Voice recognition method and device based on AI, equipment and medium

Abstract

The invention relates to the technical field of artificial intelligence and natural language processing, and relates to an AI-based voice recognition method, device, equipment and medium, the method comprises the steps of obtaining first information and second information, the first information comprising customer data of a bank, and the second information comprising voice information of a customer when a bank customer service communicates with the customer; performing desensitization processing on the first information to obtain client data after desensitization processing; establishing a repayment scheme prediction model according to the desensitized customer data; performing voice recognition on the second information, and matching the client identity according to a voice recognition result to obtain a matching result; determining a repayment scheme of the customer according to a matching result and a repayment scheme prediction model; the repayment scheme of the customer is converted into the third information, the third information is the voice information corresponding to the repayment scheme, the identity of the customer is matched by performing voice recognition on the second information, so that a personalized repayment mode is generated for the customer, and the experience feeling of the customer is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of artificial intelligence and natural language processing, and in particular to an AI-based speech recognition method, device, equipment and medium. Background Art

[0002] In today's era of rapid technological development, the field of speech recognition has achieved many remarkable achievements. Among them, the clever application of artificial intelligence technology to speech recognition is undoubtedly an innovative measure of great significance. Through the powerful algorithms and models of artificial intelligence, the system can process and analyze the input voice information very quickly, and can accurately translate it into corresponding text content with high efficiency, which has brought great convenience to people's lives and work to a great extent. However, when speech recognition is applied to bank customer service work, since bank customers come from all over the world, there is a problem of low recognition accuracy when performing speech recognition for some low-resource languages, which makes it difficult for customer service to generate recommendation solutions to satisfy customers and reduce customer experience. Summary of the invention

[0003] The purpose of the present invention is to provide an AI-based speech recognition method, device, equipment and medium to improve the above-mentioned problems.

[0004] In order to achieve the above objectives, the present application provides the following technical solutions:

[0005] On the one hand, an embodiment of the present application provides an AI-based speech recognition method, the method comprising:

[0006] Acquire first information and second information, wherein the first information includes customer data of the bank, and the second information includes voice information of the customer when the bank customer service communicates with the customer;

[0007] Performing desensitization processing on the first information to obtain desensitized customer data;

[0008] Establishing a repayment plan prediction model based on the desensitized customer data;

[0009] Performing voice recognition on the second information, and matching the customer identity according to the result of the voice recognition to obtain a matching result;

[0010] Determining the repayment plan of the customer based on the matching result and the repayment plan prediction model;

[0011] The repayment plan of the customer is converted into third information, where the third information is voice information corresponding to the repayment plan.

[0012] In a second aspect, an embodiment of the present application provides an AI-based speech recognition device, the device comprising:

[0013] An acquisition module, used to acquire first information and second information, wherein the first information includes the customer data of the bank, and the second information includes the voice information of the customer when the bank customer service communicates with the customer;

[0014] A first processing module, used for performing desensitization processing on the first information to obtain desensitized customer data;

[0015] A second processing module is used to establish a repayment plan prediction model based on the desensitized customer data;

[0016] A third processing module, configured to perform voice recognition on the second information, and match the customer identity according to the result of the voice recognition to obtain a matching result;

[0017] A fourth processing module, used to determine the repayment plan of the customer according to the matching result and the repayment plan prediction model;

[0018] The fifth processing module is used to convert the customer's repayment plan into third information, where the third information is the voice information corresponding to the repayment plan.

[0019] In a third aspect, an embodiment of the present application provides an AI-based speech recognition device, the device comprising a memory and a processor. The memory is used to store a computer program; the processor is used to implement the steps of the above-mentioned AI-based speech recognition method when executing the computer program.

[0020] In a fourth aspect, an embodiment of the present application provides a readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above-mentioned AI-based speech recognition method are implemented.

[0021] The beneficial effects of the present invention are:

[0022] The present invention performs voice recognition on the second information and matches the customer's identity according to the recognition result. It not only automatically and quickly realizes customer identity matching, but also obtains the customer's data in the bank according to the matching result, and determines the customer's personalized repayment plan according to the obtained data. Finally, the repayment plan is translated into voice information and sent to the customer. It not only provides customers with accurate repayment plans in an efficient and automatic manner, but also improves the customer's user experience.

[0023] Other features and advantages of the present invention will be described in the following description, and partly become apparent from the description, or understood by implementing the embodiments of the present invention. The purpose and other advantages of the present invention can be realized and obtained by the structures particularly pointed out in the written description and the drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0025] Figure 1 Schematic diagram of the flow of the AI-based speech recognition method described in an embodiment of the present invention.

[0026] Figure 2 Schematic diagram of the structure of the AI-based speech recognition device described in an embodiment of the present invention.

[0027] Figure 3 Schematic diagram of the structure of the AI-based speech recognition device described in an embodiment of the present invention.

[0028] Labels in the figure: 800, AI-based speech recognition device; 801, processor; 802, memory; 803, multimedia component; 804, I / O interface; 805, communication component; 901, acquisition module; 902, first processing module; 903, second processing module; 904, third processing module; 905, fourth processing module; 906, fifth processing module. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0030] It should be noted that similar reference numerals and letters represent similar items in the following drawings, so once an item is defined in one drawing, it does not need to be further defined and explained in the subsequent drawings. At the same time, in the description of the present invention, the terms "first", "second", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0031] Embodiment 1:

[0032] This embodiment provides an AI-based speech recognition method. It can be understood that a scenario can be laid out in this embodiment, for example: when a customer communicates with a bank customer service by phone, the bank customer service needs to formulate a repayment plan based on the customer's needs.

[0033] See also Figure 1 , the figure shows that the method includes step S1, step S2, step S3, step S4, step S5 and step S6.

[0034] Step S1, obtaining first information and second information, wherein the first information includes the customer data of the bank, and the second information includes the voice information of the customer when the bank customer service communicates with the customer;

[0035] Step S2: desensitizing the first information to obtain desensitized customer data;

[0036] In this step, the bank's customer data is usually in encrypted form, so it needs to be desensitized before further processing.

[0037] Step S3: establishing a repayment plan prediction model based on the desensitized customer data;

[0038] The step S3 also includes step S31, step S32 and step S33, which specifically include:

[0039] Step S31, using an exploratory data analysis method to analyze the desensitized customer data to obtain analysis results;

[0040] In this step, the customer data after desensitization processing includes multiple data fields, including but not limited to credit score, income level, credit history, consumer behavior data, etc. Descriptive statistical analysis is performed on each data field, such as mean, median, standard deviation, maximum value, minimum value, etc. By observing these descriptive statistics, we can preliminarily understand the distribution of each characteristic variable. According to the data type and analysis purpose, different visualization charts are selected to show the relationship between the data. For example, for the credit score of customers at different income levels, a box plot can be selected; for the relationship between categorical data and numerical data, such as the average debt of customers of different occupational types, a bar chart can be selected. The analysis results are composed based on multiple visualization charts, and the characteristic information that has an important impact on the customer's repayment plan is further screened out through the analysis results.

[0041] Step S32: screening the desensitized customer data according to the analysis result to obtain first characteristic information, where the first characteristic information includes credit score information, income level information, and debt information;

[0042] In this step, the information obtained from the analysis results is used to preliminarily screen the characteristic variables. For example, when it is found that the values ​​of a characteristic variable are almost the same among all customers (such as the standard deviation of a data column is close to zero), then the variable may not be very useful in predicting the customer's repayment plan and it is excluded; or if a category of a categorical variable appears very rarely and the analysis results show that the category has no obvious special relationship with other variables, it can also be considered to be excluded. Therefore, in this step, the first characteristic information finally screened out includes credit score information, income level information, and debt information.

[0043] Step S33: establishing a sample set according to the first feature information and training a preset neural network model according to the sample set to obtain a repayment plan prediction model.

[0044] In this step, there is no limitation on the neural network model, and using a sample set to train a neural network is a technical solution well known to those skilled in the art, so it will not be described in detail here.

[0045] In this embodiment, the customer data after desensitization is screened by exploratory data analysis to obtain the first characteristic information, and a personalized repayment plan can be quickly and accurately formulated for the customer based on the customer's first characteristic information. This method of formulating a personalized repayment plan based on the customer's first characteristic information has a significant effect on improving the customer's experience of using bank customer service. When communicating with bank customer service about repayment matters, customers will no longer face stereotyped and lack of targeted repayment suggestions, but will be able to feel that the bank has truly created an exclusive repayment plan for them based on their actual situation. This not only makes the repayment arrangement more in line with the customer's economic ability and pace of life, and reduces the anxiety and distress caused by excessive repayment pressure or unreasonable plans, but also allows customers to deeply appreciate the bank's attention and care for them, thereby further enhancing customers' satisfaction and loyalty to bank customer service and even the entire bank service.

[0046] Step S4, performing voice recognition on the second information, and matching the customer identity according to the result of the voice recognition to obtain a matching result;

[0047] The step S4 also includes step S41, step S42 and step S43, which specifically include:

[0048] Step S41, preprocessing the second information to obtain a preprocessed speech signal;

[0049] Step S42: extracting features from the preprocessed speech signal to obtain voiceprint features;

[0050] Step S43: determine whether there is a corresponding customer match in the bank database based on the voiceprint feature to obtain a matching result.

[0051] In this step, the matching degree information is obtained by calculating the matching degree between the voiceprint feature corresponding to the second information and the voiceprint feature in the bank database; it is determined whether the matching degree information is greater than the preset matching degree threshold. When the matching degree information is greater than the preset matching degree threshold, it is determined that the two voiceprint features match; when the matching degree information is less than the preset matching degree threshold, it is determined that the two voiceprint features do not match.

[0052] Step S5: determining the customer's repayment plan according to the matching result and the repayment plan prediction model;

[0053] The step S5 also includes step S51, step S52, step S53 and step S54, which specifically include:

[0054] Step S51: when the matching result is that there is a corresponding customer in the bank database, obtain customer data;

[0055] Step S52: Generate first repayment plan information according to the customer data;

[0056] In this step, feature extraction is performed based on the customer data to obtain the first feature information corresponding to the customer; the feature information is sent to the repayment plan prediction model to generate the first repayment plan information. The first repayment plan information is a personalized repayment plan formulated by the bank customer service for the customer based on the customer's first feature information, which can effectively meet the customer's needs.

[0057] Step S53: when the matching result is that there is no corresponding customer in the bank database, converting the second information into corresponding text information;

[0058] In this step, when the matching result is that there is no corresponding customer in the bank database, it is necessary to further determine the customer's repayment tendency based on the communication content between the customer and the voice customer service, so as to formulate the required repayment plan for the customer.

[0059] The step S53 also includes step S531, step S532, step S533 and step S534, which specifically include:

[0060] Step S531, constructing a dialect corpus;

[0061] In this step, data from various dialects are collected to establish a dialect corpus to make up for the deficiencies of the existing corpus.

[0062] Step S532, expanding the voice information in the other party's speech corpus to obtain expanded voice information;

[0063] Since low-resource language collection is difficult and requires a lot of manpower, in this step, the voice information in the other party's language corpus is expanded to solve the problem of corpus scarcity. The specific methods for expanding the voice information in the other party's language corpus include: perturbing the speech speed of the voice information; enhancing the volume of the voice signal; and enhancing the noise of the voice signal.

[0064] Step S533, extracting features from the expanded voice information to obtain second feature information;

[0065] In this step, the specific process of extracting features from the expanded speech information includes: preprocessing the speech signal to obtain a preprocessed speech signal; performing fast Fourier transform on the preprocessed speech signal to obtain spectrum information; calculating the power spectrum information based on the spectrum information; processing the power spectrum information using a Mel filter group, and taking the logarithm of the output of each filter to obtain second feature information, thereby enhancing information with smaller energy to obtain a distinctive speech feature.

[0066] Step S534: Send the second feature information to a speech recognition model to obtain text information.

[0067] In this step, the processing process of sending the second feature information to the speech recognition model is: obtaining a preset number of dilated convolution stacking layers; using the preset number of dilated convolution stacking layers to perform dilated convolution on the second feature information to obtain down-sampled information; sending the down-sampled information to the encoder to obtain text information. Since bank customer service needs to respond to customers quickly, the traditional speech recognition model uses a large amount of calculation parameters, which may result in a long time for bank customer service to respond to customers. Therefore, in this step, using the preset number of dilated convolution stacking layers to perform dilated convolution on the second feature information can reduce space loss and information loss while extracting comprehensive features, so that the amount of calculation parameters is not excessively enhanced, which not only ensures the accuracy of text information translation but also reduces the time for bank customer service to respond to customers.

[0068] Since the bank's customers come from different places, there is a problem of low recognition accuracy when performing voice recognition for some low-resource languages, which leads to mistranslation of text information and makes it impossible to use text information to generate accurate repayment plans for customers, greatly affecting customers' experience of using bank customer service. Therefore, in this embodiment, the voice information in the other party's language corpus is expanded and then feature extraction is performed on the expanded voice information to obtain a second feature signal, which is finally sent to the voice recognition model, which can effectively improve the recognition speed and accuracy of low-resource languages.

[0069] Step S54: Determine the customer's second repayment plan information based on the text information.

[0070] The step S54 also includes step S541, step S542, step S543, step S544, step S545 and step S546, which specifically include:

[0071] Step S541, preprocessing the text information to obtain preprocessed text information;

[0072] Step S542: performing semantic recognition on the preprocessed text information to obtain field information;

[0073] Step S543, obtaining a preset text feature vector;

[0074] In this step, the preset text feature vector is a text feature vector corresponding to professional banking vocabulary related to repayment, wherein the professional banking vocabulary related to repayment includes but is not limited to principal, interest, repayment period, repayment method and repayment progress, etc.

[0075] Step S544: matching the preset text feature vector with the feature vector corresponding to each field in the field information to obtain similarity information;

[0076] In this step, there is no limitation on the method of obtaining the similarity information.

[0077] Step S545, screening is performed according to the similarity information to obtain key field information;

[0078] In this step, the key field information related to repayment in the text information can be identified through similarity information, so as to quickly capture the customer's repayment tendency.

[0079] Step S546: Generate the second repayment plan information according to the key field information.

[0080] In this step, demands related to repayment pressure can be extracted based on key field information, such as heavy repayment pressure; demands related to repayment methods can be extracted, such as equal principal and interest repayment method or interest-first-principal-later repayment method; demands for repayment frequency, repayment flexibility, etc. can be extracted, and the corresponding second repayment method information can be quickly generated for the customer based on the demands of various aspects, and then translated into corresponding voice messages to reply to the customer.

[0081] Step S6: convert the customer's repayment plan into third information, where the third information is voice information corresponding to the repayment plan.

[0082] In this step, the bank customer service staff responds to the customer by converting the customer's repayment plan into voice, thus achieving full automation of the process. The systematic process of the present invention quickly and automatically generates the customer's repayment plan, effectively saving the user's time and enhancing the user experience.

[0083] Embodiment 2:

[0084] like Figure 2 As shown, this embodiment provides an AI-based speech recognition device, which includes an acquisition module 901, a first processing module 902, a second processing module 903, a third processing module 904, a fourth processing module 905 and a fifth processing module 906, which specifically include:

[0085] An acquisition module 901 is used to acquire first information and second information, wherein the first information includes customer data of a bank, and the second information includes voice information of a customer when a customer service representative of the bank communicates with the customer;

[0086] A first processing module 902, configured to perform desensitization processing on the first information to obtain desensitized customer data;

[0087] The second processing module 903 is used to establish a repayment plan prediction model based on the desensitized customer data;

[0088] The third processing module 904 is used to perform voice recognition on the second information, and match the customer identity according to the result of the voice recognition to obtain a matching result;

[0089] A fourth processing module 905 is used to determine the repayment plan of the customer according to the matching result and the repayment plan prediction model;

[0090] The fifth processing module 906 is used to convert the customer's repayment plan into third information, where the third information is the voice information corresponding to the repayment plan.

[0091] In a specific embodiment of the present disclosure, the second processing module further includes a first processing unit, a second processing unit and a third processing unit, which specifically include:

[0092] A first processing unit is used to analyze the desensitized customer data using an exploratory data analysis method to obtain an analysis result;

[0093] A second processing unit is used to screen the desensitized customer data according to the analysis result to obtain first characteristic information, where the first characteristic information includes credit score information, income level information, and debt information;

[0094] The third processing unit is used to establish a sample set according to the first feature information and train a preset neural network model according to the sample set to obtain a repayment plan prediction model.

[0095] In a specific embodiment of the present disclosure, the third processing module further includes a fourth processing unit, a fifth processing unit and a sixth processing unit, which specifically include:

[0096] a fourth processing unit, configured to preprocess the second information to obtain a preprocessed speech signal;

[0097] A fifth processing unit, configured to extract features from the preprocessed speech signal to obtain voiceprint features;

[0098] The sixth processing unit is used to determine whether there is a corresponding customer match in the bank database according to the voiceprint feature, and obtain a matching result.

[0099] It should be noted that, regarding the device in the above embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment of the method, and will not be elaborated here.

[0100] Embodiment 3:

[0101] Corresponding to the above method embodiment, this embodiment also provides an AI-based speech recognition device. The AI-based speech recognition device described below and the AI-based speech recognition method described above can refer to each other.

[0102] Figure 3 8 is a block diagram of an AI-based speech recognition device 800 according to an exemplary embodiment. Figure 3 As shown, the AI-based speech recognition device 800 may include: a processor 801 and a memory 802. The AI-based speech recognition device 800 may also include one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0103] The processor 801 is used to control the overall operation of the AI-based speech recognition device 800 to complete all or part of the steps in the above-mentioned AI-based speech recognition method. The memory 802 is used to store various types of data to support the operation of the AI-based speech recognition device 800, which may include, for example, instructions for any application or method operating on the AI-based speech recognition device 800, and application-related data, such as contact data, sent and received messages, pictures, audio, video, etc. The memory 802 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, disk or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signal may be further stored in the memory 802 or sent via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. The I / O interface 804 provides an interface between the processor 801 and other interface modules, and the above-mentioned other interface modules can be keyboards, mice, buttons, etc. These buttons can be virtual buttons or physical buttons. The communication component 805 is used for wired or wireless communication between the AI-based speech recognition device 800 and other devices. Wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G or 4G, or a combination of one or more of them, so the corresponding communication component 805 can include: Wi-Fi module, Bluetooth module, NFC module.

[0104] In an exemplary embodiment, the AI-based speech recognition device 800 can be implemented by one or more application-specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to execute the above-mentioned AI-based speech recognition method.

[0105] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, and when the program instructions are executed by the processor, the steps of the above-mentioned AI-based speech recognition method are implemented. For example, the computer-readable storage medium can be the above-mentioned memory 802 including program instructions, and the above-mentioned program instructions can be executed by the processor 801 of the AI-based speech recognition device 800 to complete the above-mentioned AI-based speech recognition method.

[0106] Embodiment 4:

[0107] Corresponding to the above method embodiment, a readable storage medium is also provided in this embodiment. The readable storage medium described below and the AI-based speech recognition method described above can refer to each other.

[0108] A readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the AI-based speech recognition method of the above method embodiment.

[0109] The readable storage medium may specifically be a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or other readable storage medium that can store program codes.

[0110] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

[0111] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A speech recognition method based on AI, characterized in that: include: Acquire first information and second information, wherein the first information includes customer data of the bank, and the second information includes voice information of the customer when the bank customer service communicates with the customer; Performing desensitization processing on the first information to obtain desensitized customer data; Establishing a repayment plan prediction model based on the desensitized customer data; Performing voice recognition on the second information, and matching the customer identity according to the result of the voice recognition to obtain a matching result; Determining the repayment plan of the customer based on the matching result and the repayment plan prediction model; The repayment plan of the customer is converted into third information, where the third information is voice information corresponding to the repayment plan.

2. The AI-based speech recognition method according to claim 1, characterized in that: A repayment plan prediction model is established based on the desensitized customer data, including: Analyze the desensitized customer data using an exploratory data analysis method to obtain analysis results; Screening the desensitized customer data according to the analysis results to obtain first characteristic information, where the first characteristic information includes credit score information, income level information, and debt information; A sample set is established according to the first feature information, and a preset neural network model is trained according to the sample set to obtain a repayment plan prediction model.

3. The AI-based speech recognition method according to claim 1, characterized in that: Performing voice recognition on the second information, and matching the customer identity according to the result of the voice recognition to obtain a matching result, including: Preprocessing the second information to obtain a preprocessed speech signal; Extracting features from the preprocessed speech signal to obtain voiceprint features; It is determined based on the voiceprint features whether there is a corresponding customer match in the bank database to obtain a matching result.

4. The AI-based speech recognition method according to claim 1, characterized in that: Determining the customer's repayment plan according to the matching result and the repayment plan prediction model includes: When the matching result is that there is a corresponding customer in the bank database, obtaining customer data; generating first repayment plan information according to the customer data; When the matching result is that there is no corresponding customer in the bank database, converting the second information into corresponding text information; The customer's second repayment plan information is determined based on the text information.

5. The AI-based speech recognition method according to claim 4, characterized in that: Converting the second information into corresponding text information includes: Constructing dialect corpus; Expanding the voice information in the other party's speech corpus to obtain expanded and processed voice information; Extracting features from the expanded voice information to obtain second feature information; The second feature information is sent to a speech recognition model to obtain text information.

6. A speech recognition device based on AI, characterized in that: include: An acquisition module, used to acquire first information and second information, wherein the first information includes the customer data of the bank, and the second information includes the voice information of the customer when the bank customer service communicates with the customer; A first processing module, used for performing desensitization processing on the first information to obtain desensitized customer data; A second processing module is used to establish a repayment plan prediction model based on the desensitized customer data; A third processing module, configured to perform voice recognition on the second information, and match the customer identity according to the result of the voice recognition to obtain a matching result; A fourth processing module, used to determine the repayment plan of the customer according to the matching result and the repayment plan prediction model; The fifth processing module is used to convert the customer's repayment plan into third information, where the third information is the voice information corresponding to the repayment plan.

7. The AI-based speech recognition device according to claim 6, characterized in that: The second processing module comprises: A first processing unit is used to analyze the desensitized customer data using an exploratory data analysis method to obtain an analysis result; A second processing unit is used to screen the desensitized customer data according to the analysis result to obtain first characteristic information, where the first characteristic information includes credit score information, income level information, and debt information; The third processing unit is used to establish a sample set according to the first feature information and train a preset neural network model according to the sample set to obtain a repayment plan prediction model.

8. The AI-based speech recognition device according to claim 6, characterized in that: The third processing module comprises: a fourth processing unit, configured to preprocess the second information to obtain a preprocessed speech signal; A fifth processing unit, configured to extract features from the preprocessed speech signal to obtain voiceprint features; The sixth processing unit is used to determine whether there is a corresponding customer match in the bank database according to the voiceprint feature, and obtain a matching result.

9. An AI-based speech recognition device, characterized in that: include: Memory for storing computer programs; A processor, configured to implement the steps of the AI-based speech recognition method as claimed in any one of claims 1 to 5 when executing the computer program.

10. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the AI-based speech recognition method as claimed in any one of claims 1 to 5 are implemented.

Images