Video interview system for docking employer and job seeker

Abstract

The invention, which relates to the technical field of information, discloses a video interview system for docking an employer and a job seeker, comprising: an identity verification module for a user to register through a mobile terminal APP and select the job seeker and the employer to verify the identity information of the user; and the intelligent matching module generates initial credit points according to the verified user identity information, issues post information to a database and generates a unique post ID based on the post information issued by the employer, and preliminarily matches a resume file with the post information based on uploading of the resume file of the job seeker to obtain an original matching score. According to the method, multi-dimensional data such as video streams and audio streams are collected in real time, dual verification of face feature comparison and voiceprint analysis is achieved through the identity forgery detection engine, real-time consistency verification is conducted on resume content in combination with the NLP technology, a dynamic identity authenticity evaluation system is constructed, and the authenticity and reliability of the interview process are effectively guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a video interview system for connecting employers and job seekers. Background Art

[0002] With the rapid development of Internet technology, online recruitment platforms have become an important channel for connecting employers and job seekers. Traditional recruitment systems mainly rely on resume text matching and static information screening. The job recommendation algorithm based on keyword retrieval parses the structured data such as skills and experience in the job seeker's resume, calculates the similarity with the job description (JD) published by the employer, and generates matching results. To a certain extent, it improves recruitment efficiency and reduces manual screening costs. However, the existing system still has certain limitations. Many recruitment websites are recruited by intermediaries, who cannot directly negotiate with employers and charge high fees.

[0003] In existing technologies, the credit systems of recruitment platforms are mostly based on one-way evaluation or simple behavioral statistics, and lack a multi-dimensional dynamic evaluation mechanism. The platforms use a static credit score model, calculating credit values ​​only based on the completeness of the materials submitted by the user or the historical delivery records, and are unable to combine dynamic factors such as identity verification results and language expression ability in video interviews in real time for comprehensive scoring. This credit assessment method makes it difficult to reflect the credibility of users in real-world interaction scenarios, resulting in low efficiency in building trust between employers and job seekers. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a video interview system for connecting employers and job seekers to solve the problems of insufficient identity authenticity verification and lack of interactive credibility assessment caused by static credit scoring in existing recruitment platforms.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] The present invention provides a video interview system for connecting employers and job seekers, which includes an identity verification module. Users register through a mobile APP and select job seekers and employers, and the user's identity information is verified;

[0008] The intelligent matching module generates an initial credit score based on the verified user identity information. Based on the job information released by the employer, the module publishes the job information to the database and generates a unique job ID. Based on the uploaded resume of the job seeker, the module performs a preliminary match between the resume and the job information to obtain the original matching score.

[0009] The appointment management module uses NLP technology and collaborative filtering algorithms to match positions, generate interview request status and appointment information, and prompt users to pay a small fee to improve matching quality;

[0010] The dual authenticity assessment module allows job seekers and employers to confirm and jump to the video call interface for a video conference. This module collects multi-dimensional data, runs the identity forgery detection engine and resume authenticity assessment in real time, and obtains a dual authenticity index.

[0011] In the evaluation module, job seekers and employers provide mutual feedback, update credit scores and generate interview evaluation reports.

[0012] As a preferred solution of the video interview system for connecting employers and job seekers of the present invention, the user registers through the mobile APP and selects the job seeker and employer, and verifies the user's identity information, including the following steps:

[0013] Users enter the registration page through the mobile app and select roles as job seekers or employers. Users upload certification materials and use the OCR interface to extract document information.

[0014] Use facial liveness detection on job applicants, compare authentication materials with real-time facial features, and output facial feature matching results;

[0015] Conduct AI review on the certification materials uploaded by the employer to obtain the certification status.

[0016] As a preferred solution of the video interview system for connecting employers and job seekers of the present invention, generating initial credit points based on the verified user identity information includes the following steps:

[0017] Based on the verified user identity information, check the user's historical behavior data and analyze third-party related data;

[0018] The initial credit score is calculated using a logistic regression algorithm.

[0019] As a preferred solution of the video interview system for connecting employers and job seekers of the present invention, the following steps are included: based on the job information released by the employer, the job information is published to the database and a unique job ID is generated:

[0020] The employer enters basic job information, performs compliance verification on the submitted job information, and generates standardized job data;

[0021] Use the cloth ID generation service to send it to the database to generate a unique position ID.

[0022] As a preferred solution of the video interview system for connecting employers and job seekers described in the present invention, the following steps are included: based on the uploaded resume file of the job seeker, the resume file is preliminarily matched with the job information to obtain the original matching score:

[0023] Based on the resume files uploaded by job seekers, we use the resume parsing engine to extract structured data and generate a resume feature set in JSON format;

[0024] The JD features of the target position are obtained from the position information, matched with the resume feature set, and the original matching score is obtained.

[0025] As a preferred solution of the video interview system for connecting employers and job seekers described in the present invention, the system matches positions based on NLP technology and collaborative filtering algorithms, generates interview request status and appointment information, and includes the following steps:

[0026] Use NLP technology to calculate the text similarity between the resume feature set and the job description to obtain a text matching score;

[0027] Collaborative filtering algorithms analyze user historical behavior data and generate collaborative recommendation weights;

[0028] The text matching score and collaborative recommendation weight are integrated to calculate and sort the comprehensive recommendation score, and generate the interview request status and appointment information.

[0029] As a preferred solution of the video interview system for connecting employers and job seekers of the present invention, wherein: prompting users to pay a small fee to improve the matching quality includes the following steps:

[0030] Upon detecting that a user has initiated a video interview request, the credit score of the user will be automatically checked;

[0031] Get a suggested payment amount based on the user's credit rating and current match quality;

[0032] Remove restrictions on interview requests and increase the matching priority of interview requests.

[0033] As a preferred solution of the video interview system for connecting employers and job seekers of the present invention, the job seeker and the employer confirm and jump to the video call interface to conduct a video conference, including the following steps:

[0034] Automatically generate meeting room numbers and access tokens based on confirmation by job seekers and employers;

[0035] Initialize session parameters and load user device information through the audio and video SDK;

[0036] Conduct a one-on-one video interview based on the conference room number and access token and activate the audio and video streaming channel.

[0037] As a preferred solution of the video interview system for connecting employers and job seekers described in the present invention, wherein: collecting multi-dimensional data, running the identity forgery detection engine and resume authenticity assessment in real time to obtain a dual authenticity index, including the following steps:

[0038] Initiate a video interview to collect real-time video stream, audio stream, device fingerprint, and IP address information;

[0039] The identity forgery detection engine extracts the current face image features, compares them with the face feature matching results, and obtains the face matching score;

[0040] The resume authenticity assessment module analyzes real-time voice content and uses NLP technology to compare resume text descriptions to obtain a resume consistency score.

[0041] Based on the dynamic fusion engine, the scores of each dimension are weighted and calculated to generate the identity consistency index;

[0042] The identity consistency index is combined with the resume consistency score weight to output a dual authenticity index.

[0043] As a preferred solution of the video interview system for connecting employers and job seekers of the present invention, the job seekers and employers conduct mutual evaluation and feedback, update credit points and generate interview evaluation reports, including the following steps:

[0044] Based on the dual authenticity index, the mutual evaluation process is triggered, and the evaluation interface is pushed to job seekers and employers to obtain the mutual evaluation score;

[0045] Weighting the mutual evaluation scores of job seekers and employers to obtain the interactive credit change value;

[0046] Combine the mutual evaluation score and the credit change value and update it to the credit score;

[0047] The resume consistency scores and mutual evaluation scores are extracted to obtain an evaluation report.

[0048] The beneficial effects of the present invention are as follows: by real-time collection of multi-dimensional data such as video streams and audio streams, the identity forgery detection engine realizes dual verification of facial feature comparison and voiceprint analysis, combines NLP technology to perform real-time consistency verification of resume content, and constructs a dynamic identity authenticity evaluation system, which effectively guarantees the authenticity and reliability of the interview process. By combining the dual authenticity index with mutual evaluation feedback data, a closed-loop credit score update mechanism is established, so that the credit evaluation can reflect the user's true performance during the interview process in real time. On the one hand, the identity authenticity of the interview participants is ensured through multimodal data collection. On the other hand, the dynamic credit evaluation realizes the accurate quantification of user behavior, and together builds a trustworthy and efficient intelligent video interview ecosystem, which fundamentally improves the quality and efficiency of recruitment matching. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0050] Figure 1 A schematic diagram of a video interview system for connecting employers and job seekers.

[0051] Figure 2 Schematic diagram of user authentication. DETAILED DESCRIPTION

[0052] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0053] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0054] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0055] Reference Figure 1 and Figure 2, which is the first embodiment of the present invention, provides a video interview system for connecting employers and job seekers, including the following steps:

[0056] Identity verification module: users register through the mobile app and select job seekers and employers to verify user identity information.

[0057] Users register through the mobile APP and select job seekers and employers to verify their identity information.

[0058] Users enter the registration page through the mobile APP and select roles as job seekers and employers. Users upload certification materials and use the OCR interface to extract document information.

[0059] Furthermore, the user selects his or her role type as a job seeker or employer on the registration page of the mobile APP, and then enters the identity information filling link. According to the selected role type, the corresponding certification materials are uploaded respectively; if the role type selected by the user is a job seeker, the ID card photo is uploaded as the certification material; if the role type selected by the user is an employer, a copy of the company's business license and the legal person's authorization letter are uploaded as the certification material, and the certification material is format checked to determine whether it is a valid document image file, for example, whether it is a clear document photo in JPG or PNG format. After the verification is passed, the OCR interface is called to recognize the document image content in the certification material.

[0060] Use facial liveness detection on job applicants, compare authentication materials with real-time facial features, and output facial feature matching results.

[0061] Furthermore, after the user completes the identity information and uploads the ID card photo, the face collection phase begins. The face liveness detection process is initiated, and the device camera is called through the mobile app to collect the user's frontal face image in real time. The user is guided to complete actions such as blinking, opening the mouth, and turning the head left and right according to the prompts to ensure that the user is truly alive. The face detection algorithm is used to locate the face area in the video frame and extract the facial key point information for subsequent feature encoding.

[0062] The face recognition algorithm is used to perform feature encoding on the collected real-time face image to generate a real-time face feature vector. At the same time, the ID face image is extracted from the uploaded ID card photo and feature encoded using the face recognition algorithm to generate an ID face feature vector. The real-time face feature vector and the ID face feature vector are similarly calculated, and the cosine similarity method is used to compare the degree of match between the two. For example, the similarity value example is 0.93. The match is determined to be successful based on the preset similarity threshold. For example, if the threshold is set to 0.85 and the calculated result is greater than this value, it is determined to be a successful match. Finally, the face feature matching result is output.

[0063] Conduct AI review on the certification materials uploaded by the employer to obtain the certification status.

[0064] Furthermore, after completing the role selection on the registration page, the employer enters the identity information filling stage and uploads a copy of the company's business license and a legal person's authorization letter as certification materials. The certification materials are formatted and checked to determine whether they are valid image files or PDF documents, such as whether they are clear JPG, PNG, or standard PDF format files. The OCR interface is called to recognize the content of the document image in the certification materials. The OCR interface locates and extracts key fields in the certification materials, such as the company name, unified social credit code, legal representative, and business term information in the company's business license, or the name of the authorized person, the name of the authorized unit, the scope of authorization, and signature information in the legal person's authorization letter.

[0065] After the OCR interface completes recognition, it outputs the extracted key fields in the form of structured text and performs a consistency comparison with the company information manually entered by the user. For example, it determines whether the company name extracted from the business license is consistent with the company name filled in by the user, or whether the unified social credit code extracted from the legal person's authorization letter matches the company code entered by the user. The image quality detection algorithm is used to judge the integrity of the certification materials, such as detecting blur, incomplete cropping, and reflection issues. If the consistency comparison result meets the preset rules and the image quality meets the standards, the certification status of the employer is marked as certified;

[0066] The intelligent matching module generates an initial credit score based on the verified user identity information. Based on the job information released by the employer, the job information is published to the database and a unique job ID is generated. Based on the uploaded resume file of the job seeker, the resume file is preliminarily matched with the job information to obtain the original matching score.

[0067] Generate initial credit points based on the verified user identity information.

[0068] Based on the verified user identity information, check the user's historical behavior data and analyze third-party related data.

[0069] Furthermore, after completing the user identity information verification, the user's unique identity identification information is obtained, and used as an index to query the user's historical behavior records, including but not limited to the number of video interviews initiated or accepted in the past, mutual evaluation feedback scores, mid-way interview cancellation records, small fee payments and refunds. At the same time, a third-party data interface is called to obtain external behavior data associated with the user, and the employer's credit rating results are obtained from the enterprise credit information system; the obtained historical behavior data and third-party related data are classified and processed, and key indicators that can be used for credit assessment are extracted. Statistics are made on whether there are frequent modifications or illegal releases in the job information released by the employer, and the extracted key indicators are correlated with the current user identity verification status for analysis to determine whether it meets the behavioral compliance standards set by the platform, such as determining whether the job seeker has failed to connect for video interviews multiple times, or determining whether the employer has a large number of low-scoring feedback records.

[0070] The initial credit score is calculated using a logistic regression algorithm.

[0071] Furthermore, after completing the user identity information verification and historical behavior data collection, the extracted key indicators are used as input variables, including whether the user has completed real-name authentication, the role type is a job seeker or an employer, the interview completion rate, the average score of mutual evaluation, whether there is a violation record, the third-party credit score and other feature fields; all input variables are standardized so that they are in a unified dimension range, Boolean variables are converted to 0 or 1, and percentage variables are normalized to the [0, 1] interval. The standardized input variables are linearly weighted summed using a logistic regression algorithm, and the linear results are mapped to the [0, 1] probability interval through the Sigmoid function to obtain the probability value of the user's credit rating; according to the preset credit rating classification rules, the probability value is mapped to the corresponding credit score.

[0072] Based on the job information released by the employer, the job information is published to the database and a unique job ID is generated.

[0073] The employer enters basic job information, performs compliance verification on the submitted job information, and generates standardized job data.

[0074] Furthermore, employers enter the job posting page in the mobile APP and manually fill in or select to enter the basic information fields of the position, including job title, salary range, work location, job responsibilities, skill requirements, years of experience, whether remote work is accepted, etc., and start the job information compliance verification process, and perform format checks and logical verification on the content of each field in turn. For the skill requirement field, a keyword whitelist mechanism is used to determine whether there are sensitive words or non-professional terms. If any field fails the compliance verification, the user is prompted to modify it and the abnormal field is highlighted; if all fields pass the verification, the job information is organized into a data record in a unified format according to the preset standardized structure, and standardized job data that complies with platform specifications is output as the data input basis for the subsequent job information release to the database and generation of a unique job ID process.

[0075] Use the cloth ID generation service to send it to the database to generate a unique position ID.

[0076] Furthermore, after the job information submitted by the employer completes compliance verification and generates standardized job data, the unique job ID generation process is started; the distributed ID generation service is called, which generates a globally unique and increasing integer identifier based on the timestamp, node identifier and serial number combination, for example, using the Snowflake algorithm implementation method, where the timestamp part represents the millisecond value of the generation time, the node identifier part represents the current server node number, and the serial number part is used to distinguish multiple generation requests under the same timestamp. The generated distributed ID is used as the primary key field of the job record and is submitted to the database write interface together with the standardized job information. After the database receives the write request, it inserts a new record in the target job information table and uses the generated distributed ID as the unique identifier of the job information for persistent storage; the final output is a complete job information record containing a unique job ID.

[0077] Based on the uploaded resume file of the job seeker, the resume file is preliminarily matched with the job information to obtain the original matching score.

[0078] Based on the resume files uploaded by job seekers, the resume parsing engine is used to extract structured data and generate a resume feature set in JSON format.

[0079] Furthermore,

[0080] The JD features of the target position are obtained from the position information, matched with the resume feature set, and the original matching score is obtained.

[0081] Furthermore, after completing identity information verification, job seekers enter the resume upload interface in the mobile APP, select the locally stored PDF or Word format resume file to submit, start the resume parsing process, and call the resume parsing engine to identify the content of the uploaded file and extract fields. The resume parsing engine uses predefined field templates and natural language processing technology to extract key information such as personal information, educational background, work experience, professional skills, language ability, etc. in the resume. For unstructured description content in text paragraphs, the named entity recognition method is used to identify structured elements such as time, company name, position title, skill keywords, etc., and organize them into key-value pairs in a unified format. The extracted structured fields are mapped to the standard resume feature field set, and finally output JSON format data records containing complete resume feature information. The resume and job information are preliminarily matched and scored on a percentage basis to obtain the original matching score.

[0082] The appointment management module matches positions based on NLP technology and collaborative filtering algorithms, generates interview request status and appointment information, and prompts users to pay a small fee to improve matching quality.

[0083] Match positions based on NLP technology and collaborative filtering algorithms, and generate interview request status and appointment information.

[0084] The text similarity between the resume feature set and the job description is calculated using NLP technology to obtain a text matching score.

[0085] Furthermore, standardized job description features are extracted from the job information released by the employer, including fields such as job title, salary range, work location, job responsibilities, skill requirements, and years of experience, and the above fields are organized into a target job feature set with a JSON structure in a unified format; at the same time, corresponding personal information, educational background, work experience, professional skills, language ability and other feature fields are extracted from the resume feature set in JSON format uploaded and parsed by the job seeker, and the keyword overlap calculation method is used to match the skill requirement field in the target job feature set with the professional skill field in the resume feature set, and the number of overlapping keywords is counted to obtain a text matching score.

[0086] The collaborative filtering algorithm analyzes user historical behavior data and generates collaborative recommendation weights.

[0087] Furthermore, user historical behavior data is extracted from the historical interaction records of job seekers and employers, including but not limited to job browsing records, resume submission records, video interview initiation and acceptance, mutual evaluation feedback scores, small fee payment behavior and other fields, and organized into a structured data set in chronological order. The user historical behavior data is normalized, and different types of interaction behaviors are converted into numerical representations of unified dimensions. The similarity between users is calculated using a user-based collaborative filtering method, and the user behavior vectors are compared using the cosine similarity algorithm to obtain the similarity scores between users. At the same time, the similarity between jobs is calculated using a job-based collaborative filtering method, and the job feature vectors are also compared using the cosine similarity algorithm to obtain the similarity scores between jobs. Based on the user-user similarity score and the job-job similarity score, combined with the current user's recent behavior preferences, the collaborative recommendation weight of each non-interacted job is obtained.

[0088] The text matching score and collaborative recommendation weight are integrated to calculate and sort the comprehensive recommendation score, and generate the interview request status and appointment information.

[0089] Furthermore, after obtaining the original text matching score between the job seeker's resume feature set and the position feature set, combined with the collaborative recommendation weight generated based on the user's historical behavior data, the two are weighted and merged according to the preset fusion rules, and the two are added to obtain the comprehensive recommendation score corresponding to each position. The comprehensive recommendation scores of all recommended positions are normalized so that they are in a unified scoring range, for example, converted into a percentage scoring form, and all positions are sorted in descending order according to the normalized comprehensive recommendation scores to generate a position recommendation ranking list; for positions with high rankings, job seekers or employers are allowed to initiate video interview requests. When either party clicks the initiate video interview button, the current time, position ID, initiator's identity and target position information are recorded, and an interview request status field is generated. For example, the initial status is set to initiated-waiting for confirmation, and an appointment option is provided for setting the expected interview time. Finally, a complete interaction record containing the interview request status and appointment information is output.

[0090] Prompt users to pay a small fee to improve matching quality.

[0091] When a user initiates a video interview request, the credit score of the user is automatically checked.

[0092] Furthermore, when a job seeker or employer clicks the button to initiate a video interview in the mobile APP, the current operation time, the initiator's identity, the target position ID and the other party's user identity are recorded, and the credit score query process is triggered; based on the initiator's identity as the query index, the credit score record corresponding to the user is retrieved from the credit score database, and the credit score record includes but is not limited to fields such as the initial credit score, historical behavior adjustment score, and third-party related data score. The latest valid credit score value is extracted from the credit score record, and it is determined whether the credit score meets the minimum credit requirements set by the platform. If the credit score meets the requirements, the video interview request process is allowed to continue. If the credit score does not meet the requirements, the user is prompted that the current credit status does not meet the initiation conditions, and operational instructions for improving the credit score are provided. The credit score query result is output as the data basis for subsequently prompting the user to pay a small fee and control the interview request authority.

[0093] Based on the user's credit rating and the current matching quality, a suggested payment amount is obtained.

[0094] Furthermore, after the user initiates a video interview request and completes the credit score query, the user's credit rating information is obtained. The credit rating information is mapped to a credit rating range based on the credit score value. The comprehensive recommendation score between the current target position and the job seeker is obtained from the job matching process. This is used as the core indicator to measure the matching quality. The preset tiered billing rules are used to determine small payments based on the two dimensions of credit rating and matching quality.

[0095] Remove restrictions on interview requests and increase the matching priority of interview requests.

[0096] Furthermore, after the user completes the credit score query and dynamically calculates the recommended payment amount, the user completes the payment according to the recommended amount or the platform determines that the user's credit rating meets the automatic restriction lifting standard, and the credit rating is A and the historical behavior record is good, then the restrictions on the user initiating a video interview request will be lifted, including canceling the daily upper limit on the number of interview requests, allowing requests to be made to multiple positions at the same time, etc.; at the same time, in the job matching sorting process, the user's interview request is marked as a high-priority processing object, and its comprehensive recommendation score is increased by a preset weight based on the original calculation, thereby improving its candidate ranking order displayed on the employer side; finally, the restriction lifting status update information and the adjusted matching priority are output.

[0097] In the dual authenticity assessment module, job seekers and employers confirm and jump to the video call interface to conduct a video conference, collect multi-dimensional data, run the identity forgery detection engine and resume authenticity assessment in real time, and obtain a dual authenticity index.

[0098] Job seekers and employers confirm and jump to the video call interface to conduct a video conference.

[0099] Based on the confirmation of the job seeker and the employer, the meeting room number and access token are automatically generated.

[0100] Furthermore, after the job seeker and the employer complete the interview request confirmation, the video conference initialization process is started. The conference room number is generated based on the current timestamp, the initiator's identity, the target position ID and the randomly generated unique serial number combination. The security token generation service is called, and the HMAC-SHA256 algorithm is combined with the pre-shared key to encrypt and sign the conference room number and related session parameters to generate a time-sensitive access token. The generated conference room number and access token are pushed to the mobile apps of the job seeker and the employer respectively, and the link address to jump to the video call interface is attached in the notification message. Finally, the complete session initialization information including the conference room number and access token is output.

[0101] Initialize session parameters through the audio and video SDK and load user device information.

[0102] Furthermore, after the job seeker and the employer confirm entering the video call interface, the audio and video communication process is started according to the automatically generated conference room number and access token, and the initialization interface provided by the audio and video SDK is called, and the conference room number, access token, local user identity and current network environment type are passed in as initialization parameters. After receiving the initialization request, the audio and video SDK automatically detects and loads the current user's device information, including camera model, microphone status, speaker availability, operating system version, device unique identifier and network connection quality indicators such as upload bandwidth and delay time. The audio and video SDK automatically selects the appropriate codec, frame rate and resolution parameters according to the device capabilities and platform configuration, and establishes a signaling channel with the remote user to complete the session handshake; and finally outputs the initialized audio and video session instance.

[0103] Conduct a one-on-one video interview based on the conference room number and access token and activate the audio and video streaming channel.

[0104] Furthermore, after the job seeker and the employer complete the verification of the conference room number and access token in the mobile APP, the conference joining interface provided by the audio and video SDK is started, and the conference room number, access token and local user identity are submitted as parameters to trigger the establishment process of the audio and video stream transmission channel. The audio and video SDK completes the signaling interaction between the local device and the remote server, negotiates the media transmission protocol, codec format and network routing path, and automatically turns on the local camera and microphone to collect audio and video data; at the same time, it dynamically adjusts the audio and video encoding parameters, frame rate, resolution and bit rate according to the device capabilities and network status of both parties to ensure communication quality, sends the collected real-time audio and video stream to the remote device through the established transmission channel, and receives the audio and video stream from the remote end for local rendering and playback, realizing two-way real-time audio and video communication between the job seeker and the employer, and outputting a continuously running one-to-one video interview session.

[0105] By integrating multi-dimensional data, the identity forgery detection engine and resume authenticity assessment are run in real time to obtain a dual authenticity index.

[0106] Initiate a video interview to collect real-time video stream, audio stream, device fingerprint, and IP address information.

[0107] Furthermore, after the job seeker and the employer confirm entering the video call interface and complete the audio and video SDK initialization, click the Start Video Interview button to trigger the collection process. The audio and video SDK automatically turns on the local camera and microphone, collects real-time video stream and audio stream data from both parties, and compresses them according to the preset encoding format, using H.264 video encoding and AAC audio encoding. The system interface is called on the device side to obtain the current device fingerprint information. The device fingerprint information includes but is not limited to the device model, operating system version, unique device identifier, browser User-Agent string and other fields. The current user's public IP address and the name of the Internet service provider used for the connection are obtained through the network interface, and the geographic location information of the IP address is recorded.

[0108] The identity forgery detection engine extracts the current face image features, compares them with the face feature matching results, and obtains the face matching score.

[0109] Furthermore, during the video interview process, key frame images are extracted from the real-time video stream at set time intervals, and a face detection algorithm is executed on each frame image to locate the face area in the picture. Subsequently, the face recognition algorithm is used to perform feature encoding on the detected face area, and a face feature vector in the current video stream is generated. The historical record of the face feature matching result is called, which is the ID face feature vector generated during the registration phase through face liveness detection. The face feature vector extracted from the current video is compared with the ID face feature vector, and the cosine similarity method is used to compare the degree of matching between the two, and finally the face matching score is output.

[0110] The resume authenticity assessment module analyzes real-time voice content and uses NLP technology to compare resume text descriptions to obtain a resume consistency score.

[0111] Furthermore, during the video interview process, real-time audio streams from job seekers are collected and transmitted to the resume authenticity assessment module. The speech recognition interface is called to convert the real-time audio content into a sequence of interview sentences in text form. Natural language processing technology is used to perform entity extraction and semantic analysis on the identified interview sentences to extract key information such as work experience, project experience, and skill mastery. At the same time, the corresponding resume text description fields are obtained from the JSON format resume feature set generated by the previous parsing. The semantic similarity between the key information extracted from the interview sentences and the corresponding fields in the resume text description is calculated. The word vector cosine similarity matching method is used to determine the degree of matching between the two. The weighted summary of the matching results of all fields is used to generate a resume consistency score.

[0112] Based on the dynamic fusion engine, the scores of each dimension are weighted and calculated to generate the identity consistency index.

[0113] Furthermore, during the video interview process, multiple evaluation dimension scores related to user identity authentication are obtained, including but not limited to face matching score, resume consistency score, device fingerprint consistency score and IP address attribution stability score. The above-mentioned dimension scores are weighted and merged according to the preset weight distribution, and the dimension scores are proportionally allocated the remaining weights. All weighted results are summed and normalized to a unified interval to form the final identity consistency index. This index is used to quantitatively evaluate whether the user currently participating in the interview is consistent with his or her registered identity information, which triggers the risk warning mechanism and records abnormal behavior data, and finally outputs the identity consistency index.

[0114] The identity consistency index is combined with the resume consistency score weight to output a dual authenticity index.

[0115] Furthermore, after the identity consistency index is calculated, the resume consistency scores already included in the index and their original weight ratio in the identity consistency index are obtained. The resume consistency score is extracted separately as an independent evaluation dimension, and its target weight in the dual authenticity index is set according to business rules. The identity consistency index and the adjusted resume consistency score are fused and calculated using a linear weighting method. After the two are added, they are mapped to a unified scoring range through normalization processing to output the dual authenticity index.

[0116] Specifically, the expression is,

[0117] D = (I × W) + (R × E);

[0118] Among them, D is the dual authenticity index, I is the identity consistency index, W is the fusion weight coefficient of the identity consistency index, R is the resume consistency score, and E is the fusion weight coefficient of the resume consistency score.

[0119] In the evaluation module, job seekers and employers provide mutual feedback, update credit scores and generate interview evaluation reports.

[0120] Job seekers and employers provide mutual feedback, update credit scores and generate interview evaluation reports.

[0121] Based on the dual authenticity index, the mutual evaluation process is triggered, and the scoring interface is pushed to job seekers and employers to obtain the mutual evaluation score.

[0122] Furthermore, after the video interview is over, it is determined whether a valid dual authenticity index has been generated. If the index reaches the preset mutual evaluation trigger threshold, the mutual evaluation process is automatically started, and a mutual evaluation task record is constructed on the platform side, including fields such as the job ID, job seeker identity, employer identity and interview timestamp corresponding to this interview. Scoring notifications are sent to job seekers and employers respectively through the message push interface of the mobile APP, and a preset scoring interface is loaded. The scoring interface includes multiple scoring dimension input controls, and each entry provides a five-point scoring selector. After the user completes the scoring and submits it, the scoring results of both parties are recorded as the job seeker mutual evaluation score and the employer mutual evaluation score respectively.

[0123] The mutual evaluation scores of job seekers and employers are weighted to obtain the interactive credit change value.

[0124] Furthermore, after the video interview, the mutual evaluation process obtains the applicant's mutual evaluation score and the employer's mutual evaluation score. The applicant's mutual evaluation score is submitted by the employer on the evaluation interface and is used to evaluate the applicant's interview performance. The employer's mutual evaluation score is submitted by the applicant and is used to evaluate the employer's communication and feedback quality. According to the credit assessment rules set by the platform, weighted coefficients are assigned to the two types of mutual evaluation scores. The two types of mutual evaluation scores are multiplied by the corresponding weights and summed to obtain a weighted comprehensive mutual evaluation score. This comprehensive mutual evaluation score is compared with the historical credit score benchmark value to calculate the credit change difference, and finally output the interactive credit change value.

[0125] Specifically, the expression is,

[0126]

[0127] Among them, C is the credit change value, A is the mutual evaluation score of job seekers, M is the mutual evaluation score of employers, U is the average evaluation weight coefficient, and J is the difference penalty weight coefficient.

[0128] Combine the mutual evaluation score and the credit change value and update the credit score.

[0129] Furthermore, after obtaining the mutual evaluation scores submitted by both the job seeker and the employer, a credit change value is calculated according to the preset score weighting rules. The credit change value represents the impact of this video interview interaction on the user's credit score. The user's current credit score baseline value is read from the credit score database, and the credit change value is added to the baseline value. The source information of this credit score change is recorded, including fields such as interview time, position ID, mutual evaluation score details and credit change value. It is stored as a credit score change log in the user behavior record table, and the updated credit score is written to the credit score database and synchronized to the relevant business processes.

[0130] The resume consistency scores and mutual evaluation scores are extracted to obtain an evaluation report.

[0131] Furthermore, after the video interview process is completed, the resume consistency score and mutual evaluation score are extracted from the evaluation data generated in the early stage. The resume consistency score comes from the matching result obtained by comparing the real-time voice content with the resume text description through natural language processing technology. The mutual evaluation score includes the two-way evaluation records of the job seeker to the employer and the employer to the job seeker. The above scores are organized according to the preset evaluation report template structure, and combined with other relevant evaluation indicators such as identity consistency index, dual authenticity index, credit change value, etc. to form a complete evaluation data set. The evaluation report generation service is called to map all evaluation indicators to the corresponding positions in the report content template by field, and output a structured evaluation report.

[0132] In summary, the present invention achieves dual verification of facial feature comparison and voiceprint analysis through real-time collection of multi-dimensional data such as video streams and audio streams by the identity forgery detection engine, and performs real-time consistency verification of resume content in combination with NLP technology, thereby building a dynamic identity authenticity evaluation system, effectively ensuring the authenticity and reliability of the interview process, and establishing a closed-loop credit score update mechanism by combining the dual authenticity index with mutual evaluation feedback data, so that the credit evaluation can reflect the user's true performance during the interview process in real time. On the one hand, the identity authenticity of the interview participants is ensured through multimodal data collection, and on the other hand, the dynamic credit evaluation realizes the accurate quantification of user behavior, jointly building a trustworthy and efficient intelligent video interview ecosystem, fundamentally improving the quality and efficiency of recruitment matching.

[0133] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A video interview system for connecting employers and job seekers, characterized by: include, Identity verification module: users register through the mobile app and select job seekers and employers to verify their identity information; The intelligent matching module generates an initial credit score based on the verified user identity information. Based on the job information released by the employer, the module publishes the job information to the database and generates a unique job ID. Based on the uploaded resume of the job seeker, the module performs a preliminary match between the resume and the job information to obtain the original matching score. The appointment management module uses NLP technology and collaborative filtering algorithms to match positions, generate interview request status and appointment information, and prompt users to pay a small fee to improve matching quality; The dual authenticity assessment module allows job seekers and employers to confirm and jump to the video call interface for a video conference. This module collects multi-dimensional data, runs the identity forgery detection engine and resume authenticity assessment in real time, and obtains a dual authenticity index. In the evaluation module, job seekers and employers provide mutual feedback, update credit scores and generate interview evaluation reports.

2. The video interview system for connecting employers and job seekers according to claim 1, characterized in that: Users register through the mobile APP and select job seekers and employers to verify their identity information, including the following steps: Users enter the registration page through the mobile app and select roles as job seekers or employers. Users upload certification materials and use the OCR interface to extract document information. Use facial liveness detection on job applicants, compare authentication materials with real-time facial features, and output facial feature matching results; Conduct AI review on the certification materials uploaded by the employer to obtain the certification status.

3. The video interview system for connecting employers and job seekers according to claim 2, wherein: Generate initial credit points based on verified user identity information. The following steps are included: Based on the verified user identity information, check the user's historical behavior data and analyze third-party related data; The initial credit score is calculated using a logistic regression algorithm.

4. The video interview system for connecting employers and job seekers according to claim 3, wherein: Based on the job information released by the employer, the job information is published to the database and a unique job ID is generated, including the following steps: The employer enters basic job information, performs compliance verification on the submitted job information, and generates standardized job data; Use the cloth ID generation service to send it to the database to generate a unique position ID.

5. The video interview system for connecting employers and job seekers according to claim 4, characterized in that: Based on the uploaded resume of the job seeker, the resume file is preliminarily matched with the job information to obtain the original matching score. The following steps are included: Based on the resume files uploaded by job seekers, we use the resume parsing engine to extract structured data and generate a resume feature set in JSON format; The JD features of the target position are obtained from the position information, matched with the resume feature set, and the original matching score is obtained.

6. The video interview system for connecting employers and job seekers according to claim 5, characterized in that: Based on NLP technology and collaborative filtering algorithm to match positions, generate interview request status and appointment information, including the following steps: Use NLP technology to calculate the text similarity between the resume feature set and the job description to obtain a text matching score; Collaborative filtering algorithms analyze user historical behavior data and generate collaborative recommendation weights; The text matching score and collaborative recommendation weight are integrated to calculate and sort the comprehensive recommendation score, and generate the interview request status and appointment information.

7. The video interview system for connecting employers and job seekers according to claim 6, characterized in that: Prompt users to pay a small fee to improve matching quality. The following steps are included: Upon detecting that a user has initiated a video interview request, the credit score of the user will be automatically checked; Get a suggested payment amount based on the user's credit rating and the current matching quality; Remove restrictions on interview requests and increase the matching priority of interview requests.

8. The video interview system for connecting employers and job seekers according to claim 7, characterized in that: Job seekers and employers confirm and jump to the video call interface for video conferencing. The following steps are included: Automatically generate meeting room numbers and access tokens based on confirmation by job seekers and employers; Initialize session parameters and load user device information through the audio and video SDK; Conduct a one-on-one video interview based on the conference room number and access token and activate the audio and video streaming channel.

9. The video interview system for connecting employers and job seekers according to claim 8, characterized in that: By integrating multi-dimensional data, running the identity forgery detection engine and resume authenticity assessment in real time, we can obtain a dual authenticity index. The following steps are included: Initiate a video interview to collect real-time video stream, audio stream, device fingerprint, and IP address information; The identity forgery detection engine extracts the current face image features, compares them with the face feature matching results, and obtains the face matching score; The resume authenticity assessment module analyzes real-time voice content and uses NLP technology to compare resume text descriptions to obtain a resume consistency score. Based on the dynamic fusion engine, the scores of each dimension are weighted and calculated to generate the identity consistency index; The identity consistency index is combined with the resume consistency score weight to output a dual authenticity index.

10. The video interview system for connecting employers and job seekers according to claim 9, characterized in that: Job seekers and employers conduct mutual evaluation and feedback, update credit scores and generate interview evaluation reports. The following steps are included: Based on the dual authenticity index, the mutual evaluation process is triggered, and the evaluation interface is pushed to job seekers and employers to obtain the mutual evaluation score; Weighting the mutual evaluation scores of job seekers and employers to obtain the interactive credit change value; Combine the mutual evaluation score and the credit change value and update it to the credit score; The resume consistency scores and mutual evaluation scores are extracted to obtain an evaluation report.

Images