CN114495982A

CN114495982A - Risk detection method, device and equipment

Info

Publication number: CN114495982A
Application number: CN202210068338.5A
Authority: CN
Inventors: 顾艳梅; 王志铭
Original assignee: Alipay Hangzhou Information Technology Co Ltd
Current assignee: Alipay Hangzhou Information Technology Co Ltd
Priority date: 2022-01-20
Filing date: 2022-01-20
Publication date: 2022-05-13

Abstract

The embodiment of the specification provides a risk detection method, a risk detection device and risk detection equipment, wherein the method comprises the following steps: acquiring a target voice stream to be detected, carrying out speaker separation processing on the target voice stream, and determining speaker information corresponding to the target voice stream; based on the speaker information, dividing the target voice stream into a plurality of target voice segments corresponding to the speakers, and performing text conversion processing on the target voice segments to obtain corresponding target text data; performing risk detection on the target text data based on risk text data in a pre-constructed risk text library to obtain a risk detection result aiming at the target voice fragment; and determining whether the target voice stream is a voice stream with risk or not based on the speaker information and the risk detection result of the target text data.

Description

Risk detection method, device and equipment

Technical Field

The present disclosure relates to the field of risk detection technologies, and in particular, to a risk detection method, apparatus, and device.

Background

With the rapid development of computer technology, in order to protect the information and property security of a user, the real identity of the user can be verified through verification modes such as account number and password, face recognition, fingerprint recognition and the like, and corresponding services are provided for the user after the verification is passed.

When the user is held, the user may be forced to input an account password or input biometric data, which may result in information leakage or property safety damage. Generally, after information leakage or property safety is damaged, a holder can be pursued according to the alarm information of a user, but the following tracking mode cannot make up for the information leakage of the user or recover the property loss of the user. Therefore, it is desirable to provide a solution capable of identifying a risk scenario and improving the detection accuracy of the risk scenario.

Disclosure of Invention

An object of the embodiments of the present specification is to provide a risk detection method, apparatus, and device, so as to provide a solution that can identify a risk scenario and improve detection accuracy of the risk scenario.

In order to implement the above technical solution, the embodiments of the present specification are implemented as follows:

in a first aspect, an embodiment of the present specification provides a risk detection method, including: acquiring a target voice stream to be detected, carrying out speaker separation processing on the target voice stream, and determining speaker information corresponding to the target voice stream; based on the speaker information, dividing the target voice stream into a plurality of target voice segments corresponding to the speakers, and performing text conversion processing on the target voice segments to obtain corresponding target text data; performing risk detection on the target text data based on risk text data in a pre-constructed risk text library to obtain a risk detection result aiming at the target voice fragment; and determining whether the target voice stream is a voice stream with risk or not based on the speaker information and the risk detection result of the target text data.

In a second aspect, embodiments of the present specification provide a risk detection apparatus, the apparatus comprising: the data acquisition module is used for acquiring a target voice stream to be detected, carrying out speaker separation processing on the target voice stream and determining speaker information corresponding to the target voice stream; the text conversion module is used for dividing the target voice stream into a plurality of target voice segments corresponding to the speaker based on the speaker information and performing text conversion processing on the target voice segments to obtain corresponding target text data; the result acquisition module is used for carrying out risk detection on the target text data based on risk text data in a pre-constructed risk text library to obtain a risk detection result aiming at the target voice fragment; and the risk detection module is used for determining whether the target voice stream is a voice stream with risk or not based on the speaker information and the risk detection result of the target text data.

In a third aspect, an embodiment of the present specification provides a risk detection device, where the risk detection device is a device in a blockchain system, and the risk detection device includes: a processor; and a memory arranged to store computer executable instructions that, when executed, cause the processor to: acquiring a target voice stream to be detected, carrying out speaker separation processing on the target voice stream, and determining speaker information corresponding to the target voice stream; based on the speaker information, dividing the target voice stream into a plurality of target voice segments corresponding to the speakers, and performing text conversion processing on the target voice segments to obtain corresponding target text data; performing risk detection on the target text data based on risk text data in a pre-constructed risk text library to obtain a risk detection result aiming at the target voice fragment; and determining whether the target voice stream is a voice stream with risk or not based on the speaker information and the risk detection result of the target text data.

In a fourth aspect, embodiments of the present specification provide a storage medium for storing computer-executable instructions, which when executed implement the following processes: acquiring a target voice stream to be detected, carrying out speaker separation processing on the target voice stream, and determining speaker information corresponding to the target voice stream; based on the speaker information, dividing the target voice stream into a plurality of target voice segments corresponding to the speakers, and performing text conversion processing on the target voice segments to obtain corresponding target text data; performing risk detection on the target text data based on risk text data in a pre-constructed risk text library to obtain a risk detection result aiming at the target voice fragment; and determining whether the target voice stream is a voice stream with risk or not based on the speaker information and the risk detection result of the target text data.

Drawings

In order to more clearly illustrate the embodiments of the present specification or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, it is obvious that the drawings in the following description are only some embodiments described in the present specification, and for those skilled in the art, other drawings can be obtained according to the drawings without any creative effort.

FIG. 1A is a flow chart of one embodiment of a risk detection method of the present disclosure;

FIG. 1B is a schematic processing diagram of an embodiment of a risk detection method of the present disclosure;

FIG. 2 is a schematic diagram of a method for determining a target speech segment according to the present disclosure;

FIG. 3 is a schematic processing diagram of another embodiment of a risk detection method of the present disclosure;

FIG. 4 is a schematic diagram of another method for determining a target speech segment according to the present disclosure;

FIG. 5 is a schematic processing diagram of another embodiment of a risk detection method of the present disclosure;

FIG. 6 is a schematic structural diagram of an embodiment of a risk detection device according to the present disclosure;

fig. 7 is a schematic structural diagram of a risk detection device according to the present disclosure.

Detailed Description

The embodiment of the specification provides a risk detection method, a risk detection device and risk detection equipment.

In order to make those skilled in the art better understand the technical solutions in the present specification, the technical solutions in the embodiments of the present specification will be clearly and completely described below with reference to the drawings in the embodiments of the present specification, and it is obvious that the described embodiments are only a part of the embodiments of the present specification, and not all of the embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present specification without any creative effort shall fall within the protection scope of the present specification.

Example one

As shown in fig. 1A and fig. 1B, an execution subject of the method may be a server, and the server may be an independent server or a server cluster composed of a plurality of servers.

The method may specifically comprise the steps of:

in S102, a target voice stream to be detected is acquired, speaker separation processing is performed on the target voice stream, and speaker information corresponding to the target voice stream is determined.

The target voice stream may be any voice stream containing voice data, for example, the target voice stream may be a voice stream containing voice data collected by an audio collection device in a preset scene, the preset scene may be a resource transfer scene, a privacy data acquisition scene, and other scenes where a safety risk may exist, and the speaker information may include information such as the number of speakers in the target voice stream, the speaking start time and the speaking stop time of each speaker.

In implementation, with the rapid development of computer technology, in order to protect the information and property security of a user, the real identity of the user can be verified through verification modes such as account number and password, face recognition, fingerprint recognition and the like, and corresponding services are provided for the user after the verification is passed. When the user is held, the user may be forced to input an account password or input biometric data, which may result in information leakage or property safety damage. Generally, after information leakage or property safety is damaged, a holder can be pursued according to the alarm information of a user, but the following tracking mode cannot make up for the information leakage of the user or recover the property loss of the user. Therefore, a solution is needed to improve the accuracy of detection of risk scenarios. Therefore, the embodiments of the present disclosure provide a technical solution that can solve the above problems, and refer to the following specifically.

Taking a target voice stream as a voice stream containing voice data acquired by an audio acquisition device in a resource transfer scene as an example, a terminal device (such as a resource transfer device like an ATM cash dispenser) can provide a resource transfer service for a user, and the resource transfer device can be provided with an audio acquisition device. The resource transfer equipment can acquire a voice stream containing voice data based on the audio acquisition equipment under the condition that a triggering instruction of a user for resource transfer service is received, and the resource transfer equipment can send the voice stream acquired by the audio acquisition equipment to the server as a target voice stream, namely, the server can acquire the target voice stream to be detected.

In addition, in order to improve the data processing efficiency of the server and reduce the resource processing pressure of the server, the resource transfer device may trigger the audio acquisition device to acquire the voice stream when determining that a plurality of users are present in the photographable area based on the camera, and send the acquired voice stream to the server as a target voice stream when the acquired voice stream includes the human voice data.

The method for acquiring the target voice stream to be detected is an optional and realizable acquisition method, and in an actual application scenario, there may also be a plurality of different acquisition methods, for example, the target voice stream may be a voice stream (for example, a user call voice stream in about 1 week) and the like that are pre-stored in a server within a preset risk detection period, the server may acquire the voice stream within the preset risk detection period and use the acquired voice stream as the target voice stream to be detected, the acquisition method of the target voice stream may be different according to different actual application scenarios, and this specification does not specifically limit this.

After the server obtains the target voice stream, the server may perform speaker separation processing on the target voice stream to determine speaker information (e.g., the number of speakers) corresponding to the target voice stream.

For example, the server may perform speaker separation processing on the target voice stream based on a pre-trained speaker separation model to obtain the number of speakers corresponding to the target voice stream, where the speaker separation model may be obtained by training the speaker separation model constructed by the machine learning algorithm based on the historical target voice stream.

In addition, there may be many different speaker separation methods, which may be different according to different practical application scenarios, and this specification is not limited in this respect.

In S104, the target speech stream is divided into a plurality of target speech segments corresponding to the speakers based on the speaker information, and text conversion processing is performed on the target speech segments to obtain corresponding target text data.

In an implementation, the speaker information determined by the server may include a speaking start time and a speaking stop time corresponding to each speaker, and the target voice stream may be divided into a plurality of target voice segments corresponding to the speakers according to the speaking start time and the speaking stop time corresponding to each speaker.

For example, as shown in fig. 2, it is assumed that the server recognizes that the target speech stream includes speaker 1 and speaker 2, and that speaker 1 starts speaking from time 1 and ends at time 3 (i.e., the speaking start time of speaker 1 is time 1 and the speaking end time is time 3), and speaker 2 starts speaking from time 2 and ends at time 4. The server may divide the target voice stream into a plurality of target voice segments corresponding to the speaker, i.e., divide the target voice stream into a target voice segment 1 corresponding to the speaker 1 (i.e., a voice segment corresponding to time 1 to time 3) and a target voice segment 2 corresponding to the speaker 2 (i.e., a voice segment corresponding to time 2 to time 4).

The determination method of the target speech segment is an optional and realizable determination method, and in an actual application scenario, there may be a plurality of different determination methods, which may be different according to different actual application scenarios, and this is not specifically limited in this embodiment of the present specification.

After the server divides the target voice stream into a plurality of target voice segments, text conversion processing may be performed on each target voice segment based on an Automatic Speech Recognition Algorithm (ASR) to obtain corresponding target text data.

In addition, in the risk scenario, there may not be cross segments between the voice segments corresponding to the speakers, and thus, by dividing the target voice stream into the target voice segments corresponding to the speakers, the accuracy of subsequent data processing and the accuracy of risk detection may be improved.

In S106, risk detection is performed on the target text data based on the risk text data in the pre-constructed risk text library, so as to obtain a risk detection result for the target speech segment.

In implementation, the target text data may be matched based on the risk text data in the pre-constructed risk text library, and the risk detection result for the target speech segment may be determined according to the matching result.

For example, the risk text data in the pre-built risk text library may include: "hang-off", "know-nothing", and the like, the word segmentation processing or the keyword extraction processing may be performed on the target text data. And matching the extracted result or the word segmentation processing result with the risk text data to obtain a matching result. If the matching result indicates that words exceeding a preset number threshold exist in the target text data and are matched with the risk text data, the risk detection result of the target voice segment can be determined to be that a risk exists.

The above method for determining the risk detection result for the target speech segment is an optional and realizable determination method, and in an actual application scenario, there may be a plurality of different determination methods, for example, the risk detection result for the target speech segment may be determined according to the similarity between the risk text data and the target text data, and the risk detection result for the target speech segment may be different according to different actual application scenarios, which is not specifically limited in the embodiments of the present specification.

In S108, it is determined whether the target voice stream is a risky voice stream based on the speaker information and the risk detection result of the target text data.

In an implementation, for example, as shown in fig. 2, it is assumed that the server determines that the target speech stream includes a target speech segment 1 corresponding to the speaker 1 and a target speech segment 2 corresponding to the speaker 2, and based on the target speech segment 1 and the target speech segment 2, it can be determined that the speaker 1 and the speaker 2 are the same speaker (for example, if the similarity between the voiceprint features corresponding to the target speech segment 1 and the target speech segment 2 is greater than a preset similarity threshold, the speaker 1 and the speaker 2 can be considered as the same speaker). If the risk detection results corresponding to the target voice segment 1 and the target voice segment 2 are both risk-free, it may be determined that the target voice stream is a risk-free voice stream.

Or, assuming that the speaker 1 and the speaker 2 can be determined to be the same speaker based on the target voice segment 1 and the target voice segment 2, and the risk detection result corresponding to the target voice segment 1 and/or the target voice segment 2 is that there is a risk, the target voice stream may be sent to a corresponding manual detection server for risk detection, and a risk detection result of the manual detection server on the target voice stream is obtained to determine whether the target voice stream is a voice stream with a risk.

Still alternatively, if it is determined that the speaker 1 and the speaker 2 are different speakers based on the target speech segment 1 and the target speech segment 2, and the risk detection result corresponding to the target speech segment 1 and/or the target speech segment 2 is risk, it is determined that the target speech stream is a risk speech stream.

The method for determining whether the target voice stream is a risky voice stream is an optional and realizable determination method, and in an actual application scenario, there may be a plurality of different determination methods, which may be different according to different actual application scenarios, and this is not specifically limited in the embodiment of the present specification.

The embodiment of the specification provides a risk detection method, which includes the steps of obtaining a target voice stream to be detected, carrying out speaker separation processing on the target voice stream, determining speaker information corresponding to the target voice stream, dividing the target voice stream into a plurality of target voice segments corresponding to speakers based on the speaker information, carrying out text conversion processing on the target voice segments to obtain corresponding target text data, carrying out risk detection on the target text data based on risk text data in a pre-constructed risk text library to obtain a risk detection result aiming at the target voice segments, and determining whether the target voice stream is a voice stream with risk or not based on the risk detection result of the speaker information and the target text data. Therefore, a target voice segment corresponding to the speaker can be obtained, the risk detection is carried out on the target text data corresponding to the target voice segment, the problem that the corresponding speaking content cannot be accurately identified due to the fact that the speaker with a short speaking segment and a small speaking volume possibly exists in the speaker is avoided, in addition, the risk detection is carried out on the target text data corresponding to the target voice segment, the problem that the risk detection is inaccurate due to the fact that more interference information (such as environmental noise and the like) exists in the target voice stream is avoided, namely, the accurate speaker information and the target text data can be obtained through speaker separation processing and text conversion processing, the risk detection is accurately carried out on the target voice stream according to the risk detection results of the speaker information and the target text data, and whether the target voice stream is a voice stream with risk or not is determined, the detection accuracy for the risk scene is improved.

Example two

As shown in fig. 3, an execution subject of the method may be a server, and the server may be an independent server or a server cluster composed of multiple servers. The method may specifically comprise the steps of:

in S302, a target voice stream to be detected is acquired.

For the specific processing procedure of S302, reference may be made to relevant contents of S102 in the first embodiment, which is not described herein again.

In S304, the target voice stream is divided into a plurality of first voice segments based on a preset time interval.

In practice, the processing manner of S304 may be varied in practical applications, and an alternative implementation manner is provided below, which may specifically refer to the following steps one to two:

step one, carrying out voice data extraction processing on a target voice stream to obtain an extracted target voice stream.

In implementation, the Voice data extraction processing may be performed on the target Voice stream based on a Voice Activity Detection (VAD) algorithm to obtain an extracted target Voice stream, where the VAD algorithm may be used to identify and eliminate a silent period from the target Voice stream, that is, the VAD algorithm may be used to perform the Voice data extraction processing on the target Voice stream to make the extracted target Voice stream be a Voice stream including Voice data.

And secondly, dividing the extracted target voice stream into a plurality of first voice segments based on a preset time interval.

The preset time interval may be any time interval, for example, the preset time interval may be 0.5s, 0.7s, and the like.

In S306, a voiceprint feature corresponding to the first speech segment is obtained.

In implementation, the voiceprint feature corresponding to the first speech segment may be determined based on a preset voiceprint recognition algorithm, where the voiceprint recognition algorithm may be an algorithm determined based on Mel cepstral coefficients (MFCC), Linear Prediction Coefficients (LPC), Linear Prediction Cepstral Coefficients (LPCC), Discrete Wavelet Transform (DWT), and the like, and different voiceprint recognition algorithms may be selected according to different actual application scenarios, which is not specifically limited by the present specification.

In S308, clustering is performed on the voiceprint features to obtain categories corresponding to the voiceprint features.

In implementation, clustering processing may be performed on the voiceprint features based on a preset clustering algorithm to obtain categories corresponding to the voiceprint features, where the preset clustering algorithm may be a K-means algorithm, a DBSCAN algorithm, an AHC algorithm, or the like.

In S310, speaker information corresponding to the target speech stream is generated based on the category corresponding to the voiceprint feature.

In the implementation, for example, as shown in fig. 4, it is assumed that the target voice stream is divided into a first voice segment 1, a first voice segment 2, and a first voice segment 3 based on a preset time interval, and a voiceprint recognition process is performed on each first voice segment, so as to obtain a voiceprint feature corresponding to each first voice segment.

As shown in fig. 4, clustering the voiceprint features can obtain a class 1 including the voiceprint features corresponding to the first speech segment 1 and the first speech segment 3, and a class 2 including the voiceprint features corresponding to the first speech segment 2, and can determine the class 1 as the class corresponding to the speaker 1 and the class 2 as the class corresponding to the speaker 2.

In S312, the voiceprint feature of the speaker is determined based on the voiceprint feature and the category to which the voiceprint feature corresponds.

In implementation, as shown in fig. 4, the voiceprint features corresponding to the first speech segment 1 and the first speech segment 3 included in the category 1 may be determined as the voiceprint feature of the speaker 1, and the voiceprint feature corresponding to the first speech segment 2 included in the category 2 may be determined as the voiceprint feature of the speaker 2.

In S314, the target speech stream is divided into a plurality of target speech segments corresponding to the speaker based on the voiceprint characteristics of the speaker.

In implementation, as shown in fig. 4, the target speech stream may be divided into a target speech segment 1 and a target speech segment 3 corresponding to the speaker 1, and a target speech segment 3 corresponding to the speaker 2 according to the voiceprint characteristics of the speakers 1 and 2.

Therefore, the target voice stream is segmented again based on the voiceprint characteristics of the speaker, so that the matching degree of the obtained target voice segment and the speaker is high, and the segmentation accuracy of the target voice segment is high.

In S316, text conversion processing is performed on the target speech segment to obtain corresponding target text data.

In S318, risk detection is performed on the target text data based on the risk text data in the pre-constructed risk text library, so as to obtain a risk detection result for the target speech segment.

For the specific processing procedure of S316 to S318, reference may be made to the relevant contents of S104 to S106 in the first embodiment, which are not described herein again.

In S320, in a case where it is determined that a plurality of speakers exist based on the speaker information and the target speech segment includes a target speech segment in which the risk detection result is a risk, the target speech stream is determined to be a risk speech stream.

In implementation, if it is determined that a plurality of speakers exist based on the speaker information and the similarity between the voiceprint features corresponding to the speakers is not greater than the preset similarity threshold, it may be determined that the plurality of speakers are different speakers. Meanwhile, if the target voice segment contains the target voice segment with the risk detection result being the risk, the target voice stream can be determined to be the voice stream with the risk.

Alternatively, in the case where it is determined that there are a plurality of speakers based on the speaker information and there are a plurality of target speech segments for which the risk detection result is a risk, the target speech stream may be determined to be a risk speech stream.

The embodiment of the specification provides a risk detection method, which includes the steps of obtaining a target voice stream to be detected, carrying out speaker separation processing on the target voice stream, determining speaker information corresponding to the target voice stream, dividing the target voice stream into a plurality of target voice segments corresponding to speakers based on the speaker information, carrying out text conversion processing on the target voice segments to obtain corresponding target text data, carrying out risk detection on the target text data based on risk text data in a pre-constructed risk text library to obtain a risk detection result aiming at the target voice segments, and determining whether the target voice stream is a voice stream with risk or not based on the risk detection result of the speaker information and the target text data. Therefore, the target voice segment corresponding to the speaker can be obtained, the risk detection can be carried out on the target text data corresponding to the target voice segment, the problem that the corresponding speaking content can not be accurately identified due to the fact that the speaker with a short speaking segment and a small speaking volume possibly exists in the speaker is avoided, in addition, the risk detection can be carried out on the target text data corresponding to the target voice segment, the problem that the risk detection is inaccurate due to the fact that more interference information (such as environmental noise and the like) exists in the target voice stream is avoided, namely, the accurate speaker information and the target text data can be obtained through speaker separation processing and text conversion processing, the risk detection can be accurately carried out on the target voice stream according to the risk detection results of the speaker information and the target text data, and whether the target voice stream is a voice stream with risk or not is determined, the detection accuracy for the risk scene is improved.

EXAMPLE III

As shown in fig. 5, an execution subject of the method may be a server, and the server may be an independent server or a server cluster composed of multiple servers. The method may specifically comprise the steps of:

in S502, a target voice stream to be detected is acquired.

In S504, speaker separation processing is performed on the target voice stream based on the pre-trained speaker separation model and the target voice stream, so as to obtain speaker information corresponding to the target voice stream.

The speaker separation model can be obtained by training a model constructed by a preset machine learning algorithm based on a historical target voice stream.

In implementation, a historical target voice stream and a historical voice segment corresponding to a speaker included in the historical target voice stream can be acquired, the historical target voice stream can be input into a pre-constructed speaker separation model to obtain a plurality of voice segments, and the speaker separation model is trained according to the voice segment output by the model, the acquired historical voice segment corresponding to the speaker and a preset error function to obtain a pre-trained speaker separation model.

The server may input the target voice stream into the pre-trained speaker separation model to obtain a plurality of voice segments, and determine the number of speakers according to the obtained voice segments, for example, if 3 voice segments are obtained according to the pre-trained speaker separation model and the target voice stream, the number of speakers corresponding to the target voice stream may be determined to be 3.

In addition, voiceprint feature extraction can be performed on the obtained voice segments to determine the voiceprint features corresponding to each speaker.

In S506, based on the speaker information, the target speech stream is divided into a plurality of target speech segments corresponding to the speaker, and text conversion processing is performed on the target speech segments to obtain corresponding target text data.

In S508, risk detection is performed on the target text data based on the risk text data in the pre-constructed risk text library, so as to obtain a risk detection result for the target speech segment.

For the specific processing procedures of S506 to S508, reference may be made to the relevant contents of S104 to S106 in the first embodiment, which are not described herein again.

In S510, when it is determined that there are no multiple speakers based on the speaker information and the target speech segment includes a target speech segment whose risk detection result is a risk, a risk detection result for the target speech stream is determined based on a pre-trained risk detection model and the target text data, and whether the target speech stream is a risk speech stream is determined according to the risk detection result.

The risk detection model can be obtained by training a model constructed by a machine learning algorithm based on historical text data.

In implementation, for example, the risk detection model constructed by the neural network algorithm may be trained through the historical text data, resulting in a pre-trained risk detection model.

Under the condition that a plurality of speakers are determined to be absent based on the speaker information and the target speech segment contains the target speech segment with the risk detection result of the risk, the target text data can be input into a risk detection model trained in advance to obtain the risk detection result.

In addition, the target text data can be integrated according to the timestamp of the target voice segment and the identity tag of the speaker (such as the speaker 1, the speaker 2 and the like), semantic analysis is carried out on the obtained target integrated data to obtain a semantic analysis result, and whether the target voice stream is the voice stream with the risk or not is determined according to the semantic analysis result and the risk detection result.

Under the condition that the target voice stream is determined to be a voice stream with risk, the server can also obtain preset alarm information and send the preset alarm information to the terminal equipment to prompt the user that the risk exists in the current scene, or the server can also directly send the preset alarm information, the equipment identifier of the terminal equipment (or the user identifier corresponding to the terminal equipment) and the target voice stream to the preset alarm party so that the preset alarm party can process the target voice stream to protect the information safety of the user.

Example four

Based on the same idea, the risk detection method provided in the embodiment of the present specification further provides a risk detection device, as shown in fig. 6.

The risk detection device includes: a data obtaining module 601, a text conversion module 602, a result obtaining module 603, and a risk detection module 604, wherein:

a data obtaining module 601, configured to obtain a target voice stream to be detected, perform speaker separation processing on the target voice stream, and determine speaker information corresponding to the target voice stream;

a text conversion module 602, configured to divide the target speech stream into a plurality of target speech segments corresponding to speakers based on the speaker information, and perform text conversion processing on the target speech segments to obtain corresponding target text data;

a result obtaining module 603, configured to perform risk detection on the target text data based on risk text data in a risk text library that is constructed in advance, to obtain a risk detection result for the target speech segment;

a risk detection module 604, configured to determine whether the target voice stream is a voice stream with a risk based on the speaker information and a risk detection result of the target text data.

In this embodiment of the present specification, the data obtaining module 601 includes:

a dividing unit, configured to divide the target voice stream into a plurality of first voice segments based on a preset time interval;

the feature acquisition unit is used for acquiring the voiceprint features corresponding to the first voice fragment;

the category determining unit is used for clustering the voiceprint features to obtain categories corresponding to the voiceprint features;

and the information generating unit is used for generating the speaker information corresponding to the target voice stream based on the category corresponding to the voiceprint characteristics.

In an embodiment of this specification, the dividing unit is configured to:

performing voice data extraction processing on the target voice stream to obtain an extracted target voice stream;

and dividing the extracted target voice stream into a plurality of first voice segments based on a preset time interval.

In an embodiment of this specification, the dividing unit is configured to:

determining the voiceprint characteristics of the speaker based on the voiceprint characteristics and the corresponding categories of the voiceprint characteristics;

and dividing the target voice stream into a plurality of target voice segments corresponding to the speaker based on the voiceprint characteristics of the speaker.

In this embodiment of the present specification, the data obtaining module 601 is configured to:

and carrying out speaker separation processing on the target voice stream based on a pre-trained speaker separation model and the target voice stream to obtain speaker information corresponding to the target voice stream, wherein the speaker separation model is obtained by training a model constructed by a preset machine learning algorithm based on a historical target voice stream.

In this embodiment of the present specification, the risk detection module 604 is configured to:

and under the condition that a plurality of speakers exist based on the speaker information and the target voice segment with the risk detection result as the risk exists is included, determining the target voice stream as the risk voice stream.

In an embodiment of this specification, the apparatus further includes:

and the risk determining module is used for determining a risk detection result aiming at the target voice stream based on a pre-trained risk detection model and the target text data under the condition that a plurality of speakers do not exist based on the speaker information and the target voice segment with risk detection results is included, and determining whether the target voice stream is a voice stream with risk or not according to the risk detection result, wherein the risk detection model is obtained by training a model constructed by a machine learning algorithm based on historical text data.

The embodiment of the present specification provides a risk detection apparatus, which acquires a target voice stream to be detected, performs speaker separation processing on the target voice stream, determines speaker information corresponding to the target voice stream, divides the target voice stream into a plurality of target voice segments corresponding to speakers based on the speaker information, performs text conversion processing on the target voice segments to obtain corresponding target text data, performs risk detection on the target text data based on risk text data in a pre-constructed risk text library to obtain a risk detection result for the target voice segments, and determines whether the target voice stream is a voice stream with risk based on the risk detection result of the speaker information and the target text data. Therefore, the target voice segment corresponding to the speaker can be obtained, the risk detection can be carried out on the target text data corresponding to the target voice segment, the problem that the corresponding speaking content can not be accurately identified due to the fact that the speaker with a short speaking segment and a small speaking volume possibly exists in the speaker is avoided, in addition, the risk detection can be carried out on the target text data corresponding to the target voice segment, the problem that the risk detection is inaccurate due to the fact that more interference information (such as environmental noise and the like) exists in the target voice stream is avoided, namely, the accurate speaker information and the target text data can be obtained through speaker separation processing and text conversion processing, the risk detection can be accurately carried out on the target voice stream according to the risk detection results of the speaker information and the target text data, and whether the target voice stream is a voice stream with risk or not is determined, the detection accuracy for the risk scene is improved.

EXAMPLE five

Based on the same idea, the embodiments of the present specification further provide a risk detection device, as shown in fig. 7.

The risk detection device may vary significantly depending on configuration or performance, and may include one or more processors 701 and memory 702, where one or more stored applications or data may be stored in memory 702. Memory 702 may be, among other things, transient storage or persistent storage. The application program stored in memory 702 may include one or more modules (not shown), each of which may include a series of computer-executable instructions for the risk detection device. Still further, processor 701 may be configured to communicate with memory 702 to execute a series of computer-executable instructions in memory 702 on the risk detection device. The risk detection apparatus may also include one or more power supplies 703, one or more wired or wireless network interfaces 704, one or more input-output interfaces 705, one or more keyboards 706.

In particular, in this embodiment, the risk detection device includes a memory, and one or more programs, wherein the one or more programs are stored in the memory, and the one or more programs may include one or more modules, and each module may include a series of computer-executable instructions for the risk detection device, and the one or more programs configured to be executed by the one or more processors include computer-executable instructions for:

acquiring a target voice stream to be detected, carrying out speaker separation processing on the target voice stream, and determining speaker information corresponding to the target voice stream;

based on the speaker information, dividing the target voice stream into a plurality of target voice segments corresponding to the speakers, and performing text conversion processing on the target voice segments to obtain corresponding target text data;

performing risk detection on the target text data based on risk text data in a pre-constructed risk text library to obtain a risk detection result aiming at the target voice fragment;

and determining whether the target voice stream is a voice stream with risk or not based on the speaker information and the risk detection result of the target text data.

Optionally, the performing speaker separation processing on the target voice stream to determine speaker information corresponding to the target voice stream includes:

dividing the target voice stream into a plurality of first voice segments based on a preset time interval;

acquiring a voiceprint feature corresponding to the first voice fragment;

clustering the voiceprint features to obtain categories corresponding to the voiceprint features;

and generating the speaker information corresponding to the target voice stream based on the category corresponding to the voiceprint feature.

Optionally, the dividing the target voice stream into a plurality of first voice segments based on a preset time interval includes:

performing the extraction processing of the voice data of the target voice stream to obtain an extracted target voice stream;

Optionally, the dividing the target speech stream into a plurality of target speech segments corresponding to speakers based on the speaker information includes:

Optionally, the determining whether the target voice stream is a risky voice stream based on the speaker information and the risk detection result of the target text data includes:

Optionally, the method further comprises:

and under the condition that a plurality of speakers do not exist based on the speaker information and the target voice segment contains a target voice segment with a risk detection result of risk, determining the risk detection result aiming at the target voice stream based on a pre-trained risk detection model and the target text data, and determining whether the target voice stream is the voice stream with the risk according to the risk detection result, wherein the risk detection model is obtained by training a model constructed by a machine learning algorithm based on historical text data.

The embodiment of the present specification provides a risk detection device, which acquires a target voice stream to be detected, performs speaker separation processing on the target voice stream, determines speaker information corresponding to the target voice stream, divides the target voice stream into a plurality of target voice segments corresponding to speakers based on the speaker information, performs text conversion processing on the target voice segments to obtain corresponding target text data, performs risk detection on the target text data based on risk text data in a pre-constructed risk text library to obtain a risk detection result for the target voice segments, and determines whether the target voice stream is a voice stream with risk based on the risk detection result of the speaker information and the target text data. Therefore, the target voice segment corresponding to the speaker can be obtained, the risk detection can be carried out on the target text data corresponding to the target voice segment, the problem that the corresponding speaking content can not be accurately identified due to the fact that the speaker with a short speaking segment and a small speaking volume possibly exists in the speaker is avoided, in addition, the risk detection can be carried out on the target text data corresponding to the target voice segment, the problem that the risk detection is inaccurate due to the fact that more interference information (such as environmental noise and the like) exists in the target voice stream is avoided, namely, the accurate speaker information and the target text data can be obtained through speaker separation processing and text conversion processing, the risk detection can be accurately carried out on the target voice stream according to the risk detection results of the speaker information and the target text data, and whether the target voice stream is a voice stream with risk or not is determined, the detection accuracy for the risk scene is improved.

EXAMPLE six

The embodiments of the present disclosure further provide a computer-readable storage medium, where a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, the computer program implements the processes of the above-mentioned embodiment of the risk detection method, and can achieve the same technical effects, and in order to avoid repetition, the details are not repeated here. The computer-readable storage medium may be a Read-Only Memory (ROM), a Random Access Memory (RAM), a magnetic disk or an optical disk.

The embodiment of the specification provides a computer-readable storage medium, which acquires a target voice stream to be detected, performs speaker separation processing on the target voice stream, determines speaker information corresponding to the target voice stream, divides the target voice stream into a plurality of target voice segments corresponding to speakers based on the speaker information, performs text conversion processing on the target voice segments to obtain corresponding target text data, performs risk detection on the target text data based on risk text data in a pre-established risk text library to obtain a risk detection result for the target voice segments, and determines whether the target voice stream is a voice stream with risk based on the risk detection result of the speaker information and the target text data. Therefore, the target voice segment corresponding to the speaker can be obtained, the risk detection can be carried out on the target text data corresponding to the target voice segment, the problem that the corresponding speaking content can not be accurately identified due to the fact that the speaker with a short speaking segment and a small speaking volume possibly exists in the speaker is avoided, in addition, the risk detection can be carried out on the target text data corresponding to the target voice segment, the problem that the risk detection is inaccurate due to the fact that more interference information (such as environmental noise and the like) exists in the target voice stream is avoided, namely, the accurate speaker information and the target text data can be obtained through speaker separation processing and text conversion processing, the risk detection can be accurately carried out on the target voice stream according to the risk detection results of the speaker information and the target text data, and whether the target voice stream is a voice stream with risk or not is determined, the detection accuracy for the risk scene is improved.

The foregoing description has been directed to specific embodiments of this disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.

In the 90 s of the 20 th century, improvements in a technology could clearly distinguish between improvements in hardware (e.g., improvements in circuit structures such as diodes, transistors, switches, etc.) and improvements in software (improvements in process flow). However, as technology advances, many of today's process flow improvements have been seen as direct improvements in hardware circuit architecture. Designers almost always obtain the corresponding hardware circuit structure by programming an improved method flow into the hardware circuit. Thus, it cannot be said that an improvement in the process flow cannot be realized by hardware physical modules. For example, a Programmable Logic Device (PLD), such as a Field Programmable Gate Array (FPGA), is an integrated circuit whose Logic functions are determined by programming the Device by a user. A digital system is "integrated" on a PLD by the designer's own programming without requiring the chip manufacturer to design and fabricate application-specific integrated circuit chips. Furthermore, nowadays, instead of manually making an Integrated Circuit chip, such Programming is often implemented by "logic compiler" software, which is similar to a software compiler used in program development and writing, but the original code before compiling is also written by a specific Programming Language, which is called Hardware Description Language (HDL), and HDL is not only one but many, such as abel (advanced Boolean Expression Language), ahdl (alternate Hardware Description Language), traffic, pl (core universal Programming Language), HDCal (jhdware Description Language), lang, Lola, HDL, laspam, hardward Description Language (vhr Description Language), vhal (Hardware Description Language), and vhigh-Language, which are currently used in most common. It will also be apparent to those skilled in the art that hardware circuitry that implements the logical method flows can be readily obtained by merely slightly programming the method flows into an integrated circuit using the hardware description languages described above.

The controller may be implemented in any suitable manner, for example, the controller may take the form of, for example, a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, an Application Specific Integrated Circuit (ASIC), a programmable logic controller, and an embedded microcontroller, examples of which include, but are not limited to, the following microcontrollers: the ARC625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320, the memory controller may also be implemented as part of the control logic for the memory. Those skilled in the art will also appreciate that, in addition to implementing the controller as pure computer readable program code, the same functionality can be implemented by logically programming method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Such a controller may thus be considered a hardware component, and the means included therein for performing the various functions may also be considered as a structure within the hardware component. Or even means for performing the functions may be regarded as being both a software module for performing the method and a structure within a hardware component.

The systems, devices, modules or units illustrated in the above embodiments may be implemented by a computer chip or an entity, or by a product with certain functions. One typical implementation device is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smartphone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

For convenience of description, the above devices are described as being divided into various units by function, and are described separately. Of course, the functionality of the various elements may be implemented in the same one or more software and/or hardware implementations in implementing one or more embodiments of the present description.

As will be appreciated by one skilled in the art, embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, one or more embodiments of the present description may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, one or more embodiments of the present description may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

Embodiments of the present description are described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the description. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable risk detection device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable risk detection device, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable risk detection device to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable risk detection device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer implemented process such that the instructions which execute on the computer or other programmable device provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

In a typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include forms of volatile memory in a computer readable medium, Random Access Memory (RAM) and/or non-volatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of a computer-readable medium.

Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media 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 memory technology, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.

It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.

One or more embodiments of the present description may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. One or more embodiments of the specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.

The above description is only an example of the present specification, and is not intended to limit the present specification. Various modifications and alterations to this description will become apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present specification should be included in the scope of the claims of the present specification.

Claims

1. A method of risk detection, comprising:

2. The method according to claim 1, wherein the performing speaker separation processing on the target voice stream to determine speaker information corresponding to the target voice stream comprises:

acquiring a voiceprint feature corresponding to the first voice fragment;

3. The method according to claim 2, wherein the segmenting the target speech stream into a plurality of first speech segments based on a preset time interval, comprises:

4. The method of claim 2, the segmenting the target speech stream into a plurality of target speech segments corresponding to speakers based on the speaker information, comprising:

5. The method according to claim 1, wherein the performing speaker separation processing on the target voice stream to determine speaker information corresponding to the target voice stream comprises:

6. The method of claim 1, wherein determining whether the target speech stream is a risky speech stream based on the speaker information and the risk detection result of the target text data comprises:

7. The method of claim 6, further comprising:

8. A risk detection device, comprising:

the data acquisition module is used for acquiring a target voice stream to be detected, carrying out speaker separation processing on the target voice stream and determining speaker information corresponding to the target voice stream;

the text conversion module is used for dividing the target voice stream into a plurality of target voice segments corresponding to the speaker based on the speaker information and performing text conversion processing on the target voice segments to obtain corresponding target text data;

the result acquisition module is used for carrying out risk detection on the target text data based on risk text data in a pre-constructed risk text library to obtain a risk detection result aiming at the target voice fragment;

and the risk detection module is used for determining whether the target voice stream is a voice stream with risk or not based on the speaker information and the risk detection result of the target text data.

9. A risk detection device, the risk detection device comprising:

a processor; and

a memory arranged to store computer executable instructions that, when executed, cause the processor to:

10. A storage medium for storing computer executable instructions which, when executed by a processor, implement the following flow: