Call bill retrieval method and device, equipment, medium and product

Abstract

The invention relates to the technical field of communication, and particularly provides a ticket retrieval method and device, equipment, a medium and a product. The method comprises the following steps: acquiring voiceprint features of a to-be-retrieved object; based on the voiceprint features of the to-be-retrieved object and a pre-trained voiceprint clustering model, performing fuzzy retrieval in the candidate call tickets to obtain a target call ticket set; and screening based on the target call ticket set to obtain a target call ticket. The candidate call bills are subjected to fuzzy retrieval through the pre-trained voiceprint clustering model, so that the search range for searching the target call bills can be remarkably narrowed, that is, the voiceprint clustering model does not need to be accurately screened and only needs to be subjected to clustering analysis of voiceprint features, and the call bill troubleshooting difficulty can be reduced while the complexity of model calculation is reduced; on the basis, secondary screening is further carried out on the target phone bill set based on the voiceprint features, and the target phone bill can be accurately searched. Therefore, the call ticket retrieval accuracy and retrieval efficiency can be improved.

Description

Technical Field

[0001] This disclosure belongs to the field of communication technology, specifically relating to a call detail record (CDR) retrieval method, apparatus, device, medium, and product. Background Technology

[0002] An intermediate number is a temporary or dedicated number generated through technical means to establish a communication connection between a user and a third party while concealing the real phone numbers of both parties. In recent years, with the rise of telecommunications fraud, scammers often use intermediate numbers to communicate with users.

[0003] Generally, when a user files a complaint, the voiceprints of the complaint call and historical calls can be compared for similarity to retrieve the corresponding call records. However, this retrieval method suffers from poor performance and low efficiency when the voiceprint database is large, thus reducing the accuracy of call detail record (CDR) retrieval. Summary of the Invention

[0004] This disclosure addresses some of the deficiencies mentioned in the background art by providing a call detail record (CDR) retrieval method, apparatus, device, medium, and product that can improve the accuracy of CDR retrieval.

[0005] In a first aspect, embodiments of this disclosure provide a call detail record (CDR) retrieval method, comprising: Obtain the voiceprint features of the object to be retrieved; Based on the voiceprint features of the object to be retrieved and the pre-trained voiceprint clustering model, fuzzy retrieval is performed in the candidate call records to obtain the target call record set. The target call detail records (CDRs) are obtained by filtering based on the target CDR set.

[0006] Optionally, the step of performing fuzzy retrieval in candidate call detail records (CDRs) based on the voiceprint features of the object to be retrieved and a pre-trained voiceprint clustering model to obtain a target CDR set includes: The voiceprint clustering model is used to perform voiceprint clustering analysis on the voiceprint features to obtain the voiceprint clustering results of the object to be retrieved. The voiceprint clustering results are used to indicate the target call detail record set.

[0007] Optionally, the object to be searched includes at least one of the following: current call, historical abnormal call; The acquisition of the voiceprint features of the object to be retrieved includes at least one of the following: When the object to be retrieved is the current call, extract the voiceprint features from the current call; When the object to be retrieved is a historical abnormal call, the voiceprint features in the corresponding record of the historical abnormal call are obtained.

[0008] Optionally, the method further includes: If the current call meets the preset voiceprint recording rules, extract and record the voiceprint features in the current call; The voiceprint recording rules include at least one of the following: The historical call behavior of the caller in the current call is abnormal; The current call received a call forwarding signaling message but did not carry call forwarding information; The current call did not receive call transfer signaling, and the location of the initially called network element is different from the location of the network element of the recipient in the current call; The user core network identifier carried in any call signaling of the current call is inconsistent with the user core network identifier of the initiator or the user core network identifier of the current call.

[0009] Optionally, the method further includes: Obtain voiceprint feature samples from historical call records; The voiceprint clustering model is trained using the voiceprint feature samples to obtain the voiceprint clustering model; The input of the voiceprint clustering model includes voiceprint features; the output of the voiceprint clustering model includes at least one of the following: a set of call detail records (CDRs) and clustering tags of the CDR sets.

[0010] Optionally, obtaining the voiceprint features of the object to be retrieved includes: Obtain the audio file of the object to be retrieved; The audio file is subjected to sound activity detection to obtain voiceprint information carrying human voice. The voiceprint information carrying human voice is separated and processed to obtain human voice voiceprint information; The voiceprint information is extracted to obtain the voiceprint features.

[0011] Optionally, the step of filtering based on the target call detail record (CDR) set to obtain the target CDR includes: The call details in the target call detail record set are filtered out to obtain the target call detail records, wherein the filtering process includes one or more of the following: The audio files of each call detail record (CDR) are transcribed to obtain the text information of each CDR; a similarity comparison is performed between the text information of each CDR and the text information of the object to be retrieved to obtain a similarity result; and the CDRs with similarity results greater than a preset threshold are taken as the target CDRs. Based on the signaling terminal information of the object to be retrieved, the signaling terminal information in each of the call detail records (CDRs) is checked, and the CDRs with consistent signaling terminal information are taken as the target CDRs.

[0012] In a second aspect, embodiments of this disclosure provide a call detail record (CDR) retrieval device, comprising: The acquisition module is used to acquire the voiceprint features of the object to be retrieved; The retrieval module is used to perform fuzzy retrieval in the candidate call records based on the voiceprint features of the object to be retrieved and the pre-trained voiceprint clustering model to obtain the target call record set. The filtering module is used to filter based on the target call detail record set to obtain the target call detail records.

[0013] In a third aspect, embodiments of this disclosure provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described call detail record (CDR) retrieval method.

[0014] In a fourth aspect, embodiments of this disclosure provide a computer-readable storage medium having a computer program stored thereon, the program being executed by a processor to implement the above-described call detail record (CDR) retrieval method.

[0015] In a fifth aspect, embodiments of this disclosure provide a computer program product including computer-readable code, or a non-volatile computer-readable storage medium carrying computer-readable code, wherein when the computer-readable code is run in a processor of an electronic device, the processor in the electronic device executes the above-described call detail record (CDR) retrieval method.

[0016] In this disclosure, the voiceprint features of the object to be retrieved are obtained; based on the voiceprint features of the object to be retrieved and a pre-trained voiceprint clustering model, fuzzy retrieval is performed on candidate call detail records (CDRs) to obtain a set of target CDRs; and the target CDRs are then filtered based on the target CDR set to obtain the target CDRs. By using the pre-trained voiceprint clustering model for fuzzy retrieval of candidate CDRs, the search scope for the target CDRs can be significantly narrowed. That is, the voiceprint clustering model does not require precise filtering, but only needs to perform clustering analysis of voiceprint features, reducing the computational complexity of the model and the difficulty of CDR retrieval. Furthermore, a secondary filtering of the target CDR set based on voiceprint features allows for precise retrieval of the target CDRs. Therefore, the accuracy and efficiency of CDR retrieval can be improved.

[0017] It should be understood that both the foregoing general description and the following detailed description are exemplary and intended to provide further illustration of the claimed technology. Attached Figure Description

[0018] Figure 1 This is a flowchart of a call detail record (CDR) retrieval method provided in this disclosure.

[0019] Figure 2 Another flowchart of a call detail record (CDR) retrieval method provided in this disclosure.

[0020] Figure 3 This is a schematic diagram of the structure of a call detail record (CDR) retrieval device provided in this disclosure.

[0021] Figure 4 This is a hardware block diagram of an electronic device provided in this disclosure.

[0022] Figure 5 This is a schematic diagram of a computer program product provided in this disclosure. Detailed Implementation

[0023] To enable those skilled in the art to better understand the technical solution of this application, the application scenario of this application will be described first below.

[0024] An intermediate number is a temporary or dedicated number generated through technical means to establish a communication connection between a user and a third party while concealing the real phone numbers of both parties. In recent years, the situation of telecommunications fraud has become increasingly serious, with fraudsters often using intermediate numbers to communicate with users. Due to the privacy protection capabilities of intermediate numbers, fraudsters utilize them in various ways to commit fraud, increasing the difficulty of tracing the source. Issues such as irregular behavior of some network elements and historical installation reasons may prevent intermediate number network elements from obtaining certain call information. Alternatively, other reasons may prevent the complaint number from having call records, such as missing call forwarding information or the complainant using a number other than the original called number when making a complaint while using dual SIM dual standby. All of these factors can cause the intermediate number platform to be unable to find the corresponding call records through the complaint information, i.e., it cannot accurately find the call details.

[0025] Taking the lack of call forwarding information as an example, a user complains about number X, but the called party information for B cannot be found in X's call log. This is because intermediate numbers are often reused, meaning there may be a large number of call logs for number X within the complaint period, making it impossible to find the call details simply by searching the number. Furthermore, the complainant may have misremembered number X, resulting in a situation where the user provides the correct number but cannot find the correct number X. In such cases, a voiceprint comparison scheme can be introduced to compare the complainant's voiceprint information with the voiceprint information in relevant records, greatly accelerating the investigation of such problems.

[0026] In related technologies, when a user files a complaint, the similarity between the voiceprints of the complaint call and historical calls can be compared to extract audio feature vector information, i.e., voiceprint features. Modeling methods such as Gaussian models and i-vector models are used to build the model. A similarity score is obtained by comparing the voiceprint features. If multiple voiceprint comparisons exist, they must be compared one by one to retrieve the corresponding call record. However, this retrieval method suffers from poor performance and low execution efficiency when the voiceprint database is large. Alternatively, deep neural networks can be used to replace similarity comparison with vector distance. However, constructing an end-to-end voiceprint matching system using deep neural networks requires a large amount of labeled audio data, resulting in high costs, inconvenient storage, low execution efficiency, and difficulties in subsequent updates. Therefore, this reduces the accuracy of call detail record retrieval.

[0027] To address the aforementioned technical problems, this disclosure provides an inventive concept: based on extracted voiceprint features, fuzzy retrieval of candidate call detail records (CDRs) is performed, thereby narrowing the search scope for the target CDR. In this disclosure, the voiceprint clustering model does not require precise filtering; it only needs to perform clustering analysis of voiceprint features, reducing the computational complexity of the model and lowering the difficulty of CDR retrieval. Furthermore, a secondary filtering of the target CDR set based on voiceprint features can accurately locate the target CDR. Therefore, the accuracy and efficiency of CDR retrieval can be improved.

[0028] The present disclosure will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present disclosure and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to this application / disclosure are shown in the drawings, not the entire structure.

[0029] Figure 1 This is a flowchart of a call detail record (CDR) retrieval method provided in this disclosure. Figure 1 As shown, the method includes: S101: Obtain the voiceprint features of the object to be retrieved.

[0030] Specifically, in this embodiment, the object to be retrieved includes at least one of the following: user alarm calls, user complaint calls, calls between users and business personnel, etc. The object to be retrieved can be a real-time call or a historical call. The audio file of the object to be retrieved can be preprocessed to obtain the voiceprint information of the human voice in the audio file. Feature extraction can be performed on the voiceprint information to obtain the voiceprint features of the object to be retrieved. The extracted voiceprint features are relatively pure human voice information in the audio file that contains the core dialogue content of the call. The voiceprint features obtained in this way can improve the accuracy and matching degree of call detail record retrieval.

[0031] S102: Based on the voiceprint features of the object to be retrieved and the pre-trained voiceprint clustering model, perform fuzzy retrieval in the candidate call records to obtain the target call record set.

[0032] Specifically, the acquired voiceprint features of the target object can be normalized, and the normalized voiceprint features are then input into a pre-trained voiceprint clustering model. The voiceprint clustering model performs fuzzy retrieval on candidate call records (CDRs), outputting a set of target CDRs with high similarity to the voiceprint features of the target object. Specifically, the voiceprint clustering model determines voiceprint features similar to the target object's voiceprint features by measuring the distance and density between the target object's voiceprint features and those in candidate CDRs, grouping similar features into one category to obtain the corresponding set of target CDRs. In this embodiment, the voiceprint clustering model is used for auxiliary filtering, eliminating some obviously different CDRs from the candidate CDRs, thereby narrowing the search range and reducing the difficulty of CDR filtering.

[0033] S103: Filter based on the target call detail record set to obtain the target call detail record.

[0034] Specifically, after selecting the target call detail records (CDRs) from the candidate CDRs, further filtering can be performed to accurately obtain the target CDRs. For example, semantic filtering, terminal configuration information filtering, and user self-selection can be applied to the target CDR set. These filtering methods can be implemented independently or in combination. It is understood that the above filtering methods are merely illustrative, and this disclosure does not impose specific limitations on the filtering methods.

[0035] In this disclosure, the voiceprint features of the object to be retrieved are obtained; based on the voiceprint features of the object to be retrieved and a pre-trained voiceprint clustering model, fuzzy retrieval is performed on candidate call detail records (CDRs) to obtain a set of target CDRs; and the target CDRs are then filtered based on the target CDR set to obtain the target CDRs. By using the pre-trained voiceprint clustering model for fuzzy retrieval of candidate CDRs, the search scope for the target CDRs can be significantly narrowed. That is, the voiceprint clustering model does not require precise filtering, but only needs to perform clustering analysis of voiceprint features, reducing the computational complexity of the model and the difficulty of CDR retrieval. Furthermore, a secondary filtering of the target CDR set based on voiceprint features allows for precise retrieval of the target CDRs. Therefore, the accuracy and efficiency of CDR retrieval can be improved.

[0036] In one possible implementation, an exemplary method for obtaining the voiceprint features of the object to be retrieved includes: The process involves: acquiring the audio file of the object to be searched; performing sound activity detection on the audio file to obtain voiceprint information carrying human voices; separating the voiceprint information carrying human voices to obtain human voice voiceprint information; and extracting the human voice voiceprint information to obtain voiceprint features.

[0037] Specifically, in this embodiment, the preprocessed audio files are typically source files in AMR-WB, AMR-NB, or PCMA formats, and an audio file with a sampling rate of 16kHz is selected for processing. Since subsequent processing for human voice extraction is required, and the main energy of human voices is concentrated in the 85Hz-2.5kHz range (male voice) and 165Hz-2.5kHz range (female voice), a 16kHz sampling rate can completely cover the core frequency band of human voices, avoiding distortion caused by the loss of high-frequency information. Typically, AMR-WB format source files have a sampling rate of 16kHz, while AMR-NB and PCMA format source files have a sampling rate of 8kHz. Therefore, the source files need to be resampled to obtain an audio file with a sampling rate of 16kHz.

[0038] The acquired audio file undergoes sound activity detection. Audio segments with a valid duration exceeding a first preset duration are selected to obtain voiceprint information carrying human voice. The valid duration refers to the selected audio segment being a speech segment, avoiding interference from non-speech segments (such as silence, background noise, or music) on voiceprint information extraction. In this embodiment, the first preset duration can be 15 seconds. Sound activity detection can be implemented using VAD (Voice Activity Detection) technology. Furthermore, if the selected audio exceeds 15 seconds, the first 15 seconds of audio can be extracted to obtain voiceprint information carrying human voice.

[0039] Voiceprint information carrying human voice is a mixed signal including human voice and background noise. Human voice separation processing can be performed on the acquired voiceprint information to obtain pure human voice voiceprint information. In this embodiment, Fastica blind source separation technology can be used to separate independent source signals from the mixed signal. Setting nComponent to 2 allows for the separation of the main human voice voiceprint information and the background noise voiceprint information. The obtained main human voice voiceprint information and background noise voiceprint information are then normalized, and the normal distribution values ​​of the two types of voiceprint information are calculated. Since the main energy of human voice is concentrated in the lower frequency band, the lower score can be selected as the human voice voiceprint information.

[0040] Next, MFCC (Mel-scale Frequency Cepstral Coefficients) features are extracted from the human voiceprint information. The MFCC feature matrix is ​​then normalized, that is, the MFCC feature values ​​are mapped to a fixed range, such as (0,1), to obtain the final voiceprint features. In this embodiment, the librosa library can be used for feature extraction.

[0041] It is understood that the above methods are merely illustrative examples, and this disclosure does not impose any specific limitations on them.

[0042] In one possible implementation, an exemplary method for obtaining a set of target call records by performing fuzzy retrieval in candidate call records based on the voiceprint features of the object to be retrieved and a pre-trained voiceprint clustering model includes: Voiceprint clustering analysis was performed on voiceprint features using a voiceprint clustering model to obtain the voiceprint clustering results of the object to be retrieved.

[0043] Specifically, the voiceprint clustering results are used to indicate the target call detail record (CDR) set. Voiceprint features are input into the voiceprint clustering model, and voiceprint clustering analysis is performed in the voiceprint database constructed from candidate CDRs, thereby outputting voiceprint clustering results that are similar to the voiceprint features of the object to be retrieved.

[0044] In this embodiment, the voiceprint clustering model can use the unsupervised clustering algorithm OPTICS (Ordering Points To Identify the Clustering Structure). The initial value of the minimum number of samples min_simples required to determine whether a voiceprint feature can become a core point is set to 5, and the steepness threshold xi used to define the identification criteria of the cluster boundary in the voiceprint clustering process is set to 0.05. The voiceprint clustering model can be trained based on the above model parameters.

[0045] An example method for pre-training the voiceprint clustering model is shown below: Obtain voiceprint feature samples from historical call records; use the voiceprint feature samples to train the voiceprint clustering model to obtain the voiceprint clustering model.

[0046] Specifically, after obtaining voiceprint feature samples from historical call detail records, a portion of these voiceprint feature samples is randomly selected and input into the voiceprint clustering model. The voiceprint clustering model is then trained based on the initial values ​​of the model parameters determined above. Subsequently, a portion of voiceprint feature samples is randomly selected and input into the voiceprint clustering model multiple times to gradually adjust the minimum number of samples and the steepness threshold.

[0047] In the voiceprint clustering model, the silhouette coefficient is the core indicator for evaluating the clustering effect. It quantifies intra-cluster density and inter-cluster separation. The silhouette coefficient can be used to determine the adjusted clustering effect, and simultaneously records the maximum radius threshold `max_eps` used to control the upper limit of the neighborhood search. The minimum number of samples and the steepness threshold can be adjusted with step sizes of 3 and 0.1 respectively. An example formula for calculating the silhouette coefficient is shown below:

[0048] in, For the profile coefficient, Intra-cluster dissimilarity This represents the dissimilarity between clusters.

[0049] After obtaining the optimal model parameters, the voiceprint feature samples of historical call detail records are input into the voiceprint clustering model for cluster analysis. The voiceprint clustering model outputs at least one of the call detail record set and the clustering label of the call detail record set.

[0050] In one possible implementation, obtaining the voiceprint features of the object to be retrieved includes at least one of the following: When the object to be retrieved is the current call, extract the voiceprint features from the current call; when the object to be retrieved is a historical abnormal call, obtain the voiceprint features from the corresponding records of the historical abnormal call.

[0051] This disclosure can be used for online retrieval of target call detail records (CDRs) in real-time calls, facilitating users to promptly find related CDRs during real-time calls, such as complaint calls and alarm calls, to meet user needs. It can also be used for correlation analysis of historical CDRs to find target CDRs related to a specific historical CDR. Specifically, the retrieval objects include at least one of the following: current calls and historical abnormal calls. For historical abnormal calls, voiceprint features have already been extracted and recorded based on preset voiceprint recording rules during the call. Using historical abnormal calls as the retrieval object allows direct access to the previously extracted voiceprint features. For current calls, it is not necessary to determine whether the current call is abnormal; voiceprint features can be extracted for any current call. Alternatively, it can be determined whether the current call is abnormal based on preset voiceprint recording rules. If the rules are met, voiceprint features are extracted and recorded, thus reducing the waste of resources from extracting voiceprint features for normal calls.

[0052] In one possible implementation, the method further includes: If the current call meets the preset voiceprint recording rules, extract and record the voiceprint features in the current call.

[0053] Specifically, if the current call meets the preset voiceprint recording rules, it means that during the call establishment process, by monitoring the call signaling, it was found that there may be an abnormal situation in the current call. It is necessary to extract and record the voiceprint features in the current call to facilitate the use of voiceprint features for clustering processing during subsequent call detail record retrieval.

[0054] Among them, the voiceprint recording rules include at least one of the following (1) to (4).

[0055] (1) The calling user in the current call has abnormal historical calling behavior.

[0056] Specifically, when making a call using the intermediate number platform, the platform can classify the caller's historical call behavior. This classification result can be used to determine whether the caller has exhibited abnormal call behavior. When a call established in the current session enters the intermediate number platform, the classification result of the caller's historical call behavior can be obtained. If the classification result indicates anomalies, voiceprint features can be extracted and recorded.

[0057] For example, if a calling user's calling behavior shows a pattern of high-frequency calls and low-frequency connections, then it is determined that the calling user has abnormal calling behavior.

[0058] (2) The current call has received a call transfer signaling message but does not carry call transfer information.

[0059] Specifically, if during the call setup process, a 181 signaling message that does not contain the path of call forwarding is received, it indicates that the current call has received call forwarding signaling but not call forwarding information, and the call detail record (CDR) cannot be found using the called number. Therefore, this current call may be abnormal, and voiceprint features can be extracted and recorded.

[0060] (3) The current call has not received a call transfer signaling and the location of the network element of the initial called party is inconsistent with the location of the network element of the current call recipient.

[0061] Specifically, during the current call signaling negotiation process, if no call transfer signaling is received, and it cannot be determined that no call transfer has occurred, it may be because the network element failed to transmit the call transfer signaling back to the intermediate number platform's network element. In this case, it is also necessary to determine whether the location of the initial called party's network element is consistent with that of the actual recipient of the call. The location of the initial called party's network element, i.e., the destination province, can be determined based on the term-ioi charging information field in the p-charing-vector field (carrying the call's billing identifier field) of the called party's response. Furthermore, the location of the current call's network element, i.e., the actual destination province, can be determined based on the sbc-domin information (the network element information closest to C) in the p-access-network-info field of the response. If no call transfer signaling is received, and the destination province of the initial called party and the actual called party are inconsistent, it indicates that the current call may be abnormal, and voiceprint features can be extracted and recorded.

[0062] (4) The user core network identifier carried in any call signaling of the current call is inconsistent with the user core network identifier of the initiator or the user core network identifier of the recipient of the current call.

[0063] Specifically, during the current call signaling negotiation process, if the user core network identifier code (i.e., p-assert-identity information) carried by any call signaling is inconsistent with the user core network identifier code of the initiator or the recipient of the current call, that is, inconsistent with the From or To field in the call signaling, it indicates that the current call may be abnormal, and voiceprint features can be extracted and recorded.

[0064] For example, when the calling user A calls the actual called party C through an intermediate number, call forwarding is set up to hide the real calling user A. Call forwarding is set up at the called party B. At this time, the p-assert-identity information is inconsistent with the To field, so voiceprint features need to be extracted and recorded.

[0065] In one possible implementation, an exemplary method for obtaining target call detail records (CDRs) by filtering based on a target CDR set includes: The call details in the target call detail record set are filtered out to obtain the target call detail records.

[0066] Specifically, as mentioned earlier, fuzzy search using a voiceprint clustering model can narrow down the selection range from candidate call detail records (CDRs) to obtain a target CDR set. The target CDR set includes CDRs with at least one cluster of information. Further filtering from this smaller target CDR set allows for more precise identification of the target CDRs.

[0067] The screening process includes one or more of the following: (1) to (2): (1) Transcribe the audio files of each call detail record (CDR) to obtain the text information of each CDR; compare the similarity between the text information of each CDR and the text information of the target object to obtain the similarity result, and take the CDR with the similarity result greater than the preset threshold as the target CDR.

[0068] Specifically, each call detail record (CDR) in the target call detail record (CDR) set can be transcribed from audio to text to obtain the text information of each CDR. Simultaneously, the audio file of the target object can be transcribed into text. Then, the text information of each CDR can be compared with the text information of the target object to determine the target CDR.

[0069] (2) Based on the signaling terminal information of the object to be retrieved, the signaling terminal information in each call detail record is checked, and the call detail record with consistent signaling terminal information is taken as the target call detail record.

[0070] Specifically, the signaling terminal information of the target call detail record (CDR) and the signaling terminal information of each call detail record (CDR) are obtained. Since the calling or called party of the target call detail record usually matches the calling or called party of the target CDR, if a CDR with matching signaling terminal information is found, it is used as the target CDR. For example, if the target call is a complaint call, and the calling party is the complainant who makes the complaint to customer service, then if there is a CDR in the target CDR set whose calling or called party signaling terminal information matches the complainant's signaling terminal information, that CDR can be used as the target CDR. The signaling terminal information includes, for example, the model of the terminal used in the call.

[0071] In addition, the call time range and number of the target call detail record can be obtained from the content of the audio file of the object to be searched, so as to assist in filtering and thus obtain the target call detail record.

[0072] It is understood that the above screening methods are merely examples, and the above screening methods can be used alone or in combination. This disclosure does not impose any specific limitations on them.

[0073] Figure 2 Another flowchart is provided for a call detail record (CDR) retrieval method according to this disclosure. Figure 2 As shown, the method includes: S201: Perform audio processing on the current call to obtain human voiceprint information.

[0074] Specifically, when the current call is a user's complaint call or a police report call, if the desired call detail record cannot be found through the complaint number X or the user's number, the user's historical audio files or complaint audio files can be used to extract the voiceprint with the user's authorization.

[0075] The system acquires the audio file of the current call and performs various enhancement techniques, including Gaussian filtering, amplitude modulation, frequency modulation, overlaying background vocals, and encoding conversion (PCMA, AMR-WB, etc.). Multiple enhanced audio files are then acquired sequentially. These files are then formatted into a uniform 16K PCM audio format. Sound activity detection is performed on the audio files, selecting valid voiceprint information carrying human voices. Finally, voice separation processing is applied to this voiceprint information to obtain the individual human voiceprint data.

[0076] S202: Extract features from human voiceprint information to obtain voiceprint features.

[0077] Specifically, MFCC features are extracted from human voiceprint information to obtain MFCC features. The MFCC feature matrix is ​​then normalized to obtain the final voiceprint features. The processed voiceprint features can improve the recall rate of subsequent voiceprint matching.

[0078] S203: Use the voiceprint clustering model to perform cluster analysis on voiceprint features and obtain voiceprint clustering results.

[0079] Specifically, voiceprint features are input into a voiceprint clustering model. Cluster analysis is then performed using the voiceprint clustering model and the voiceprint features of candidate call records (CDRs) to achieve fuzzy retrieval within the candidate CDRs, resulting in the voiceprint clustering results, which is the target CDR set. During the previous training of the voiceprint clustering model, audio MFCC features were extracted, and the optics algorithm was used to establish the voiceprint clustering model. Profilometry was then used for parameter tuning, ultimately achieving the goal of establishing a voiceprint database for convenient storage of clustering results, significantly accelerating subsequent voiceprint similarity matching. The voiceprint clustering results can include information from at least one cluster.

[0080] S204: Screen the voiceprint clustering results to obtain the target call detail record.

[0081] Specifically, further screening and elimination can be carried out from the smaller range of voiceprint cluster results. This can be done by methods such as audio transcription of text and verification of signaling terminal information to obtain the target call detail records.

[0082] Figure 3 This is a schematic diagram of the structure of a call detail record (CDR) retrieval device provided in this disclosure. Figure 3 As shown, the device 300 includes: an acquisition module 310, a retrieval module 320, and a filtering module 330.

[0083] The acquisition module 310 is used to acquire the voiceprint features of the object to be retrieved; The retrieval module 320 is used to perform fuzzy retrieval in the candidate call records based on the voiceprint features of the object to be retrieved and the pre-trained voiceprint clustering model to obtain the target call record set. The filtering module 330 is used to filter based on the target call detail record set to obtain the target call detail record.

[0084] Optionally, the retrieval module is used for: The voiceprint clustering model is used to perform voiceprint clustering analysis on the voiceprint features to obtain the voiceprint clustering results of the object to be retrieved. The voiceprint clustering results are used to indicate the target call detail record set.

[0085] Optionally, the device further includes: If the current call meets the preset voiceprint recording rules, extract and record the voiceprint features in the current call; The voiceprint recording rules include at least one of the following: The historical call behavior of the caller in the current call is abnormal; The current call received a call forwarding signaling message but did not carry call forwarding information; The current call did not receive call transfer signaling, and the location of the initially called network element is different from the location of the network element of the recipient in the current call; The user core network identifier carried in any call signaling of the current call is inconsistent with the user core network identifier of the initiator or the user core network identifier of the current call.

[0086] Optionally, the device further includes: Obtain voiceprint feature samples from historical call records; The voiceprint clustering model is trained using the voiceprint feature samples to obtain the voiceprint clustering model; The input of the voiceprint clustering model includes voiceprint features; the output of the voiceprint clustering model includes at least one of the following: a set of call detail records (CDRs) and clustering tags of the CDR sets.

[0087] Optionally, the acquisition module is used for: Obtain the audio file of the object to be retrieved; The audio file is subjected to sound activity detection to obtain voiceprint information carrying human voice. The voiceprint information carrying human voice is separated and processed to obtain human voice voiceprint information; The voiceprint information is extracted to obtain the voiceprint features.

[0088] Optionally, the filtering module is used for: The call details in the target call detail record set are filtered out to obtain the target call detail records, wherein the filtering process includes one or more of the following: The audio files of each call detail record (CDR) are transcribed to obtain the text information of each CDR; a similarity comparison is performed between the text information of each CDR and the text information of the object to be retrieved to obtain a similarity result; and the CDRs with similarity results greater than a preset threshold are taken as the target CDRs. Based on the signaling terminal information of the object to be retrieved, the signaling terminal information in each of the call detail records (CDRs) is checked, and the CDRs with consistent signaling terminal information are taken as the target CDRs.

[0089] This application also provides an electronic device for performing the above-described call detail record (CDR) retrieval method. Please refer to... Figure 4 It illustrates a schematic diagram of an electronic device provided by some embodiments of this application. For example... Figure 4As shown, the electronic device 4 includes: a processor 400, a memory 401, a bus 402, and a communication interface 403. The processor 400, the communication interface 403, and the memory 401 are connected via the bus 402. The memory 401 stores a computer program that can run on the processor 400. When the processor 400 runs the computer program, it executes the call detail record (CDR) retrieval method provided in any of the foregoing embodiments of this application.

[0090] The memory 401 may include high-speed random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Communication between this device network element and at least one other network element is achieved through at least one communication interface 403 (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc.

[0091] Bus 402 can be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. The memory 401 is used to store programs. After receiving an execution instruction, the processor 400 executes the program. The call detail record (CDR) retrieval method disclosed in any of the foregoing embodiments of this application can be applied to the processor 400, or implemented by the processor 400.

[0092] The processor 400 may be an integrated circuit chip with signal processing capabilities. In implementation, each step of the above method can be completed by the integrated logic circuitry in the hardware of the processor 400 or by instructions in software form. The processor 400 may be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it may also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), an off-the-shelf programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor may be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly embodied in the execution of a hardware decoding processor, or executed by a combination of hardware and software modules in the decoding processor. The software modules may reside in random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, or other mature storage media in the art. The storage medium is located in memory 401. The processor 400 reads the information in memory 401 and, in conjunction with its hardware, completes the steps of the above method.

[0093] The electronic device provided in this application embodiment and the call detail record (CDR) retrieval method provided in this application embodiment are based on the same inventive concept and have the same beneficial effects as the methods they adopt, operate, or implement.

[0094] This application also provides a computer-readable storage medium corresponding to the call detail record (CDR) retrieval method provided in the foregoing embodiments. The computer-readable storage medium shown can be an optical disc, on which a computer program is stored. When the computer program is run by a processor, it executes the CDR retrieval method provided in any of the foregoing embodiments.

[0095] It should be noted that examples of the computer-readable storage medium may also include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other optical and magnetic storage media, which will not be elaborated here.

[0096] The computer-readable storage medium provided in the above embodiments of this application and the call detail record (CDR) retrieval method provided in the embodiments of this application are based on the same inventive concept and have the same beneficial effects as the methods adopted, run or implemented by the applications stored therein.

[0097] This application also provides a computer program product 500, such as... Figure 5 As shown. This computer program product carries a computer program 501. The instructions included in the program code can be used to execute the steps of the call detail record (CDR) retrieval method described in the above method embodiments. For details, please refer to the above method embodiments, which will not be repeated here.

[0098] The aforementioned computer program product can be implemented through hardware, software, or a combination thereof. In one optional embodiment, the computer program product is specifically embodied in a computer storage medium; in another optional embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0099] The basic principles of this disclosure have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in this disclosure are merely examples and not limitations, and should not be considered as essential features of each embodiment of this disclosure. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the scope of this disclosure to the necessity of employing the aforementioned specific details for implementation.

[0100] The block diagrams of devices, apparatuses, devices, and systems disclosed herein are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, apparatuses, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0101] Additionally, as used herein, the "or" used in a list of items beginning with "at least one" indicates a separate list, such that a list of, for example, "at least one of A, B, or C" means A or B or C, or AB or AC or BC, or ABC (i.e., A and B and C). Furthermore, the word "exemplary" does not imply that the described example is preferred or better than other examples.

[0102] It should also be noted that in the systems and methods of this disclosure, the components or steps can be decomposed and / or recombined. These decompositions and / or recombinations should be considered as equivalent solutions to this disclosure.

[0103] Various changes, substitutions, and modifications can be made to the technology described herein without departing from the teachings defined by the appended claims. Furthermore, the scope of the claims of this disclosure is not limited to the specific aspects of the processes, machines, manufactures, events, means, methods, and actions described above. Currently existing or later-developed processes, machines, manufactures, events, means, methods, or actions that perform substantially the same function or achieve substantially the same result as the corresponding aspects described herein can be utilized. Therefore, the appended claims include such processes, machines, manufactures, events, means, methods, or actions within their scope.

[0104] The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use this disclosure. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects without departing from the scope of this disclosure. Therefore, this disclosure is not intended to be limited to the aspects shown herein, but rather to be carried out within the widest scope consistent with the principles and novel features disclosed herein.

[0105] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of this disclosure to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations therein.

Claims

1. A call detail record (CDR) retrieval method, characterized in that, include: Obtain the voiceprint features of the object to be retrieved; Based on the voiceprint features of the object to be retrieved and the pre-trained voiceprint clustering model, fuzzy retrieval is performed in the candidate call records to obtain the target call record set. The target call detail records (CDRs) are obtained by filtering from the target CDR set.

2. The method according to claim 1, characterized in that, The method, based on the voiceprint features of the object to be retrieved and a pre-trained voiceprint clustering model, performs fuzzy retrieval in the candidate call detail records (CDRs) to obtain a set of target CDRs, including: The voiceprint clustering model is used to perform voiceprint clustering analysis on the voiceprint features to obtain voiceprint clustering results, which are used to indicate the target call detail record set in the object to be retrieved.

3. The method according to claim 1, characterized in that, The object to be searched includes at least one of the following: current call, historical abnormal call; The acquisition of the voiceprint features of the object to be retrieved includes at least one of the following: When the object to be retrieved is the current call, extract the voiceprint features from the current call; When the object to be retrieved is a historical abnormal call, the voiceprint features in the corresponding record of the historical abnormal call are obtained.

4. The method according to claim 1, characterized in that, The method further includes: If the current call meets the preset voiceprint recording rules, extract and record the voiceprint features in the current call; The voiceprint recording rules include at least one of the following: The historical call behavior of the caller in the current call is abnormal; The current call received a call forwarding signaling message but did not carry call forwarding information; The current call did not receive call transfer signaling, and the location of the initially called network element is different from the location of the network element of the recipient in the current call; The user core network identifier carried in any call signaling of the current call is inconsistent with the user core network identifier of the initiator or the user core network identifier of the current call.

5. The method according to claim 2, characterized in that, The method further includes: Obtain voiceprint feature samples from historical call records; The voiceprint clustering model is trained using the voiceprint feature samples to obtain the voiceprint clustering model; The input of the voiceprint clustering model includes voiceprint features; the output of the voiceprint clustering model includes at least one of the following: a set of call detail records (CDRs) and clustering tags of the CDR sets.

6. The method according to claim 1, characterized in that, The process of obtaining the voiceprint features of the object to be retrieved includes: Obtain the audio file of the object to be retrieved; The audio file is subjected to sound activity detection to obtain voiceprint information carrying human voice. The voiceprint information carrying human voice is separated and processed to obtain human voice voiceprint information; The voiceprint information is extracted to obtain the voiceprint features.

7. The method according to any one of claims 1-6, characterized in that, The step of filtering based on the target call detail record (CDR) set to obtain the target CDRs includes: The call details in the target call detail record set are filtered out to obtain the target call detail records, wherein the filtering process includes one or more of the following: The audio files of each call detail record (CDR) are transcribed to obtain the text information of each CDR; a similarity comparison is performed between the text information of each CDR and the text information of the object to be retrieved to obtain a similarity result; and the CDRs with similarity results greater than a preset threshold are taken as the target CDRs. Based on the signaling terminal information of the object to be retrieved, the signaling terminal information in each of the call detail records (CDRs) is checked, and the CDRs with consistent signaling terminal information are taken as the target CDRs.

8. A call detail record (CDR) retrieval device, characterized in that, include: The acquisition module is used to acquire the voiceprint features of the object to be retrieved; The retrieval module is used to perform fuzzy retrieval in the candidate call records based on the voiceprint features of the object to be retrieved and the pre-trained voiceprint clustering model to obtain the target call record set. The filtering module is used to filter based on the target call detail record set to obtain the target call detail records.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, The processor executes the computer program to implement the method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by a processor to implement the method as described in any one of claims 1-7.

11. A computer program product, characterized in that, Includes computer-readable code, or a non-volatile computer-readable storage medium carrying computer-readable code, wherein when the computer-readable code is run in a processor of an electronic device, the processor in the electronic device performs the method as described in any one of claims 1-7.

Images