Method, device and equipment for adaptive fax transmission between PBX and eFax platforms

Abstract

The invention provides a method, device and equipment for self-adaptive fax transmission between a PBX and an eFax platform, the method is applied to the technical field of information, and the method comprises the following steps: collecting and fusing PBX side features, eFax platform side features, real-time network state features and historical fax performance features, and generating context feature vectors; inputting the context feature vector into an adaptive strategy decision model for training, outputting a reference fax strategy set, performing weighted optimization on the reference fax strategy set, and generating a finally executed target fax strategy; and after executing the target fax strategy, reversely updating the strategy income estimation of the adaptive strategy decision model based on the error between the actual transmission result index and the transmission result predicted by the adaptive strategy decision model. According to the method, the adaptive strategy decision module can be adaptively optimized according to transmission data between the PBX and the eFax platform, so that self-learning and continuous optimization of the whole system are realized.

Description

Technical Field

[0001] This application relates to the field of information technology, and more specifically, to a method, apparatus, and device for adaptive fax transmission between PBX and eFax platforms. Background Technology

[0002] Currently, Private Branch Exchanges (PBXs) are widely used in corporate environments. Within an enterprise, PBXs and eFax services interoperate using a common set of protocols. As enterprise communications shift to a unified system on the internet or local area network (IP network), all multimedia data is transmitted over the same network. For PBXs and eFax platforms to work together, they need to use the same protocols. However, different eFax platforms have different standard protocols. For example, different eFax platforms differ in their Session Initiation Protocol (SIP), Session Description Protocol (SDP), and network stability, which can easily lead to high fax negotiation failure rates, high implementation costs, high operational complexity, and difficulties in compatibility and maintenance.

[0003] In related technologies, static configuration files or manual adjustments are used to solve the communication problem between PBX and eFax platforms. In other words, the communication between PBX and eFax platforms lacks a dynamic, intelligent, and self-learning mechanism, which makes it impossible to continuously improve the fax transmission success rate. Summary of the Invention

[0004] This application provides a method for adaptive fax transmission between PBX and eFax platforms. This method can adaptively optimize the adaptive strategy decision module based on the transmission data between PBX and eFax platforms, thereby achieving self-learning and continuous optimization of the system as a whole.

[0005] Firstly, a method for adaptive fax transmission between a PBX and an eFax platform is provided, the method comprising: In response to fax session requests, the system collects and integrates device characteristics from the user-level switch (PBX) side, compatibility characteristics from the eFax platform side, real-time network status characteristics, and historical fax performance characteristics to generate a structured context feature vector. The context feature vector is input into the adaptive policy decision model, and the adaptive policy decision is used to map an initial fax policy set, which includes at least one of protocol selection policy, routing policy, retransmission policy and encoding parameter policy. A reference fax policy set is determined from the initial fax policy based on the reward function, which is used to measure the performance of the context feature vector after executing the reference fax policy set. The reference fax strategy set is input into a protocol converter and a network predictor in parallel; wherein, the protocol converter is used to predict the protocol success probability of the reference fax strategy set under the target eFax platform based on a protocol compatibility verification model; the network predictor is used to predict the network reliability and network latency of the transmission link corresponding to the execution of the reference fax strategy set based on time-series network prediction. Based on the success probability of the protocol, the reliability of the network, and the latency of the network, the reference fax policy set is weighted and optimized to generate the final target fax policy to be executed. After executing the final target fax strategy, the actual transmission result indicators are collected, and the actual reward function is calculated based on the transmission result indicators. The error between the actual reward function and the expected reward function predicted by the adaptive strategy decision model is calculated. Based on the error, the strategy reward estimate of the adaptive strategy decision model, the model parameters of the network predictor, and the feature weights in the feature fusion process are updated synchronously, thereby realizing the overall self-learning and continuous optimization of the system.

[0006] The above approach involves collecting and fusing features from user-level PBX devices, eFax platform compatibility features, real-time network status features, and historical fax performance features to generate a structured context feature vector. This vector is then used to train an adaptive policy decision model, outputting a reference fax policy set. This set represents the policy set with the highest compatibility and transmission success rate. The reference fax policy set is then weighted and optimized based on protocol success probability, network reliability, and network latency to generate the final target fax policy. After executing the target fax policy, actual transmission result metrics are collected, and the actual reward function is calculated based on these metrics. The error between the actual reward function and the expected reward function predicted by the adaptive policy decision model is calculated. Based on this error, the policy reward estimate of the adaptive policy decision model, the model parameters of the network predictor, and the feature weights in the feature fusion process are updated synchronously. In other words, the fax sending and receiving strategy is automatically determined by the adaptive strategy decision model, and the fax strategy is optimized based on the protocol success probability, network reliability and network latency to obtain the optimal target fax strategy. Finally, the adaptive strategy decision model is optimized in reverse based on the predicted result value and the actual transmission result index. This not only improves the reliability and transmission efficiency of fax communication, but also enables adaptive fax control.

[0007] In conjunction with the first aspect, in some possible implementations, the collection and fusion of user-level switch (PBX) side device characteristics, eFax service platform side compatibility characteristics, real-time network status characteristics, and historical fax performance characteristics generates a structured context feature vector, including: Collect the PBX's device identifier, firmware version, protocol characteristics, and network parameters; the session description protocols and compatibility modes supported by the eFax platform; the round-trip latency, packet loss rate, jitter, and bandwidth of the real-time network status characteristics; and the historical success rate and error types of the historical fax performance characteristics. Supplement the missing feature data in the PBX-side device characteristics, the eFax platform-side compatibility characteristics, the real-time network status characteristics, and the historical fax performance characteristics; The supplemented PBX-side device characteristics, eFax platform-side compatibility characteristics, real-time network status characteristics, and historical fax performance characteristics are standardized. The standardized features of the PBX-side equipment, the compatibility features of the eFax platform, the real-time network status features, and the historical fax performance features are encoded using one-hot encoding or embedded encoding. Based on the first preset weight, the encoded PBX-side device features, eFax platform-side compatibility features, real-time network status features, and historical fax performance features are weighted and fused to obtain a structured context feature vector; wherein, the first preset weight is a pre-configured feature dimension weight coefficient.

[0008] In conjunction with the first aspect, in some possible implementations, before the weighted fusion of the encoded PBX-side device features, the eFax platform-side compatibility features, the real-time network status features, and the historical fax performance features based on a first preset weight to obtain a structured context feature vector, the method further includes: Collect the call signaling, session description protocol parameters, and session direction of the PBX; Based on the call signaling, the session description protocol parameters, the session direction, and the device identifier of the PBX, the corresponding target eFax platform is matched.

[0009] In conjunction with the first aspect, in some possible implementations, the protocol compatibility verification model predicts the protocol success probability of the reference fax strategy set under the target eFax platform, including: Based on the device identifier of the PBX, the compatibility mode of the eFax platform, and the target protocol mapping table matching the transmission channel characteristics between the PBX and the eFax platform; the target protocol mapping table is used to verify the compatibility of the reference fax policy set; Based on the preset classifier in the protocol compatibility verification model, predict the success probability of the target protocol mapping table under the target eFax platform. If the success probability of the protocol is greater than a first preset threshold, the success probability of the protocol is determined and output.

[0010] In conjunction with the first aspect, in some possible implementations, the timing-based prediction network predicts the network reliability and network latency of the transmission link corresponding to the execution of the reference fax policy set, including: Obtain the context feature vector state and historical network observation sequences; The context feature vector state and the historical network observation sequence are input into the time series prediction network to predict the transmission success rate and average round-trip delay changes of the transmission link.

[0011] In conjunction with the first aspect, in some possible implementations, the network reliability includes transmission success rate, the network latency includes average round-trip time variation, and the weighted optimization of the reference fax policy set based on the protocol success probability, the network reliability, and the network latency to generate the final target fax policy includes: An objective function is constructed based on the success probability of the protocol, the reliability of the network, and the network latency; the formula for calculating the objective function is as follows: ; This is the reference fax strategy set; This represents the probability of the protocol succeeding. This is the transmission success rate; This represents the change in average round-trip time; , as well as The second preset weight; The objective function is optimized to generate the target fax strategy.

[0012] In conjunction with the first aspect, in some possible implementations, the method further includes: If the maximum value of the optimized objective function is less than the second preset threshold, then the objective function is resampled, recalculated, and optimized. If the maximum value of the objective function after multiple resampling, calculation and optimization is less than the second preset threshold, the set of strategies with the maximum value of the objective function after multiple resampling, calculation and optimization will be used as the target fax strategy.

[0013] In conjunction with the first aspect, in some possible implementations, the synchronous updating of the policy return estimate of the adaptive policy decision model, the model parameters of the network predictor, and the feature weights in the feature fusion process based on the error includes: Based on this error, the expected return estimate after executing the initial set of fax strategies in the adaptive strategy decision model is corrected; The error is used as the gradient of the loss function to update the third preset weight of the temporal prediction network in the network predictor; Based on this error, the first preset weight in the context feature vector fusion process is updated in reverse. The optimized adaptive policy decision model, the third preset weight of the network predictor, and the first preset weight in the feature fusion process are synchronously updated to the PBX.

[0014] Secondly, an apparatus for adaptive fax transmission between a PBX and an eFax platform is provided, the apparatus comprising: The data acquisition and fusion module is used to respond to fax session requests by acquiring and fusing device features on the user-level switch PBX side, compatibility features on the fax service eFax platform side, real-time network status features, and historical fax performance features to generate a structured context feature vector. The mapping and determination module is used to input the context feature vector into the adaptive policy decision model, and to map an initial fax policy set from the adaptive policy decision. The initial fax policy set includes at least one of protocol selection policy, routing selection policy, retransmission policy and encoding parameter policy. A reference fax policy set is determined from the initial fax policy based on a reward function, which is used to measure the performance of the context feature vector after executing the reference fax policy set. The input and prediction module is used to input the reference fax strategy set into the protocol converter and the network predictor in parallel; wherein, the protocol converter is used to predict the protocol success probability of the reference fax strategy set under the target eFax platform based on the protocol compatibility verification model; the network predictor is used to predict the network reliability and network latency of the transmission link corresponding to the execution of the reference fax strategy set based on time-series network prediction. The generation module is used to perform weighted optimization on the reference fax policy set based on the success probability of the protocol, the reliability of the network, and the network latency, and generate the final target fax policy to be executed. The calculation and update module is used to collect actual transmission result indicators after executing the final target fax strategy, and calculate the actual reward function based on the transmission result indicators; calculate the error between the actual reward function and the expected reward function predicted by the adaptive strategy decision model; based on the error, synchronously update the strategy return estimate of the adaptive strategy decision model, the model parameters of the network predictor, and the feature weights in the feature fusion process, thereby realizing the overall self-learning and continuous optimization of the system.

[0015] Thirdly, an apparatus is provided, including a memory and a processor. The memory is used to store executable program code, and the processor is used to call and run the executable program code from the memory, causing the apparatus to perform the methods described above for adaptive fax transmission between PBX and eFax platforms.

[0016] Fourthly, a computer program product is provided, comprising: computer program code, which, when run on a computer, causes the computer to perform the method described above for adaptive fax transmission between PBX and eFax platforms.

[0017] Fifthly, a computer-readable storage medium is provided that stores computer program code, which, when executed on a computer, causes the computer to perform the method described above for adaptive fax transmission between PBX and eFax platforms. Attached Figure Description

[0018] Figure 1This is a schematic diagram of the implementation environment of a method for adaptive fax transmission between a PBX and an eFax platform provided in an embodiment of this application; Figure 2 This is a schematic flowchart of a method for adaptive fax transmission between a PBX and an eFax platform provided in an embodiment of this application; Figure 3 This is a schematic flowchart of another method for adaptive fax transmission between a PBX and an eFax platform provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an apparatus for adaptive fax transmission between a PBX and an eFax platform, provided in an embodiment of this application. Figure 5 This is a schematic diagram of the structure of a device provided in an embodiment of this application. Detailed Implementation

[0019] The technical solutions in this application will be clearly and thoroughly described below with reference to the accompanying drawings. In the description of the embodiments of this application, unless otherwise stated, " / " means "or," for example, A / B can mean A or B. "And / or" in the text is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Furthermore, in the description of the embodiments of this application, "multiple" refers to two or more than two.

[0020] Hereinafter, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature.

[0021] Before introducing the methods of the embodiments of this application, the technical terms that may be involved in the embodiments of this application will be explained first.

[0022] A Private Branch Exchange (PBX) is a dedicated telephone network system for an enterprise. It manages calls by sharing external lines between extensions and is also known as a PBX or PBX. Its core functions include handling internal extension communications and connecting to the Public Switched Telephone Network (PSTN), integrating telephone, fax, and other services, and automatically dialing external lines.

[0023] eFax is an internet-based paperless online office tool that allows users to send and receive faxes simply by logging in via a webpage, eliminating the need for traditional fax machines. It connects to a server that converts electronic documents into fax signals for transmission to the target fax machine, offering both environmental friendliness and cross-regional transmission capabilities.

[0024] The Session Description Protocol (SDP) is an application-layer control protocol used to describe multimedia sessions.

[0025] The Session Initiation Protocol (SIP) is a multimedia communication protocol developed by the IETF (Internet Engineering Task Force). It is a text-based application-layer control protocol used to create, modify, and release sessions of one or more participants. SIP is an IP voice session control protocol originating from the Internet, characterized by its flexibility, ease of implementation, and scalability.

[0026] In the following description of the embodiments of this application, it is used as... Figure 1 Taking an example, the implementation environment of the embodiments of this application will be introduced.

[0027] For example, such as Figure 1 As shown, the implementation environment includes server 110 and PBX 120.

[0028] The eFax platform 110 is an internet-based paperless online office tool. The eFax platform 110 includes a network communication unit, a processor, and memory, among other things. In some embodiments, the eFax platform 110 is used to send faxes to or receive call faxes from a PBX system.

[0029] PBX120 is a dedicated telephone network system for internal enterprise use. In some embodiments, when a fax call is received, the PBX invites the eFax platform of server 110 to participate in the session.

[0030] Figure 2 This is a schematic flowchart illustrating a method for adaptive fax transmission between a PBX and a third-party eFax platform, provided in an embodiment of this application.

[0031] For example, such as Figure 2 As shown, taking a PBX as the execution subject as an example, this application describes a method for adaptive fax transmission between a PBX and a third-party eFax platform. The method 200 includes the following steps.

[0032] Step 201: In response to a fax session request, collect and fuse PBX-side device features, eFax platform-side compatibility features, real-time network status features, and historical fax performance features to generate a structured context feature vector.

[0033] A fax session request is a communication process of "handshake" and "negotiation" between two fax machines before they officially begin transmitting documents. The fax session request is used to check the reliability of the communication link. A PBX is a dedicated internal telephone network system for enterprises. PBX-side device characteristics include the PBX's device identifier, firmware version, network parameters, and T.38 protocol characteristics for IP networks. The eFax platform is an internet-based paperless online office tool. eFax platform-side compatibility characteristics include the session description protocols and compatibility modes supported by the eFax platform. Real-time network status characteristics include data such as packet loss rate, jitter, round-trip time, and bandwidth. Historical fax performance characteristics include the historical success rate and error types of faxes between the PBX and the eFax platform. A context feature vector refers to converting raw data into a numerical vector with a fixed length, explicit order, and clear meaning.

[0034] It should be understood that since machine learning algorithms expect inputs to be numerical vectors in a uniform dimension and continuous space, the original data needs to be transformed and fused to obtain contextual feature vectors.

[0035] In one possible implementation, the device identifier, firmware version, protocol characteristics, and network parameters of the PBX are collected; the session description protocols and compatibility modes supported by the eFax platform are collected; the round-trip latency, packet loss rate, jitter, and bandwidth of the real-time network status characteristics are collected; and the historical success rate and error type of the historical fax performance characteristics are collected. Missing feature data in the PBX-side device characteristics, the eFax platform-side compatibility characteristics, the real-time network status characteristics, and the historical fax performance characteristics are supplemented. The supplemented PBX-side device characteristics, eFax platform-side compatibility characteristics, real-time network status characteristics, and historical fax performance characteristics are standardized. The standardized PBX-side device characteristics, eFax platform-side compatibility characteristics, real-time network status characteristics, and historical fax performance characteristics are encoded based on one-hot encoding or embedded encoding. The encoded PBX-side device characteristics, eFax platform-side compatibility characteristics, real-time network status characteristics, and historical fax performance characteristics are weighted and fused based on a first preset weight to obtain a structured context feature vector. The first preset weight is a pre-configured feature dimension weight coefficient.

[0036] It should be understood that during the process of collecting features from the PBX side and the eFax platform side, the features from the PBX side and the eFax platform side are mostly discrete data. However, in practical applications, the system needs numerical vectors with a unified dimension and in a continuous space. Therefore, it is necessary to process the features from the PBX side and the eFax platform side to obtain structured context feature vectors.

[0037] One-hot coding is a method for converting categorical data (non-numeric) into a numerical format that computers can understand. Embedding coding is a technique for converting high-dimensional, sparse categorical data (such as words and ID numbers) into low-dimensional, dense, and semantically rich real-valued vectors.

[0038] In this implementation, since the features of the PBX-side devices and the eFax platform are mostly discrete data, while the system can process numerical vectors of a uniform dimension in a continuous space, it is necessary to process and fuse the features of the PBX-side devices and the eFax platform to obtain a structured context feature vector. This context feature vector is then used to train an adaptive policy decision model. To provide a more detailed explanation of the above embodiments, the following description is divided into several parts.

[0039] The first part describes the device identifier, firmware version, protocol characteristics, and network parameters of the PBX; the session description protocols and compatibility modes supported by the eFax platform; the round-trip latency, packet loss rate, jitter, and bandwidth of the real-time network status characteristics; and the historical success rate and error types of the historical fax performance characteristics.

[0040] The device identifier for this PBX is used to identify the PBX device, and the firmware version is the embedded software version number stored in the PBX hardware. The firmware version contains key information such as device function support, protocol compatibility, and vulnerability fixes. Protocol characteristics include the types, versions, and configuration parameters of communication protocols supported by the PBX, including the session initiation protocol used for call establishment. Network parameters are the key parameters required for the PBX to access the network. Network parameters include IP address, subnet mask, and gateway.

[0041] The eFax platform supports a Session Description Protocol (STP), which is used to describe multimedia sessions (such as fax and voice calls). The STP's function is to agree on key information between communicating parties, such as media type (fax data / audio), transmission protocol, port number, and encoding format, ensuring correct data transmission. The eFax platform supports a compatibility mode designed to adapt to different manufacturers' PBXs, traditional fax machines, and older protocols. The eFax platform includes protocol compatibility, device compatibility, and version compatibility.

[0042] Round-trip time (RTT) is the total delay time for a network data packet to travel from the sender (e.g., PBX) to the receiver (e.g., eFax server) and back to the sender, measured in milliseconds (ms). Packet loss rate is the percentage of data packets lost during transmission out of the total number of transmitted data packets. Jitter is the difference (standard deviation) between multiple round-trip times, measured in milliseconds (ms), reflecting the stability of network latency. Bandwidth is the maximum data transmission rate of the network link, measured in megabits per second (Mbps).

[0043] Historical fax performance characteristics: Historical success rate is the percentage of successfully completed faxes within the statistical period out of the total number of faxes initiated. Error types are the various fault reasons that cause fax failures, common types include: protocol error, network error, device error, format error, and number error.

[0044] The second part explains the missing feature data in the PBX-side device characteristics, the eFax platform-side compatibility characteristics, the real-time network status characteristics, and the historical fax performance characteristics.

[0045] It should be understood that during the acquisition of the PBX-side device characteristics, the eFax platform-side compatibility characteristics, the real-time network status characteristics, and the historical fax performance characteristics, due to incompatibility in data transmission or acquisition, there may be missing features in these categories. In such cases, it is necessary to supplement the missing feature data in these categories.

[0046] Part Three explains the standardization process for the supplemented PBX-side device characteristics, eFax platform-side compatibility characteristics, real-time network status characteristics, and historical fax performance characteristics.

[0047] It should be understood that in fax sessions, feature data comes from different PBXs and eFax platforms. The collected feature data has problems with inconsistent formats, units, and naming rules. Furthermore, the numerical ranges of the collected PBX-side device features, eFax platform-side compatibility features, real-time network status features, and historical fax performance features deviate from the standard range of the adaptive policy decision model. In this case, it is necessary to standardize the supplemented PBX-side device features, eFax platform-side compatibility features, real-time network status features, and historical fax performance features. The standardized data should be formatted according to a unified rule for easy management.

[0048] In some embodiments, the value of the PBX-side device feature is limited to a first preset range, the eFax platform-side compatibility feature is limited to a second preset range, the real-time network status feature is limited to a third preset range, and the historical fax performance feature is limited to a fourth preset range.

[0049] The first preset range, the second preset range, the third preset range, and the fourth preset range are numerical ranges automatically determined by the PBX, and this application embodiment does not limit them.

[0050] For example, if the upper limit of the value of the PBX-side device feature is greater than the upper limit of the first preset range, then the upper limit of the first preset range is determined as the upper limit of the value of the PBX-side device feature. If the lower limit of the value of the PBX-side device feature is less than the lower limit of the first preset range, then the lower limit of the first preset range is determined as the lower limit of the value of the value of the PBX-side device feature.

[0051] Part Four explains the encoding of the standardized PBX-side device characteristics, eFax platform-side compatibility characteristics, real-time network status characteristics, and historical fax performance characteristics based on one-hot encoding or embedded encoding.

[0052] It should be understood that the standardized feature data contains a large amount of textual data. However, the adaptive policy decision model can only understand and learn data in the form of numerical inputs. Therefore, the standardized feature data needs to be encoded.

[0053] One-hot encoding is suitable for low-cardinality, unordered categorical features. It generates a unique binary vector for each categorical feature. Embedding encoding is suitable for high-cardinality, semantically related categorical features (such as firmware version, error subclass, device model). It learns low-dimensional dense vectors through neural networks, preserving the similarity between categories.

[0054] Part Five explains the content of the structured context feature vector obtained by weighted fusion of the encoded PBX-side device features, eFax platform-side compatibility features, real-time network status features, and historical fax performance features based on the first preset weight.

[0055] It should be understood that the encoded features come from different feature sources, and the importance of different feature data varies. Therefore, it is necessary to weight and fuse the encoded PBX-side device features, eFax platform-side compatibility features, real-time network status features, and historical fax performance features based on a first preset weight to obtain a structured context feature vector. Since the first preset weight is set by the decision-making needs of different business scenarios, the multi-source features are fused according to the needs of different business scenarios, making the context feature vector more adaptable to business requirements.

[0056] The structured context feature vector format is as follows:

[0057] in, Characteristics of PBX-side equipment; For eFax platform-side compatibility features; For round-trip time; Packet loss rate; For shaking; This represents the historical success rate.

[0058] It should be understood that before performing the above steps to weight and fuse the encoded PBX-side device features, eFax platform-side compatibility features, real-time network status features, and historical fax performance features based on the first preset weights to obtain a structured context feature vector, the following steps may also be performed.

[0059] In practical applications, enterprises may have multiple PBXs of different models and multiple eFax platforms. Therefore, it is necessary to determine the target eFax platform based on the PBX's call signaling, session description protocol parameters, and session direction.

[0060] In one possible implementation, the call signaling, session description protocol parameters, and session direction of the PBX are collected; the corresponding target eFax platform is matched based on the call signaling, session description protocol parameters, session direction, and the device identifier of the PBX.

[0061] The call signaling is Session Initiation Protocol (SIP) signaling. In some embodiments, if the fax is initiated by the PBX, the call signaling is SIPINVITE signaling; if the call is initiated by the eFax platform and the PBX is the receiving party, the call signaling is 200 OK.

[0062] The Session Description Protocol (SDP) parameters include encoding parameters, port parameters, and negotiation methods. The encoding parameters in the SDP define the fax encoding protocol, and the code parameter specifies the transmission standard for fax data; the only core encodings in the fax scenario are T.30 and T.38. The port parameter specifies the media transmission port. SDP negotiation methods are implemented through SIP signaling interaction, and there are two types of negotiation methods, with "proposal / response" being the primary method in the fax scenario.

[0063] The session direction is used to confirm whether the PBX is the sender.

[0064] The target eFax platform is the platform that establishes a communication link with the PBX.

[0065] In some embodiments, a fax session is triggered to confirm that the PBX sends a SIPINVITE signaling message to the target eFax platform and that the SDP parameters are consistent with the rule base; the 200 OK signaling message replied by the target eFax platform is checked to confirm that the SDP response is consistent with the encoding and port proposed by the PBX; and the fax is confirmed to have been successfully sent / received.

[0066] In this implementation, the target eFax platform is matched based on the call signaling, the session description protocol parameters, the session direction, and the device identifier of the PBX. That is, the target eFax platform is determined based on the parameters of the PBX so as to train an adaptive policy decision model based on the target eFax platform, thereby achieving adaptation between the PBX and the third-party eFax platform.

[0067] Step 202: Input the context feature vector into the adaptive policy decision model, and map an initial fax policy set from the adaptive policy decision. The initial fax policy set includes at least one of protocol selection policy, routing policy, retransmission policy, and encoding parameter policy. Determine a reference fax policy set from the initial fax policy based on the reward function, which is used to measure the performance of the context feature vector after executing the reference fax policy set.

[0068] The adaptive policy decision model is used to achieve optimal sending and receiving decisions for fax services under different network conditions, device types, and transmission channels, thereby improving the overall fax success rate and system throughput performance. The fax policy set stores the policy options that the adaptive policy decision model can execute.

[0069] It should be understood that in practical applications, the context feature vector comes from the Feature Encoder module, while the execution module of the adaptive policy decision model is the Strategy Engine module. Since the Feature Encoder module and the Strategy Engine module are two different modules with different structures, the context feature vector may not match the adaptive policy decision model. In this case, it is necessary to first convert the context feature vector into the state representation within the Strategy Engine module.

[0070] In some embodiments, the specific formula for converting the context feature vector into a state representation within the policy decision module is as follows.

[0071]

[0072] in, This represents the state of the context feature vector within the policy decision module. For context feature vectors; Let be the state mapping function, which is used to ensure that the dimension of the context feature vector matches the adaptive policy decision model.

[0073] In some embodiments, the mapping formula for mapping the initial fax strategy set from the adaptive strategy decision model is as follows.

[0074]

[0075] in, This is the initial set of fax strategies; For adaptive policy decision-making models, Select a policy for the protocol; for example, the policy could be whether to enable T.38, G.711, or ECM. For example, the routing strategy can be the channel or relay routing strategy of the eFax platform; For example, the retransmission strategy can be the retry interval and timeout strategy for transmission between the PBX and the eFax platform. For example, the encoding parameter strategy can be transmission redundancy and encoding compression parameters.

[0076] In one possible implementation, a reference fax strategy set is determined from an initial fax strategy set using a reward function, based on a multi-armed slot machine algorithm.

[0077] The reference fax strategy set is the fax strategy set with the best compatibility and the highest transmission success rate among the initial fax strategy sets.

[0078] In some embodiments, the formula for calculating the reference fax policy set is as follows.

[0079]

[0080] in, This represents the state of the context feature vector within the policy decision module. The fax policies that can be executed in the initial fax policy set; For the reward function; This includes the transmission success rate, average latency, and jitter after executing the executable fax strategy; average latency and jitter can assess bandwidth utilization during transmission. Estimate the expected revenue for each initial fax strategy; output a reference fax strategy set, representing the expected revenue under the current... The set of protocol parameters that maximizes the success rate of fax transmission.

[0081] In this implementation, a reference fax strategy set is determined from the initial fax strategy set based on the multi-armed slot machine algorithm and the reward function. That is, the protocol probability of each fax strategy in the initial fax strategy set is calculated in real time, and the optimally configured strategy is selected from the initial fax strategy set, thereby improving the fax success rate and reducing the failure rate.

[0082] Step 203: Input the reference fax strategy set into the protocol converter and the network predictor in parallel; wherein, the protocol converter is used to predict the protocol success probability of the reference fax strategy set under the target eFax platform based on the protocol compatibility verification model; the network predictor is used to predict the network reliability and network latency of the transmission link corresponding to the execution of the reference fax strategy set based on time-series network prediction.

[0083] It should be understood that in practical applications, after the policy decision module calculates the reference fax policy set, it will perform feasibility prediction on the reference fax policy set. That is, the reference fax policy set will be input into the protocol converter and network predictor in parallel.

[0084] The Protocol Translator is used to predict the protocol success probability of the reference fax policy set under the target eFax platform based on the protocol compatibility verification model; the Network Predictor is used to predict the network reliability and network latency of the transmission link corresponding to the execution of the reference fax policy set based on time-series network prediction.

[0085] In practical applications, the reference fax policy set is mapped to an executable transport protocol configuration set, and compatibility on the target eFax platform is verified.

[0086] In one possible implementation, a target protocol mapping table is matched based on the device identifier of the PBX, the compatibility mode of the eFax platform, and the transmission channel characteristics between the PBX and the eFax platform; the target protocol mapping table is used to verify the compatibility of the reference fax policy set; and the protocol success probability under the target eFax platform is predicted based on a preset classifier in the protocol compatibility verification model.

[0087] The device identifier is used to uniquely identify the PBX. A target protocol mapping table, based on the PBX's device identifier, the eFax platform's compatibility mode, and the transmission channel characteristics between the PBX and the eFax platform, represents the selection of the most compatible and highest-success-rate protocol configuration from available protocol combinations (such as T.38+UDP, T.30+TCP, SIP 2.0+T.38, etc.) based on the PBX's hardware characteristics, the eFax platform's vendor protocol specifications, and the network conditions of the transmission channel (such as bandwidth, latency, and packet loss rate). This protocol compatibility verification model is used to predict the protocol success probability of the target protocol mapping table. Preset classifiers include logistic regression classifiers or gradient tree model classifiers.

[0088] The first preset threshold is a threshold automatically set by the PBX, and this application embodiment does not limit it.

[0089] In some embodiments, the formula for calculating the probability of protocol success of the target protocol mapping table under the target eFax platform based on the logistic regression classifier or gradient tree model classifier in the protocol compatibility verification model is as follows.

[0090]

[0091] in, The probability of protocol success; For the preset classifier; For reference fax strategy set, This represents the state of the context feature vector within the policy decision module.

[0092] It should be understood that after calculating the protocol success rate based on the above formula, it is necessary to determine the compatibility of the target protocol mapping table and the transmission success rate based on this success rate. That is, a first preset threshold needs to be set to determine the compatibility of the target protocol mapping table and the transmission success rate.

[0093] In some embodiments, if the success probability of the protocol is greater than a first preset threshold, the success probability of the protocol is determined and output; if the success probability of the protocol is less than or equal to the first preset threshold, the reference fax policy set is recalculated based on the policy decision module.

[0094] It should be understood that if the success probability of the protocol is greater than the first preset threshold, it means that the target mapping table has the best compatibility and the highest transmission success rate. In this case, the target protocol mapping table is the best protocol mapping table, so the success probability of the protocol is determined to be output.

[0095] It should also be understood that if the success probability of the protocol is less than or equal to the first preset threshold, it indicates that the compatibility of the target mapping table is poor or the transmission success rate is not high. In this case, the target mapping table is not the optimal protocol mapping table. Therefore, in order to obtain the optimal target protocol mapping table, it is necessary to recalculate the reference fax policy set based on the policy decision module.

[0096] In this implementation, the protocol configuration scheme with the best compatibility and the highest transmission success rate is determined by predicting whether the target protocol mapping table that matches the transmission channel characteristics between the PBX and the eFax platform is the best-matched and most successful-transmission protocol configuration scheme based on the protocol converter, so as to obtain the optimal target protocol mapping table.

[0097] It should be understood that in practical applications, it is also necessary to ensure that the policy decision module can still make stable decisions under real-time network fluctuations. Therefore, the reference fax policy set is input into the network predictor to preset the reliability and performance of the transmission link of the reference fax policy set.

[0098] In one possible implementation, the context feature vector state and the historical network observation sequence are obtained; the context feature vector state and the historical network observation sequence are input into the time series prediction network to predict the transmission success rate and average round-trip delay changes of the transmission link.

[0099] Among them, the context feature vector state is the state representation of the context feature vector within the policy decision module; the historical network observation sequence is the historical sequence of the context feature vector state.

[0100] In some embodiments, the formulas for predicting the transmission success rate and average round-trip delay variation of the transmission link based on the timing prediction network are as follows.

[0101]

[0102] in, For time series prediction networks; It is a Long Short-Term Memory (LSTM) network. The output is the transmission success rate of the transmission link ( ). ) and changes in average round-trip time ( ).

[0103] In this implementation, the transmission success rate and average round-trip delay variation of the transmission link are predicted based on the time-series prediction network, which can ensure that the policy decision module can still make stable decisions under real-time network fluctuations, so that the transmission link of the reference fax policy set has reliability.

[0104] Step 204: Based on the success probability of the protocol, the reliability of the network, and the latency of the network, perform weighted optimization on the reference fax policy set to generate the final target fax policy to be executed.

[0105] Among them, network reliability includes the transmission success rate of the transmission link predicted by the time-series prediction network; network latency includes the average round-trip latency variation predicted by the time-series prediction network.

[0106] It should be understood that after obtaining the results from the protocol converter and network predictor, the fax strategy needs to be optimized to obtain the optimal target fax strategy.

[0107] In one possible implementation, an objective function is constructed based on the success probability of the protocol, the reliability of the network, and the network latency; the objective function is then optimized to generate a target fax strategy.

[0108] In some embodiments, the formula for calculating the objective function is as follows.

[0109]

[0110] in, The objective function is... This is the reference fax strategy set; This represents the probability of the protocol succeeding. This is the transmission success rate; This represents the change in average round-trip time; , as well as This is the second preset weight.

[0111] The second preset weight is the weight automatically determined by the PBX, and this application embodiment does not limit this.

[0112] In some embodiments, the fax strategy with the maximum value in the objective function is used as the target fax strategy, and the calculation formula for generating the target fax strategy is as follows.

[0113]

[0114] in, Targeted fax strategy; This is the reference fax strategy set; The objective function is denoted as .

[0115] In this implementation, the optimal fax strategy is selected as the target fax strategy based on the success probability of the protocol, the reliability of the network, and the network latency, thereby ensuring the compatibility of the target fax strategy in business scenarios and the stability of network transmission.

[0116] It should be understood that in practical applications, although the target fax strategy is the optimal fax strategy, its compatibility and transmission success rate may not be high. Therefore, it is necessary to determine whether the target fax strategy is feasible based on the second preset threshold.

[0117] In one possible implementation, if the maximum value of the optimized objective function is less than a second preset threshold, the objective function is resampled, recalculated, and optimized. If the maximum value of the objective function after multiple resampling, recalculation, and optimization is less than the second preset threshold, the set of strategies with the maximum values ​​of the objective function after multiple resampling, recalculation, and optimization is taken as the target fax strategy.

[0118] The second preset threshold is a threshold automatically determined by the PBX, and this application embodiment does not limit it.

[0119] It should be understood that if the maximum value of the objective function is less than the second preset threshold, it means that although the compatibility or transmission success rate of the strategy set corresponding to the objective function is the best among all reference fax strategy sets, it still cannot meet the business requirements. Therefore, it is necessary to resample and re-optimize the reference fax strategy set.

[0120] In some embodiments, if the maximum value of the optimized objective function is less than a second preset threshold, the device features of the user-level switch PBX side, the compatibility features of the fax service eFax platform side, the real-time network status features, and the historical fax performance features are re-collected and fused to generate a structured context feature vector, so as to recalculate and optimize the objective function based on the context feature vector.

[0121] It is understandable that the specific implementation of re-collecting and fusing the device characteristics of the user-level switch PBX side, the compatibility characteristics of the fax service eFax platform side, the real-time network status characteristics, and the historical fax performance characteristics to generate a structured context feature vector, so as to recalculate and optimize the objective function based on the context feature vector, can be found in the relevant descriptions of steps 201-204 above, and will not be repeated here.

[0122] It should be understood that although the maximum value of the objective function after multiple resampling, calculation and optimization is less than the second preset threshold, the policy set corresponding to the current objective function is the optimal policy set. Therefore, the policy set with the maximum value in the current objective function is taken as the optimal solution for the target fax policy.

[0123] In some embodiments, when the set of strategies that take the maximum value of the objective function after multiple resampling, calculation and optimization is taken as the target fax strategy, the maximum value of the objective function is taken as the second preset threshold.

[0124] It should be understood that the set of strategies with the maximum value in the objective function after multiple resampling, calculation and optimization is taken as the optimal solution for the target fax strategy. That is, if the second preset threshold is too large, then using the maximum value of the objective function as the second preset threshold can adjust the standard for judging whether the target fax strategy is feasible in a timely manner.

[0125] In this implementation, even if the calculated target fax strategy is not feasible, the objective function can be resampled, calculated, and optimized. The optimal solution is determined from the objective function after multiple resampling, calculation, and optimization, thereby obtaining the target fax strategy with the best compatibility and the highest transmission success rate, thus improving the decision-making power of the strategy decision module.

[0126] Step 205: After executing the final target fax strategy, collect the actual transmission result indicators and calculate the actual reward function based on the transmission result indicators; calculate the error between the actual reward function and the expected reward function predicted by the adaptive strategy decision model; based on the error, synchronously update the strategy return estimate of the adaptive strategy decision model, the model parameters of the network predictor, and the feature weights in the feature fusion process, thereby realizing the overall self-learning and continuous optimization of the system.

[0127] The transmission result metrics include the actual transmission success rate, actual average latency, and actual jitter of the transmission link corresponding to the execution of the target fax strategy.

[0128] In one possible implementation, after executing the final target fax strategy, the actual transmission success rate, actual average latency, and actual jitter of the outgoing link corresponding to the target fax strategy are collected; and the actual reward function is calculated based on the actual transmission success rate, actual average latency, and actual jitter.

[0129] In some embodiments, the formula for calculating the actual return function is as follows.

[0130]

[0131] in, For actual return function; This represents the actual transmission success rate. This represents the actual average latency; This refers to actual jitter; , as well as These are weights configured by the strategy decision module and used to reflect business priorities. For example, if actual transmission success rate takes precedence, then... If the actual average latency is relatively large, then... Relatively large.

[0132] In one possible implementation, the error between the actual return function and the expected return function predicted by the adaptive policy decision model is calculated.

[0133] In some embodiments, the prediction reward function is calculated based on the transmission success rate of the transmission link, the average round-trip delay variation, and the predicted jitter predicted by the network predictor in the adaptive policy decision model.

[0134] The formula for calculating the predicted return function is the same as the formula for calculating the actual return function.

[0135] In some embodiments, the formula for calculating the error between the actual return function and the predicted return function is as follows.

[0136]

[0137] in, This represents the error between the actual return function and the predicted return function. For actual return function; This is the function for predicting returns.

[0138] It should be understood that the error between the actual return function and the predicted return function indicates that there is an error between the predicted return and the actual return of the current adaptive strategy decision model. In this case, the prediction accuracy in the adaptive strategy decision model can be optimized in reverse based on this error, and the selection of the target fax strategy can be adjusted.

[0139] In one possible implementation, the expected return estimate after executing the initial fax strategy set in the adaptive policy decision model is corrected based on the error; the error is used as the gradient of the loss function to update the third preset weight of the temporal prediction network in the network predictor; the first preset weight in the context feature vector fusion process is updated in reverse based on the error; and the optimized adaptive policy decision model, the third preset weight of the network predictor, and the first preset weight in the feature fusion process are synchronously updated to the PBX.

[0140] The expected return estimate is used to select a reference fax strategy set from the initial fax strategy set.

[0141] In some embodiments, the expected return estimate is calculated based on the following formula.

[0142]

[0143] in, This represents the learning rate of the new expected return estimate on the old expected return estimate in the multi-armed slot machine algorithm, which is also the correction magnitude of the new expected return estimate on the old expected return estimate in the multi-armed slot machine algorithm. This is the current actual return function.

[0144] It should be understood that after updating the expected return estimate, PBX will dynamically adjust the probability of selecting the reference fax strategy set from the initial fax strategy set.

[0145] In some embodiments, the temporal prediction network can be a long short-term memory network, which includes learning weights, i.e., third preset weights. In order to make the predicted values ​​of the long short-term memory network closer to the actual values, the learning weights of the long short-term memory network can be adjusted in reverse based on the error. The weights of the learning weights of the long short-term memory network are adjusted as follows.

[0146]

[0147] in, The adjusted learning weights; These are the initial learning weights in the Long Short-Term Memory network; The learning rate in a long short-term memory network; loss function Regarding the gradient of the learning weights; This is an error-based loss function.

[0148] In some embodiments, during the context feature vector fusion process, the context feature vector includes multiple feature vectors, and the first preset weight includes multiple weight coefficients, adjusting the importance of the multiple feature vectors based on the following formula.

[0149]

[0150] in, The weight coefficients for the i-th input feature; The learning rate is used to control the magnitude of each adjustment to the weight coefficients, avoiding deviations caused by adjustments that are too fast or too slow. This represents the error between the actual return function and the predicted return function. For the reward function R, the i-th input feature / function f i The partial derivatives of the function f, where f is the partial derivative of the function f. i It is the i-th feature vector in the context feature vector.

[0151] It should be understood that in practical applications, after updating the adaptive policy decision model based on the error back-update, the third preset weight of the network predictor, and the first preset weight in the feature fusion process, the above information is synchronously updated to the PBX.

[0152] In this implementation, the adaptive policy decision model is updated in reverse based on the error, thereby precisely adjusting the parameters in the adaptive policy decision model.

[0153] This application provides a method for adaptive fax transmission between a PBX and a third-party eFax platform. The method collects and fuses device characteristics from the PBX side, compatibility characteristics from the eFax platform, real-time network status characteristics, and historical fax performance characteristics to generate a structured context feature vector. Based on this context feature vector, an adaptive policy decision model is trained, outputting a reference fax policy set. This reference fax policy set represents the policy set with the highest compatibility and transmission success rate. The reference fax policy set is then weighted and optimized based on protocol success probability, network reliability, and network latency to generate the final target fax policy. After executing the target fax policy, actual transmission result indicators are collected, and an actual reward function is calculated based on these indicators. The error between the actual reward function and the expected reward function predicted by the adaptive policy decision model is calculated. Based on this error, the policy reward estimate of the adaptive policy decision model, the model parameters of the network predictor, and the feature weights in the feature fusion process are updated synchronously. In other words, the fax sending and receiving strategy is automatically determined by the adaptive strategy decision model, and the fax strategy is optimized based on the protocol success probability, network reliability and network latency to obtain the optimal target fax strategy. Finally, the adaptive strategy decision model is optimized in reverse based on the predicted result value and the actual transmission result index. This not only improves the reliability and transmission efficiency of fax communication, but also enables adaptive fax control.

[0154] Figure 3 This is a schematic flowchart illustrating a method for adaptive fax transmission between a PBX and a third-party eFax platform, provided in an embodiment of this application.

[0155] For example, such as Figure 3 As shown, taking a PBX as the execution subject as an example, this application describes a method for adaptive fax transmission between a PBX and a third-party eFax platform. The method 300 includes the following steps.

[0156] Step 301: In response to the fax request trigger, the PBX identifies the target eFax platform.

[0157] It is understood that the specific implementation of step 301 can be found in the relevant description in step 201 above, and will not be repeated here.

[0158] Step 302: The context acquisition module generates feature data.

[0159] The feature data includes standardized PBX-side device features, eFax platform-side compatibility features, real-time network status features, and historical fax performance features.

[0160] It is understood that the specific implementation of step 302 can be found in the relevant description in step 201 above, and will not be repeated here.

[0161] Step 303: Generate context feature vectors.

[0162] It is understood that the specific implementation of step 303 can be found in the relevant description in step 201 above, and will not be repeated here.

[0163] Step 304: The adaptive strategy decision model maps the initial fax strategy set.

[0164] It is understood that the specific implementation of step 304 can be found in the relevant description in step 202 above, and will not be repeated here.

[0165] Step 305: Determine the reference fax policy set from the initial fax policy set.

[0166] It is understood that the specific implementation of step 305 can be found in the relevant description in step 202 above, and will not be repeated here.

[0167] Step 306: The protocol converter and network predictor predict the protocol success probability, network reliability, and network latency.

[0168] It is understandable that the specific implementation of step 306 can be found in the relevant description in step 203 above, and will not be repeated here.

[0169] Step 307: Generate the target fax strategy.

[0170] It is understood that the specific implementation of step 307 can be found in the relevant description in step 204 above, and will not be repeated here.

[0171] Step 308: Initiate a fax and monitor the status.

[0172] It should be understood that initiating a fax and monitoring its status indicates the execution of the final target fax strategy and monitoring the transmission status between the PBX and the eFax platform.

[0173] It is understandable that the specific implementation of step 308 can be found in the relevant description in step 205 above, and will not be repeated here.

[0174] Step 309: Generate fax session results.

[0175] It should be understood that generating a fax session result indicates that the PBX and eFax platform have confirmed and begun transmitting the fax session.

[0176] Step 310: Collect actual transmission result indicators.

[0177] It is understandable that the specific implementation of step 310 can be found in the relevant description in step 205 above, and will not be repeated here.

[0178] Step 311: Update the adaptive policy decision model.

[0179] It is understandable that the specific implementation of step 311 can be found in the relevant description in step 205 above, and will not be repeated here.

[0180] Step 312: Synchronize and update to the PBX instance.

[0181] It is understandable that the specific implementation of step 312 can be found in the relevant description in step 205 above, and will not be repeated here.

[0182] Figure 4 This is a schematic diagram of the structure of an apparatus for adaptive fax transmission between a PBX and a third-party eFax platform, provided in an embodiment of this application.

[0183] For example, the device 400 includes: The acquisition and fusion module 401 is used to collect and fuse device features on the user-level switch PBX side, compatibility features on the fax service eFax platform side, real-time network status features, and historical fax performance features in response to fax session requests, and generate a structured context feature vector. The mapping and determination module 402 is used to input the context feature vector into the adaptive policy decision model, and to map an initial fax policy set from the adaptive policy decision. The initial fax policy set includes at least one of protocol selection policy, routing selection policy, retransmission policy and encoding parameter policy. A reference fax policy set is determined from the initial fax policy based on a reward function, which is used to measure the performance of the context feature vector after executing the reference fax policy set. The input and prediction module 403 is used to input the reference fax strategy set into the protocol converter and the network predictor in parallel; wherein, the protocol converter is used to predict the protocol success probability of the reference fax strategy set under the target eFax platform based on the protocol compatibility verification model; the network predictor is used to predict the network reliability and network latency of the transmission link corresponding to the execution of the reference fax strategy set based on time-series network prediction. The generation module 404 is used to perform weighted optimization on the reference fax policy set based on the success probability of the protocol, the reliability of the network, and the network latency, and generate the final target fax policy to be executed. The calculation and update module 405 is used to collect actual transmission result indicators after executing the final target fax strategy, and calculate the actual reward function based on the transmission result indicators; calculate the error between the actual reward function and the expected reward function predicted by the adaptive strategy decision model; and based on the error, synchronously update the strategy reward estimate of the adaptive strategy decision model, the model parameters of the network predictor, and the feature weights in the feature fusion process, thereby realizing the overall self-learning and continuous optimization of the system.

[0184] In one possible implementation, the device 400 includes: The data acquisition and fusion module 401 is specifically used to acquire the device identifier, firmware version, protocol characteristics, and network parameters of the PBX; the session description protocols and compatibility modes supported by the eFax platform; the round-trip latency, packet loss rate, jitter, and bandwidth of the real-time network status characteristics; and the historical success rate and error type of the historical fax performance characteristics. The data acquisition and fusion module 401 is specifically used to supplement the missing feature data in the PBX-side device features, the eFax platform-side compatibility features, the real-time network status features, and the historical fax performance features. The acquisition and fusion module 401 is specifically used to standardize the supplemented PBX-side device characteristics, eFax platform-side compatibility characteristics, real-time network status characteristics, and historical fax performance characteristics. The acquisition and fusion module 401 is specifically used to encode the standardized device features of the PBX side, the compatibility features of the eFax platform side, the real-time network status features, and the historical fax performance features based on one-hot encoding or embedded encoding. The acquisition and fusion module 401 is specifically used to perform weighted fusion of the encoded PBX-side device features, eFax platform-side compatibility features, real-time network status features, and historical fax performance features based on a first preset weight, to obtain a structured context feature vector; wherein, the first preset weight is a pre-configured feature dimension weight coefficient.

[0185] In one possible implementation, the device 400 includes: The acquisition and fusion module 401 is also used to acquire the call signaling, session description protocol parameters and session direction of the PBX; The acquisition and fusion module 401 is also used to match the corresponding target eFax platform based on the call signaling, the session description protocol parameters, the session direction, and the device identifier of the PBX.

[0186] In one possible implementation, the device 400 includes: The input and prediction module 403 is used to match the target protocol mapping table based on the device identifier of the PBX, the compatibility mode of the eFax platform, and the transmission channel characteristics between the PBX and the eFax platform; the target protocol mapping table is used to verify the compatibility of the reference fax policy set. The input and prediction module 403 is used to predict the success probability of the target protocol mapping table under the target eFax platform based on the preset classifier in the protocol compatibility verification model. The input and prediction module 403 is used to determine the output success probability of the protocol when the success probability of the protocol is greater than a first preset threshold.

[0187] In one possible implementation, the device 400 includes: Input and prediction module 403 is used to obtain the context feature vector state and historical network observation sequence; The input and prediction module 403 is used to input the context feature vector state and the historical network observation sequence into the time series prediction network to predict the transmission success rate and average round-trip delay changes of the transmission link.

[0188] In one possible implementation, the device 400 includes: Generation module 404 is used to construct an objective function based on the success probability of the protocol, the reliability of the network, and the network latency; wherein, the calculation formula of the objective function is as follows: ; This is the reference fax strategy set; This represents the probability of the protocol succeeding. This is the transmission success rate; This represents the change in average round-trip time; , as well as The second preset weight; The generation module 404 is used to optimize the objective function and generate the objective fax strategy.

[0189] In one possible implementation, the device 400 includes: The generation module 404 is used to resample, calculate and optimize the objective function if the maximum value of the optimized objective function is less than the second preset threshold. The generation module 404 is used to take the set of strategies with the maximum values ​​of the objective function after multiple resampling, calculation and optimization as the target fax strategy when the maximum value of the objective function after multiple resampling, calculation and optimization is less than the second preset threshold.

[0190] In one possible implementation, the device 400 includes: Calculate and update module 405 to correct the expected return estimate after executing the initial set of fax strategies in the adaptive strategy decision model based on the error; The calculation and update module 405 is used to update the third preset weights of the temporal prediction network in the network predictor using the error as the gradient of the loss function. The calculation and update module 405 is used to back-update the first preset weights in the context feature vector fusion process based on the error. The calculation and update module 405 is used to synchronously update the optimized adaptive policy decision model, the third preset weight of the network predictor, and the first preset weight in the feature fusion process to the PBX.

[0191] Figure 5 This is a schematic diagram of the structure of a device provided in an embodiment of this application.

[0192] For example, such as Figure 5 As shown, the device 500 includes a memory 501 and a processor 502, wherein the memory 501 stores executable program code 5011, and the processor 502 is used to call and execute the executable program code 503 to perform a method for adaptive fax transmission between PBX and eFax platforms.

[0193] Furthermore, embodiments of this application also protect an apparatus that may include a memory and a processor, wherein the memory stores executable program code, and the processor is used to call and execute the executable program code to perform a method for adaptive fax transmission between a PBX and an eFax platform provided in embodiments of this application.

[0194] This embodiment can divide the device into functional modules based on the above method example. For example, each module can correspond to a separate function, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware. It should be noted that the module division in this embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods.

[0195] It should be understood that the apparatus provided in this embodiment is used to execute the above-described method for adaptive fax transmission between PBX and eFax platforms, and therefore can achieve the same effect as the above-described implementation method.

[0196] When using integrated units, the device may include a processing module and a storage module. When applied to a workpiece, the processing module can be used to control and manage the workpiece's operations. The storage module can be used to support the execution of relevant program code by the workpiece.

[0197] The processing module may be a processor or a controller, which can implement or execute various exemplary logic blocks, modules, and circuits shown in conjunction with the disclosure of this application. The processor may also be a combination of functions that implement computing capabilities, such as a combination of one or more microprocessors, a combination of digital signal processing (DSP) and microprocessors, etc., and the storage module may be a memory.

[0198] In addition, the apparatus provided in the embodiments of this application may specifically be a chip, component or module. The chip may include a connected processor and a memory. The memory is used to store instructions. When the processor calls and executes the instructions, the chip can execute the method for adaptive fax transmission between PBX and eFax platforms provided in the above embodiments.

[0199] This embodiment also provides a computer-readable storage medium storing computer program code. When the computer program code is run on a computer, the computer executes the above-described method steps to implement the method for adaptive fax transmission between PBX and eFax platforms provided in the above embodiment.

[0200] This embodiment also provides a computer program product that, when run on a computer, causes the computer to perform the aforementioned related steps to implement the method for adaptive fax transmission between PBX and eFax platforms provided in the above embodiment.

[0201] In this embodiment, the device, computer-readable storage medium, computer program product, or chip are all used to execute the corresponding methods provided above. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods provided above, and will not be repeated here.

[0202] Through the above description of the embodiments, those skilled in the art will understand that, for the sake of convenience and brevity, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above.

[0203] In the embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0204] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for adaptive fax transmission between PBX and eFax platforms, characterized in that, include: In response to fax session requests, the system collects and integrates device characteristics from the user-level switch (PBX) side, compatibility characteristics from the eFax platform side, real-time network status characteristics, and historical fax performance characteristics to generate a structured context feature vector. The context feature vector is input into the adaptive policy decision model, and the adaptive policy decision is used to map an initial fax policy set, which includes at least one of protocol selection policy, routing policy, retransmission policy and encoding parameter policy; a reference fax policy set is determined from the initial fax policy based on a reward function, which is used to measure the performance of the context feature vector after executing the reference fax policy set. The reference fax strategy set is input in parallel into a protocol converter and a network predictor; wherein, the protocol converter is used to predict the protocol success probability of the reference fax strategy set under the target eFax platform based on a protocol compatibility verification model; the network predictor is used to predict the network reliability and network latency of the transmission link corresponding to the execution of the reference fax strategy set based on time-series network prediction. Based on the success probability of the protocol, the network reliability, and the network latency, the reference fax strategy set is weighted and optimized to generate the final target fax strategy to be executed. After executing the final target fax strategy, the actual transmission result indicators are collected, and the actual reward function is calculated based on the transmission result indicators. The error between the actual reward function and the expected reward function predicted by the adaptive strategy decision model is calculated. Based on the error, the strategy reward estimate of the adaptive strategy decision model, the model parameters of the network predictor, and the feature weights in the feature fusion process are updated synchronously, thereby realizing the overall self-learning and continuous optimization of the system.

2. The method according to claim 1, characterized in that, The process involves collecting and fusing features from the user-level switch (PBX) side, compatibility features from the eFax platform side, real-time network status features, and historical fax performance features to generate a structured context feature vector, including: The system collects the PBX's device identifier, firmware version, protocol characteristics, and network parameters; the session description protocols and compatibility modes supported by the eFax platform; the round-trip latency, packet loss rate, jitter, and bandwidth of the real-time network status characteristics; and the historical success rate and error types of the historical fax performance characteristics. Supplement the missing feature data in the PBX-side device characteristics, the eFax platform-side compatibility characteristics, the real-time network status characteristics, and the historical fax performance characteristics; The supplemented PBX-side device characteristics, eFax platform-side compatibility characteristics, real-time network status characteristics, and historical fax performance characteristics are standardized. The standardized PBX-side device characteristics, eFax platform-side compatibility characteristics, real-time network status characteristics, and historical fax performance characteristics are encoded using one-hot encoding or embedded encoding. Based on the first preset weight, the encoded PBX-side device features, eFax platform-side compatibility features, real-time network status features, and historical fax performance features are weighted and fused to obtain a structured context feature vector; wherein, the first preset weight is a pre-configured feature dimension weight coefficient.

3. The method according to claim 2, characterized in that, Before the step of weightedly fusing the encoded PBX-side device features, eFax platform-side compatibility features, real-time network status features, and historical fax performance features based on a first preset weight to obtain a structured context feature vector, the method further includes: Collect the call signaling, session description protocol parameters, and session direction of the PBX; The target eFax platform is matched based on the call signaling, the session description protocol parameters, the session direction, and the device identifier of the PBX.

4. The method according to claim 1, characterized in that, The protocol compatibility verification model predicts the protocol success probability of the reference fax strategy set under the target eFax platform, including: Based on the device identifier of the PBX, the compatibility mode of the eFax platform, and the target protocol mapping table matching the transmission channel characteristics between the PBX and the eFax platform; the target protocol mapping table is used to verify the compatibility of the reference fax policy set; Based on the preset classifier in the protocol compatibility verification model, predict the protocol success probability of the target protocol mapping table under the target eFax platform; If the success probability of the protocol is greater than a first preset threshold, the success probability of the protocol is determined and output.

5. The method according to claim 1, characterized in that, The time-series prediction network predicts the network reliability and network latency of the transmission link corresponding to the execution of the reference fax strategy set, including: Obtain the context feature vector state and historical network observation sequences; The context feature vector state and the historical network observation sequence are input into the time-series prediction network to predict the transmission success rate and average round-trip delay changes of the transmission link.

6. The method according to claim 1, characterized in that, The network reliability includes transmission success rate, and the network latency includes average round-trip time variation. The weighted optimization of the reference fax strategy set based on the protocol success probability, network reliability, and network latency to generate the final target fax strategy includes: An objective function is constructed based on the protocol success probability, network reliability, and network latency; wherein the calculation formula for the objective function is: ; The reference fax strategy set; The success probability of the protocol; The transmission success rate; This refers to the average round-trip time variation; , as well as The second preset weight; The objective function is optimized to generate the target fax strategy.

7. The method according to claim 6, characterized in that, The method further includes: If the maximum value of the optimized objective function is less than the second preset threshold, then the objective function is resampled, recalculated, and optimized. If the maximum value of the objective function after multiple resampling, calculation and optimization is less than the second preset threshold, the set of strategies with the maximum value of the objective function after multiple resampling, calculation and optimization will be used as the target fax strategy.

8. The method according to claim 1, characterized in that, The step of synchronously updating the policy return estimate of the adaptive policy decision model, the model parameters of the network predictor, and the feature weights in the feature fusion process based on the error includes: Based on the error, the expected return estimate after executing the initial fax strategy set in the adaptive strategy decision model is corrected; The error is used as the gradient of the loss function to update the third preset weight of the temporal prediction network in the network predictor; The first preset weight in the context feature vector fusion process is updated in reverse based on the error. The optimized adaptive policy decision model, the third preset weight of the network predictor, and the first preset weight in the feature fusion process are synchronously updated to the PBX.

9. An apparatus for adaptive fax transmission between a PBX and an eFax platform, characterized in that, The device includes: The data acquisition and fusion module is used to respond to fax session requests by acquiring and fusing device features on the user-level switch PBX side, compatibility features on the fax service eFax platform side, real-time network status features, and historical fax performance features to generate a structured context feature vector. The mapping and determination module is used to input the context feature vector into the adaptive policy decision model, and map an initial fax policy set from the adaptive policy decision. The initial fax policy set includes at least one of protocol selection policy, routing selection policy, retransmission policy, and encoding parameter policy. A reference fax policy set is determined from the initial fax policy based on a reward function, which is used to measure the performance of the context feature vector after executing the reference fax policy set. An input and prediction module is used to input the reference fax strategy set into a protocol converter and a network predictor in parallel; wherein, the protocol converter is used to predict the protocol success probability of the reference fax strategy set under the target eFax platform based on a protocol compatibility verification model; the network predictor is used to predict the network reliability and network latency of the transmission link corresponding to the execution of the reference fax strategy set based on time-series network prediction. The generation module is used to perform weighted optimization on the reference fax strategy set based on the protocol success probability, the network reliability, and the network latency, and generate the final target fax strategy to be executed. The calculation and update module is used to collect actual transmission result indicators after executing the final target fax strategy, and calculate the actual reward function based on the transmission result indicators; calculate the error between the actual reward function and the expected reward function predicted by the adaptive strategy decision model; and based on the error, synchronously update the strategy return estimate of the adaptive strategy decision model, the model parameters of the network predictor, and the feature weights in the feature fusion process, thereby realizing the overall self-learning and continuous optimization of the system.

10. A device, characterized in that, The device includes: Memory, used to store executable program code; A processor for calling and running the executable program code from the memory, causing the device to perform the method as described in any one of claims 1 to 8.

Images