A method and apparatus for determining the result of a SIP call based on joint analysis of signaling and media

Abstract

The application relates to a judgment method and device for SIP call result based on signaling and media joint analysis, belongs to the communication technical field, and solves the call result judgment problem; the method comprises the following steps: obtaining SIP signaling and RTP media data; converting original media flow into a media file by capturing and analyzing the RTP media data; extracting features from the media file and generating a corresponding text file; obtaining SIP signaling messages by capturing and analyzing the SIP signaling; dividing stages of the call according to the SIP messages; combining the SIP signaling messages and the text file information extracted from the RTP media data to analyze and judge in each stage, obtaining the call result, and accurately positioning the call failure cause. The application improves the accuracy of the call result judgment by segmentally analyzing the SIP signaling process and combining and analyzing the SIP signaling and the RTP media flow.

Description

Technical Field

[0001] This invention belongs to the field of communication technology, specifically relating to a method and apparatus for determining SIP call results based on the joint analysis of signaling and media. Background Technology

[0002] SIP protocol is a signaling control protocol widely used in voice call methods such as IMS, VoLTE, and VoNR. The signaling monitoring system collects and analyzes signaling between interfaces in the aforementioned communication network to reconstruct the entire process of calls, registrations, and other services. The signaling monitoring system generates a Call Detail Record (CDR) for a call by parsing the signaling protocol. The recorded fields include: start and end time, calling number, called number, source network element address, destination network element address, call duration, call result, and release reason. The CDR allows for convenient and direct understanding of the call process. Except for the call result field, all other fields are directly analyzed and judged using signaling direction, parameters, and time. The call result field, however, is indirectly calculated and analyzed.

[0003] The call result field, especially the call loss result determination for abnormal call processes, is the most important field in CDR. It can directly reflect whether a call was successful or failed, and if it failed, what type of failure it was. The call result field requires a comprehensive judgment of the sequence, direction, signaling cause value, message parameters, and other information of all signaling messages throughout the entire call process.

[0004] Call results can provide network maintenance personnel with information on network quality, end-user call quality, and customer service personnel with information to answer end-user call complaints.

[0005] Current SIP call result assessments only include a few simple categories, typically including call success, called party busy, connection unavailable, erroneous call, and call timeout. This limited range of assessments only meets the basic needs of network maintenance and customer service personnel, failing to accurately reflect the call result status type. Current assessments rely entirely on signaling messages to determine the message flow of a call, utilizing SIP error codes for analysis. While the SIP protocol defines various error codes, these codes are sometimes unclear or have multiple meanings. This can lead to misunderstandings or incorrect determinations of the cause of call failure. SIP calls may involve communication between multiple networks and operators, increasing the complexity of cross-network assessments. Different operators may interpret the protocol differently, resulting in different error codes for a single abnormal call. Therefore, relying solely on signaling to determine call results can lead to inconsistencies.

[0006] In summary, while existing technologies and tools can help determine SIP call outcomes, they still have limitations, such as being insufficiently comprehensive, only allowing for broad call outcome determinations, ambiguity in signaling error codes, and inconsistencies in determining cross-network issues. Therefore, ensuring accurate call outcome determination requires comprehensive consideration of these factors and the adoption of appropriate methods and strategies. Summary of the Invention

[0007] In view of the above analysis, the present invention aims to disclose a method and apparatus for determining SIP call results based on the joint analysis of signaling and media; by performing segmented analysis of the SIP signaling process and by combining the analysis of SIP signaling and RTP media streams, the accuracy of call result determination is improved.

[0008] This invention discloses a method for determining SIP call results based on joint analysis of signaling and media, comprising:

[0009] Step S1: Obtain SIP signaling and RTP media data;

[0010] Step S2: By capturing and analyzing the RTP media data, the original media stream is converted into a media file; features are extracted from the media file and corresponding text files are generated;

[0011] Step S3: Obtain SIP signaling messages by capturing and parsing the SIP signaling; divide the call into stages according to the SIP signaling messages;

[0012] Step S4: In each stage, the SIP signaling messages and the text file information extracted from the RTP media data are combined for analysis and judgment to obtain the call result and accurately locate the cause of the call failure.

[0013] Further, step S2 includes:

[0014] Step S201: Capture and analyze the real-time RTP media data provided by the operator's network element using a media probe; encode the voice data in the media stream, including abnormal notification tones, into audio formats including WAV and MP3, and save them as audio media files;

[0015] Step S202: Extract features from the media file to generate a text file containing additional information, including punctuation marks and timestamps.

[0016] Further, step S202 includes:

[0017] 1) Extract the audio signal from the audio media file and convert it into a Mel spectrum as input features;

[0018] 2) The encoder is used to process the Mel spectrum of the input representation to extract features and generate a latent representation;

[0019] The Mel spectrum is input into the Transformer-based model for encoding, generating the encoded latent representation:

[0020] 3) An autoregressive decoder is used to progressively generate text sequences from the generated implicit representation;

[0021] At each step of the autoregressive decoding, a new text token is generated based on the output of the previous decoder and the encoder's hidden representation Z; and a cross-attention mechanism is used to ensure that the generated text is consistent with the audio content.

[0022] Furthermore, in step S4, the call is divided into three stages using step S3: call initiation, pre-ringing stage, ringing stage, and answering stage; in each stage, the call result is determined based on the caller's call perception, whether they hear a ringback tone or a caller ID tone, and whether they have already spoken with the called party.

[0023] Furthermore, the judgment of each call result in the pre-ringing stage before the call begins includes:

[0024] 1) For call results where the user does not exist or has an empty number, is absent, is busy, has no service on the called side, or has service restrictions or call restrictions, the judgment and classification are based on the obtained SIP signaling and parameters as the signaling judgment basis and the keywords identified in the text as the media judgment basis. The two are combined or used alone as the basis for call result classification.

[0025] 2) For call results of call cancellation / early release of the calling party, request timeout, and network failure, the judgment and classification are based on the obtained SIP signaling and parameters as the signaling judgment criteria;

[0026] 3) For call results of service interruption and overdue fees, the keywords identified in the text are used as the basis for media judgment and classification.

[0027] Furthermore, the judgment of each call result during the ringing phase includes:

[0028] 1) For call results that are rejected or do not respond after a long time, the judgment and classification are based on the obtained SIP signaling and parameters as the signaling judgment basis and the keywords identified in the text as the media judgment basis. The two are combined or used alone as the basis for call result classification.

[0029] 2) For call results of call waiting and ringing release, the SIP signaling and parameters obtained are used as the basis for judgment and classification.

[0030] Furthermore, the judgment of each call result included in the response phase includes:

[0031] 1) For call results that are held or dropped, the judgment and classification are based on the obtained SIP signaling and parameters as the signaling judgment basis and the keywords identified in the text as the media judgment basis. The two are combined or used alone as the basis for call result classification.

[0032] 2) For call results with a bang response, the SIP signaling and parameters obtained are used as the basis for judgment and classification.

[0033] This invention also discloses a determination device for SIP call results based on joint analysis of signaling and media, comprising:

[0034] Carrier network elements, RTP media probes, voice-to-text servers, and signaling probes;

[0035] The operator network element is used to provide SIP signaling and RTP media data;

[0036] The RTP media probe is used to process the media data of abnormal notification sounds and convert the raw media stream into a media file.

[0037] The speech-to-text server is used to process the RTP audio stream in the media file output by the RTP media probe, map it into a text sequence, and output a text file.

[0038] The signaling probe is used to monitor and analyze SIP signaling; capture and parse SIP signaling messages transmitted over the network to analyze the establishment, maintenance, and release processes of calls; divide the call into different stages based on the SIP signaling messages; receive text files converted by a speech-to-text server; and analyze and locate call results in different SIP call stages by combining SIP signaling message parameters and text files, obtain call results, and accurately locate the cause of call failure before outputting them to the user.

[0039] Furthermore, the speech-to-text server includes: a feature extraction module, an encoder module, and a decoder module;

[0040] The feature extraction module is used to extract audio signals from the audio media files output by the RTP media probe and convert them into Mel spectrum as input features.

[0041] The encoder module is used to process the Mel spectrum of the input representation, encode it based on the Transformer model, and generate a latent representation after feature extraction;

[0042] The decoder module is used to perform autoregressive decoding on the generated implicit representation to gradually generate a text sequence;

[0043] At each step of the autoregressive decoding, a new text token is generated based on the output of the previous decoder and the encoder's hidden representation Z; and a cross-attention mechanism is used to ensure that the generated text is consistent with the audio content.

[0044] Furthermore, while capturing and parsing SIP signaling, the signaling probe also receives media data captured by the RTP media probe, which is converted into text files by a speech-to-text server; and combines the SIP signaling messages and text files to perform call result analysis and location.

[0045] Using the SIP call pre- and post-call process, a call is divided into three stages: call initiation, pre-ringing stage, ringing stage, and answering stage. At each stage, the call result is judged based on the caller's call perception, whether they hear a ringback tone or a caller ID tone, and whether they have already spoken with the called party.

[0046] The present invention can achieve the following beneficial effects:

[0047] 1) This invention reasonably combines signaling and media analysis to determine call results, enabling the precise determination of call outcomes that were previously impossible using only signaling analysis. This allows for accurate conclusions regarding previously unidentifiable call failures, providing a more accurate and intuitive representation of call results for communication network maintenance professionals.

[0048] 2) Currently, the industry only has a few simple call result judgments. This invention divides the call into three stages for call result judgment through signaling analysis, avoiding confusion in the judgment results. It also subdivides the call results into dozens of types, effectively allowing network maintenance personnel and customer service personnel to obtain more detailed call failure conclusions and locate the problem. This is beneficial for daily network operation and maintenance, evaluating network call quality, and can better guarantee network services for users. Attached Figure Description

[0049] The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Throughout the drawings, the same reference numerals denote the same parts.

[0050] Figure 1 This is a flowchart of the method for determining SIP call results based on joint analysis of signaling and media in an embodiment of the present invention;

[0051] Figure 2 This is a schematic diagram of the SIP call process segmentation in an embodiment of the present invention;

[0052] Figure 3 This is a schematic diagram showing the connection of the determination device for SIP call results based on the joint analysis of signaling and media in an embodiment of the present invention. Detailed Implementation

[0053] Preferred embodiments of the present invention will now be described in detail with reference to the accompanying drawings, which form part of this application and, together with the embodiments of the present invention, serve to illustrate the principles of the present invention.

[0054] Example 1

[0055] One embodiment of the present invention discloses a method for determining SIP call results based on joint analysis of signaling and media, such as... Figure 1 As shown, it includes:

[0056] Step S1: Obtain SIP signaling and RTP media data;

[0057] Step S2: By capturing and analyzing the RTP media data, the original media stream is converted into a media file; features are extracted from the media file and corresponding text files are generated;

[0058] Step S3: Obtain SIP signaling messages by capturing and parsing the SIP signaling; divide the call into stages according to the SIP signaling messages;

[0059] Step S4: In each stage, the SIP signaling messages and the text file information extracted from the RTP media data are combined for analysis and judgment to obtain the call result and accurately locate the cause of the call failure.

[0060] Specifically, in step S1, the obtained SIP signaling and RTP media data are provided by the operator's network element. The data provided by the operator's network element includes data on voice call methods such as fixed network IMS, VoLTE, and VoNR.

[0061] Specifically, step S2 includes:

[0062] Step S201: Capture and analyze the real-time RTP media data provided by the operator's network element using a media probe; encode the voice data in the media stream, including abnormal notification tones, into audio formats including WAV and MP3, and save them as audio media files;

[0063] The media probe is mainly used to process the media data of abnormal notification sounds. The media probe can convert the original media stream into a media file. In this process, the media probe encodes the voice data in the media data into an appropriate audio format and saves it as an audio file, such as WAV, MP3 and other formats.

[0064] Step S202: Extract features from the media file to generate a text file containing additional information, including punctuation marks and timestamps;

[0065] Specifically, it includes:

[0066] 1) Extract the audio signal from the audio media file and convert it into a Mel spectrum as input features;

[0067] The audio signal S(t) is converted into a Mel-spectrogram, which is then used as the input representation:

[0068] X = Mel-Spectrogram(S(t));

[0069] Where S(t) represents RTP audio data and X represents the Mel spectrum representation;

[0070] 2) The encoder is used to process the Mel spectrum of the input representation to extract features and generate a latent representation;

[0071] The Mel spectrum X is input into the Transformer-based model for encoding, generating the encoded latent representation Z:

[0072] The encoding process can be represented by the following formula:

[0073] Z = Encoder(X);

[0074] 3) An autoregressive decoder is used to progressively generate text sequences from the generated implicit representation;

[0075] In each step of the autoregressive decoding, a new text token is generated based on the output of the previous decoder and the encoder's hidden representation Z; and a cross-attention mechanism is used to ensure that the generated text is consistent with the audio content.

[0076] The decoding process can be represented by the following formula:

[0077] T(t) = Decoder(Z).

[0078] Specifically, in step S3, the SIP signaling is captured and parsed using a signaling probe to obtain SIP signaling messages;

[0079] The signaling probe is a device used to monitor and analyze SIP signaling. It can capture and parse SIP signaling messages transmitted over the network in order to analyze the call establishment, maintenance, and release processes.

[0080] In this embodiment, the signaling probe analyzes both media files and text files to determine the call result. The signaling probe then sends the call result to upper-layer application users, such as network maintenance personnel and customer service personnel.

[0081] Specifically, in step S4, the call is divided into three stages using step S3: call start, pre-ringing stage, ringing stage, and answering stage; in each stage, the call result is judged based on the caller's call perception, whether a ringback tone or caller ID tone is heard, and whether a conversation has already taken place with the called party.

[0082] like Figure 2 As shown, in the three phases of a single call,

[0083] Initiating the call, the pre-ringing stage;

[0084] During this phase, the calling user sends a SIP INVITE request to the called user, requesting the establishment of a call connection. This phase also includes the exchange of SDP (Session Description Protocol) information, used to negotiate the call's media parameters and codec information.

[0085] Ringing Phase: Once the called party accepts the call invitation, the called terminal will return a 180 Ring message, indicating that the called terminal has rung and will send a ringback tone back to the calling party or the calling party will hear a custom ringback tone. This indicates that the call has been connected and the routing path from the caller end to each node has been established.

[0086] Answering Phase: After the ringing, the called party answers or presses the answer button, and the call officially begins. During the call, the calling and called parties exchange real-time audio and video media data via the RTP (Real-time Transport Protocol).

[0087] RTP can transmit both call media data between the calling and called parties, as well as call failure notification tones. When an RTP notification tone is delivered to the calling user, it is typically used to inform the user of a new event or to remind the user to perform certain actions. These notification tones can be short sound clips, such as a "beep" or "beep," or custom voice prompts, such as "The number you dialed is busy" or "The number you dialed is switched off." In general, RTP notification tones play an important role in real-time communication systems, helping users to promptly understand important information, call status, or event reminders, and improving the user experience.

[0088] Call results are judged separately according to three call stages, which can effectively avoid confusion in the judgment of different types of results, and make each call result clearer and more intuitive.

[0089] Specifically, the judgment of each call result in the pre-ringing stage before the call begins includes:

[0090] 1) For call results where the user does not exist or has an empty number, is absent, is busy, has no service on the called side, or has service restrictions or call restrictions, the judgment and classification are based on the obtained SIP signaling and parameters as the signaling judgment basis and the keywords identified in the text as the media judgment basis. The two are combined or used alone as the basis for call result classification.

[0091] 2) For call results of call cancellation / early release of the calling party, request timeout, and network failure, the judgment and classification are based on the obtained SIP signaling and parameters as the signaling judgment criteria;

[0092] 3) For call results of service interruption and overdue fees, the keywords identified in the text are used as the basis for media judgment and classification.

[0093] The classification and judgment criteria for call results in the initial call and pre-ringing stages are shown in Table 1:

[0094] Table 1

[0095]

[0096]

[0097]

[0098] Specifically, the judgment of each call result during the ringing phase includes:

[0099] 1) For call results that are rejected or do not respond after a long time, the judgment and classification are based on the obtained SIP signaling and parameters as the signaling judgment basis and the keywords identified in the text as the media judgment basis. The two are combined or used alone as the basis for call result classification.

[0100] 2) For call results of call waiting and ringing release, the SIP signaling and parameters obtained are used as the basis for judgment and classification.

[0101] The table below shows the classification and judgment criteria for call results during the ringing phase:

[0102] Table 2

[0103]

[0104] Specifically, the judgment of each call result in the response phase includes:

[0105] 1) For call results that are held or dropped, the judgment and classification are based on the obtained SIP signaling and parameters as the signaling judgment basis and the keywords identified in the text as the media judgment basis. The two are combined or used alone as the basis for call result classification.

[0106] 2) For call results with a bang response, the SIP signaling and parameters obtained are used as the basis for judgment and classification.

[0107] The table below shows the classification and judgment criteria for call results during the response phase:

[0108] Table 3

[0109]

[0110] In summary, the embodiments of the present invention reasonably combine signaling and media analysis to determine call results. This enables the determination of call results that were previously impossible to accurately determine using only signaling, through media file analysis. This provides accurate conclusions for previously unidentifiable call failures, allowing communication call results to be presented more accurately and intuitively to communication network maintenance professionals.

[0111] Currently, the industry only offers a few simple call result judgment methods. This invention uses signaling analysis to divide the call into three stages for call result judgment, avoiding confusion in the judgment results. It also subdivides the call results into dozens of types, effectively allowing network maintenance personnel and customer service personnel to obtain more detailed call failure conclusions and pinpoint the location. This is beneficial for daily network operation and maintenance, evaluating network call quality, and providing better network service guarantees for users.

[0112] Example 2

[0113] One embodiment of the present invention discloses a determination device for SIP call results based on joint analysis of signaling and media, such as... Figure 3 As shown, it includes: operator network elements, RTP media probes, voice-to-text servers, and signaling probes;

[0114] Carrier network elements, RTP media probes, voice-to-text servers, and signaling probes;

[0115] The operator network element is used to provide SIP signaling and RTP media data;

[0116] The RTP media probe is used to process the media data of abnormal notification sounds and convert the raw media stream into a media file.

[0117] The speech-to-text server is used to process the RTP audio stream in the media file output by the RTP media probe, map it into a text sequence, and output a text file.

[0118] The signaling probe is used to monitor and analyze SIP signaling; capture and parse SIP signaling messages transmitted over the network to analyze the establishment, maintenance, and release processes of calls; divide the call into different stages based on the SIP signaling messages; receive text files converted by a speech-to-text server; and analyze and locate call results in different SIP call stages by combining SIP signaling message parameters and text files, obtain call results, and accurately locate the cause of call failure before outputting them to the user.

[0119] Specifically, the data provided by the operator network element includes SIP signaling and RTP media data for voice call methods including fixed-line IMS, VoLTE, and VoNR.

[0120] Specifically, the RTP media probe captures and analyzes real-time RTP media data provided by the operator's network element; it encodes the voice data in the media stream, including abnormal notification tones, into audio formats including WAV and MP3, and saves them as audio media files.

[0121] Specifically, the speech-to-text server includes: a feature extraction module, an encoder module, and a decoder module;

[0122] The feature extraction module is used to extract audio signals from the audio media files output by the RTP media probe and convert them into Mel spectrum as input features.

[0123] The audio signal S(t) is converted into a Mel-spectrogram, which is then used as the input representation:

[0124] X = Mel-Spectrogram(S(t));

[0125] Where S(t) represents RTP audio data and X represents the Mel spectrum representation;

[0126] The encoder module is used to process the Mel spectrum of the input representation, encode it based on the Transformer model, and generate a latent representation after feature extraction;

[0127] The Mel spectrum X is input into the Transformer-based model for encoding, generating the encoded latent representation Z:

[0128] The encoding process can be represented by the following formula:

[0129] Z = Encoder(X);

[0130] The decoder module is used to perform autoregressive decoding on the generated implicit representation to gradually generate a text sequence;

[0131] At each step of the autoregressive decoding, a new text token is generated based on the output of the previous decoder and the encoder's hidden representation Z; and a cross-attention mechanism is used to ensure that the generated text is consistent with the audio content.

[0132] The decoding process can be represented by the following formula:

[0133] T(t) = Decoder(Z).

[0134] Specifically, the signaling probe is a device used to monitor and analyze SIP signaling. It can capture and parse SIP signaling messages transmitted over the network in order to analyze the call establishment, maintenance, and release processes.

[0135] In this embodiment, while capturing and parsing SIP signaling, the signaling probe also receives media data captured by the RTP media probe, which is converted into text files by a speech-to-text server; and combines the SIP signaling messages and text files to perform call result analysis and location.

[0136] Using the SIP call pre- and post-call process, a call is divided into three stages: call initiation, pre-ringing stage, ringing stage, and answering stage. At each stage, the call result is judged based on the caller's call perception, whether they hear a ringback tone or a caller ID tone, and whether they have already spoken with the called party.

[0137] Specifically, the judgment of each call result in the pre-ringing stage before the call begins includes:

[0138] 1) For call results where the user does not exist or has an empty number, is absent, is busy, has no service on the called side, or has service restrictions or call restrictions, the judgment and classification are based on the obtained SIP signaling and parameters as the signaling judgment basis and the keywords identified in the text as the media judgment basis. The two are combined or used alone as the basis for call result classification.

[0139] 2) For call results of call cancellation / early release of the calling party, request timeout, and network failure, the judgment and classification are based on the obtained SIP signaling and parameters as the signaling judgment criteria;

[0140] 3) For call results of service interruption and overdue fees, the keywords identified in the text are used as the basis for media judgment and classification.

[0141] The judgment of each call result during the ringing phase includes:

[0142] 1) For call results that are rejected or do not respond after a long time, the judgment and classification are based on the obtained SIP signaling and parameters as the signaling judgment basis and the keywords identified in the text as the media judgment basis. The two are combined or used alone as the basis for call result classification.

[0143] 2) For call results of call waiting and ringing release, the SIP signaling and parameters obtained are used as the basis for judgment and classification.

[0144] The judgment of each call result included in the response phase includes:

[0145] 1) For call results that are held or dropped, the judgment and classification are based on the obtained SIP signaling and parameters as the signaling judgment basis and the keywords identified in the text as the media judgment basis. The two are combined or used alone as the basis for call result classification.

[0146] 2) For call results with a bang response, the SIP signaling and parameters obtained are used as the basis for judgment and classification.

[0147] For more specific technical details in this embodiment, please refer to Embodiment 1.

[0148] In summary, the embodiments of the present invention reasonably combine signaling and media analysis to determine call results. This enables the determination of call results that were previously impossible to accurately determine using only signaling, through media file analysis. This provides accurate conclusions for previously unidentifiable call failures, allowing communication call results to be presented more accurately and intuitively to communication network maintenance professionals.

[0149] Currently, the industry only offers a few simple call result judgment methods. This invention uses signaling analysis to divide the call into three stages for call result judgment, avoiding confusion in the judgment results. It also subdivides the call results into dozens of types, effectively allowing network maintenance personnel and customer service personnel to obtain more detailed call failure conclusions and pinpoint the location. This is beneficial for daily network operation and maintenance, evaluating network call quality, and providing better network service guarantees for users.

[0150] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for determining SIP call results based on joint analysis of signaling and media, characterized in that, include: Step S1: Obtain SIP signaling and RTP media data; Step S2: By capturing and analyzing the RTP media data, the original media stream is converted into a media file; Extract features from media files and generate corresponding text files; Step S3: Obtain SIP signaling messages by capturing and parsing the SIP signaling; The call will be phased according to SIP signaling messages; Step S4: In each stage, the SIP signaling messages and the text file information extracted from the RTP media data are combined for analysis and judgment to obtain the call result and accurately locate the cause of the call failure. In step S4, the call is divided into three stages using the method described in step S3: call initiation, pre-ringing stage, ringing stage, and answering stage. At each stage, the call result is determined based on the caller's call perception, whether they hear a ringback tone or a caller ID tone, and whether they have already spoken with the called party. The judgment of each call result in the pre-ringing stage before the call begins includes: 1) For call results where the user does not exist or has an empty number, is absent, is busy, has no service on the called side, or has service restrictions or call restrictions, the judgment and classification are based on the obtained SIP signaling and parameters as the signaling judgment basis and the keywords identified in the text as the media judgment basis. The two are combined or used alone as the basis for call result classification. 2) For call results of call cancellation / early release of the calling party, request timeout, and network failure, the judgment and classification are based on the obtained SIP signaling and parameters as the signaling judgment criteria; 3) For call results of service interruption and overdue fees, the keywords identified in the text are used as the basis for media judgment and classification.

2. The method for determining SIP call results based on joint analysis of signaling and media as described in claim 1, characterized in that, Step S2 includes: Step S201: Capture and analyze the real-time RTP media data provided by the operator's network element using a media probe; encode the voice data in the media stream, including abnormal notification tones, into audio formats including WAV and MP3, and save them as audio media files; Step S202: Extract features from the media file to generate a text file containing additional information, including punctuation marks and timestamps.

3. The method for determining SIP call results based on joint analysis of signaling and media as described in claim 2, characterized in that, Step S202 includes: 1) Extract the audio signal from the audio media file and convert it into a Mel spectrum as input features; 2) The encoder is used to process the Mel spectrum of the input representation to extract features and generate a latent representation; The Mel spectrum is input into the Transformer-based model for encoding, generating the encoded latent representation: 3) An autoregressive decoder is used to progressively generate text sequences from the generated implicit representation; At each step of the autoregressive decoding, a new text token is generated based on the output of the previous decoder and the encoder's hidden representation Z; and a cross-attention mechanism is used to ensure that the generated text is consistent with the audio content.

4. The method for determining SIP call results based on joint analysis of signaling and media as described in claim 1, characterized in that, The ringing phase includes the following judgments on the results of each call: 1) For call results that are rejected or do not respond after a long time, the judgment and classification are based on the obtained SIP signaling and parameters as the signaling judgment basis and the keywords identified in the text as the media judgment basis. The two are combined or used alone as the basis for call result classification. 2) For call results of call waiting and ringing release, the SIP signaling and parameters obtained are used as the basis for judgment and classification.

5. The method for determining SIP call results based on joint analysis of signaling and media as described in claim 1, characterized in that, The judgment of each call result in the response phase includes: 1) For call results that are held or dropped, the judgment and classification are based on the obtained SIP signaling and parameters as the signaling judgment basis and the keywords identified in the text as the media judgment basis. The two are combined or used alone as the basis for call result classification. 2) For call results with a bang response, the SIP signaling and parameters obtained are used as the basis for judgment and classification.

6. A device for determining SIP call results based on joint analysis of signaling and media, implementing the determination method for SIP call results based on joint analysis of signaling and media as described in any one of claims 1-5, characterized in that, include: Carrier network elements, RTP media probes, voice-to-text servers, and signaling probes; The operator network element is used to provide SIP signaling and RTP media data; The RTP media probe is used to process the media data of abnormal notification sounds and convert the raw media stream into a media file. The speech-to-text server is used to process the RTP audio stream in the media file output by the RTP media probe, map it into a text sequence, and output a text file. The signaling probe is used to monitor and analyze SIP signaling; It captures and parses SIP signaling messages transmitted over the network to analyze the call establishment, maintenance, and release process; divides the call into different stages based on the SIP signaling messages; receives text files converted by a voice-to-text server; and analyzes and locates call results in different SIP call stages by combining SIP signaling message parameters and text files, obtains call results, and accurately locates the reasons for call failures before outputting them to the user.

7. The determination device for SIP call results based on joint analysis of signaling and media according to claim 6, characterized in that, The speech-to-text server includes: a feature extraction module, an encoder module, and a decoder module; The feature extraction module is used to extract audio signals from the audio media files output by the RTP media probe and convert them into Mel spectrum as input features. The encoder module is used to process the Mel spectrum of the input representation, encode it based on the Transformer model, and generate a latent representation after feature extraction; The decoder module is used to perform autoregressive decoding on the generated implicit representation to gradually generate a text sequence; At each step of the autoregressive decoding, a new text token is generated based on the output of the previous decoder and the encoder's hidden representation Z; and a cross-attention mechanism is used to ensure that the generated text is consistent with the audio content.

8. The determination device for SIP call results based on joint analysis of signaling and media according to claim 6, characterized in that, While capturing and parsing SIP signaling, the signaling probe also receives media data captured by the RTP media probe, which is converted into text files by the voice-to-text server; and combines the SIP signaling messages and text files to analyze and locate call results. A call can be divided into three phases using the SIP call pre- and post-call process: call initiation, pre-ringing phase, ringing phase, and answering phase. At each stage, the call result is determined based on the caller's call perception, whether they hear a ringback tone or a caller ID tone, and whether they have already spoken with the called party.

Images