Communication voice quality inspection AS voice code conversion method based on cascading

Abstract

The invention relates to the technical field of communication, and discloses a cascade-based communication voice quality inspection AS voice code conversion method. The method is based on an IMS network architecture, is applied to a communication voice quality inspection scene, and realizes voice code conversion. The method comprises a first-level quality inspection AS media server, a second-level quality inspection AS signaling server, a cascade connection module and an intelligent task scheduling module. The first-stage quality inspection AS media server receives a voice sampling file and executes main transcoding processing, the second-stage quality inspection AS signaling server receives a file transmitted by the first-stage server and executes auxiliary transcoding processing, and the cascade connection module adopts an SFTP protocol to realize file transmission between the two-stage server. The intelligent task scheduling module dynamically adjusts a processing strategy according to a system load; according to the invention, the real-time bottleneck of large-scale voice quality inspection is solved, high-concurrency, high-reliability and extensible voice code conversion is realized, and technical support is provided for operator fraud risk control.

Description

Technical Field

[0001] This invention relates to the field of communication technology, and specifically to a cascaded communication voice quality inspection AS voice coding conversion method. Background Technology

[0002] In the field of voice quality inspection, with the increasing demand from operators for identifying fraudulent risk numbers, the amount of voice data that voice quality inspection systems need to process is growing exponentially. Traditional voice encoding and conversion uses a single-server processing model, the core problem of which lies in the significant bottleneck of processing capacity. The hardware resources of a single server are limited, and when faced with large-scale concurrent voice transcoding tasks, it cannot respond to all requests in a timely manner, leading to increased transcoding latency and severely impacting the real-time performance of voice quality inspection. In this model, a large number of transcoding tasks accumulate in the task queue of a single server, and once the server's processing limit is exceeded, task loss or transcoding failures may occur, failing to meet the business needs of operators for rapid quality inspection of massive amounts of voice data.

[0003] Based on the above problems, in order to ensure the timeliness and accuracy of fraud risk identification, there is an urgent need for a speech coding conversion technology solution that can overcome the limitations of processing capacity and achieve high-concurrency processing. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a cascaded AS (Autonomous System) speech encoding and conversion method for communication speech quality inspection. Based on the IMS network architecture, this method is applied to communication speech quality inspection scenarios to achieve speech encoding and conversion. The basic steps include receiving speech sampling files, performing transcoding processing, and outputting transcoding results. The invention is characterized by further comprising: a first-level quality inspection AS media server, a second-level quality inspection AS signaling server, a cascaded connection module, and an intelligent task scheduling module. The first-level quality inspection AS media server receives the speech sampling files and performs the main transcoding processing. The second-level quality inspection AS signaling server receives the files transmitted by the first-level quality inspection AS media server and performs auxiliary transcoding processing. The cascaded connection module enables file transfer between the two levels of servers. The intelligent task scheduling module dynamically adjusts the processing strategy according to the system load to ensure efficient and stable transcoding processes.

[0005] Preferably, the intelligent task scheduling module includes a dynamic threshold control unit, a task diversion unit, and a multi-threaded parallel processing unit; the dynamic threshold control unit dynamically adjusts the task queue threshold according to the system load, and the threshold calculation formula is threshold = baseline value × (1 + CPU utilization × 0.6 + queue time × 0.4); when the task queue of the first-level quality inspection AS media server exceeds the set threshold, the task diversion unit automatically diverts the excess tasks to the second-level quality inspection AS signaling server for processing; the multi-threaded parallel processing unit is configured with multiple working threads to realize the parallel execution of transcoding tasks.

[0006] Preferably, the cascaded connection module uses the SFTP protocol to implement file transfer, and the data is encrypted using the AES-256 encryption algorithm during the transfer process, and the transmission security is ensured through a dynamic key negotiation mechanism; the SFTP protocol is based on the SSH protocol to build a transmission channel, and supports data integrity verification and breakpoint resume function during file transfer.

[0007] Preferably, the method uses a shared memory directory as a temporary file storage medium, with the shared memory directory path being / dev / shm / ; the shared memory directory directly maps to physical memory space, enabling millisecond-level read and write operations on files and avoiding performance loss caused by disk I / O.

[0008] Preferably, the first-level quality inspection AS media server captures VoLTE call voice sampling files through the IFC rules in the IMS network; the IFC rules are pre-configured in the S-CSCF network element, and when a VoLTE call is established, the S-CSCF triggers the call media stream to the first-level quality inspection AS media server according to the IFC rules to ensure accurate acquisition of voice sampling files.

[0009] Preferably, before performing auxiliary transcoding processing, the second-level quality inspection AS signaling server first performs integrity verification on the SIP signaling associated with the received file; the verification content includes call initiation information, call duration information, and terminal identification information in the SIP signaling, to ensure that the transcoded file accurately matches the corresponding call scenario.

[0010] Preferably, the task queue baseline threshold of the first-level quality inspection AS media server is set to 1000, and the task queue baseline threshold of the second-level quality inspection AS signaling server is set to 3000; the dynamic threshold control unit collects the CPU utilization and task queue time of the two-level servers in real time, and dynamically adjusts the actual threshold according to the set formula.

[0011] Preferably, both the first-level quality inspection AS media server and the second-level quality inspection AS signaling server are configured with 20 worker threads; the worker threads adopt a thread pool management mechanism, and idle threads remain in a ready state, responding immediately when a transcoding task arrives, avoiding resource consumption caused by thread creation and destruction.

[0012] Preferably, the shared memory directory adopts a hierarchical storage structure, with the first-level directory divided by call date, the second-level directory divided by terminal number range, and the third-level directory divided by call time period; the hierarchical storage structure enables fast file retrieval and classification management, and improves file access efficiency.

[0013] Preferably, the method is compatible with the conversion of three voice coding formats: AMR narrowband, AMR wideband, and A-LAW. The method connects with the provincial IMS aggregation subsystem to forward the transcoded voice files to the voice quality inspection platform, supporting the intelligent identification and control of fraudulent numbers.

[0014] Technical Effects: This invention creatively solves the processing bottleneck problem of traditional single servers by using a cascaded architecture of a first-level quality inspection AS media server and a second-level quality inspection AS signaling server, combined with an intelligent task scheduling module and a shared memory transmission scheme. The two-level servers work collaboratively to achieve task distribution, dynamic threshold control adapts to load changes, and shared memory and SFTP protocol ensure transmission efficiency and security, guaranteeing the real-time performance and reliability of large-scale voice transcoding tasks, and providing efficient technical support for operators' fraud risk control. Attached Figure Description

[0015] Figure 1 This is a flowchart of the AS speech coding conversion method for communication speech quality inspection based on cascading, as described in this application. Figure 2 This is a system block diagram of the cascaded architecture of this application; Figure 3 This is a detailed flowchart of the speech encoding conversion process in this application. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0017] Traditional technical solutions have the following technical problems: Traditional voice quality inspection systems use a single server to process voice encoding conversion, which has limited processing capacity and cannot cope with the high concurrency requirements of large-scale voice quality inspection. This leads to increased transcoding delays, task loss, and affects the real-time performance of quality inspection. At the same time, a single point of failure can easily paralyze the entire system, resulting in insufficient reliability.

[0018] Based on this, please refer to Figures 1-3 This embodiment provides a cascaded communication voice quality inspection AS voice coding conversion method, including: based on the IMS network architecture, applied to a communication voice quality inspection scenario, to realize voice coding conversion, including the basic steps of receiving a voice sampling file, performing transcoding processing, and outputting the transcoding result, and also including: S1: First-level quality inspection AS media server, second-level quality inspection AS signaling server, cascading connection module and intelligent task scheduling module; S2: The first-level quality inspection AS media server receives the audio sampling file and performs the main transcoding process; S3: The second-level quality inspection AS signaling server receives the file transmitted by the first-level quality inspection AS media server and performs auxiliary transcoding processing; S4: The cascaded connection module enables file transfer between two-level servers; S5: The intelligent task scheduling module dynamically adjusts the processing strategy according to the system load to ensure that the transcoding process is efficient and stable.

[0019] It's worth noting that this technical solution constructs a two-tier cascaded server architecture, clearly defining the functions and collaborative logic of each server level. The first-tier quality inspection AS media server, as the core transcoding node, directly connects to the IMS network, receiving VoLTE call voice sampling files triggered by IFC rules, focusing on key transcoding processing, including encoding format parsing and core transcoding algorithm execution. The second-tier quality inspection AS signaling server, as an auxiliary transcoding node, receives intermediate transcoding files transmitted from the first-tier server, handling the remaining transcoding process and final output tasks, forming a complementary two-tier processing mode. The cascaded connection module supports file transfer between the two-tier servers, ensuring smooth and secure data transmission. The intelligent task scheduling module is crucial for achieving high-concurrency processing. By monitoring the load status of the two-tier servers in real time, including indicators such as CPU utilization, memory usage, and task queue length, it dynamically adjusts processing strategies. When the load on a certain level server is too high, it automatically adjusts the task allocation ratio to avoid overloading a single node. The entire solution is designed around the IMS network architecture, fully adapting to the operator's existing network environment, and achieving seamless integration of voice encoding conversion and quality inspection processes.

[0020] The technical effects achieved by the above embodiments include: the two-tier server cascade architecture overcomes the processing capacity limitations of a single server, distributes transcoding pressure, and improves the system's concurrent processing capabilities; the intelligent task scheduling module ensures load balancing and reduces transcoding latency and task loss; and the cascaded structure avoids the impact of single points of failure, improves system reliability, and provides stable technical support for large-scale voice quality inspection.

[0021] Traditional technical solutions have the following technical problems: In existing technologies, the task scheduling of speech transcoding mostly adopts fixed allocation or simple load balancing methods, which cannot adjust the strategy according to the dynamic load of the system, resulting in uneven load distribution, with some server resources being idle while others are overloaded, affecting the overall transcoding efficiency.

[0022] Based on this, the intelligent task scheduling module includes a dynamic threshold control unit, a task diversion unit, and a multi-threaded parallel processing unit. The dynamic threshold control unit dynamically adjusts the task queue threshold according to the system load, and the threshold calculation formula is: threshold = baseline value × (1 + CPU utilization × 0.6 + queue time × 0.4). When the task queue of the first-level quality inspection AS media server exceeds the set threshold, the task diversion unit automatically diverts the excess tasks to the second-level quality inspection AS signaling server for processing. The multi-threaded parallel processing unit is configured with multiple working threads to realize the parallel execution of transcoding tasks.

[0023] It's worth noting that the intelligent task scheduling module consists of three functional units working together to form a complete scheduling system. The dynamic threshold control unit is the core of the scheduling. It dynamically adjusts the threshold by setting a scientific threshold calculation formula. In the formula, the baseline value is a preset initial threshold, CPU utilization is weighted at 0.6, and queue time is weighted at 0.4. The actual threshold is obtained through weighted calculation, ensuring that the threshold adjustment accurately reflects the server load status. CPU utilization is collected in real time by the server's built-in monitoring tool, updating data every 100 milliseconds. Queue time is the average waiting time of all pending tasks in the current task queue, calculated by statistically analyzing the time difference between a task entering the queue and starting processing. The task distribution unit is responsible for the dynamic allocation of tasks. It monitors the task queue length of the first-level server in real time. When the queue length exceeds the actual threshold calculated by the dynamic threshold control unit, the distribution mechanism is immediately activated, distributing the excess tasks sequentially to the second-level server. During the distribution process, the integrity and order of tasks are maintained, preventing task disorder. The multi-threaded parallel processing unit improves the processing efficiency of a single server through a multi-threaded mechanism. Each server is configured with multiple independent worker threads to handle multiple transcoding tasks simultaneously. Threads are scheduled through a task queue to ensure that each thread has a balanced load and avoids thread idleness or overload.

[0024] The technical effects achieved by the above embodiments include: dynamic threshold control enables precise matching of load status and scheduling strategy; task splitting unit avoids overload of a single server; multi-threaded parallelism improves the processing efficiency of a single node; overall intelligent load balancing is achieved, improving the overall transcoding efficiency and stability of the system.

[0025] Traditional technical solutions have the following technical problems: existing voice transcoding file transmission mostly uses ordinary file transmission protocols, which lack effective encryption mechanisms and integrity guarantees, making them prone to data leakage or transmission errors. At the same time, they do not support interrupted transmission, and need to be retransmitted after network interruption, which affects transmission efficiency.

[0026] Based on this, the cascaded connection module uses the SFTP protocol to realize file transfer. During the transfer process, the data is encrypted using the AES-256 encryption algorithm, and the transmission security is ensured through a dynamic key negotiation mechanism. The SFTP protocol is based on the SSH protocol to build a transmission channel, supporting data integrity verification and breakpoint resume function during file transfer.

[0027] It's worth noting that the cascading connection module uses the SFTP protocol as its core to build the transmission system. SFTP, based on the secure channel of the SSH protocol, enables file transfer. First, a secure connection is established between the two servers via SSH, and authentication is performed during the connection process to ensure the legitimacy of both parties. Transmitted data is encrypted using the AES-256 encryption algorithm. AES-256 employs symmetric encryption with a 256-bit key length. The encryption process includes data block generation and round-robin transformations to ensure that data cannot be stolen or tampered with during transmission. A dynamic key negotiation mechanism is used for key generation and updates. Each time a transmission connection is established, a session key is dynamically generated using the Diffie-Hellman algorithm. The key is only valid within the current transmission session and is automatically destroyed after the session ends, avoiding security risks from key leakage. During file transfer, the SFTP protocol automatically performs data integrity checks by calculating the CRC32 checksum of the file. The sending end transmits the checksum along with the file data. The receiving end recalculates the checksum and compares it with the sending end's checksum. If they do not match, it indicates a transmission error, triggering a retransmission mechanism. The resume function works by recording the position of the transmitted file segment. When the network is interrupted or the transmission is paused, the transmission resumes from the end of the transmitted segment when the transmission is resumed, without having to retransmit the entire file, which greatly improves the transmission efficiency, especially suitable for the transmission of large voice files.

[0028] The technical effects achieved by the above embodiments include: AES-256 encryption and dynamic key negotiation ensure file transmission security, prevent data leakage and tampering, integrity verification ensures the accuracy of transmitted data, breakpoint resume improves transmission efficiency, and meets the needs of secure and efficient transmission of voice-encoded files.

[0029] Traditional technical solutions have the following technical problems: Traditional voice transcoding file storage relies on disk storage, which has a slow disk I / O speed, resulting in high file read and write latency and affecting the overall efficiency of the transcoding process. Especially in high-concurrency scenarios, disk I / O can easily become a performance bottleneck.

[0030] Based on this, the method uses a shared memory directory as a temporary storage medium for files, with the shared memory directory path being / dev / shm / . The shared memory directory directly maps to physical memory space, enabling millisecond-level read and write operations for files and avoiding performance loss caused by disk I / O.

[0031] It's worth noting that this method uses a shared memory directory as temporary storage during the speech transcoding process. The shared memory directory path is fixed at ` / dev / shm / `, which is the default shared memory mount point in Linux systems, directly mapping to the server's physical memory space rather than a disk storage area. After the speech sample file is received by the first-level server, it is first stored in this shared memory directory. File read operations required during transcoding retrieve data directly from memory, without going through disk I / O. Intermediate files generated during transcoding are also temporarily stored in this directory, and are then transferred to the second-level server or persistently stored on disk as needed after transcoding is complete. The storage capacity of the shared memory directory is configured according to the server's physical memory size, reserving 40% of the total physical memory as available space for the shared memory directory to ensure sufficient space for storing temporary files during concurrent transcoding. Simultaneously, a file cleanup mechanism is established to automatically delete temporary files that have been transcoded and transferred or persistent, freeing up shared memory space and preventing space exhaustion. File read and write operations in the shared memory directory interact directly with physical memory through system calls, achieving millisecond-level read and write speeds, which are tens or even hundreds of times faster than disk I / O.

[0032] The technical effects achieved by the above embodiments include: the shared memory directory significantly improves file read and write speed, eliminates disk I / O performance bottlenecks, shortens the overall time of transcoding process, and at the same time, the file cleanup mechanism ensures the effective use of shared memory space and improves the stability of the system in high-concurrency scenarios.

[0033] Traditional technical solutions have the following technical problems: existing voice transcoding systems have difficulty accurately capturing VoLTE call voice sampling files and lack a coordination mechanism with IMS network core elements, resulting in untimely or missed acquisition of voice files, affecting the comprehensiveness of voice quality inspection.

[0034] Based on this, the first-level quality inspection AS media server captures VoLTE call voice sampling files through the IFC rules in the IMS network; the IFC rules are pre-configured in the S-CSCF network element, and when a VoLTE call is established, the S-CSCF triggers the call media stream to the first-level quality inspection AS media server according to the IFC rules to ensure accurate acquisition of voice sampling files.

[0035] It's worth noting that the Level 1 Quality Inspection AS Media Server establishes a collaborative mechanism with the S-CSCF network element of the IMS network. The core of this mechanism is the accurate capture of VoLTE call voice data through IFC rules. IFC rules are a set of rules in the IMS network that define the triggering conditions for business logic. These rules are pre-configured into the S-CSCF network element through the network management platform. The rules include trigger events, triggering conditions, and execution actions. The trigger event is set to a VoLTE call establishment event, triggered the moment a VoLTE call is successfully established when a user initiates or receives one. The triggering condition is set to all VoLTE calls, ensuring no voice data is missed. The execution action is set to forward the call media stream to the Level 1 Quality Inspection AS Media Server, carrying basic call information such as the terminal number and call duration. As the core control node of the IMS network, the S-CSCF network element monitors the call status in real time during VoLTE call establishment. When it detects that the trigger event meets the triggering conditions, it immediately executes the preset action, forwarding the call media stream to the Level 1 Quality Inspection AS Media Server through a designated interface. The first-level quality inspection AS media server is configured with a dedicated interface to receive media streams and perform real-time sampling processing to generate audio sampling files, preparing for subsequent transcoding processing.

[0036] The technical effects achieved by the above embodiments include: the collaboration between IFC rules and S-CSCF network elements enables accurate and timely capture of VoLTE call voice sampling files, ensuring that the voice data required for voice quality inspection is comprehensive and complete, providing complete data support for subsequent transcoding and quality inspection.

[0037] Traditional technical solutions have the following technical problems: In the existing voice transcoding process, there is no effective correlation and verification mechanism between the transcoded file and the corresponding call scenario, which may result in a mismatch between the transcoded file and the call information, affecting the accuracy of the voice quality inspection results.

[0038] Based on this, before performing auxiliary transcoding processing, the second-level quality inspection AS signaling server first performs integrity verification on the SIP signaling associated with the received file; the verification content includes call initiation information, call duration information, and terminal identification information in the SIP signaling, to ensure that the transcoded file accurately matches the corresponding call scenario.

[0039] It is worth mentioning that the Level 2 Quality Inspection AS signaling server establishes a SIP signaling integrity verification mechanism, completing the verification process before performing auxiliary transcoding. First, when the Level 1 Quality Inspection AS media server transmits the intermediate transcoding file, it simultaneously transmits the corresponding SIP signaling data. This SIP signaling data is session control signaling generated during a VoLTE call and contains key call information. After receiving the file, the Level 2 Quality Inspection AS signaling server first extracts the synchronously transmitted SIP signaling data and then verifies its integrity. The verification includes call initiation information, call duration information, and terminal identification information. Call initiation information includes the caller's number, the called party's number, and the call initiation time. This information must be confirmed to be correctly formatted and complete, without missing or incorrect information. The call duration information must be compared with the duration of the transcoded file to ensure consistency, allowing for an error of ±1 second; anything exceeding this range is considered a mismatch. Terminal identification information includes the device identifiers of the initiating and called parties' terminals. This identification information must be confirmed to be valid and conform to the IMS network's terminal identification specifications. During the verification process, if all verification items meet the requirements, the verification is deemed successful and auxiliary transcoding processing is initiated; if any item fails to meet the requirements, the verification is deemed unsuccessful, and the first-level quality inspection AS media server is notified of the verification failure information, requesting the retransmission of the file and the corresponding SIP signaling data.

[0040] The technical effects achieved by the above embodiments include: the SIP signaling integrity verification mechanism ensures that the transcoded file and the call scenario are accurately matched, avoids errors in quality inspection results caused by mismatch between the file and the information, and improves the accuracy and reliability of voice quality inspection.

[0041] Traditional technical solutions have the following technical problems: existing dynamic threshold adjustment lacks clear benchmark thresholds and scientific adjustment parameters, resulting in unreasonable threshold settings, inability to effectively adapt to the processing capabilities of different servers, and affecting the effectiveness of task scheduling.

[0042] Based on this, the task queue baseline threshold of the first-level quality inspection AS media server is set to 1000, and the task queue baseline threshold of the second-level quality inspection AS signaling server is set to 3000; the dynamic threshold control unit collects the CPU utilization and task queue time of the two-level servers in real time, and dynamically adjusts the actual threshold according to the set formula.

[0043] It's worth noting that a clear baseline threshold for the task queues of the two server tiers was defined. Different baseline values ​​were set based on the functional positioning and processing capabilities of the two tiers of servers. The first-tier quality inspection AS media server primarily handles core transcoding processing, which has higher complexity; therefore, its baseline threshold is set at 1000. The second-tier quality inspection AS signaling server handles auxiliary transcoding processing, which has relatively lower complexity and needs to be able to handle offloading tasks; therefore, its baseline threshold is set at 3000. A dynamic threshold control unit dynamically adjusts the baseline thresholds according to a preset formula, collecting two key parameters in real time: CPU utilization and task queue latency. CPU utilization is obtained through the server's CPU monitoring interface, collected every 100 milliseconds, representing the overall CPU utilization and covering the load of all cores. Task queue latency is calculated by statistically analyzing the waiting time of each task in the task queue, which is the difference between the time the task entered the queue and the current time. The average queue latency is calculated every 50 milliseconds. The collected CPU utilization and task queue latency are then input into the formula to calculate the actual thresholds. The first and second tier servers calculate their own actual thresholds based on their own parameters, without affecting each other. At the same time, upper and lower limits for thresholds are set. The actual lower limit for the first-level server is 500 and the upper limit is 1500, while the actual lower limit for the second-level server is 1500 and the upper limit is 4500, to avoid excessive threshold adjustments that deviate from a reasonable range.

[0044] The technical effects achieved by the above embodiments include: a clear benchmark threshold to adapt to the processing capability differences of the two-level servers; a scientific parameter collection and calculation method to ensure that the actual threshold accurately reflects the load status; and upper and lower limits of the threshold to ensure the rationality of adjustment, thereby improving the accuracy and effectiveness of task scheduling.

[0045] Traditional technical solutions have the following technical problems: existing transcoding servers lack an effective mechanism for thread management, and the frequent creation and destruction of threads leads to excessive resource consumption. At the same time, the unreasonable configuration of the number of threads either results in insufficient threads leading to task backlog or excessive threads leading to resource contention, which affects transcoding efficiency.

[0046] Based on this, both the first-level quality inspection AS media server and the second-level quality inspection AS signaling server are configured with 20 worker threads; the worker threads adopt a thread pool management mechanism, and idle threads remain in a ready state, responding immediately when a transcoding task arrives, thus avoiding resource consumption caused by thread creation and destruction.

[0047] It's worth noting that both server tiers are uniformly configured with 20 worker threads. This number was determined through multiple tests based on server hardware configuration and transcoding task complexity, resulting in an optimal value that meets high-concurrency transcoding requirements without wasting resources. The worker threads are managed using a thread pool mechanism. The thread pool initializes and creates 20 worker threads upon server startup, placing them in a ready state to await task allocation. The thread pool has an internal task queue; transcoding tasks arrive first and are then queued. The thread pool scheduler allocates tasks to idle worker threads according to a first-in, first-out (FIFO) principle. Each worker thread automatically returns to the ready state after completing its current task, awaiting the next task allocation, avoiding the destruction and re-creation of threads after task completion and reducing system resource consumption from thread creation and destruction. Simultaneously, the thread pool establishes a load monitoring mechanism to monitor the load of each worker thread in real time. If a thread's task processing time exceeds 1.5 times the average processing time three consecutive times, the thread is considered overloaded, and the scheduler suspends assigning new tasks to that thread until it completes its current task and returns to an idle state, ensuring load balance among threads.

[0048] The technical effects achieved by the above embodiments include: a fixed number of worker threads to adapt to transcoding task requirements, a thread pool management mechanism to reduce resource consumption, load monitoring to ensure thread load balancing, improved transcoding efficiency and stability of a single server, and support for high-concurrency processing of the system.

[0049] Traditional technical solutions suffer from the following problems: files in shared memory directories lack an effective classification and management mechanism, resulting in disorganized file storage and low efficiency in file retrieval and retrieval. In particular, under high-concurrency scenarios, the accumulation of a large number of files further exacerbates the difficulty of retrieval.

[0050] Based on this, the shared memory directory adopts a hierarchical storage structure, with the first-level directory divided by call date, the second-level directory divided by terminal number range, and the third-level directory divided by call time period; the hierarchical storage structure enables fast file retrieval and classification management, improving file access efficiency.

[0051] It's worth noting that a three-tiered storage structure is built within the shared memory directory. Each level of the directory is divided according to specific rules to ensure clear file categorization. The first-level directory is divided by call date, with directory names using the format "YYYYMMDD," such as 20250101 representing call files from January 1, 2025. A corresponding first-level directory is automatically created daily, storing all voice sampling files and temporary transcoding files for that day. The second-level directory, under the first-level directory, is divided by terminal number range. When the terminal number is a mobile phone number, it's divided by the first 7 digits, such as 1380013, 1391234, etc. Each number range corresponds to one second-level directory, and call files generated by terminals within the same number range are stored in the corresponding second-level directory. The third-level directory, under the second-level directory, is divided by call time period. A 24-hour day is divided into 12 time periods, each lasting 2 hours. Directory names use the format "HHHH-HHHH," such as 0000-0200, 0200-0400, etc. Files whose call times fall within the corresponding time period are stored in this third-level directory. File names follow the format "Terminal Number_Call Start Time_File Type" to ensure each file name is unique and contains key identification information. Simultaneously, a directory indexing mechanism is established to record the storage path and key information of each file, facilitating rapid retrieval. During a search, users can locate the corresponding third-level directory by calling date, terminal number range, and call duration, and then quickly find the target file by its filename.

[0052] The technical effects achieved by the above embodiments include: the three-level hierarchical storage structure enables orderly classification and management of files, the directory indexing mechanism improves file retrieval speed, avoids retrieval difficulties caused by messy file storage, significantly improves file retrieval efficiency, and ensures the smooth progress of the transcoding process.

[0053] Traditional technical solutions have the following technical problems: existing voice transcoding methods have limited compatibility with encoding formats, which cannot meet the voice encoding conversion needs of different communication networks and terminals. At the same time, their compatibility with voice quality inspection platforms is insufficient, affecting the overall voice quality inspection process.

[0054] Based on this, the method is compatible with the conversion of three voice coding formats: AMR narrowband, AMR wideband, and A-LAW. The method connects with the provincial IMS aggregation subsystem to forward the transcoded voice files to the voice quality inspection platform, supporting the intelligent identification and control of fraudulent numbers.

[0055] It's worth noting that this method adapts to three mainstream voice coding formats, enabling mutual conversion between different formats. The AMR narrowband coding format is mainly used for ordinary voice calls, with a coding rate ranging from 4.75kbps to 12.2kbps. During transcoding, the coding rate and quantization parameters need to be adjusted according to the target format. The AMR wideband coding format is suitable for high-definition voice services, with a coding rate ranging from 6.60kbps to 23.85kbps. High-definition voice characteristics are prioritized during transcoding. The A-LAW coding format uses non-linear quantization technology with a bit rate of 64kbps. Its quantization rules must be followed during transcoding to ensure that the voice quality is not compromised. The method incorporates transcoding algorithm libraries for the three coding formats, automatically calling the corresponding transcoding algorithm based on the input file's coding format and the target coding format. Simultaneously, this method establishes a standard interface with the provincial IMS aggregation subsystem. After transcoding, the voice file is named and packaged in a unified format and transmitted to the IMS aggregation subsystem via the interface. The IMS aggregation subsystem then forwards the file to the voice quality inspection platform. During transmission, the method carries call information corresponding to the file, such as terminal number, call time, and encoding format, facilitating correlation analysis by the voice quality inspection platform and enabling intelligent identification and control of fraudulent numbers. This method also supports adjusting transcoding output parameters to adapt to the data format requirements of the voice quality inspection platform.

[0056] The technical effects achieved by the above embodiments include: adapting to three mainstream voice coding formats to meet the transcoding needs of different scenarios, seamlessly connecting with the IMS aggregation subsystem and quality inspection platform, realizing seamless connection between transcoding and quality inspection processes, and providing strong support for the identification and control of fraudulent numbers.

[0057] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A cascaded communication voice quality inspection AS voice coding conversion method, based on the IMS network architecture, applied to communication voice quality inspection scenarios, to realize voice coding conversion, including the basic steps of receiving voice sampling files, performing transcoding processing, and outputting transcoding results, characterized in that, Also includes: The system includes a first-level quality inspection AS media server, a second-level quality inspection AS signaling server, a cascading connection module, and an intelligent task scheduling module. The first-level quality inspection AS media server receives the voice sampling file and performs the main transcoding process; the second-level quality inspection AS signaling server receives the file transmitted by the first-level quality inspection AS media server and performs auxiliary transcoding process; the cascading connection module realizes file transfer between the two-level servers; the intelligent task scheduling module dynamically adjusts the processing strategy according to the system load to ensure that the transcoding process is efficient and stable.

2. The AS speech coding and conversion method for communication speech quality inspection based on cascading as described in claim 1, characterized in that, The intelligent task scheduling module includes a dynamic threshold control unit, a task diversion unit, and a multi-threaded parallel processing unit. The dynamic threshold control unit dynamically adjusts the task queue threshold according to the system load, and the threshold calculation formula is: threshold = baseline value × (1 + CPU utilization × 0.6 + queue time × 0.4). When the task queue of the first-level quality inspection AS media server exceeds the set threshold, the task diversion unit automatically diverts the excess tasks to the second-level quality inspection AS signaling server for processing. The multi-threaded parallel processing unit is configured with multiple working threads to realize the parallel execution of transcoding tasks.

3. The AS speech coding and conversion method for communication speech quality inspection based on cascading as described in claim 1, characterized in that, The cascaded connection module uses the SFTP protocol to transfer files. During the transfer process, the data is encrypted using the AES-256 encryption algorithm, and the transmission security is ensured through a dynamic key negotiation mechanism. The SFTP protocol is based on the SSH protocol to build a transmission channel, supporting data integrity verification and breakpoint resume functions during file transfer.

4. The AS speech coding and conversion method for communication speech quality inspection based on cascading as described in claim 1, characterized in that, The method uses a shared memory directory as a temporary storage medium for files, with the shared memory directory path being / dev / shm / . The shared memory directory directly maps to physical memory space, enabling millisecond-level read and write operations for files and avoiding performance loss caused by disk I / O.

5. The AS speech coding and conversion method for communication speech quality inspection based on cascading as described in claim 1, characterized in that, The first-level quality inspection AS media server captures VoLTE call voice sampling files through the IFC rules in the IMS network. The IFC rules are pre-configured in the S-CSCF network element. When a VoLTE call is established, the S-CSCF triggers the call media stream to the first-level quality inspection AS media server according to the IFC rules to ensure accurate acquisition of voice sampling files.

6. The AS speech coding conversion method for communication speech quality inspection based on cascading as described in claim 1, characterized in that, Before performing auxiliary transcoding processing, the second-level quality inspection AS signaling server first performs integrity verification on the SIP signaling associated with the received file; the verification content includes call initiation information, call duration information, and terminal identification information in the SIP signaling, to ensure that the transcoded file accurately matches the corresponding call scenario.

7. The AS speech coding and conversion method for communication speech quality inspection based on cascading as described in claim 2, characterized in that, The task queue baseline threshold of the first-level quality inspection AS media server is set to 1000, and the task queue baseline threshold of the second-level quality inspection AS signaling server is set to 3000. The dynamic threshold control unit collects the CPU utilization and task queue time of the two-level servers in real time and dynamically adjusts the actual threshold according to the set formula.

8. The AS speech coding and conversion method for communication speech quality inspection based on cascading as described in claim 2, characterized in that, The first-level quality inspection AS media server and the second-level quality inspection AS signaling server are each configured with 20 worker threads. The worker threads adopt a thread pool management mechanism, and idle threads remain in a ready state. When a transcoding task arrives, they respond immediately to avoid resource consumption caused by thread creation and destruction.

9. The AS speech coding and conversion method for communication speech quality inspection based on cascading as described in claim 4, characterized in that, The shared memory directory adopts a hierarchical storage structure, with the first-level directory divided by call date, the second-level directory divided by terminal number range, and the third-level directory divided by call time period; The hierarchical storage structure enables fast file retrieval and categorized management, improving file access efficiency.

10. The AS speech coding conversion method for communication speech quality inspection based on cascading as described in claim 1, characterized in that, The method is compatible with the conversion of three voice coding formats: AMR narrowband, AMR wideband, and A-LAW. By interfacing with the provincial IMS aggregation subsystem, the method forwards the transcoded voice files to the voice quality inspection platform, supporting the intelligent identification and control of fraudulent numbers.

Images