. Based on the lightweight codec parameters on the terminal side, the audio and video data are feature extracted and processed through a three-layer pulse neural network SNN to obtain audio and video feature data, where the number of neurons in each layer of the three-layer pulse neural network SNN is 64, 32, and 16 respectively; Performing dynamic time step encoding processing on audio and video feature data to obtain pulse coded data; Input the pulse coding data and the cloud high-performance coding and decoding parameters into the cloud high-performance coding and decoding module for data processing to obtain initial coding data; Redundant encoding is performed on the initial coded data through a HARQ mechanism to obtain adaptive redundant coded data, wherein the HARQ mechanism includes an adaptive redundant coding generation unit and a feedback control unit; Data optimization is performed based on the adaptive redundant coding data and feedback information of the feedback control unit to obtain coded communication data.

[0030] Specifically, the optimized communication parameter combination is parsed so as to be adapted on the terminal side and the cloud side respectively. The optimized communication parameter combination includes the coding rate , frame length , sampling rate , quantization bit number and transmission protocol type, these parameters are hierarchically parsed according to computing resources, bandwidth conditions and delay requirements. For the terminal side, due to its limited computing resources, lightweight encoding and decoding parameters such as lower coding complexity, shorter frame length and adaptive sampling rate are adopted to reduce computing overhead and delay; while for the cloud side, due to its stronger computing power, more efficient deep encoding and decoding algorithms and high-precision sampling strategies are adopted to optimize audio and video quality. After parsing the communication parameters, the terminal side uses a three-layer spike neural network (SNN) to perform feature extraction processing on the audio and video data to reduce redundant data and extract key information. SNN transmits information through discrete pulse signals and uses time dependence for calculation, so it has great advantages in low-power computing scenarios. SNN consists of a three-layer neural network with 64, 32 and 16 neurons in each layer, respectively, to achieve step-by-step dimensionality reduction and extract key features of audio and video data. In terms of mathematical modeling, SNN uses the Leaky Integrate-and-Fire (LIF) model to simulate the change of neuronal membrane potential over time, and its update formula is: in, Indicates Neurons at time The membrane potential at the moment, is the attenuation factor, is the input signal, is the discharge threshold. Exceed , the neuron will generate a pulse signal and reset the membrane potential. Through this mechanism, SNN can efficiently extract audio and video feature data in a low-bandwidth, low-power environment and convert it into pulse-coded data. Dynamic time step encoding is performed on audio and video feature data to adapt to network changes and optimize transmission efficiency. Dynamic time step encoding adjusts the time step according to the data change rate. When the data changes drastically, a shorter step is used to retain details, and when the data changes slowly, a longer step is used to reduce the amount of data. Its mathematical expression is: in, represents the adaptive time step, is the regulating factor, is the rate of change of audio and video feature data over time. When the step length is larger becomes shorter to preserve more detail, and when When the step length is small, The pulse coded data is then transmitted to the cloud high-performance codec module for data processing to generate initial coded data. The cloud codec module is based on a deep neural network (DNN) or a variational autoencoder (VAE) to enhance the transmitted data, such as noise reduction, dereverberation, and frame compensation, to improve the audio and video quality. The coding and decoding process is expressed as: in, is the input pulse code data, For the encoder, For the decoder, is the neural network parameter. This module optimizes parameters through end-to-end training to provide the best audio and video reconstruction quality under different network environments. After generating the initial coded data, the HARQ (Hybrid Automatic Repeat Request) mechanism is used to perform redundant coding on it to improve the data's ability to resist packet loss. The HARQ mechanism includes an adaptive redundant coding generation unit and a feedback control unit, wherein the redundant coding generation unit is based on a low-density parity check code (LDPC) or a convolutional code to improve the data's recoverability when the channel state is poor. Assuming that the transmitted data Generate coded data after redundant encoding , and its calculation formula is: in, is the generation matrix, which represents the redundant coding strategy. When the receiving end detects packet loss or too high error bits, the feedback control unit will send a retransmission request signal and dynamically adjust the redundancy according to the current channel state information to optimize the transmission efficiency. Based on the adaptive redundant coding data and the feedback information of the feedback control unit, the data is further optimized to finally obtain the encoded communication data. The optimization process is based on the objective function of minimizing the bit error rate and maximizing the throughput: in, represents the bit error rate, represents the throughput, is the redundant coding ratio. Adaptive adjustment is performed using reinforcement learning or Bayesian optimization, so that in a high packet loss environment, the redundant coding ratio is increased, and in a low packet loss environment, the redundancy is reduced to improve bandwidth utilization. The encoded communication data finally generated has the characteristics of high anti-packet loss capability, low latency and strong adaptability, so that VOIP can still provide a smooth communication experience in various complex network environments.

[0031] The HARQ mechanism performs redundant coding on the initial coded data to obtain adaptive redundant coded data. The HARQ mechanism includes an adaptive redundant coding generation unit and a feedback control unit, including: performing data packet segmentation processing on the initial coded data to obtain a coded data packet sequence, wherein the length of each data packet is determined according to the size of the network transmission unit; performing cyclic redundancy check CRC coding processing on each data packet in the coded data packet sequence to obtain a CRC check sequence; inputting the CRC check sequence into the adaptive redundant coding generation unit for convolution coding processing to obtain initial redundant coded data; performing transmission quality detection processing on the initial redundant coded data through the feedback control unit to obtain transmission quality parameters, wherein the transmission quality parameters include packet loss rate, bit error rate and number of retransmissions; dynamically adjusting the redundant coding ratio based on the transmission quality parameters to obtain a target redundant ratio, wherein the redundant ratio is increased when the transmission quality parameter is lower than a preset threshold, and the redundant ratio is reduced when the transmission quality parameter is higher than the preset threshold; re-encoding the initial redundant coded data according to the target redundant ratio to obtain adaptive redundant coded data.

[0032] In a specific embodiment, the process of executing step S400 may specifically include the following steps: Based on the Lagrange duality principle, the encoded communication data is divided into blocks to obtain multiple data blocks, and the access frequency statistics and storage cost calculation of the data blocks are performed to obtain the data block access frequency and storage cost; The priority of the data block is calculated based on the access frequency and storage cost of the data block to obtain the cache priority of the data block. The cache priority calculation formula is P=f / c, where f is the access frequency and c is the storage cost; The distance and signal strength of adjacent communication nodes are calculated in real time to obtain the node distance value and signal strength value, and the node distance value is compared with the preset distance threshold D, and the signal strength value is compared with the preset signal strength threshold S to obtain a comprehensive comparison result; Based on the comprehensive comparison results and cache priority, the task allocation calculation is performed through the dynamic programming algorithm to obtain the task allocation plan, which satisfies the constraint condition of minimizing the total delay of the system; The data blocks are stored and allocated according to the task allocation plan to obtain the storage location and communication node of the data blocks.

[0033] Specifically, the encoded communication data is divided into blocks to improve data access efficiency and optimize the distributed storage strategy. The data block problem is modeled as an optimization problem, whose goal is to minimize the total storage overhead and transmission delay of the system, while ensuring data integrity and optimizing access efficiency. Assuming the size of the original communication data is , the set of available storage nodes is , where each node With storage capacity and access rate , the optimization goal of data partitioning is expressed as: in, Indicates allocation to a node The data block size, Represents the access rate of the node. represents the storage cost, is the regularization coefficient that weighs the storage cost and access latency. In order to solve this optimization problem, the Lagrangian duality principle is used to construct the Lagrangian function: in, is the Lagrange multiplier used to constrain the total size of all data blocks to be equal to the original communication data size By solving the KKT (Karush-Kuhn-Tucker) condition, the optimal data partitioning scheme is obtained, so that the distribution of multiple data blocks after partitioning between different storage nodes reaches the optimal balance. After completing the data partitioning, the access frequency statistics and storage cost calculation of the data blocks are performed to obtain the data block access frequency and storage cost. Access frequency Reflects the number of times a data block is requested and the storage cost Depends on the type of storage media (such as SSD, HDD or distributed cache) and the amount of storage space occupied. The access frequency is calculated using a sliding time window model: in, It's time The frequency of visits, is the number of data accesses at the current moment, is the attenuation coefficient (usually taken as , used to smooth historical access data. Storage cost By data block size and storage device unit cost Decide: Based on the access frequency and storage cost, the cache priority is calculated, that is, the storage importance of the data block in the cache. The calculation formula of the cache priority is as follows: in, is the cache priority, Represents the access frequency of the data block, represents the storage cost. When it is larger, it means that the access frequency of the data block is high and the storage cost is low. It should be cached first to reduce the system I / O overhead and improve access efficiency. After determining the cache priority, the distance and signal strength of adjacent communication nodes are calculated in real time to optimize the data storage location and task allocation strategy. Assuming that the adjacent communication nodes and The physical distance between , received signal strength It is expressed as: in, is the baseline signal strength, is the path loss exponent (usually between 2 and 4), For Node The communication distance of the node is compared with the preset distance threshold. And the signal strength value and the preset signal strength threshold , get the comprehensive comparison results, and select the appropriate storage nodes. On this basis, the dynamic programming algorithm is used to perform task allocation calculations to solve the optimal data block storage location and data transmission path. The goal is to minimize the total system delay, that is: in, Represents a data block Store to Node The probability of Representative Node Processing data blocks The delay, Represents the storage cost of the node, is the regularization coefficient of storage cost. The dynamic programming problem is solved recursively using the Bellman equation to obtain the global optimal task allocation scheme. According to the task allocation scheme, the data blocks are allocated for storage, and the storage location and communication nodes of the data blocks are determined, thereby optimizing the data access efficiency of the VOIP system and improving the stability and reliability of network transmission.

[0034] The encoded communication data is processed into blocks based on the Lagrangian duality principle to obtain a plurality of data blocks, including: calculating the amount of data on the encoded communication data to obtain a ratio N of the total amount of data D and a preset block size S; rounding up the ratio N to obtain an initial number of blocks M, and constructing a Lagrangian dual function L(x,λ) based on the initial number of blocks M; taking a derivative of the Lagrangian dual function L(x,λ) to obtain an optimal block boundary sequence {b1,b2,...,bM-1}, where bi represents the end position of the i-th block; segmenting the encoded communication data based on the optimal block boundary sequence to obtain M data blocks; performing size balancing on the M data blocks to obtain a plurality of data blocks with a size close to 64KB; and performing boundary adjustment on the data blocks to obtain a plurality of data blocks with a size of precisely 64KB.

[0035] In a specific embodiment, the process of executing step S500 may specifically include the following steps: The computing power and network connection quality of the communication nodes are evaluated to obtain node evaluation parameters, and the communication nodes are divided into master nodes and slave nodes based on the node evaluation parameters; Collect status information of the master node and slave nodes to obtain node status data, which includes computing load, cache capacity and link quality; Based on the node status data, a connection topology graph G between nodes is constructed. The connection topology graph G is subjected to multi-objective optimization calculation by the P-SCA algorithm to obtain an optimization parameter group. The multi-objective optimization includes minimizing system delay, maximizing network throughput, and balancing node load. Perform path planning for the optimized parameter group to obtain the initial transmission path. The master node is responsible for global communication control and data forwarding, while the slave nodes cooperate to perform data processing and relay transmission. The validity of the initial transmission path is verified to obtain a path verification result. When a node failure or link interruption is detected, path reconstruction is triggered, and the initial transmission path is optimized based on the path verification result to obtain a data transmission path.

[0036] Specifically, node evaluation parameters used to measure communication node performance are defined, including key indicators such as computing power, cache capacity, and network connection quality. Calculated by the node's CPU computing resources (such as main frequency in GHz) and parallel processing capabilities (such as the number of cores): in, is the CPU frequency, is the number of cores. Cache capacity Represents the available storage resources of the node, defined as: in, is the total storage space, is the used storage space. The quality of the network connection is measured by bandwidth , Delay and packet loss rate Calculate the bandwidth Defined as: in, is the amount of data transmitted per unit time, is the transmission time. Represents the time difference between sending and receiving a data packet: in, and The timestamps of receiving and sending respectively. Packet loss rate Defined as the ratio of lost packets to total transmitted packets: in, is the number of packets lost, is the total number of data packets. Based on these evaluation parameters, thresholds are set to divide the master node and slave node. The nodes with strong computing power, sufficient cache capacity, and stable network connection are selected as master nodes, while the remaining nodes participate in auxiliary computing and data forwarding as slave nodes. After completing the division of master nodes and slave nodes, their status information is collected to form node status data, including computing load, cache capacity, and link quality. Computational load Reflects the real-time processing pressure of the node, expressed by CPU usage: in, The currently used computing resources. Cache capacity data The link quality is obtained through the storage management module, and the link quality is obtained through the bandwidth , Delay and packet loss rate After collecting the status data of all nodes, a connection topology diagram is constructed based on this data. , the graph consists of communication nodes and communication links composition: in, Represents all nodes, Represents the connection between nodes. Edge weights of topological graphs Determined by the quality of the communication link, calculated by latency and bandwidth: in, For Node arrive The transmission delay of is the available bandwidth, and is the weight coefficient. After constructing the topology graph, the P-SCA algorithm is used to perform multi-objective optimization calculations to obtain the optimization parameter group. This optimization involves three goals: minimizing system delay, maximizing network throughput, and balancing node load. The optimization model is expressed as: P-SCA uses Pareto frontier search to find the optimal solution set under all optimization objectives through iteration. , its core optimization formula is: in, is the target weight parameter, which is dynamically tuned based on genetic algorithm or particle swarm optimization to find the optimal path and load balancing solution. After obtaining the optimized parameter group, the communication path is planned to generate the initial transmission path. Responsible for global communication control and data forwarding, from node Responsible for data processing and relay transmission. Assume is the optimized path set, then the path optimization equation is expressed as: The path planning process uses the Dijkstra algorithm or the A* search algorithm to find the path combination with the minimum delay. The validity of the initial transmission path is verified, that is, whether the path is available, whether there are failed nodes or link interruptions. Path verification is achieved through network detection and heartbeat signals, which are defined as follows: When detected When If the path fails, path reconstruction is required. Path reconstruction is dynamically adjusted through greedy algorithms or deep reinforcement learning, with the goal of finding a new optimal path. : During the path optimization process, historical data is used for prediction, such as using LSTM (Long Short-Term Memory Network) or GCN (Graph Neural Network) for time series prediction, to perceive network failures in advance and adjust the path in advance.

[0037] Among them, a connection topology graph G between nodes is constructed based on the node status data, and a multi-objective optimization calculation is performed on the connection topology graph G through the P-SCA algorithm to obtain an optimization parameter group, wherein the multi-objective optimization includes minimizing system delay, maximizing network throughput and node load balancing, including: constructing an adjacency matrix M based on the link quality data in the node status data, and weighting the adjacency matrix M to obtain a node connection topology graph G; calculating and processing the node degree and link weight in the connection topology graph G to obtain an initial optimization objective function, wherein the initial optimization objective function includes a system delay function D, a network throughput function T and a load balancing function L; inputting the initial optimization objective function into the P-SCA algorithm for decomposition processing to obtain a convex subproblem sequence, wherein each convex subproblem corresponds to a local optimal solution; iteratively solving the convex subproblem sequence to obtain multiple local optimal solutions, and weighting the local optimal solutions to obtain a global optimal solution; based on the global optimal solution, parameter optimization processing is performed on system delay, network throughput and load balancing to obtain an optimization coefficient matrix; the optimization coefficient matrix is ​​analyzed to obtain an optimization parameter group, and the optimization parameter group includes the transmission power, channel allocation and routing strategy of each node.

[0038] In a specific embodiment, the multi-scenario adaptive communication method for executing a VOIP product further includes the following steps: Calculate the QoS score based on end-to-end delay, packet loss rate, jitter amplitude and audio clarity to obtain the call quality score; The user feedback data is processed through the fuzzy comprehensive evaluation algorithm to obtain the user experience score; Monitor CPU usage, memory usage, and bandwidth usage to obtain system resource utilization; Perform weighted fusion on call quality score, user experience score and system resource utilization to obtain performance evaluation results; The performance evaluation result is compared with the preset performance threshold to obtain a parameter adjustment signal, and the benchmark communication parameters in the scenario parameter library are updated according to the parameter adjustment signal to complete the feedback of the performance evaluation result.

[0039] Specifically, key network parameters are monitored and modeled. End-to-end delay ( ) represents the time interval from when the data packet leaves the sender to when it reaches the receiver, and is calculated as follows: in, is the receiving timestamp of the data packet, is the sending timestamp of the data packet. Too large will cause echo and asynchrony in VOIP voice communication. Packet loss rate ( ) reflects the proportion of data packets that fail to reach the destination successfully during transmission: in, Represents the number of packets lost, Represents the total number of transmitted packets. An increase in packet loss rate can cause discontinuous voice or even call interruption. Jitter amplitude ( ) measures the fluctuation of packet arrival intervals: in, It is The arrival interval of packets, is the average inter-arrival interval, The number of packets. Too much jitter can cause intermittent voice, which is smoothed by Jitter Buffer. Audio clarity ( ) is determined by factors such as the signal-to-noise ratio (SNR), coding method, and forward error correction mechanism, and is calculated using the speech quality evaluation algorithm PESQ: in, is the voice signal data. Combining the above four parameters, the QoS score is calculated ( ): in, is the weight, which represents the influence of each parameter on the QoS score. The user feedback data is processed by the fuzzy comprehensive evaluation algorithm to obtain the user experience score. ) combines user subjective ratings (such as user ratings, satisfaction surveys) and objective data (such as call duration, number of call interruptions). Assume that user feedback data includes evaluation index, define the fuzzy membership matrix : in, Representative The evaluation index is The membership of the level score. Combined with the weight vector , calculate the user experience score: This method can fuzzy process user feedback data, making the final score more consistent with the actual experience. At the same time, in order to measure the health of system resources, monitor CPU occupancy, memory usage and bandwidth occupancy to obtain system resource utilization. CPU occupancy ( ) is defined as: in, Represents the current CPU usage, Represents the total available CPU resources. Memory usage ( ) is calculated in a similar way: in, Represents the used memory, Represents total memory. Bandwidth usage ( ) represents the ratio of currently used bandwidth to total available bandwidth: in, is the current flow, is the maximum available bandwidth. Calculate system resource utilization ( ): in, is the weight factor of different resources. The call quality score, user experience score and system resource utilization are weighted and integrated to obtain the performance evaluation result. Define the performance evaluation score ( ): in, It is a weight parameter used to balance QoS, user experience and system resource utilization. By adjusting these parameters, the VOIP system is more inclined to optimize user experience, reduce resource consumption or improve call quality. In order to enable the system to adaptively adjust communication parameters, the performance evaluation results are compared with the preset performance thresholds to obtain parameter adjustment signals, and the benchmark communication parameters in the scenario parameter library are updated based on the signals. Set performance thresholds As the minimum standard for call quality, Lower than , trigger parameter optimization: in, The communication parameters need to be adjusted. Indicates that the current communication status is normal. If the parameter adjustment signal is triggered, the baseline communication parameters need to be updated in the scene parameter library. For example, if the end-to-end delay is too high, the encoding complexity is reduced and the encoding rate is adjusted. : If the packet loss rate is too high, add forward error correction (FEC): If system resource usage is too high, reduce the frame rate or sampling rate: See also Figure 2 , Figure 2 A schematic block diagram of the structure of a multi-scenario adaptive communication system 200 for a VOIP product provided in an embodiment of the present application, such as Figure 2 As shown, the multi-scenario adaptive communication system 200 of VOIP products includes: The acquisition module 210 is used to collect and extract features of network status parameters, obtain scene categories, and retrieve reference communication parameters corresponding to the scene categories from a scene parameter library; A calculation module 220, configured to calculate a composite cumulative distribution function based on the reference communication parameters to obtain an optimized communication parameter combination; The codec module 230 is used to apply the optimized communication parameter combination to the edge-cloud collaborative codec framework, and perform codec processing on the audio and video data to obtain encoded communication data; The block module 240 is used to process the encoded communication data in blocks and determine the storage location and communication node of the data block; The multi-objective optimization module 250 is used to divide the communication nodes into master nodes and slave nodes, and perform multi-objective optimization on the master nodes and the slave nodes to obtain a data transmission path.

[0040] Through the collaboration of the above components, by constructing the adaptive adjustment mechanism of communication parameters of the composite distribution function and Fox H function optimization algorithm, this solution can accurately evaluate the network status, realize dynamic optimization of communication parameters, enable VOIP products to have scene adaptation capabilities, and ensure call quality in different scenarios. Based on the edge-cloud collaborative coding and decoding framework and the feature extraction method of the pulse neural network SNN, combined with the adaptive redundant coding of the HARQ mechanism, this solution realizes the reasonable allocation of computing resources and the reliability of data transmission, and solves the problem of low processing efficiency in traditional methods. Using the distributed caching strategy based on the Lagrange duality principle and the task allocation mechanism based on dynamic programming, this solution realizes efficient storage and rapid recovery of data, effectively reducing system latency. Through the direct communication technology between devices and the multi-objective optimization of the P-SCA algorithm, this solution establishes an efficient multi-node collaboration framework, which can respond promptly when the node fails or the link is interrupted, and improves the fault tolerance of the system. Based on the multi-dimensional performance evaluation system, including the comprehensive evaluation of QoS score, user experience score and system resource utilization, this solution realizes the precise adjustment of communication parameters and ensures the stable operation of VOIP products in various scenarios. Through the fuzzy comprehensive evaluation algorithm and dynamic parameter update mechanism, this solution has established a complete performance feedback adjustment mechanism, which can continuously optimize the system parameters according to the actual communication effect, thereby improving the communication service quality of VOIP products.

[0041] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0042] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0043] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A multi-scenario adaptive communication method for VOIP products, characterized in that: include: Collecting and extracting features of network status parameters to obtain scene categories, and retrieving reference communication parameters corresponding to the scene categories from a scene parameter library; Calculate a composite cumulative distribution function based on the reference communication parameters to obtain an optimized communication parameter combination; Applying the optimized communication parameter combination to the edge-cloud collaborative coding and decoding framework, and performing coding and decoding processing on the audio and video data to obtain coded communication data; Processing the encoded communication data in blocks, and determining storage locations and communication nodes of the data blocks; The communication nodes are divided into master nodes and slave nodes, and multi-objective optimization is performed on the master nodes and slave nodes to obtain a data transmission path.

2. The multi-scenario adaptive communication method of VOIP products according to claim 1, characterized in that: The collecting and feature extracting of network status parameters to obtain a scene category, and retrieving a reference communication parameter corresponding to the scene category from a scene parameter library, includes: Collect bandwidth utilization, delay jitter, packet loss rate and signal strength to obtain the original data of network status; Performing deep learning on the original network status data to obtain a network status feature vector, and performing time series feature classification on the network status feature vector to obtain feature time series data; Probability mapping is performed on the feature time series data to obtain a scene category probability distribution, wherein the scene categories include high-speed rail mobile scenes, indoor fixed scenes, outdoor roaming scenes, underground communication scenes, and elevator scenes; The corresponding scene category is determined according to the scene category probability distribution, and parameters are retrieved from the scene parameter library based on the scene category to obtain benchmark communication parameters, which include coding rate, frame length, sampling rate, number of quantization bits and transmission protocol type.

3. The multi-scenario adaptive communication method of VOIP products according to claim 1, characterized in that: The step of calculating a composite cumulative distribution function based on the reference communication parameters to obtain an optimized communication parameter combination includes: Performing cumulative probability calculation on the bandwidth utilization and delay jitter in the reference communication parameters to obtain a bandwidth utilization cumulative distribution function and a delay jitter cumulative distribution function; Performing joint distribution calculation on the bandwidth utilization cumulative distribution function and the delay jitter cumulative distribution function to obtain a composite cumulative distribution function; Inputting the composite cumulative distribution function into a Fox H function for feature decomposition to obtain a Fox H function feature sequence, wherein the Fox H function uses Mellin transformation for feature extraction; By recursively iterating the Fox H function characteristic sequence, an optimization mapping matrix of n×m dimensions is constructed, wherein n represents the number of discrete levels of bandwidth utilization, and m represents the number of discrete levels of delay jitter; The optimization mapping matrix is ​​iteratively solved by the Fox H function optimization algorithm to obtain the optimal communication parameter value; The optimal communication parameter value is parsed to obtain an optimized communication parameter combination, wherein the optimized communication parameter combination includes an updated coding rate, frame length, sampling rate, quantization bit number and transmission protocol type.

4. The multi-scenario adaptive communication method of VOIP products according to claim 1, characterized in that: The step of applying the optimized communication parameter combination to the edge-cloud collaborative coding and decoding framework and performing coding and decoding processing on the audio and video data to obtain coded communication data includes: Applying the optimized communication parameter combination to the edge-cloud collaborative coding and decoding framework, performing parameter parsing on the optimized communication parameter combination to obtain lightweight coding and decoding parameters on the terminal side and high-performance coding and decoding parameters on the cloud side; Based on the lightweight codec parameters on the terminal side, feature extraction processing is performed on the audio and video data through a three-layer pulse neural network SNN to obtain audio and video feature data, wherein the number of neurons in each layer of the three-layer pulse neural network SNN is 64, 32, and 16 respectively; Performing dynamic time step encoding processing on the audio and video feature data to obtain pulse coded data; Inputting the pulse coding data and the cloud high-performance coding parameters into the cloud high-performance coding module for data processing to obtain initial coding data; Performing redundant coding on the initial coded data through a HARQ mechanism to obtain adaptive redundant coded data, wherein the HARQ mechanism includes an adaptive redundant coding generation unit and a feedback control unit; Data optimization is performed based on the adaptive redundant coding data and feedback information of the feedback control unit to obtain coded communication data.

5. The multi-scenario adaptive communication method of VOIP products according to claim 1, characterized in that: The step of processing the encoded communication data into blocks and determining the storage location and communication node of the data blocks includes: Based on the Lagrange duality principle, the encoded communication data is processed into blocks to obtain multiple data blocks, and access frequency statistics and storage cost calculations are performed on the data blocks to obtain data block access frequency and storage cost; A priority calculation is performed based on the data block access frequency and the storage cost to obtain a cache priority of the data block, and the cache priority calculation formula is P=f / c, where f is the access frequency and c is the storage cost; Calculate the distance and signal strength of adjacent communication nodes in real time to obtain a node distance value and a signal strength value, and compare the node distance value with a preset distance threshold D, and the signal strength value with a preset signal strength threshold S to obtain a comprehensive comparison result; Based on the comprehensive comparison result and the cache priority, a task allocation calculation is performed through a dynamic programming algorithm to obtain a task allocation scheme, wherein the task allocation scheme satisfies the constraint condition of minimizing the total delay of the system; The data blocks are stored and allocated according to the task allocation scheme to obtain storage locations and communication nodes of the data blocks.

6. The multi-scenario adaptive communication method of VOIP products according to claim 1, characterized in that: The step of dividing the communication nodes into master nodes and slave nodes, and performing multi-objective optimization on the master nodes and slave nodes to obtain a data transmission path includes: Evaluate the computing power and network connection quality of the communication nodes to obtain node evaluation parameters, and divide the communication nodes into master nodes and slave nodes based on the node evaluation parameters; Collecting status information of the master node and the slave node to obtain node status data, wherein the node status data includes computing load, cache capacity and link quality; Based on the node status data, a connection topology graph G between nodes is constructed, and a multi-objective optimization calculation is performed on the connection topology graph G through a P-SCA algorithm to obtain an optimization parameter group, wherein the multi-objective optimization includes minimizing system delay, maximizing network throughput, and balancing node load; Performing path planning on the optimization parameter group to obtain an initial transmission path, wherein the master node is responsible for global communication control and data forwarding, and the slave nodes cooperate to perform data processing and relay transmission; The validity of the initial transmission path is verified to obtain a path verification result, and path reconstruction is triggered when a node failure or a link interruption is detected. The initial transmission path is optimized based on the path verification result to obtain a data transmission path.

7. The multi-scenario adaptive communication method of VOIP products according to claim 1, characterized in that: The multi-scenario adaptive communication method of the VOIP product also includes: Calculate the QoS score based on end-to-end delay, packet loss rate, jitter amplitude and audio clarity to obtain the call quality score; The user feedback data is processed through the fuzzy comprehensive evaluation algorithm to obtain the user experience score; Monitor CPU usage, memory usage, and bandwidth usage to obtain system resource utilization; Performing weighted fusion on the call quality score, the user experience score and the system resource utilization to obtain a performance evaluation result; The performance evaluation result is compared with a preset performance threshold to obtain a parameter adjustment signal, and the reference communication parameters in the scenario parameter library are updated according to the parameter adjustment signal to complete the feedback of the performance evaluation result.

8. A multi-scenario adaptive communication system for VOIP products, characterized in that: A multi-scenario adaptive communication method for a VOIP product according to any one of claims 1 to 7, comprising: A collection module, used to collect and extract features of network status parameters, obtain scene categories, and retrieve reference communication parameters corresponding to the scene categories from a scene parameter library; A calculation module, used for calculating a composite cumulative distribution function based on the reference communication parameters to obtain an optimized communication parameter combination; A coding and decoding module, used to apply the optimized communication parameter combination to the edge-cloud collaborative coding and decoding framework, and perform coding and decoding processing on the audio and video data to obtain coded communication data; A block division module, used for performing block processing on the encoded communication data and determining the storage location and communication node of the data block; The multi-objective optimization module is used to divide the communication nodes into master nodes and slave nodes, and perform multi-objective optimization on the master nodes and slave nodes to obtain a data transmission path.

Images