A business processing method and device

Abstract

The present invention discloses a business processing method and device, including: using a speech recognition model to recognize the collected business instruction voice, generate text information, and extract environmental information (such as emotional state) through speech feature analysis; parsing text information to generate payment, transfer, inquiry and other instruction information; integrating instruction information and environmental information to generate analysis results, and guiding users to execute or cancel operations accordingly. In order to improve security, the present invention introduces voiceprint identity authentication technology to ensure the authenticity of the user's identity, and provides confirmation inquiries before payment operations. The voice collection effect is enhanced by beamforming technology, and the accuracy of the analysis results is optimized by multimodal data fusion technology. The device of the present invention has a modular design, including speech recognition, analysis and execution modules, which is easy to integrate and expand. It is suitable for scenarios such as intelligent customer service and voice payment. It has the characteristics of high efficiency, accuracy and safety, and has wide application value.

Description

Technical Field

[0001] The present invention relates to the technical field of speech recognition and business processing, and in particular to a business processing method and device based on speech recognition and environmental information analysis. Background Art

[0002] In today's digital age, the demand for convenience and intelligence in business processing is growing. Traditional business processing methods, especially when it comes to payment operations, often rely on manual input of information, which is inconvenient for some special scenarios or specific groups of people (such as people with limited vision, people whose hands are busy and cannot operate equipment, etc.). At the same time, when processing payment services, existing business processing systems usually only focus on the content of the payment instructions themselves, and rarely consider other factors related to payment, such as the user's environmental information and emotional state. This may lead to unreasonable payment decisions in some cases, affecting user experience and even bringing certain payment risks. In addition, the current system may face problems such as environmental noise interference and low efficiency of identity authentication during the execution of voice recognition and payment operations, and urgently needs smarter and safer business processing methods.

[0003] Publication number CN118246910A discloses a conversational online payment method, system, medium, device and program product, including: obtaining a user's audio signal; for the audio signal, obtaining a noise-reduced audio signal through a sound noise reduction model; in response to a selected language category, calling a speech recognition model corresponding to the selected language category, and converting the noise-reduced audio signal into text; based on the text, parsing the user's intention through a large language model; based on the text, obtaining the user's emotion category through a sentiment analysis model; if the user's intention is to inquire about the fees, then a bill query is performed, and intervention measures are selected based on the user's emotion category to push the bill; if the user's intention is to pay the fees, then intervention measures are selected based on the user's emotion category to guide the user to complete the payment operation. Summary of the invention

[0004] The present invention provides a business processing method and device, which aims to achieve a more intelligent, convenient and secure business processing process by making full use of voice recognition, analysis technology and multimodal data fusion, especially in payment business, and can comprehensively consider various factors to make reasonable payment operation guidance, improve user experience and ensure payment security.

[0005] The present invention provides a business processing method, comprising the following specific steps:

[0006] Step S1: Voice information collection and preliminary analysis

[0007] The speech recognition model recognizes the business command voice collected by the voice collection device and converts the voice signal into processable text information. At the same time, the voice feature analysis technology is used to analyze the business command voice and extract environmental information from it. The environmental information here covers many aspects, such as the background noise of the voice, the potential information related to the user's emotions reflected by the voice intonation and speed, etc.

[0008] Step S2: Further processing of information

[0009] The text information obtained in step S1 is processed in a targeted manner to extract instruction information, which specifically includes transfer, consumption payment, inquiry, etc., as well as amount, payee information, etc. At the same time, the previously obtained environmental information is processed to obtain state information, an important part of which is emotional information, which can reflect the emotional state of the user when issuing the business instruction voice, such as calm, anxious, hesitant, etc.

[0010] Step S3: Comprehensive analysis

[0011] The environmental information and instruction information obtained in step S2 are comprehensively analyzed, and the two types of information in different modes are fully integrated through multimodal data fusion technology to fully and deeply understand the business scenario and user intention, and finally obtain the analysis results. This multimodal data fusion technology can overcome the limitations of single information analysis and more accurately grasp the overall situation of the payment business.

[0012] Step S4: Payment operation instructions

[0013] According to the analysis results obtained in step S3, the user's payment operation is guided. The payment operation mainly includes two situations: payment and cancellation of payment. The specific guidance decision is based on a comprehensive judgment of various factors reflected in the analysis results, such as the size of the payment amount, the user's emotional state, the risk level of the payment type, etc., to ensure that the payment operation is in line with the user's true intention and is reasonable and safe.

[0014] The advantages of the present invention compared with the prior art are:

[0015] (1) Business processing is more secure;

[0016] (2) A wider range of application scenarios;

[0017] (3) The ability to obtain instructions with strong anti-interference capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A flowchart of a business processing method;

[0019] Figure 2A flowchart of a business processing method in some embodiments;

[0020] Figure 3 is a block diagram of a business processing device according to some other embodiments of the present disclosure. DETAILED DESCRIPTION

[0021] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0022] Example 1

[0023] This embodiment provides a service processing method and device based on speech recognition and speech feature analysis technology, which analyzes user needs through the combination of text information and environmental information, and guides users to complete service operations. Figure 1 , the following are the detailed implementation steps:

[0024] Step S1: Speech Recognition and Environment Analysis

[0025] The user's business command voice is recorded through voice collection equipment (such as smart phones, microphone arrays). Beamforming technology is used to enhance the collection effect of voice signals and improve voice quality.

[0026] The speech recognition model is used to recognize the collected business instruction voice to obtain the corresponding text information.

[0027] Extract voice features from business command voices and analyze voice environment information, including noise level and user emotional state. Emotion recognition: Use emotion recognition technology to extract emotional information (such as anger, calmness, anxiety) to provide reference for subsequent operations.

[0028] Step S2: Information processing

[0029] The text generated by the speech recognition is input into the second speech recognition module to extract the business instruction information, including:

[0030] Payment information (such as payment amount, account number);

[0031] Transfer information (such as recipient account);

[0032] Query information (such as balance, transaction history).

[0033] Based on the voice environment information extracted by features, status information is generated (such as whether the user's emotional state is normal and whether the environment is quiet).

[0034] Step S3: Instruction and status analysis

[0035] Perform multimodal data fusion processing on text information and environmental information;

[0036] Generate an analysis report based on the fusion result, including the priority of user requirements and whether further verification is required. When the emotion information is abnormal (such as the user is angry or uneasy), trigger a confirmation query preferentially. When the environmental noise is too loud or the command voice is unclear, prompt the user to re-enter the command.

[0037] Step S4: Guide the user to operate

[0038] Provide a payment confirmation query to ensure the user's informed consent for payment or transfer operations; if the analysis result shows a high risk (such as the user's emotion is abnormal, the command is unclear), cancel the operation and prompt the user.

[0039] Perform business operations according to the analysis result (such as completing payment, transfer, or query); if the analysis result shows an abnormality, cancel the operation and suggest the user to retry or contact the customer service.

[0040] A business processing method and device based on speech recognition and speech feature analysis technology. Before guiding the user to operate, it also includes performing voiceprint identity authentication on the user's business command voice to verify the user's identity. The process is as Figure 2 shown. The implementation of voiceprint authentication includes the following steps:

[0041] Collect the voice command of the current user, and compare the features of the real-time voice with the pre-stored template:

[0042] Dynamic Time Warping (DTW): Calculate the time alignment distance between the input voice and the template:

[0043] D(i,j) = \|Q(i) - T(j)\|^2 + min\{D(i - 1,j), D(i,j - 1), D(i - 1,j - 1)\} where Q(i) and T(j) are the feature vectors of the input voice and the template respectively. Then use a Convolutional Neural Network (CNN) or an end-to-end speaker recognition network: SpeakerNet to extract deep features and perform matching. Judge whether the authentication is passed according to the similarity score S:

[0044] If S ≥ T (the set threshold), the authentication is passed.

[0045] If S < T, the authentication fails, trigger a security prompt or request the user to retry.

[0046] Embodiment 2

[0047] A business processing device based on speech recognition and speech feature analysis technology, whose structural block diagram is as Figure 3 shown. The operation process of this device includes the following steps:

[0048] Step 1: Voice collection and processing

[0049] The user inputs a command (such as "pay 200 yuan to Zhang San") through the voice collection module, and the system completes the following processing:

[0050] The first speech recognition module converts the speech signal into text information;

[0051] The first speech recognition module extracts environmental information (such as noise level, speech clarity, etc.);

[0052] The voiceprint recognition module extracts the user's voiceprint features and compares them with the voiceprint sample library.

[0053] Step 2: Voiceprint recognition and identity authentication

[0054] If the voiceprint matches successfully, the next step will be entered, and the prompt "Authentication successful, please continue" will be displayed; if the voiceprint matches fail, the operation will be terminated and the user will be prompted "Authentication failed, please try again".

[0055] The second speech recognition module parses the text information and generates instruction information:

[0056] Instruction type: payment;

[0057] Payment amount: 200 yuan;

[0058] Payee: Zhang San.

[0059] The system comprehensively considers environmental information (low noise level) and user status information (normal mood) to determine whether to continue the operation.

[0060] Step 4: Analysis and decision making

[0061] The analysis module integrates text information, voiceprint comparison results, environmental information, etc. to generate comprehensive analysis results:

[0062] If the matching result and instruction information meet the safety conditions, the execution phase will begin;

[0063] If an abnormality is detected (such as identity mismatch, abnormal emotions, etc.), the operation will be canceled.

[0064] Step 5: Implementation and Feedback

[0065] Provide payment confirmation prompt: "200 yuan will be paid to Zhang San, please confirm."

[0066] If the user confirms, the payment is completed and feedback is given that the operation is successful;

[0067] If the user does not confirm, the operation will be canceled and a message will be displayed saying “Payment canceled”.

[0068] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A business processing method, characterized in that: include: S1. The first voice module recognizes the business command voice collected by the voice collection device to obtain text information; the first voice recognition module analyzes the business command voice to obtain environmental information; S2, the second voice module processes the text information to obtain instruction information, the instruction information includes payment information, transfer information and query information; processes the environmental information to obtain state information, the state information includes emotional information; S3, the analysis module obtains analysis results for the instruction information and the status information; S4. According to the analysis result, guide the user to perform business operations, where the business operations include operation execution and operation cancellation.

2. A business processing method according to claim 1, characterized in that: Before the speech recognition model recognizes the payment instruction speech collected by the speech collection device to obtain text information, the method further includes: Data collection and preprocessing: The preprocessing module collects and cleans the data and divides the data into a training set and a test set; Model training: the speech recognition model learns the training set; Model testing: Use the test set to verify the performance of the speech recognition model.

3. A business processing method according to claim 2, characterized in that: Before guiding the user to perform a business operation according to the analysis result, wherein the business operation includes operation execution and operation cancellation, the method further includes: The user is authenticated by voiceprint according to the service instruction voice data, wherein the service instruction voice data includes user input voice data.

4. A business processing method according to claim 3, characterized in that: Prior to guiding the user to perform business operations, the method further includes: providing a confirmation inquiry for executing payment.

5. A business processing method according to claim 4, characterized in that: The voice collection includes using beamforming technology to enhance the collection effect of the payment instruction voice.

6. A business processing method according to claim 5, characterized in that: The analysis results are obtained through multimodal data fusion technology.

7. A business processing method according to any one of claims 1 to 6, characterized in that: The operation execution or the operation cancellation is performed according to the analysis result.

8. A service processing device, characterized in that: include The first speech recognition module is used for the speech recognition model to recognize the business instruction speech collected by the speech collection device to obtain text information; The voice feature analysis technology analyzes the business command voice to obtain environmental information; The second speech recognition module is used to process the text information to obtain instruction information, which includes payment information, transfer information and query information; and to process the environmental information to obtain state information, which includes emotional information. Analysis module: used to analyze the instruction information and the environmental information to obtain analysis results; guidance execution module: used to guide user business operations based on the analysis results, and the business operations include operation execution and operation cancellation.

Images