CN-121999301-A - Examination authentication method and system based on face recognition

CN121999301ACN 121999301 ACN121999301 ACN 121999301ACN-121999301-A

Abstract

The invention relates to the technical field of online examination supervision and discloses an examination authentication method and system based on face recognition, wherein the method comprises the steps of collecting initial information of behaviors and environments of an examinee, generating preliminary identity risk clues through processing, acquiring face images of the examinee, carrying out multi-stage identity verification processing on the face image data according to the risk clues, generating identity confirmation signals, acquiring behavior image data of the examinee in the examination process, synchronously inputting the identity confirmation signals and the behavior image data into a behavior analysis flow, carrying out multidimensional verification of behavior states, generating behavior abnormal state signals, integrating the identity confirmation signals and the behavior abnormal state signals, and carrying out comprehensive authentication judgment, so that a final examination authentication result is dynamically determined.

Inventors

YANG JIN
YAO NIANLONG
Di Zixu
LIU JICHEN
WANG HAO

Assignees

人民卫生电子音像出版社有限公司

Dates

Publication Date: 20260508
Application Date: 20260407

Claims (10)

1. An examination authentication method based on face recognition is characterized by comprising the following steps: initial information acquisition operation of the behaviors and the environment of the examinee is executed to form original monitoring data, the original monitoring data is subjected to preliminary processing, and a preliminary identity risk clue is generated; Performing face image acquisition operation of the examinee to form face image data, performing multi-stage identity verification processing on the face image data according to the preliminary identity risk clues, and generating an identity confirmation signal; executing image acquisition operation of examination behaviors of an examinee to form behavior image data, inputting the identity confirmation signal and the behavior image data into a behavior analysis flow synchronously, executing multidimensional verification of behavior states, and generating a behavior abnormal state signal; And integrating the identity confirmation signal and the abnormal behavior state signal, executing comprehensive authentication judgment, and dynamically determining a final examination authentication result.
2. The examination authentication method based on face recognition according to claim 1, wherein the performing an initial information collection operation of an examinee's behavior and environment to form original monitoring data, performing a preliminary processing on the original monitoring data, and generating a preliminary identity risk clue includes: in the starting stage of the examination, collecting initial biological signals of an examinee to form initial biological characteristic data; Continuously monitoring fluctuation factors in the examination environment to form an environment fluctuation data set; Recording a time track of an examination process to form time sequence data; and carrying out data fusion and context correlation analysis on the initial biological feature data, the environment fluctuation data set and the time sequence data, calculating to obtain a dynamically-changed identity risk score, and taking the identity risk score as a primary identity risk clue.
3. The examination authentication method based on face recognition according to claim 2, wherein the data fusion and context correlation analysis are performed on the initial biometric data, the environmental fluctuation dataset and the time series data, and a dynamically changing identity risk score is calculated, which comprises: establishing a risk score calculation model, wherein the risk score calculation model comprises a historical behavior pattern library, an environment baseline library and a time association rule library; Matching a historical pattern similar to the current initial biological feature data from the historical behavior pattern library, and extracting a historical risk mark; Comparing the environmental fluctuation data set from the environmental baseline library, and calculating the environmental deviation degree; Applying the time association rule base to analyze the normal range of the time sequence data in different stages of the examination and identify time abnormal points; and performing weighted aggregation operation on the historical risk marks, the environment deviation degree and the time outliers, and updating and outputting the dynamically-changed identity risk scores in real time.
4. The examination authentication method based on face recognition according to claim 1, wherein the step of performing an acquisition operation of an examinee face image to form face image data, performing a multi-stage authentication process on the face image data according to the preliminary identity risk cue, and generating an identity confirmation signal, comprises: Triggering a high-resolution image acquisition device when the preliminary identity risk clue is received, and capturing the face area of the examinee at the current moment to form the face image data; Dynamically activating a group of corresponding candidate identity verifiers according to the specific values of the preliminary identity risk clues; Inputting the face image data to each of the corresponding candidate identity verifiers in parallel, wherein each candidate identity verifier comprises a plurality of verification paths, and each verification path performs comparison verification on a specific facial feature combination; Collecting verification conclusions from all activated verification paths, each verification conclusion containing a positive or negative judgment; performing statistical analysis on the verification conclusion, calculating the confidence coefficient of each candidate identity verifier, and sorting according to the confidence coefficient; Adopting core judgment output by the candidate identity verifier with highest ranking, and carrying out logic calibration by combining the initial identity risk clues to generate a final identity confirmation signal; and performing statistical analysis on the verification conclusion, calculating the confidence coefficient of each candidate identity verifier, and sequencing according to the confidence coefficient, wherein the method comprises the following steps of: counting the number of positive judgment and the number of negative judgment in the verification conclusion outputted from all verification paths under each candidate identity verifier; Calculating the proportion of the number of positive judgment of each candidate identity verifier to the total verification conclusion number of the candidate identity verifier as the internal consistency rate of the candidate identity verifier; Combining the accuracy record of the candidate identity verifier in the history verification task, calibrating the internal consistency rate, and generating the confidence coefficient; and arranging all the candidate identity verifiers according to the confidence level from high to low, and finishing the sorting.
5. The face recognition-based examination authentication method of claim 4, wherein dynamically activating a set of corresponding candidate verifiers based on the specific values of the preliminary identity risk cues comprises: Different candidate identity verifier sets are configured for different identity risk numerical ranges in advance; comparing the specific value of the preliminary identity risk clue with a plurality of preset risk value thresholds, and determining a specific risk value range to which the specific value is belonged; And loading model parameters and verification rules of the group of corresponding candidate identity verifiers associated with the specific risk numerical range from a configuration database, and completing the dynamic activation.
6. The examination authentication method based on face recognition according to claim 1, wherein inputting the identity confirmation signal and the behavior image data into a behavior analysis flow in synchronization, performing multidimensional verification of behavior states, generating a behavior abnormality state signal, comprises: establishing a behavior analysis engine, wherein the behavior analysis engine comprises a gesture analysis module, a sight tracking module and an interaction action recognition module; Feeding the behavior image data to the gesture analysis module, the sight tracking module and the interaction action recognition module simultaneously; the gesture analysis module extracts body skeleton key points from the behavior image data, analyzes the motion trail of the key points and generates gesture anomaly; The sight tracking module locates an eye region from the behavior image data, estimates a sight direction and a focus point thereof, and generates a sight drift degree; The interaction action recognition module recognizes interaction actions of hands and the examination room from the behavior image data, and generates abnormal interaction frequency; introducing the identity confirmation signal as a weight adjustment factor, and respectively carrying out weighted correction on the attitude anomaly degree, the sight drift degree and the anomaly interaction frequency; and fusing the weighted and corrected attitude anomaly degree, the line-of-sight drift degree and the anomaly interaction frequency, and outputting the behavior anomaly state signal through a behavior state judgement device.
7. The examination authentication method based on face recognition according to claim 6, wherein introducing the identity confirmation signal as a weight adjustment factor, respectively performing weighted correction on the attitude anomaly degree, the line-of-sight drift degree and the anomaly interaction frequency, comprises: analyzing the identity confirmation signal, and mapping the identity confirmation signal into a weight coefficient, wherein the higher the credibility of the identity confirmation signal is, the larger the weight coefficient is; multiplying the attitude anomaly degree by the weight coefficient to obtain a corrected attitude anomaly degree; multiplying the sight line drift degree by the weight coefficient to obtain a corrected sight line drift degree; multiplying the abnormal interaction frequency by the weight coefficient to obtain the corrected abnormal interaction frequency.
8. The face recognition-based examination authentication method of claim 1, wherein integrating the identity confirmation signal with the behavior anomaly state signal, performing comprehensive authentication judgment, and dynamically determining a final examination authentication result, comprises: creating an authentication decision logic unit, wherein the authentication decision logic unit receives the identity confirmation signal and the abnormal behavior state signal; The authentication decision logic unit is internally provided with a decision rule matrix, and the decision rule matrix defines authentication conclusion corresponding to the combination of different grades of identity confirmation signals and different grades of behavior abnormal state signals; In the authentication decision logic unit, matching the received grade of the identity confirmation signal with the grade of the abnormal behavior state signal, inquiring the decision rule matrix, and obtaining a preliminary authentication conclusion; And (3) carrying out timeliness rechecking on the preliminary authentication conclusion according to the real-time feedback information of the examination process, outputting the preliminary authentication conclusion as the final examination authentication result if the rechecking is passed, and triggering a new round of data acquisition and processing flow if the rechecking is not passed.
9. The examination authentication method based on face recognition according to claim 8, wherein the time-efficient review of the preliminary authentication conclusion according to real-time feedback information of the examination process comprises: monitoring timing information and current task nodes of the examination system as the real-time feedback information; judging whether the time interval from the generation of the preliminary authentication conclusion to the current moment exceeds a preset validity period threshold value or not; Judging whether the current examination task node is in a key authentication stage or not; If the time interval does not exceed the validity period threshold and the current examination task node is in the key authentication stage, the rechecking is judged to pass, otherwise, the rechecking is judged not to pass.
10. A face recognition based examination authentication system comprising a memory, a processor and a computer program stored in the memory and running on the processor, characterized in that the processor, when executing the computer program, realizes the steps of the face recognition based examination authentication method according to any one of claims 1 to 9.

Description

Examination authentication method and system based on face recognition Technical Field The invention relates to the technical field of online examination supervision, in particular to an examination authentication method and system based on face recognition. Background Current online examination authentication systems generally rely on pre-registered face information for identity verification. The prior art generally employs separate authentication and behavior monitoring modules. The identity verification is carried out by carrying out one-time static face comparison when the examinee logs in, or by setting a fixed time point in the examination process to carry out periodic snapshot comparison so as to confirm that the identity of the examinee is not replaced. The behavior monitoring detects abnormal behaviors through independent video streams by using algorithms such as gesture analysis, sight tracking and the like. The results of the authentication and the analysis of the behavior monitoring are usually in a logically sequential or parallel relationship, and the two are independent from each other at the data processing level. Such a solution has drawbacks. The authentication flow is static and preset, and cannot be adjusted according to the actual condition of the prompt and perpetual change of the examination site. Identity state and behavior analysis are mutually split. When the behavior analysis model judges that the motion is abnormal, the real-time identity confidence of the current examinee cannot be known and referred to, which may lead to two misjudgments, namely, insufficient monitoring of the examinee with high risk but normal behavior on the identity, or false alarm of the examinee with undoubted identity confirmation but irregular small motion. The final judgment of the system is only the simple logic combination of two types of isolated signals at a decision layer, and the depth information fusion is lacked. The core problem faced by the examination authentication system is how to enable the identity verification process to dynamically respond to potential risks according to the field environment, and how to break the data barriers between the identity verification and the behavior monitoring, so that the identity state can be integrated into the analysis and judgment of the behavior meaning in real time and in depth, and the more accurate and more active recognition of complex fraud behaviors such as tilmicosin and cheating assistance is realized. Disclosure of Invention The invention aims to provide an examination authentication method and system based on face recognition so as to solve the problems in the background technology. In order to achieve the above object, the present invention provides an examination authentication method based on face recognition, the method comprising: initial information acquisition operation of the behaviors and the environment of the examinee is executed to form original monitoring data, the original monitoring data is subjected to preliminary processing, and a preliminary identity risk clue is generated; Performing face image acquisition operation of the examinee to form face image data, performing multi-stage identity verification processing on the face image data according to the preliminary identity risk clues, and generating an identity confirmation signal; executing image acquisition operation of examination behaviors of an examinee to form behavior image data, inputting the identity confirmation signal and the behavior image data into a behavior analysis flow synchronously, executing multidimensional verification of behavior states, and generating a behavior abnormal state signal; And integrating the identity confirmation signal and the abnormal behavior state signal, executing comprehensive authentication judgment, and dynamically determining a final examination authentication result. Preferably, an initial information acquisition operation of the behavior and the environment of the examinee is executed to form original monitoring data, the original monitoring data is subjected to preliminary processing, and a preliminary identity risk clue is generated, including: in the starting stage of the examination, collecting initial biological signals of an examinee to form initial biological characteristic data; Continuously monitoring fluctuation factors in the examination environment to form an environment fluctuation data set; Recording a time track of an examination process to form time sequence data; and carrying out data fusion and context correlation analysis on the initial biological feature data, the environment fluctuation data set and the time sequence data, calculating to obtain a dynamically-changed identity risk score, and taking the identity risk score as a primary identity risk clue. Preferably, the data fusion and context correlation analysis are performed on the initial biometric data, the environmental fluctuation dataset and the time s