WO2022142031A1 - Invalid call determination method and apparatus, computer device, and storage medium - Google Patents

Abstract

Embodiments of the present application relate to the field of artificial intelligence, are applied to the field of intelligent customer service, and relate to an invalid call determination method and apparatus, a computer device, and a storage medium. The method comprises: acquiring a first response voice segment, the first response voice segment being a segment of a peer-end response voice received when an intelligent customer service representative initiates a call; inputting the first response voice segment into a pre-trained deep learning neural network model to perform voiceprint feature extraction to obtain voiceprint features of the first response voice segment; calculating the similarity between the voiceprint features of the first response voice segment and pre-extracted electronically synthesized voice voiceprint features, and if the similarity is greater than a preset threshold, determining that the first voice segment is an electronically synthesized voice; and determining that the call initiated by the intelligent customer service representative is an invalid call. If the response voice is an electronically synthesized voice, the call is determined to be an invalid call, and the call is terminated immediately to release occupied resources, such that the waste of resources is prevented.

Description

Invalid call judgment method, device, computer equipment and storage medium

This application claims the priority of the Chinese patent application filed on December 29, 2020 with the application number 202011594212.9 and the invention title is "Invalid Call Judgment Method, Apparatus, Computer Equipment and Storage Medium", the entire contents of which are by reference Incorporated in this application.

technical field

The present application relates to the technical field of artificial intelligence, and in particular, to a method, device, computer equipment and storage medium for judging invalid calls.

Background technique

With the development of communication technology and Internet technology, intelligent customer service and computer customer service have replaced manual customer service in customer service business. The intelligent customer service AI saves a lot of manpower. The inventor found that about 10%-15% of the outbound calls of intelligent customer service are unanswered calls. The current processing method of such calls is usually to wait until the operator hangs up, and all communication operations When the user is not connected to the phone, the operator will fill in the waiting time of the user on the calling side by broadcasting electronically synthesized voice. At present, when doing AI calling projects, such electronic sounds have not been pre-judged. During the period of waiting for the customer to answer, the radio system continues to pick up the sound, and will convert it as the customer's voice into text and transmit it to the rear end.

Doing so will not only transmit such data to each system of the entire AI outbound link for storage, but also need to eliminate this type of data when performing statistics and summarization, which increases the complexity of secondary statistics. During this period of electronic broadcasting, expensive outbound line resources are occupied.

SUMMARY OF THE INVENTION

The purpose of the embodiments of the present application is to provide a method, device, computer equipment and storage medium for judging invalid calls, so as to solve the problem of waste of call resources caused by invalid calls to smart customer services.

In order to solve the above technical problems, the embodiment of the present application provides a method for judging an invalid call, which adopts the following technical solutions:

Acquiring a first response voice segment, where the first response voice segment is a segment in the opposite-end response voice received when the intelligent customer service initiates a call;

Inputting the first response voice segment into a pre-trained deep learning neural network model for voiceprint feature extraction to obtain the voiceprint feature of the first response voice segment;

Calculate the similarity between the voiceprint feature of the first response voice segment and the pre-extracted electronically synthesized voice voiceprint feature, and when the similarity is greater than a preset threshold, determine that the first voice segment is an electronically synthesized voice;

When the first voice segment is an electronically synthesized voice, it is determined that the call initiated by the intelligent customer service is an invalid call.

In order to solve the above technical problems, the embodiment of the present application also provides an invalid call judgment device, which adopts the following technical solutions:

an obtaining module, configured to obtain a first response voice fragment, where the first response voice fragment is a fragment in the opposite end response voice received when the intelligent customer service initiates a call;

an extraction module, configured to input the first response voice segment into a pre-trained deep learning neural network model for voiceprint feature extraction to obtain the voiceprint feature of the first response voice segment;

A calculation module, configured to calculate the similarity between the voiceprint feature of the first response voice segment and the pre-extracted electronically synthesized voice voiceprint feature, and when the similarity is greater than a preset threshold, determine the first voice segment for electronically synthesized speech;

A determining module, configured to determine that the call initiated by the intelligent customer service is an invalid call when the first voice segment is an electronically synthesized voice.

In order to solve the above-mentioned technical problems, the embodiment of the present application also provides a computer device, which adopts the following technical solutions:

A computer device includes a memory and a processor, wherein computer-readable instructions are stored in the memory, and the processor implements the following steps when executing the computer-readable instructions:

Acquiring a first response voice segment, where the first response voice segment is a segment in the opposite-end response voice received when the intelligent customer service initiates a call;

Inputting the first response voice segment into a pre-trained deep learning neural network model for voiceprint feature extraction to obtain the voiceprint feature of the first response voice segment;

Calculate the similarity between the voiceprint feature of the first response voice segment and the pre-extracted electronically synthesized voice voiceprint feature, and when the similarity is greater than a preset threshold, determine that the first voice segment is an electronically synthesized voice;

When the first voice segment is an electronically synthesized voice, it is determined that the call initiated by the intelligent customer service is an invalid call.

In order to solve the above technical problems, the embodiments of the present application also provide a computer-readable storage medium, which adopts the following technical solutions:

A computer-readable storage medium, where computer-readable instructions are stored on the computer-readable storage medium, and when the computer-readable instructions are executed by a processor, the processor is caused to perform the following steps:

Acquiring a first response voice segment, where the first response voice segment is a segment in the opposite-end response voice received when the intelligent customer service initiates a call;

Inputting the first response voice segment into a pre-trained deep learning neural network model for voiceprint feature extraction to obtain the voiceprint feature of the first response voice segment;

Calculate the similarity between the voiceprint feature of the first response voice segment and the pre-extracted electronically synthesized voice voiceprint feature, and when the similarity is greater than a preset threshold, determine that the first voice segment is an electronically synthesized voice;

When the first voice segment is an electronically synthesized voice, it is determined that the call initiated by the intelligent customer service is an invalid call.

Compared with the prior art, the embodiments of the present application mainly have the following beneficial effects:

According to whether the received answer voice is an electronically synthesized voice, it is judged whether it is an invalid call. If it is an invalid call, the call can be terminated immediately, the occupied resources can be released, the waste of resources can be avoided, and the service efficiency of intelligent customers can be improved.

Description of drawings

In order to illustrate the solutions in the present application more clearly, the following will briefly introduce the accompanying drawings used in the description of the embodiments of the present application. For those of ordinary skill, other drawings can also be obtained from these drawings without any creative effort.

FIG. 1 is an exemplary system architecture diagram to which the present application can be applied;

2 is a flowchart of an embodiment of a method for judging an invalid call according to the present application;

Fig. 3 is the flow chart of the specific embodiment of deep neural network training in Fig. 2;

4 is a schematic structural diagram of an embodiment of an invalid call judging device according to the present application;

FIG. 5 is a schematic structural diagram of an embodiment of a computer device according to the present application.

Detailed ways

Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field of this application; the terms used herein in the specification of the application are for the purpose of describing specific embodiments only It is not intended to limit the application; the terms "comprising" and "having" and any variations thereof in the description and claims of this application and the above description of the drawings are intended to cover non-exclusive inclusion. The terms "first", "second" and the like in the description and claims of the present application or the above drawings are used to distinguish different objects, rather than to describe a specific order.

Reference herein to an "embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor a separate or alternative embodiment that is mutually exclusive of other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

In order to make those skilled in the art better understand the solutions of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings.

As shown in FIG. 1 , the system architecture 100 may include terminal devices 101 , 102 , and 103 , a network 104 and a server 105 . The network 104 is a medium used to provide a communication link between the terminal devices 101 , 102 , 103 and the server 105 . The network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, among others.

The user can use the terminal devices 101, 102, 103 to interact with the server 105 through the network 104 to receive or send messages and the like. Various communication client applications may be installed on the terminal devices 101 , 102 and 103 , such as web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, and the like.

The terminal devices 101, 102, and 103 can be various electronic devices that have a display screen and support web browsing, including but not limited to smart phones, tablet computers, e-book readers, MP3 players (Moving Picture Experts Group Audio Layer III, dynamic Picture Experts Compression Standard Audio Layer 3), MP4 (Moving Picture Experts Group Audio Layer IV, Moving Picture Experts Compression Standard Audio Layer 4) Players, Laptops and Desktops, etc.

The server 105 may be a server that provides various services, such as a background server that provides support for the pages displayed on the terminal devices 101 , 102 , and 103 .

It should be noted that the invalid call judging method provided by the embodiment of the present application is generally executed by the server/terminal device, and accordingly, the invalid call judging apparatus is generally set in the server/terminal device.

It should be understood that the numbers of terminal devices, networks and servers in FIG. 1 are merely illustrative. There can be any number of terminal devices, networks and servers according to implementation needs.

Continuing to refer to FIG. 2 , a flowchart of an embodiment of the method for invalid call determination according to the present application is shown. The method for judging an invalid call includes the following steps:

Step S201: Obtain a first response voice segment, where the first response voice segment is a segment in the opposite end response voice received when the intelligent customer service initiates a call.

In this embodiment, the electronic device (for example, the server/terminal device shown in FIG. 1 ) on which the invalid call determination method runs can receive the first response voice segment through a wired connection or a wireless connection. It should be pointed out that the above wireless connection methods may include but are not limited to 3G/4G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or developed in the future .

The intelligent customer service includes a radio system, and the other end answers the voice when receiving a call through the radio system. Use general audio editing software to edit the answering voice of the opposite end to obtain the first answering voice fragment.

Step S202, inputting the first response voice segment into a pre-trained deep learning neural network model to extract voiceprint features to obtain the voiceprint features of the first response voice segment.

In this embodiment, voiceprint feature extraction is performed by extracting a pre-trained deep learning neural network model. The main goal of the pre-trained deep learning neural network model in this embodiment is to identify whether the received response voice is an electronically synthesized voice or a human response voice. The feature extraction method can better extract the features that reflect the difference between the two. Please refer to Figure 3 for the specific training process of the deep learning neural network model.

Step S203: Calculate the similarity between the voiceprint feature of the first response voice segment and the pre-extracted electronically synthesized voice voiceprint feature, and when the similarity is greater than a preset threshold, determine that the first voice segment is an electronic Synthesized speech.

In this embodiment, the pre-extracted electronically synthesized voice voiceprint features come from the electronic voice broadcasts of various telecommunication operators. Here, the above-mentioned deep learning neural network model is not used to directly determine whether it is an electronic voice, and the pre-extracted electronically synthesized voice is used. The voiceprint features compare the similarity, because the number of telecom operators is effective, and the types of electronic voice broadcasts are limited and can be obtained in advance. By comparing the similarity between the pre-extracted electronically synthesized voiceprint features and the voiceprint features of the response voice, it is more accurate to judge whether it is an electronically synthesized voice rather than a human response.

The calculation of feature similarity adopts the Euclidean distance algorithm, that is, the distance between two feature vectors is calculated. When the similarity is greater than the preset threshold, it is determined that the answering voice is an electronically synthesized voice.

Step S204, when the first voice segment is an electronically synthesized voice, it is determined that the call initiated by the intelligent customer service is an invalid call.

When the answering voice is electronically synthesized voice, the current call is invalid. It is judged that the current call is an invalid call, and based on this judgment, the background of the intelligent customer service ends the current call and releases the line resources, computing resources, and storage resources occupied by the current call.

This application judges whether it is an invalid call according to whether the received answer voice is an electronically synthesized voice, and if it is an invalid call, the call can be terminated immediately, the occupied resources can be released, the waste of resources can be avoided, and the service efficiency of intelligent customers can be improved.

In some optional implementations of this embodiment, after step S201, the above electronic device may further perform the following steps:

Inputting the first speech segment into a preset Mel filter to perform mel spectrum extraction to obtain the mel spectrum of the first speech segment;

The Mel spectrum of the first speech segment is input into a pre-trained deep learning neural network for voiceprint feature extraction to obtain the voiceprint feature of the first response speech segment.

The audio signal is a one-dimensional signal, and only the time domain information can be seen intuitively, but the frequency domain information cannot be seen. It can be transformed to the frequency domain through Fourier transform (FT), but the time domain information is lost, and the time-frequency relationship cannot be seen. Usually the audio information is subjected to short-time Fourier transform to obtain a spectrogram. Since the spectrogram contains a large amount of information, the spectrogram is passed through a mel filter to obtain a mel spectrum.

In this embodiment, the melspectrogram of the audio time-domain signal can be extracted by using the melspectrogram() function in the librosa library of python.

In some optional implementation manners, before the foregoing step S201, the foregoing electronic device may perform the following steps:

receive answering voice;

According to the preset response time, intercept the voice segment without the start part in the response voice to obtain the first voice segment, wherein the start part is the part of the response voice from the start to the response time .

After the smart customer service initiates the call, the customer usually needs a certain response time to answer the call. To avoid misjudgment, after the smart customer service initiates the call, the voice from the start time to the preset response time is cut off. It is judged whether the voice segment is electronically synthesized voice or not.

In some optional implementations, the above-mentioned preset response time is determined by the following steps:

Obtain the response time samples of M intelligent customer service after the call is initiated to before the call is connected, where M is a positive integer greater than 0;

Statistically analyze the probability distribution of the reaction time according to the M reaction time samples;

The preset reaction time is determined according to the reaction time probability distribution.

Assuming that the probability distribution of the response time obtained by statistical analysis of M response time samples shows that 80% of the response times are less than 10 seconds, then the preset response time is set to 10 seconds, and the radio system immediately starts to receive the voice of the peer end after the intelligent customer service initiates the call , truncates the first 10 seconds of the received voice from the peer end, to the greatest extent possible to avoid the situation where the customer is judged to be an invalid call when the call is not connected due to a reasonable response time, resulting in the end of the call.

Through big data analysis, the answer voice segment is intercepted at the best time to judge whether it is an electronically synthesized voice, and whether to end the call is determined according to the judgment. Outbound resources.

In some optional implementation manners, referring to FIG. 3 , before the foregoing step S202, the foregoing electronic device may perform the following steps:

Acquiring training samples, the training samples are N voice samples marked whether they are electronically synthesized voices, and N is a positive integer greater than 0;

Inputting the training samples into a deep learning neural network model, and obtaining N prediction results output by the deep learning neural network model in response to the training samples;

Compare the N prediction results and the annotations through a loss function, where the loss function is:

Among them, N is the number of training samples, the corresponding yi for the i-th sample is the marked result, h=(h1,h2,...,hc) is the prediction result of sample i, where C is the number of all categories;

The parameters of each node of the deep learning neural network are adjusted until the loss function reaches the minimum, and a trained deep learning neural network is obtained.

The deep learning neural network model can be regarded as a voiceprint feature extraction model connected to the output layer, where the output layer is the softmax output layer, and the softmax output layer is used to determine whether the input speech is electronically synthesized speech according to the features extracted by the aforementioned voiceprint feature extraction model. During training, compare the predicted results with the labeled results through softmaxloss. When the softmaxloss reaches the minimum value, the training of the deep learning neural network model ends.

In some optional implementation manners, before the foregoing step S204, the foregoing electronic device may perform the following steps:

When the first response voice segment is an electronically synthesized voice, a second response voice segment is intercepted from the response voice, wherein the start time of the second response voice segment lags the end time of the first voice segment ;

Inputting the second response voice segment into a pre-trained deep learning neural network for voiceprint feature extraction to obtain the voiceprint feature of the second response voice segment;

Calculate the similarity between the voiceprint feature of the second response voice segment and the pre-extracted electronically synthesized voice voiceprint feature, and when the similarity is greater than a preset threshold, determine that the second voice segment is an electronically synthesized voice;

When both the first response voice segment and the second response voice segment are electronically synthesized voices, it is determined that the call initiated by the intelligent customer service is an invalid call.

In some cases, in order to avoid misjudgment, the first and second segments of the response voice are intercepted to determine whether they are electronically synthesized voices. When both voices are electronically synthesized voices, it is determined that the current call is an invalid call.

In some optional implementation manners, after the foregoing step S204, the foregoing electronic device may perform the following steps:

End the call initiated by the intelligent customer service;

Returns the call not connected message to the preset service log database, and returns the call end message to the preset outbound call resource management platform.

When the call is invalid, end the current call and return the call unconnected message to the service log database. The service log database records the situation of the call and is used to check whether the customer's call is connected.

Return the call end message to the preset outbound call resource management platform, and notify the outbound call resource management platform to release the occupied resources. Outbound call resources are limited. When it is determined that the call is invalid, the current call is immediately terminated and the outbound call resources are released, which is beneficial to improve the utilization rate of the outbound call resources.

Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing relevant hardware through computer-readable instructions, and the computer-readable instructions can be stored in a computer-readable storage medium. , when the computer-readable instructions are executed, the processes of the above-mentioned method embodiments may be included. Wherein, the aforementioned storage medium may be a non-volatile storage medium such as a magnetic disk, an optical disk, a read-only memory (Read-Only Memory, ROM), or a random access memory (Random Access Memory, RAM) or the like.

It should be understood that although the various steps in the flowchart of the accompanying drawings are sequentially shown in the order indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless explicitly stated herein, the execution of these steps is not strictly limited to the order and may be performed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but may be executed at different times, and the execution sequence is also It does not have to be performed sequentially, but may be performed alternately or alternately with other steps or at least a portion of sub-steps or stages of other steps.

With further reference to FIG. 4 , as an implementation of the method shown in FIG. 2 above, the present application provides an embodiment of an invalid call determination device, and the device embodiment corresponds to the method embodiment shown in FIG. 2 . Can be used in various electronic devices.

As shown in FIG. 4 , the invalid call determination apparatus 400 in this embodiment includes: an acquisition module 401 , an extraction module 402 , a calculation module 403 , and a determination module 404 . in:

The obtaining module 401 is configured to obtain a first response voice segment, where the first response voice segment is a segment in the opposite-end response voice received when the intelligent customer service initiates a call;

The extraction module 402 is used for inputting the first response voice segment into a pre-trained deep learning neural network for voiceprint feature extraction to obtain the voiceprint feature of the first response voice segment;

The calculation module 403 is used to calculate the similarity between the voiceprint feature of the first response voice segment and the pre-extracted electronically synthesized voice voiceprint feature, and when the similarity is greater than a preset threshold, determine the first voice Fragments are electronically synthesized speech;

The determining module 404 is configured to determine that the call initiated by the intelligent customer service is an invalid call when the first voice segment is an electronically synthesized voice.

In this embodiment, according to whether the received answer voice is an electronically synthesized voice, it is judged whether it is an invalid call. If it is an invalid call, the call can be terminated immediately, the occupied resources can be released, the waste of resources can be avoided, and the service efficiency of intelligent customers can be improved. .

In some optional implementation manners of this embodiment, the device for determining an invalid call further includes:

a first extraction submodule, configured to input the first speech segment into a preset Mel filter to perform mel spectrum extraction to obtain the mel spectrum of the first speech segment;

The second extraction sub-module is configured to input the Mel spectrum of the first speech segment into a pre-trained deep learning neural network model for voiceprint feature extraction to obtain the voiceprint feature of the first response speech segment.

In some optional implementation manners of this embodiment, the device for determining an invalid call further includes:

a first receiving sub-module for receiving the answering voice;

The first interception sub-module is configured to intercept, according to a preset response time, a voice segment that does not contain a starting part in the response voice, to obtain a first voice segment, wherein the starting part is the response voice from the beginning. from the beginning to the portion of the reaction time.

In some optional implementation manners of this embodiment, the device for determining an invalid call further includes:

The first obtaining sub-module is used to obtain the response time samples after the M intelligent customer service initiates the call to before the call is connected, where M is a positive integer greater than 0;

The first statistical sub-module is used to statistically analyze the probability distribution of the reaction time according to the M reaction time samples;

The first determination sub-module is configured to determine the preset reaction time according to the reaction time probability distribution.

In some optional implementation manners of this embodiment, the device for determining an invalid call further includes:

The second acquisition sub-module is used for acquiring training samples, where the training samples are N speech samples marked whether they are electronically synthesized speech, and N is a positive integer greater than 0;

a first prediction submodule, configured to input the training samples into a deep learning neural network model, and obtain N prediction results output by the deep learning neural network model in response to the training samples;

The first comparison sub-module is used to compare whether the N prediction results are consistent with the label through a loss function, wherein the loss function is:

Among them, N is the number of training samples, the corresponding yi for the i-th sample is the marked result, h=(h1,h2,...,hc) is the prediction result of sample i, where C is the number of all categories;

The first adjustment sub-module is used to adjust the parameters of each node of the deep learning neural network model, and ends when the loss function reaches a minimum, and a trained deep learning neural network model is obtained.

In some optional implementation manners of this embodiment, the device for determining an invalid call further includes:

The second interception sub-module is configured to intercept the second response speech segment from the response speech when the first response speech segment is an electronically synthesized speech, wherein the start time of the second response speech segment is delayed by Describe the end time of the first voice segment;

The third extraction sub-module is used for inputting the second response voice segment into a pre-trained deep learning neural network model for voiceprint feature extraction to obtain the voiceprint feature of the second response voice segment;

The first calculation submodule is used to calculate the similarity between the voiceprint feature of the second response voice segment and the pre-extracted electronically synthesized voice voiceprint feature, and when the similarity is greater than a preset threshold, determine the first The second speech segment is electronically synthesized speech;

The second determination submodule is configured to determine that the call initiated by the intelligent customer service is an invalid call when both the first response voice segment and the second response voice segment are electronically synthesized voices.

In some optional implementation manners of this embodiment, the device for determining an invalid call further includes:

Disconnecting the sub-module for ending the call initiated by the intelligent customer service;

The message return sub-module is used to return the call not connected message to the preset service log database, and return the call end message to the preset outbound call resource management platform.

To solve the above technical problems, the embodiments of the present application also provide computer equipment. Please refer to FIG. 5 for details. FIG. 5 is a block diagram of a basic structure of a computer device according to this embodiment.

The computer device 5 includes a memory 51 , a processor 52 , and a network interface 53 that communicate with each other through a system bus. It should be pointed out that only the computer device 5 with components 51-53 is shown in the figure, but it should be understood that it is not required to implement all of the shown components, and more or less components may be implemented instead. Among them, those skilled in the art can understand that the computer device here is a device that can automatically perform numerical calculation and/or information processing according to pre-set or stored instructions, and its hardware includes but is not limited to microprocessors, special-purpose Integrated circuit (Application Specific Integrated Circuit, ASIC), programmable gate array (Field-Programmable Gate Array, FPGA), digital processor (Digital Signal Processor, DSP), embedded equipment, etc.

The computer equipment may be a desktop computer, a notebook computer, a palmtop computer, a cloud server and other computing equipment. The computer device can perform human-computer interaction with the user through a keyboard, a mouse, a remote control, a touch pad or a voice control device.

The memory 51 includes at least one type of readable storage medium, and the readable storage medium includes flash memory, hard disk, multimedia card, card-type memory (for example, SD or DX memory, etc.), random access memory (RAM), static Random Access Memory (SRAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), Programmable Read Only Memory (PROM), Magnetic Memory, Magnetic Disk, Optical Disk, etc. In some embodiments, the memory 51 may be an internal storage unit of the computer device 5 , such as a hard disk or a memory of the computer device 5 . In other embodiments, the memory 51 may also be an external storage device of the computer device 5, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, flash memory card (Flash Card), etc. Of course, the memory 51 may also include both the internal storage unit of the computer device 5 and its external storage device. In this embodiment, the memory 51 is generally used to store the operating system and various application software installed on the computer device 5 , such as computer-readable instructions for a method for determining an invalid call. In addition, the memory 51 can also be used to temporarily store various types of data that have been output or will be output.

In some embodiments, the processor 52 may be a central processing unit (Central Processing Unit, CPU), a controller, a microcontroller, a microprocessor, or other data processing chips. This processor 52 is typically used to control the overall operation of the computer device 5 . In this embodiment, the processor 52 is configured to execute computer-readable instructions or process data stored in the memory 51, for example, computer-readable instructions for executing the invalid call determination method.

The network interface 53 may include a wireless network interface or a wired network interface, and the network interface 53 is generally used to establish a communication connection between the computer device 5 and other electronic devices.

According to whether the received answer voice is an electronically synthesized voice, it is judged whether it is an invalid call. If it is an invalid call, the call can be terminated immediately, the occupied resources can be released, the waste of resources can be avoided, and the service efficiency of intelligent customers can be improved.

The present application also provides another embodiment, that is, to provide a computer-readable storage medium, where the computer-readable storage medium stores computer-readable instructions, and the computer-readable instructions can be executed by at least one processor to The at least one processor is caused to execute the steps of the above-mentioned invalid call judgment method. The computer-readable storage medium may be non-volatile or volatile.

According to whether the received answer voice is an electronically synthesized voice, it is judged whether it is an invalid call. If it is an invalid call, the call can be terminated immediately, the occupied resources can be released, the waste of resources can be avoided, and the service efficiency of intelligent customers can be improved.

From the description of the above embodiments, those skilled in the art can clearly understand that the method of the above embodiment can be implemented by means of software plus a necessary general hardware platform, and of course can also be implemented by hardware, but in many cases the former is better implementation. Based on this understanding, the technical solution of the present application can be embodied in the form of a software product in essence or in a part that contributes to the prior art, and the computer software product is stored in a storage medium (such as ROM/RAM, magnetic disk, CD-ROM), including several instructions to make a terminal device (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) execute the methods described in the various embodiments of this application.

Obviously, the above-described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. The accompanying drawings show the preferred embodiments of the present application, but do not limit the scope of the patent of the present application. This application may be embodied in many different forms, rather these embodiments are provided so that a thorough and complete understanding of the disclosure of this application is provided. Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing specific embodiments, or perform equivalent replacements for some of the technical features. . Any equivalent structures made by using the contents of the description and drawings of this application, which are directly or indirectly used in other related technical fields, are all within the scope of protection of the patent of this application.

Claims (20)

A method for judging an invalid call, comprising the following steps: Acquiring a first response voice segment, where the first response voice segment is a segment in the opposite-end response voice received when the intelligent customer service initiates a call; Inputting the first response voice segment into a pre-trained deep learning neural network model for voiceprint feature extraction to obtain the voiceprint feature of the first response voice segment; Calculate the similarity between the voiceprint feature of the first response voice segment and the pre-extracted electronically synthesized voice voiceprint feature, and when the similarity is greater than a preset threshold, determine that the first voice segment is an electronically synthesized voice; When the first voice segment is an electronically synthesized voice, it is determined that the call initiated by the intelligent customer service is an invalid call. The method for judging an invalid call according to claim 1, wherein after the step of acquiring the first answer voice segment, the method further comprises: Inputting the first speech segment into a preset Mel filter to perform mel spectrum extraction to obtain the mel spectrum of the first speech segment; The Mel spectrum of the first speech segment is input into a pre-trained deep learning neural network model for voiceprint feature extraction to obtain the voiceprint feature of the first response speech segment. The method for judging an invalid call according to claim 1, wherein before the step of acquiring the first response voice segment, the method further comprises: receive answering voice; According to the preset response time, intercept the voice segment without the start part in the response voice to obtain the first voice segment, wherein the start part is the part of the response voice from the start to the response time . The method for judging an invalid call according to claim 3, wherein before the step of receiving the answering voice, the method further comprises: Obtain the response time samples of M intelligent customer service after the call is initiated to before the call is connected, where M is a positive integer greater than 0; Statistically analyze the probability distribution of the reaction time according to the M reaction time samples; The preset reaction time is determined according to the reaction time probability distribution. The method for judging an invalid call according to claim 1, wherein, in the voiceprint feature extraction by inputting the first response voice segment into a pre-trained deep learning neural network model, and obtaining the first response voice segment Before the step of voiceprint feature, it also includes: Acquiring training samples, the training samples are N voice samples marked whether they are electronically synthesized voices, and N is a positive integer greater than 0; Inputting the training samples into a deep learning neural network model, and obtaining N prediction results output by the deep learning neural network model in response to the training samples; Compare the N prediction results and the annotations through a loss function, where the loss function is:

Among them, N is the number of training samples, the corresponding yi for the i-th sample is the marked result, h=(h1,h2,...,hc) is the prediction result of sample i, where C is the number of all categories; The parameters of each node of the deep learning neural network model are adjusted until the loss function reaches a minimum, and a trained deep learning neural network model is obtained. The method for judging an invalid call according to claim 1, wherein, before the step of determining that the call initiated by the intelligent customer service is an invalid call when the first voice segment is an electronically synthesized voice, the method further comprises: When the first response voice segment is an electronically synthesized voice, a second response voice segment is intercepted from the response voice, wherein the start time of the second response voice segment lags the end time of the first voice segment ; The second response voice fragment is input into a pre-trained deep learning neural network model for voiceprint feature extraction to obtain the voiceprint feature of the second response voice fragment; Calculate the similarity between the voiceprint feature of the second response voice segment and the pre-extracted electronically synthesized voice voiceprint feature, and when the similarity is greater than a preset threshold, determine that the second voice segment is an electronically synthesized voice; When both the first response voice segment and the second response voice segment are electronically synthesized voices, it is determined that the call initiated by the intelligent customer service is an invalid call. The method for judging an invalid call according to claim 1, wherein after the step of determining that the call initiated by the intelligent customer service is an invalid call when the first voice segment is an electronically synthesized voice, the method further comprises: End the call initiated by the intelligent customer service; Returns the call not connected message to the preset service log database, and returns the call end message to the preset outbound call resource management platform. A device for judging an invalid call, comprising: an obtaining module, configured to obtain a first response voice fragment, where the first response voice fragment is a fragment in the opposite end response voice received when the intelligent customer service initiates a call; an extraction module, configured to input the first response voice segment into a pre-trained deep learning neural network model for voiceprint feature extraction to obtain the voiceprint feature of the first response voice segment; A calculation module, configured to calculate the similarity between the voiceprint feature of the first response voice segment and the pre-extracted electronically synthesized voice voiceprint feature, and when the similarity is greater than a preset threshold, determine the first voice segment for electronically synthesized speech; A determining module, configured to determine that the call initiated by the intelligent customer service is an invalid call when the first voice segment is an electronically synthesized voice. A computer device includes a memory and a processor, wherein computer-readable instructions are stored in the memory, and the processor also implements the following steps when executing the computer-readable instructions: Acquiring a first response voice segment, where the first response voice segment is a segment in the opposite-end response voice received when the intelligent customer service initiates a call; Inputting the first response voice segment into a pre-trained deep learning neural network model for voiceprint feature extraction to obtain the voiceprint feature of the first response voice segment; Calculate the similarity between the voiceprint feature of the first response voice segment and the pre-extracted electronically synthesized voice voiceprint feature, and when the similarity is greater than a preset threshold, determine that the first voice segment is an electronically synthesized voice; When the first voice segment is an electronically synthesized voice, it is determined that the call initiated by the intelligent customer service is an invalid call. The computer device according to claim 9, wherein, after the step of obtaining the first response voice segment, the processor further implements the following steps when executing the computer-readable instructions: Inputting the first speech segment into a preset Mel filter to perform mel spectrum extraction to obtain the mel spectrum of the first speech segment; The Mel spectrum of the first speech segment is input into a pre-trained deep learning neural network model for voiceprint feature extraction to obtain the voiceprint feature of the first response speech segment. The computer device according to claim 9, wherein, before the step of obtaining the first response voice segment, the processor further implements the following steps when executing the computer-readable instructions: receive answering voice; According to the preset response time, intercept the voice segment without the start part in the response voice to obtain the first voice segment, wherein the start part is the part of the response voice from the start to the response time . The computer device according to claim 11, wherein, before the step of receiving a response voice, the processor further implements the following steps when executing the computer-readable instructions: Obtain the response time samples of M intelligent customer service after the call is initiated to before the call is connected, where M is a positive integer greater than 0; Statistically analyze the probability distribution of the reaction time according to the M reaction time samples; The preset reaction time is determined according to the reaction time probability distribution. The computer device according to claim 9, wherein, after inputting the first response voice segment into a pre-trained deep learning neural network model to perform voiceprint feature extraction, the voiceprint of the first response voice segment is obtained Before the step of the feature, the processor also implements the following steps when executing the computer-readable instructions: Acquiring training samples, the training samples are N voice samples marked whether they are electronically synthesized voices, and N is a positive integer greater than 0; Inputting the training samples into a deep learning neural network model, and obtaining N prediction results output by the deep learning neural network model in response to the training samples; Compare the N prediction results and the annotations through a loss function, where the loss function is:

Among them, N is the number of training samples, the corresponding yi for the i-th sample is the marked result, h=(h1,h2,...,hc) is the prediction result of sample i, where C is the number of all categories; The parameters of each node of the deep learning neural network model are adjusted until the loss function reaches a minimum, and a trained deep learning neural network model is obtained. The computer device according to claim 9, wherein, before the step of determining that the call initiated by the intelligent customer service is an invalid call when the first voice segment is an electronically synthesized voice, the processor executes the The computer readable instructions further implement the following steps: When the first response voice segment is an electronically synthesized voice, a second response voice segment is intercepted from the response voice, wherein the start time of the second response voice segment lags the end time of the first voice segment ; Inputting the second response voice segment into a pre-trained deep learning neural network model for voiceprint feature extraction to obtain the voiceprint feature of the second response voice segment; Calculate the similarity between the voiceprint feature of the second response voice segment and the pre-extracted electronically synthesized voice voiceprint feature, and when the similarity is greater than a preset threshold, determine that the second voice segment is an electronically synthesized voice; When both the first response voice segment and the second response voice segment are electronically synthesized voices, it is determined that the call initiated by the intelligent customer service is an invalid call. The computer device according to claim 9, wherein, after the step of determining that the call initiated by the intelligent customer service is an invalid call when the first voice segment is an electronically synthesized voice, the processor executes the The computer readable instructions also implement the following steps: End the call initiated by the intelligent customer service; Returns the call not connected message to the preset service log database, and returns the call end message to the preset outbound call resource management platform. A computer-readable storage medium, where computer-readable instructions are stored on the computer-readable storage medium, and when the computer-readable instructions are executed by a processor, the processor is caused to perform the following steps: Acquiring a first response voice segment, where the first response voice segment is a segment in the opposite-end response voice received when the intelligent customer service initiates a call; Inputting the first response voice segment into a pre-trained deep learning neural network model for voiceprint feature extraction to obtain the voiceprint feature of the first response voice segment; Calculate the similarity between the voiceprint feature of the first response voice segment and the pre-extracted electronically synthesized voice voiceprint feature, and when the similarity is greater than a preset threshold, determine that the first voice segment is an electronically synthesized voice; When the first voice segment is an electronically synthesized voice, it is determined that the call initiated by the intelligent customer service is an invalid call. The computer-readable storage medium of claim 16, wherein, after the step of obtaining the first response speech segment, the computer-readable instructions, when executed by the processor, cause the processor to further execute the following step: The first voice fragment is input into the preset Mel filter and carries out Mel spectrum extraction, and the Mel spectrum of the first voice fragment is obtained; The Mel spectrum of the first speech segment is input into a pre-trained deep learning neural network model for voiceprint feature extraction to obtain the voiceprint feature of the first response speech segment. The computer-readable storage medium of claim 16, wherein, before the step of obtaining the first response speech segment, the computer-readable instructions, when executed by the processor, cause the processor to perform the following steps : receive answering voice; According to the preset response time, intercept the voice segment without the start part in the response voice to obtain the first voice segment, wherein the start part is the part of the response voice from the start to the response time . The computer-readable storage medium of claim 18, wherein, before the step of receiving a response voice, the computer-readable instructions, when executed by the processor, cause the processor to further perform the following steps: Obtain the response time samples of M intelligent customer service after the call is initiated to before the call is connected, where M is a positive integer greater than 0; Statistically analyze the probability distribution of the reaction time according to the M reaction time samples; The preset reaction time is determined according to the reaction time probability distribution. The computer-readable storage medium according to claim 16, wherein the voiceprint feature extraction is performed by inputting the first response speech segment into a pre-trained deep learning neural network model to obtain the first response speech segment Before the step of the voiceprint feature, when the computer-readable instructions are executed by the processor, the processor further executes the following steps: Acquiring training samples, the training samples are N voice samples marked whether they are electronically synthesized voices, and N is a positive integer greater than 0; Inputting the training sample into a deep learning neural network model, and obtaining N prediction results output by the deep learning neural network model in response to the training sample; Compare the N prediction results and the annotations through a loss function, where the loss function is:

Among them, N is the number of training samples, the corresponding yi for the i-th sample is the marked result, h=(h1,h2,...,hc) is the prediction result of sample i, where C is the number of all categories; The parameters of each node of the deep learning neural network model are adjusted until the loss function reaches a minimum, and a trained deep learning neural network model is obtained.

Images