Examination seat arrangement method and system based on intelligent examination table

Abstract

The invention relates to the technical field of intelligent education equipment and examination management, in particular to an examination seat arrangement method and system based on an intelligent examination desk, and the method comprises the steps: collecting environment data through the intelligent examination desk, and generating an environment feature matrix in combination with an implicit perception technology; dividing examination room areas based on the quantum dot distribution model, optimizing seat distribution through reinforcement learning, and generating optimized seat distribution data; and completing examinee identity verification by using a multi-mode verification module, binding verification data with the optimized seat distribution data, unlocking an examination interface and loading an examination task. According to the invention, the fairness and the intelligent level of examination room seat distribution are effectively improved, the preferential distribution of examinees with special requirements and the dynamic management of the examination process are realized, and the reliability and the convenience are remarkable.

Description

Technical Field

[0001] The present invention relates to the field of intelligent educational equipment and examination management technology, and in particular to an examination seat arrangement method and system based on an intelligent examination platform. Background Art

[0002] The reasonable arrangement of examination seats plays an important role in improving the fairness and efficiency of the examination. However, the currently widely used technologies usually rely on manual or semi-automatic ranking systems, which have significant defects. For example:

[0003] The existing technology (Chinese invention patent, publication number: CN117132222A, name: A test ranking system based on intelligent tables and chairs) mainly relies on intelligent tables and chairs and ranking algorithms to arrange seats, which uses a server to import candidate information and generates matching information between candidates and seats through an automatic ranking algorithm with fixed rules. Candidate identity verification is completed through the MCU module of the tables and chairs combined with an information reading device.

[0004] The existing technologies have the following defects: the existing fixed rule ranking algorithms (such as column sorting method) cannot be intelligently adjusted according to the dynamic environmental parameters in the examination room, and it is difficult to adapt to changes in light, temperature and humidity; the existing technologies are limited to a small amount of static data (such as table and chair numbers, candidate information) in terms of environmental perception, and ignore the impact of implicit environmental data (such as air flow and micro-vibration) on the comfort of the examination room; the existing technologies are difficult to achieve priority care for candidates with special needs, for example, candidates who need specific seats due to physical conditions cannot dynamically adjust seat allocation; conflict detection and allocation optimization are relatively mechanical, and adaptive optimization technology has not been introduced, which easily leads to low resource utilization. Summary of the invention

[0005] In view of the many problems existing in the above-mentioned prior art, the present invention provides a method and system for arranging examination seats based on a smart examination platform. The present invention collects explicit and implicit environmental data, uses a spatiotemporal convolutional neural network to extract environmental features, divides the examination area in combination with a quantum dot distribution model, further optimizes seat allocation with a reinforcement learning algorithm, and implements identity authentication and task loading through a smart examination platform. The final effect is to achieve dynamic environmental adaptation, special needs care, and intelligent management of the entire examination process.

[0006] A method for arranging test seats based on an intelligent test platform comprises the following steps:

[0007] Collect environmental data of the examination room through the intelligent examination platform to generate environmental perception data; combine implicit perception technology to collect air flow and micro-vibration characteristics to generate implicit environmental data; integrate and process environmental perception data and implicit environmental data to generate an environmental feature matrix; receive and standardize the candidate information data;

[0008] Based on the environmental feature matrix and the physical layout of the examination room, the quantum dot distribution model is used to divide the examination room area and generate regional suitability data; based on the regional suitability data, candidates with special needs are preferentially allocated to suitable areas to generate regional allocation data;

[0009] Based on the regional allocation data and candidate information data, the seat state matrix is ​​initialized, and preliminary seat allocation data is generated through a dynamic programming algorithm; conflict detection and optimization are performed on the preliminary seat allocation data, and the seat allocation plan is adjusted using a reinforcement learning algorithm to generate optimized seat allocation data;

[0010] Through the multimodal verification module of the intelligent examination platform, the candidate's biometric data is collected, identity authentication is completed, and verification data is generated; the verification data is bound to the optimized seat allocation data to complete the examination interface unlocking and task loading.

[0011] Preferably, the environmental data includes light intensity, humidity, temperature and noise level, and the implicit environmental data is generated by real-time detection of air flow characteristics and micro-vibration frequency characteristics by a highly sensitive sensor configured in the intelligent examination table.

[0012] Preferably, the environmental feature matrix performs dynamic feature extraction on environmental perception data and implicit environmental data through a spatiotemporal convolutional neural network, specifically including joint analysis of spatial distribution and temporal variation patterns, and extraction of feature data on light uniformity, temperature and humidity stability, and noise variation trends.

[0013] Preferably, the quantum dot distribution model adopts a quantum energy simulation method to generate regional suitability data by performing quantum state simulation calculations on the environmental suitability parameters in the examination area. The regional suitability data includes the light stability value, noise interference value and environmental temperature and humidity balance value of each area.

[0014] Preferably, the regional allocation data is generated by a reinforcement learning algorithm, wherein the allocation rules for special needs candidates use the regional suitability score as a weight, and the matching relationship between special needs candidates and priority areas is dynamically adjusted by calculating the comprehensive suitability scores of different candidate areas.

[0015] Preferably, the preliminary seat allocation data is generated by jointly optimizing the regional allocation data and the examinee information data through a dynamic programming algorithm, and the optimization process includes calculating the priority allocation sequence of seats in each area with the goal of minimizing the distance between examinees and meeting special needs.

[0016] Preferably, the optimized seat allocation data is generated by a conflict detection module and a reinforcement learning adjustment module, which specifically includes detecting conflict factors in the seat allocation plan and dynamically optimizing the allocation strategy according to the utilization rate of examination room resources and the priority of examinee needs.

[0017] Preferably, the multimodal verification module includes a fingerprint recognition module and a face recognition module, and the verification pass data is generated by the matching results of the fingerprint feature vector and the face feature vector.

[0018] Preferably, the test interface unlocking and task loading include activating the examinee's test authority based on the verification data, loading the test task in real time, and recording the start time, completion status and abnormal operation information of the test task.

[0019] A system for implementing the test seat arrangement method based on the intelligent test platform, comprising:

[0020] Environmental sensing module for collecting environmental data;

[0021] Implicit sensing module for collecting air flow and micro-vibration characteristics;

[0022] An environmental feature analysis module for integrating environmental perception data and implicit environmental data and generating an environmental feature matrix;

[0023] Candidate information management module for receiving and standardizing candidate information data;

[0024] A regional division module for dividing the examination room area and generating regional suitability data using a quantum dot distribution model based on the environmental feature matrix and the physical layout of the examination room;

[0025] A seating arrangement module for generating regional allocation data and initializing a seat state matrix based on regional suitability data, generating preliminary seat allocation data through a dynamic programming algorithm, and generating optimized seat allocation data through a reinforcement learning algorithm;

[0026] A multimodal verification module for collecting the examinee's biometric data and generating verification pass data;

[0027] An execution module used to bind the verified data with the optimized seat allocation data and complete the test interface unlocking and task loading.

[0028] Compared with the prior art, the advantages and beneficial effects of the present invention are:

[0029] The present invention combines explicit and implicit environmental data (light, air flow, etc.) to generate an environmental feature matrix through a spatiotemporal convolutional neural network to comprehensively optimize the environmental suitability of the seating arrangement;

[0030] With the help of quantum state simulation and reinforcement learning technology, the present invention effectively improves the accuracy of matching candidates with seats, especially the priority allocation and optimization of candidates with special needs;

[0031] The multimodal verification technology in the present invention (such as the combination of fingerprint and face recognition) improves the security and accuracy of identity verification and ensures a seamless connection between the candidate and the seat binding process;

[0032] The present invention realizes the precise management of examination authority, and can record and analyze the execution status of examination tasks in real time, providing guarantee for rapid response to abnormal behavior. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a schematic diagram of the process of the present invention;

[0034] Figure 2 It is a schematic diagram of the regional suitability division and allocation in the present invention;

[0035] Figure 3 A schematic diagram for generating optimized seat allocation data in the present invention;

[0036] Figure 4 It is a schematic diagram of multimodal verification and task loading in the present invention;

[0037] Figure 5 It is a structural block diagram of the system of the present invention. DETAILED DESCRIPTION

[0038] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of the present disclosure. In the following detailed description, for ease of explanation, many specific details are set forth to provide a comprehensive understanding of the embodiments of the present disclosure. However, it is obvious that one or more embodiments may also be implemented without these specific details. In addition, in the following description, descriptions of known structures and technologies are omitted to avoid unnecessary confusion of the concepts of the present disclosure.

[0039] The terms used herein are only for describing specific embodiments and are not intended to limit the present disclosure. The terms "include", "comprising", etc. used herein indicate the existence of the features, steps, operations and / or components, but do not exclude the existence or addition of one or more other features, steps, operations or components.

[0040] All terms (including technical and scientific terms) used herein have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted as having a meaning consistent with the context of this specification, and should not be interpreted in an idealized or overly rigid manner.

[0041] like Figure 1 As shown, a method for arranging test seats based on an intelligent test platform includes the following steps:

[0042] Collect environmental data of the examination room through the intelligent examination platform to generate environmental perception data; combine implicit perception technology to collect air flow and micro-vibration characteristics to generate implicit environmental data; integrate and process environmental perception data and implicit environmental data to generate an environmental feature matrix; receive and standardize the candidate information data;

[0043] The environmental sensing modules equipped in the smart examination platform include light sensors, temperature and humidity sensors, and noise sensors, which are used to monitor the physical environmental parameters in the examination room in real time. These data are directly collected by sensors, such as light intensity is quantified as light level, temperature and humidity data are expressed as relative humidity and temperature values, and noise levels are recorded in decibel (dB) values.

[0044] Implicit environmental data refers to features that are difficult to capture directly with traditional sensors but have a significant impact on the environment, such as air flow and micro-vibration. The intelligent test platform uses a built-in high-sensitivity pressure sensor and microfluidic detector to capture air flow rate, turbulence mode, and micro-vibration signals caused by external vibration sources.

[0045] All collected data are integrated through a feature fusion module, combining information from the time and space dimensions to generate a multi-dimensional environmental perception data matrix. The spatiotemporal convolutional neural network (ST-CNN) is used to analyze these data matrices and extract key features such as light uniformity, temperature and humidity stability, and noise distribution trends. Through the convolution operation, the environmental features are represented as a two-dimensional matrix, where each unit corresponds to the environmental suitability score of a specific area.

[0046] The test system pre-enters the candidate's identity information, including name, admission ticket number, and special requirement marks (such as light sensitivity, noise sensitivity, etc.). This information is read by the identity authentication module when the candidate logs in to the smart test platform.

[0047] The candidate information data is cleaned and standardized to ensure uniform data format, eliminate redundant information, and provide structured input for subsequent area allocation and seat arrangement. For example, special needs marks are encoded as specific label sets (such as "light sensitive" is 1, "noise sensitive" is 2).

[0048] The processed environmental perception data and implicit environmental data are aggregated by region. For example, the light, noise, and air flow characteristics around each test platform are assigned to specific grid cells. Through region division and feature calculation, an environmental feature matrix is ​​formed, where each row represents a specific region and each column corresponds to a specific environmental feature.

[0049] The environmental feature matrix is ​​further dynamically analyzed through a neural network to capture the changing trends of environmental parameters over time, such as the fluctuation range of light intensity in different time periods.

[0050] In an example, in a certain examination room, the smart examination platform detects through sensors that the light intensity in area A is 300 lux, the noise level is 60 dB, and the relative humidity is 50%. At the same time, the air velocity sensor records that the air flow in the area is relatively uniform, and the vibration sensor detects that the micro-vibration level is low. Combined with the implicit environmental data generated by implicit perception technology, area A is marked as a high suitability area.

[0051] Dynamic Adaptation: A candidate who is sensitive to light is assigned to Area A, where their special needs are met, and the system dynamically updates the environmental feature matrix based on real-time changes. If the light intensity in Area A increases to 500 lux (exceeding the suitability threshold), the system will lower its suitability score and trigger a reallocation process.

[0052] Preferably, the environmental data includes light intensity, humidity, temperature and noise level, and the implicit environmental data is generated by real-time detection of air flow characteristics and micro-vibration frequency characteristics by a highly sensitive sensor configured in the intelligent examination table.

[0053] Light sensors are placed on the surface or top of the examination table to capture the light intensity of the surrounding environment. The sensor quantifies the light intensity into a numerical signal and can record the uniformity and stability of the light in real time. These data are important for the regional adaptation of light-sensitive candidates (such as myopic or photophobic candidates).

[0054] Humidity and temperature sensors are installed near the examination table to sense the local microclimate conditions in the examination room. The sensors measure the water vapor content and temperature changes in the air, transmit the data to the processing unit, and generate the local humidity and temperature distribution characteristics of the examination room.

[0055] The noise sensor is based on sound wave capture technology and can accurately measure the sound pressure level (decibel) at a specific location. This data helps identify areas with lower noise levels and provide a suitable examination environment for noise-sensitive candidates.

[0056] The smart examination platform is equipped with a microfluidic detection module, which generates air flow characteristic data in real time by sensing air flow velocity and turbulence patterns. The air flow characteristics are used to identify areas with relatively stable air flow, thereby avoiding discomfort or interference to candidates caused by air flow changes.

[0057] A highly sensitive pressure sensor is embedded in the base of the test platform to capture micro-vibration signals caused by external interference sources (such as nearby mechanical vibrations or people walking around). Vibration characteristic data is generated through frequency and amplitude analysis to determine whether a certain area has good vibration isolation.

[0058] All environmental data and implicit environmental data are integrated and processed through the sensor fusion module to form a multi-dimensional environmental feature matrix, which provides input for subsequent regional suitability analysis. Through the feature extraction algorithm, the data is normalized to ensure the consistency of the suitability analysis.

[0059] The configuration of multiple types of sensors ensures comprehensive coverage and real-time perception of environmental characteristics, and can capture key parameters such as light, temperature and humidity, noise, air flow and vibration. These data reflect the micro-environmental differences in various areas of the examination room and provide an accurate data basis for seat arrangement. The introduction of implicit environmental data makes up for the implicit environmental characteristics (such as micro-flow and vibration) that are difficult to capture with traditional sensors, significantly improving the comprehensiveness of environmental perception.

[0060] Real-time updated environmental data enables the system to quickly adjust the area suitability score when the examination room conditions change. For example, when the light at a certain examination table suddenly increases or the noise level rises, the system can identify the unsuitable state of the area and dynamically adjust the candidate allocation strategy to ensure that the examination environment is always in the best state.

[0061] Accurate monitoring of air flow and micro-vibration enables the system to effectively eliminate unstable factors in the environment. For example, areas with low vibration characteristics can be allocated to candidates who are easily distracted by the outside world, ensuring that they focus on the test content. Comprehensive analysis of temperature, humidity and light data provides customized seating arrangements for candidates with special needs, improving their comfort and test experience.

[0062] In an example, in a certain examination room, the light sensor records the light intensity in area A as 400 lux, the temperature and humidity sensor records the relative humidity as 45% and the temperature as 22°C, the noise sensor records the sound pressure level as 50 dB, the air flow rate sensor records the flow rate as stable at 0.2 m / s, and the vibration sensor detects that the vibration frequency is in a low-frequency stable state (<10 Hz).

[0063] After system analysis, area A is marked as a suitable area for light-sensitive and noise-sensitive candidates. The system assigns a light-sensitive candidate to this area and dynamically monitors light changes to ensure a stable environment.

[0064] If the light intensity in area A increases to 600 lux (exceeding the suitability threshold), the system will readjust the suitability score and prioritize transferring the candidate to area B with even lighting.

[0065] Preferably, the environmental feature matrix performs dynamic feature extraction on environmental perception data and implicit environmental data through a spatiotemporal convolutional neural network, specifically including joint analysis of spatial distribution and temporal variation patterns, and extraction of feature data on light uniformity, temperature and humidity stability, and noise variation trends.

[0066] Environmental perception data (such as light intensity, humidity, temperature, and noise level) and implicit environmental data (such as air flow characteristics and micro-vibration frequency characteristics) are divided into multiple spatial grid units, each of which represents a small area in the examination room. Through two-dimensional convolution operations, spatial features are extracted, such as the uniform distribution of light between regions, regional differences in temperature and humidity, and the local concentration of noise.

[0067] Based on the trend of data changes in each spatial unit over time, the network captures the dynamic change pattern of the time dimension through the time convolution layer. For example, the fluctuation of noise level at different time points, the periodic change of light intensity, the gradual change trend of temperature and humidity, etc.

[0068] The spatiotemporal convolutional neural network combines spatial features with temporal features through stacked convolutional layers to form a high-dimensional feature map that contains temporal dynamics and spatial distribution characteristics. This feature map is used to generate the final environmental feature matrix.

[0069] By comparing the light intensity of different spatial units and calculating the standard deviation of light distribution, we can identify areas with higher light uniformity, which are suitable for light-sensitive candidates. We analyze the temporal variation of temperature and humidity, and mark areas with smaller temperature and humidity fluctuations as high suitability areas to improve the comfort of candidates. By detecting the fluctuation pattern of noise levels (such as the frequency and amplitude of sudden noise), we can assess the degree of noise interference in the area and assign quiet areas to noise-sensitive candidates.

[0070] The high-dimensional features extracted by the spatiotemporal convolutional neural network are mapped into a two-dimensional matrix, where rows represent spatial units and columns represent different environmental features. The value of each cell in the matrix reflects the suitability score of the area for a certain environmental feature. For example, the first column of the matrix represents the light uniformity score, and the second column represents the temperature and humidity stability score.

[0071] The dynamic feature extraction capability of the spatiotemporal convolutional neural network enables the environmental feature matrix to fully reflect the dynamic environmental conditions of the examination room. For example, when the light distribution is uneven in space and there are temporal fluctuations, the system can accurately identify and mark these areas as low suitability areas. The environmental feature matrix provides a quantitative basis for subsequent seat allocation. For example, the matrix's suitability score directly guides the regional matching of candidates, achieving accurate matching of candidate needs with environmental conditions. When the examination room environment changes (such as fluctuations in light intensity or a sudden increase in noise), the feature matrix can be updated in real time to provide support for dynamic adjustment of seat allocation plans.

[0072] Through the comprehensive scoring of different environmental characteristics in the matrix, the system can give priority to meeting the suitability requirements of candidates with special needs, while ensuring that ordinary candidates are allocated to comfort areas, thereby improving the fairness of the overall examination environment.

[0073] In an example, in a certain examination room, the light sensor records the light intensity in different areas, and the data is divided into four grid units, namely A, B, C, and D. The light intensity uniformity score of area A is 0.8, area B is 0.5, area C is 0.6, and area D is 0.7. The noise sensor records that the noise fluctuation in area B is the smallest, with a score of 0.9.

[0074] Feature matrix generation: The spatiotemporal convolutional neural network integrates the above data to generate the following feature matrix:

[0075]

[0076] Based on the scores of the feature matrix, the system prioritizes light-sensitive candidates to area A and noise-sensitive candidates to area B. At the same time, it dynamically monitors light changes. If the light intensity fluctuation in area A exceeds the threshold, the system will immediately readjust the matrix scores and update the allocation plan.

[0077] like Figure 2 As shown, based on the environmental feature matrix and the physical layout of the examination room, the quantum dot distribution model is used to divide the examination room area to generate regional suitability data; according to the regional suitability data, candidates with special needs are preferentially allocated to suitable areas to generate regional allocation data;

[0078] Preferably, the quantum dot distribution model adopts a quantum energy simulation method to generate regional suitability data by performing quantum state simulation calculations on the environmental suitability parameters in the examination area. The regional suitability data includes the light stability value, noise interference value and environmental temperature and humidity balance value of each area.

[0079] The quantum energy simulation method draws on the concept of wave function in quantum physics to map the environmental suitability parameters in the examination room into the characteristic values ​​of the quantum state, and forms quantitative data by calculating the suitability scores of various environmental factors in the area.

[0080] The fluctuation of light intensity is regarded as the amplitude of light waves, and the light stability value is obtained by calculating the consistency of the amplitude; noise interference is simulated as the frequency distribution of sound waves. The higher the low-frequency stability, the lower the noise interference value; temperature and humidity are quantified as the average temperature and humidity deviation values ​​in the area. The smaller the deviation, the better the balance.

[0081] Quantization of suitability parameters, quantum state formula:

[0082] ψ(x, t) = A(x)·e i(ωt-kx)

[0083] Among them, ψ(x, t) represents the comprehensive wave function representing the suitability of the region; A(x) represents the amplitude of the environmental parameters in the region (such as light intensity); ω represents the frequency of change in the time dimension; kx represents the wave vector in the spatial dimension, reflecting the spatial variation of environmental parameters in the region. The model establishes wave functions for light, noise, temperature and humidity respectively, and combines regional specific environmental data to quantify suitability into a high-dimensional matrix.

[0084] Quantum state wave functions are generated for the light, noise, temperature and humidity of each area, and the regional suitability score is calculated by the superposition principle:

[0085] S = ∫ V |ψ(x, t)| 2 dx

[0086] Where S represents the regional suitability score; V represents the spatial volume of the target area; |ψ(x, t)| 2 Represents the overall suitability strength of regional environmental parameters. The weighted calculation of each parameter score generates the final regional suitability data.

[0087] The quantum energy simulation method is used to quantify and dynamically simulate environmental parameters, which significantly improves the accuracy of environmental suitability assessment. For example, the fluctuation of light intensity not only reflects the current environmental conditions, but also takes into account its changing trend over time. The suitability scores of different areas can intuitively reflect the differences in environmental quality of the examination rooms, providing strong support for regional allocation.

[0088] Since quantum state simulation can capture changes in spatial and temporal dimensions, regional suitability data can be updated in real time, ensuring that the system can quickly adjust the seat allocation plan when environmental conditions change. For example, when the noise level suddenly increases at a certain moment, the system will lower the suitability score of the area and prioritize other areas.

[0089] The generation of regional suitability data achieves a precise match between candidates' needs and environmental conditions. For example, light-sensitive candidates are preferentially assigned to areas with higher light stability values, and noise-sensitive candidates are assigned to areas with lower noise interference values.

[0090] In the example, a test room is divided into four areas (A, B, C, D), and the environmental parameters are as follows:

[0091] Area A: The light intensity fluctuation is low (uniform amplitude), the noise frequency distribution is mainly low frequency, and the temperature and humidity deviation is small;

[0092] Area B: The light intensity is high but unstable, the noise interference frequency is high, and the temperature and humidity fluctuate greatly;

[0093] Area C: The light is relatively uniform but dim, the noise is stable and low, and the temperature and humidity change in a moderate range;

[0094] Area D: The light is uneven, the high frequency accounts for a large proportion of the noise frequency distribution, and the temperature and humidity deviations are large.

[0095] Suitability score calculation:

[0096] The quantum state wave function of the light in region A is ψ A (x, t) = 0.8·e i(0.2t-0.1x) , the noise wave function is ψ A_noise (x, t) = 0.98·e i(0.1t-0.05x) , the temperature and humidity wave function is ψ A_temp (x, t) = 0.7·e i(0.3t-0.1x) The overall suitability score for Region A is:

[0097] S A =0.4·∫ V |ψ A (x, t)| 2 dx+0.3·∫ V |ψ A_noise (x, t)| 2 dx+0.3·∫ V |ψ A_temp (x, t)| 2 dx

[0098] The results show that the suitability score of area A is 0.85. Based on the suitability score, the system prioritizes light-sensitive and noise-sensitive candidates to area A, and ordinary candidates to area C, while areas B and D are marked as low suitability areas.

[0099] Preferably, the regional allocation data is generated by a reinforcement learning algorithm, wherein the allocation rules for special needs candidates use the regional suitability score as a weight, and the matching relationship between special needs candidates and priority areas is dynamically adjusted by calculating the comprehensive suitability scores of different candidate areas.

[0100] The reinforcement learning algorithm adopted by the present invention realizes the optimization of seat allocation in the form of "intelligent agent-environment interaction". In each allocation operation, the intelligent agent observes the current environment state (i.e., the matching of the regional suitability score and the candidates with special needs), takes actions (allocates candidates to specific areas), and obtains rewards (Reward) from the environmental feedback.

[0101] State space S: It consists of the suitability score of each area and the special needs of the examinee, such as the light uniformity score, noise interference value, temperature and humidity balance value of the area, and the examinee's demand label (such as light sensitivity, noise sensitivity, etc.) Action space A: The operation of assigning the examinee to a specific area.

[0102] Reward function:

[0103]

[0104] Among them, R represents the total reward value; w 1 Represents the weight coefficient, which is used to adjust the influence of different suitability parameters; Q i Indicates the suitability score of the region; D i Indicates the corresponding suitability parameter of the examinee's needs; Match(Q i ,D i ) indicates the matching degree between regional suitability and examinee’s needs.

[0105] The reinforcement learning agent establishes an initial matching scheme based on the initial regional suitability score and the needs of the candidates. As the environmental state changes (such as fluctuations in light intensity or increased noise in a certain area), the system recalculates the regional suitability score and updates the state space. The reinforcement learning agent readjusts the matching scheme based on the new state to maximize the reward function value. Through multiple rounds of training, the agent gradually optimizes the allocation strategy so that the matching scheme tends to the global optimal.

[0106] The overall suitability score is a weighted combination of the regional suitability parameters and the candidate's needs:

[0107]

[0108] Among them, S total represents the comprehensive suitability score; w i Indicates the weight of the suitability parameter, for example, the light uniformity weight may be higher than the noise interference weight; Q j Represents the suitability score of a parameter within a region.

[0109] Through the dynamic adjustment of the reinforcement learning algorithm, candidates with special needs can be matched to areas that meet their needs first. For example, candidates who are sensitive to light will be assigned to areas with higher light stability values, while candidates who are sensitive to noise will be assigned to areas with lower noise interference values. The introduction of comprehensive suitability scores ensures that the system takes multiple environmental parameters into consideration during the matching process, avoiding over-optimization of a single parameter.

[0110] As the examination environment changes in real time (such as a sudden increase in noise in a certain area), the system can quickly adjust the regional suitability score and dynamically optimize the allocation plan to ensure that candidates are always in a suitable examination environment.

[0111] After multiple rounds of training, the reinforcement learning algorithm can quickly generate the optimal allocation plan, significantly improving the allocation efficiency while ensuring fair consideration of the needs of different candidates.

[0112] In the example, the examination room is divided into three areas (A, B, and C), and the initial suitability score of each area is:

[0113] Area A: light uniformity 0.8, noise interference value 0.9, temperature and humidity balance value 0.7;

[0114] Area B: light uniformity 0.7, noise interference value 0.8, temperature and humidity balance value 0.6;

[0115] Area C: Light uniformity 0.6, noise interference value 0.7, temperature and humidity balance value 0.8.

[0116] Special needs candidates' needs: Candidate X: light sensitive; Candidate Y: noise sensitive.

[0117] Reinforcement learning process:

[0118] Based on the regional scores and candidate needs, the system calculates a comprehensive suitability score:

[0119] Region A: The overall score is 0.4·0.8+0.3·0.9+0.3·0.7=0.79;

[0120] Region B: The overall score is 0.4·0.7+0.3·0.8+0.3·0.6=0.72;

[0121] Area C: The comprehensive score is 0.4·0.6+0.3·0.7+0.3·0.8=0.69.

[0122] After the light intensity fluctuation caused the light uniformity in area A to drop to 0.5, the system recalculated the comprehensive score and the result was:

[0123] Region A: The overall score dropped to 0.4·0.5+0.3·0.9+0.3·0.7=0.68;

[0124] The system adjusts the allocation plan for candidate X and transfers him from area A to area B.

[0125] Final solution: Candidate X is assigned to area B with the best light uniformity; Candidate Y is assigned to area A with the lowest noise interference value.

[0126] like Figure 4 As shown, based on the regional allocation data and the examinee information data, the seat state matrix is ​​initialized, and preliminary seat allocation data is generated through a dynamic programming algorithm; conflict detection and optimization are performed on the preliminary seat allocation data, and the seat allocation plan is adjusted using a reinforcement learning algorithm to generate optimized seat allocation data;

[0127] Preferably, the preliminary seat allocation data is generated by jointly optimizing the regional allocation data and the examinee information data through a dynamic programming algorithm, and the optimization process includes calculating the priority allocation sequence of seats in each area with the goal of minimizing the distance between examinees and meeting special needs.

[0128] Dynamic Programming (DP) is a method for solving optimization problems, which is mainly achieved by decomposing complex problems into sub-problems and solving them recursively. In the present invention, the core of the dynamic programming algorithm is to establish a recursive equation to optimize the matching scheme between candidates and seats.

[0129] Let f(i, j) represent the minimum cost of assigning the first i candidates to the first j seats, where the cost includes the distance between candidates and the penalty for not meeting special needs.

[0130] Transfer equation:

[0131]

[0132] Among them, C(i, j) represents the incremental cost of assigning the i-th candidate to the j-th seat; k represents the possible seat choice of the previous candidate; and f(i-1, k) represents the optimal allocation plan for the first i-1 candidates.

[0133] Distance between candidates. Distance cost is calculated by the physical distance between candidates:

[0134]

[0135] Among them, x i ,y i represents the target seat coordinates of candidate i; x j ,y j represents the target seat coordinates of candidate j. For allocations that do not meet special requirements, a penalty term is introduced:

[0136] P(i, j) = α·(1-M(i, j))

[0137] Among them, M(i, j) represents the binary matching function of whether candidate i meets the special requirements (1 for satisfaction and 0 for non-satisfaction); α is the penalty weight coefficient.

[0138] The goal of the dynamic programming algorithm is to minimize the following total cost:

[0139]

[0140] Among them, n represents the total number of candidates and m represents the total number of seats.

[0141] Based on the regional suitability score, the dynamic programming algorithm prioritizes seat allocation within high suitability areas, ensuring that candidates with special needs can be allocated seats that best meet their needs.

[0142] The dynamic programming algorithm reduces the discomfort caused by over-concentration or over-dispersion of candidates, such as noise interference or light obstruction, by minimizing the distance between candidates. The distance between candidates is more evenly distributed, improving the comfort of the overall examination environment.

[0143] The allocation priority of candidates with special needs is fully guaranteed. For example, candidates who are sensitive to light can be allocated seats in areas with even lighting, and candidates who are sensitive to noise can be allocated seats in areas with low noise.

[0144] The dynamic programming algorithm can quickly adjust the allocation plan when environmental parameters or examinee needs change, ensuring the dynamic adaptability of the plan.

[0145] In the example, the examination room is divided into three areas (A, B, C), and the suitability scores of each area are:

[0146] Area A: 0.85; Area B: 0.72; Area C: 0.65.

[0147] Candidates with special needs include: Candidate X (noise-sensitive); Candidate Y (light-sensitive).

[0148] The initial state is defined as each candidate has not yet been assigned a seat. For candidate X, the system calculates the best seat choice in area A. The optimization formula is:

[0149]

[0150] The results show that the best seat in area A is in the center. For candidate Y, the system gives priority to seats in area B based on the light stability value. The final allocation plan: candidate X is assigned to the central seat in area A; candidate Y is assigned to the seat with the highest light stability value in area B.

[0151] Preferably, the optimized seat allocation data is generated by a conflict detection module and a reinforcement learning adjustment module, which specifically includes detecting conflict factors in the seat allocation plan and dynamically optimizing the allocation strategy according to the utilization rate of examination room resources and the priority of examinee needs.

[0152] In the present invention, the optimized seat allocation data is generated by a conflict detection module and a reinforcement learning adjustment module. The conflict detection module is used to identify potential conflict factors in the preliminary seat allocation plan, mainly including physical conflicts (such as multiple candidates are assigned to the same seat or the distance between adjacent seats is too small), demand conflicts (such as special needs candidates are not assigned to areas that meet their needs), and resource conflicts (such as excessive concentration or significant insufficiency of regional resource utilization). The conflict detection module quickly identifies problems based on the logical analysis of the allocation matrix. For example, by calculating whether each seat is assigned to multiple candidates at the same time, if the allocation mark of a seat exceeds 1, it is marked as a conflict. In addition, by checking whether the allocation matrix meets the matching rules of special needs candidates and environmental suitability parameters, the existence of demand conflicts can be judged. After detecting the conflict, the reinforcement learning adjustment module adjusts the seat allocation through a dynamic optimization strategy, gives priority to resolving high-priority conflicts, and dynamically optimizes the allocation strategy based on the utilization of examination room resources and the priority of candidate needs.

[0153] The reinforcement learning adjustment module is centered on intelligent agents and continuously optimizes the seat allocation plan through iterative interactions with the environment. The system models the current seat allocation state as a state space, including resource utilization, candidate demand satisfaction rate, and the number of conflicts. The action space includes operations such as reallocating seats and adjusting regional priorities. The reward function comprehensively considers the utilization rate of examination room resources, the degree of satisfaction of candidate needs, and the efficiency of conflict resolution. Specifically, the reward value is equal to the weighted sum of resource utilization and demand satisfaction rate minus the penalty term of the number of conflicts. During the optimization process, the reinforcement learning module dynamically adjusts the allocation strategy based on the feedback of the reward function, and minimizes conflicts and maximizes resource utilization through iterative calculations.

[0154] The optimization process starts with the input of the initial allocation plan. First, the conflict detection module analyzes the plan and marks the conflicting areas. Based on the marking results, the reinforcement learning module prioritizes the allocation plan for areas with more conflicts according to the reward function, for example, by reallocating special needs candidates to more suitable areas or adjusting the distribution of ordinary candidates in areas with surplus resources, so as to optimize the overall plan. After multiple rounds of iterations, the number of conflicts gradually decreases, and the allocation plan tends to be optimal until all conflicts are resolved or the number of conflicts is reduced to an acceptable range.

[0155] The effects of the present invention are reflected in many aspects. First, through the precise analysis of the conflict detection module and the dynamic adjustment of reinforcement learning, the system can efficiently solve the conflict problem in seat allocation and ensure the rationality of the allocation plan. For example, when multiple light-sensitive candidates are concentrated in the same area and cause conflicts, the system can identify and adjust the allocation plan in time, and reallocate some candidates to other areas with uniform light. Secondly, the present invention significantly improves the utilization rate of examination room resources. Through dynamic adjustment, the resources in low-utilization areas are reasonably allocated to ordinary candidates, avoiding resource waste, while meeting the priority use requirements of high-demand areas. In addition, the optimization strategy of reinforcement learning ensures the fairness and rationality of the allocation, the priority of candidates with special needs is fully guaranteed, and the allocation of ordinary candidates takes into account resource balance and optimization of the overall environment of the examination room.

[0156] For example, in an actual application in a certain examination room, the initial allocation plan allocated 5 ordinary candidates to area A, 3 light-sensitive candidates to area B, and 2 noise-sensitive candidates to area C. However, due to fluctuations in the light stability score of area B, the needs of some light-sensitive candidates were not met. The conflict detection module marks area B as a conflict area, and the reinforcement learning module adjusts the plan based on the reward function to reallocate some light-sensitive candidates to area A with a higher light stability value, while adjusting the distribution of ordinary candidates to ensure maximum resource utilization. The final allocation plan meets the requirements of all candidates with special needs, while achieving a balanced distribution and efficient use of examination room resources. Through the above method, the present invention significantly improves the scientificity, flexibility and accuracy of examination seat arrangement.

[0157] Through the multimodal verification module of the intelligent examination platform, the candidate's biometric data is collected, identity authentication is completed, and verification data is generated; the verification data is bound to the optimized seat allocation data to complete the examination interface unlocking and task loading.

[0158] Preferably, the multimodal verification module includes a fingerprint recognition module and a face recognition module, and the verification pass data is generated by the matching results of the fingerprint feature vector and the face feature vector.

[0159] Fingerprint recognition module: candidates input fingerprint data on the intelligent examination platform, and the fingerprint image is captured by the high-resolution fingerprint sensor, and the fingerprint feature vector (such as bifurcation point, endpoint, etc.) is extracted. The extracted feature data is matched with the fingerprint template pre-stored in the system, and a partial verification result is generated if the match is successful.

[0160] The face recognition module uses the camera of the intelligent examination platform to capture the examinee's face image and uses a deep learning model (such as convolutional neural network, CNN) to extract the face feature vector, including facial key points (such as the relative positions of eyes, nose, and mouth) and texture features. The extracted feature vector is matched with the face template stored in the database to generate a matching result.

[0161] Comprehensive verification: conduct a comprehensive analysis of the matching results of the fingerprint feature vector and the face feature vector, and generate the final verification pass data through a weighted fusion strategy. The fusion strategy calculation formula is:

[0162] V=α·F 指纹 +β·F 人脸

[0163] Where V represents the comprehensive matching score; F 指纹 represents the fingerprint matching score; F 人脸 Represents the face matching score; α and β represent the weight coefficients of fingerprint and face respectively, which are dynamically adjusted according to the scene requirements.

[0164] After the verification is passed, the system binds the candidate's verification data with their optimized seat allocation data and generates a unique identifier (such as verification binding ID). This identifier contains the candidate's identity authentication information and their seat allocation results, which serves as the basis for unlocking the test interface and loading the test tasks.

[0165] The binding data is transmitted to the management system of the smart examination platform through the local area network. After the candidate logs in, the system unlocks the corresponding examination interface based on the binding data and loads the candidate's examination task. The examination task loading includes the presentation of the task content and the monitoring settings related to the candidate's seat.

[0166] The multimodal fusion verification of fingerprints and faces avoids the misidentification or forgery problems that may occur in a single verification method. Even if the fingerprint image is partially missing or the face is affected by lighting conditions, multimodal fusion can still provide high accuracy. The joint verification of fingerprints and faces improves verification security and reduces the risk of identity fraud during the examination process.

[0167] The fingerprint and face feature extraction and matching adopt efficient algorithms to ensure that the verification process is completed quickly (usually within 1 second). The generation and transmission of binding data is completed through the local area network, which has high real-time performance. The binding of verification data and seat allocation data realizes a one-to-one correspondence, which facilitates the subsequent unlocking of the test interface and loading of tasks.

[0168] Verification ensures that candidates can only access their corresponding test interfaces and tasks through data binding, preventing task loading errors or interface confusion. The system records the time and results of each verification, which is convenient for test center management and subsequent tracing.

[0169] Example: In a certain exam, the examination room is equipped with a multimodal verification module of the intelligent examination table. After Candidate A sits in the designated seat, he places his finger on the fingerprint recognition sensor and faces the camera for face collection. The fingerprint feature vector collected by the fingerprint sensor is [0.8, 0.4, 0.6, 0.9], and the feature vector extracted by the face recognition model is [0.7, 0.5, 0.8, 0.6]. The system calculates the matching scores of fingerprint and face respectively:

[0170] F 指纹 =0.8·0.7+0.4·0.5+0.6·0.8+0.9·0.6=1.74

[0171] F 人脸 =0.7·0.8+0.5·0.6+0.8·0.9+0.6·0.7=1.90

[0172] The comprehensive matching score is calculated through the comprehensive verification strategy:

[0173] V=0.6·1.74+0.4·1.90=1.804

[0174] If the comprehensive score V>1.5 is the passing condition, the identity verification of candidate A is passed. The system binds the verification pass data with the seat allocation data of candidate A (area B seat 3) and generates a unique identifier ID-A3.

[0175] The binding data is transmitted to the examination platform management system through the local area network, unlocking the examination interface of seat 3 in area B and loading the corresponding examination task. At the same time, the system records the verification time and binding data of candidate A to ensure that the entire process of examination management is traceable.

[0176] Preferably, the test interface unlocking and task loading include activating the examinee's test authority based on the verification data, loading the test task in real time, and recording the start time, completion status and abnormal operation information of the test task.

[0177] When the verification data is transmitted to the task management module of the intelligent examination platform, the system automatically matches the verification data with the candidate information database to confirm the candidate's identity and the corresponding examination authority. The core of examination authority activation lies in the uniqueness and consistency of the verification data. For example, the bound verification data must be consistent with the seat allocation data, and only candidates who pass the verification can enter the examination interface. The system creates a unique session key (SessionKey) for the candidate based on the permission activation status. The key is used to protect the transmission security of the examination task data and prevent unauthorized access.

[0178] After the permissions are activated, the task management system extracts the corresponding test content from the task database based on the candidate information and loads it to the smart test station through the local area network. The loading process includes the initialization of test questions, test rules, timers, and interactive interfaces to ensure that the test tasks are presented complete and correct. Task loading also triggers the real-time monitoring module, which is responsible for recording the candidate's operation behavior during the test, such as page switching, answering progress, and whether there are abnormal operations (such as abnormal exit from the interface).

[0179] The system records the start time of the test task in real time when the task is loaded, ensuring that the time when the examinee starts answering questions can be accurately tracked. During the execution of the task, the system continuously monitors the task completion status, including information such as the answer completion rate and submission time. Once the test task is completed, the system automatically generates a task completion status report and stores it in the test data log. If the examinee performs abnormal operations during the task execution (such as attempting to switch task interfaces, unauthorized access to other functional modules), the system immediately captures and records the abnormal information, including the time when the abnormality occurred, the specific operation description and the impact assessment.

[0180] Unlocking the exam interface strictly relies on identity verification results, ensuring that each candidate can only access the exam tasks bound to them. The rigor of permission management effectively prevents the incorrect allocation of exam tasks and unauthorized access.

[0181] The system can load tasks immediately after verification, avoiding the delay problem of traditional exam task loading. At the same time, it ensures the accuracy and stability of test questions and interfaces through integrity verification of task loading.

[0182] The system comprehensively records key operations (such as start time, completion status, abnormal information, etc.) during task execution, providing strong guarantees for the security and fairness of the examination process. In particular, the instant capture and recording of abnormal operations helps managers quickly troubleshoot problems and take corresponding measures.

[0183] The activation of examination permissions and task loading are protected by encryption mechanisms throughout the process, ensuring the security of examination data transmission and avoiding security risks caused by data leakage or tampering.

[0184] Embodiment: In an exam based on a smart test table, after candidate A completes identity authentication through the multimodal verification module, the verification pass data generated by the system is bound to the seat allocation information (area B seat 3). After the verification pass data is transmitted to the task management system, the system activates the examination authority based on candidate A's identity information, and creates a session key to protect the loading process of the task data. Subsequently, the system extracts candidate A's test content from the task database, including questions, timer rules, and test interface configuration, and presents it on the touch screen of the smart test table after the task is loaded. At the beginning of the exam, the system records the task start time as 10:00:00, and monitors candidate A's operation behavior in real time in the background.

[0185] During the exam, Candidate A attempted to switch the task interface. The system identified this operation as abnormal and recorded detailed information, including the abnormal time 10:15:32, the operation description "attempt to enter an unauthorized module", and the preliminary assessment "did not affect the current exam task". After the task was completed, the system automatically recorded the exam completion time as 11:30:00 and generated a task completion status report, including Candidate A's answer completion rate of 100%, submission time of 11:29:45, and no major abnormal impact.

[0186] like Figure 5 As shown, a system for implementing the test seat arrangement method based on the intelligent test platform includes:

[0187] Environmental sensing module for collecting environmental data;

[0188] Implicit sensing module for collecting air flow and micro-vibration characteristics;

[0189] An environmental feature analysis module for integrating environmental perception data and implicit environmental data and generating an environmental feature matrix;

[0190] Candidate information management module for receiving and standardizing candidate information data;

[0191] A regional division module for dividing the examination room area and generating regional suitability data using a quantum dot distribution model based on the environmental feature matrix and the physical layout of the examination room;

[0192] A seating arrangement module for generating regional allocation data and initializing a seat state matrix based on regional suitability data, generating preliminary seat allocation data through a dynamic programming algorithm, and generating optimized seat allocation data through a reinforcement learning algorithm;

[0193] A multimodal verification module for collecting the examinee's biometric data and generating verification pass data;

[0194] An execution module used to bind the verified data with the optimized seat allocation data and complete the test interface unlocking and task loading.

[0195] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware.

[0196] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for arranging test seats based on an intelligent test platform, characterized in that: The following steps are involved: Collect environmental data of the examination room through the intelligent examination platform to generate environmental perception data; combine implicit perception technology to collect air flow and micro-vibration characteristics to generate implicit environmental data; integrate and process environmental perception data and implicit environmental data to generate an environmental feature matrix; receive and standardize the candidate information data; Based on the environmental feature matrix and the physical layout of the examination room, the quantum dot distribution model is used to divide the examination area and generate regional suitability data; According to the regional suitability data, candidates with special needs are allocated to suitable areas in priority to generate regional allocation data; Based on the area allocation data and candidate information data, the seat state matrix is ​​initialized, and preliminary seat allocation data is generated through a dynamic programming algorithm; Conduct conflict detection and optimization on the preliminary seat allocation data, use reinforcement learning algorithm to adjust the seat allocation plan, and generate optimized seat allocation data; Through the multimodal verification module of the intelligent examination platform, the candidate's biometric data is collected, identity authentication is completed, and verification data is generated; the verification data is bound to the optimized seat allocation data to complete the examination interface unlocking and task loading.

2. The method for arranging test seats based on an intelligent test platform according to claim 1, characterized in that: The environmental data includes light intensity, humidity, temperature and noise level. The implicit environmental data is generated by real-time detection of air flow characteristics and micro-vibration frequency characteristics by highly sensitive sensors configured in the intelligent examination table.

3. The method for arranging test seats based on an intelligent test platform according to claim 1, characterized in that: The environmental feature matrix dynamically extracts features from environmental perception data and implicit environmental data through a spatiotemporal convolutional neural network, specifically including a joint analysis of spatial distribution and temporal variation patterns, and extracts feature data on light uniformity, temperature and humidity stability, and noise variation trends.

4. The method for arranging test seats based on an intelligent test platform according to claim 1, characterized in that: The quantum dot distribution model adopts a quantum energy simulation method to generate regional suitability data by performing quantum state simulation calculations on the environmental suitability parameters in the examination area. The regional suitability data includes the light stability value, noise interference value and environmental temperature and humidity balance value of each area.

5. The method for arranging test seats based on an intelligent test platform according to claim 1, characterized in that: The regional allocation data is generated through a reinforcement learning algorithm, in which the allocation rules for special needs candidates use the regional suitability score as a weight, and dynamically adjust the matching relationship between special needs candidates and priority areas by calculating the comprehensive suitability scores of different candidate areas.

6. The method for arranging test seats based on an intelligent test platform according to claim 1, characterized in that: The preliminary seat allocation data is generated by jointly optimizing the regional allocation data and the examinee information data through a dynamic programming algorithm. The optimization process includes calculating the priority allocation sequence of seats in each area with the goal of minimizing the distance between examinees and meeting special needs.

7. The method for arranging test seats based on an intelligent test platform according to claim 1, characterized in that: The optimized seat allocation data is generated through a conflict detection module and a reinforcement learning adjustment module, which specifically includes detecting conflict factors in the seat allocation plan and dynamically optimizing the allocation strategy based on the utilization rate of examination room resources and the priority of examinee needs.

8. The method for arranging test seats based on an intelligent test platform according to claim 1, characterized in that: The multimodal verification module includes a fingerprint recognition module and a face recognition module, and the verification pass data is comprehensively generated by the matching results of the fingerprint feature vector and the face feature vector.

9. The method for arranging test seats based on an intelligent test platform according to claim 1, characterized in that: The test interface unlocking and task loading include activating the examinee's test authority based on the verification data, loading the test task in real time, and recording the start time, completion status and abnormal operation information of the test task.

10. A system for implementing the test seat arrangement method based on the intelligent test platform according to any one of claims 1 to 9, characterized in that: include: Environmental sensing module for collecting environmental data; Implicit sensing module for collecting air flow and micro-vibration characteristics; An environmental feature analysis module for integrating environmental perception data and implicit environmental data and generating an environmental feature matrix; Candidate information management module for receiving and standardizing candidate information data; A regional division module for dividing the examination room area and generating regional suitability data using a quantum dot distribution model based on the environmental feature matrix and the physical layout of the examination room; A seating arrangement module for generating regional allocation data and initializing a seat state matrix based on regional suitability data, generating preliminary seat allocation data through a dynamic programming algorithm, and generating optimized seat allocation data through a reinforcement learning algorithm; A multimodal verification module for collecting the examinee's biometric data and generating verification pass data; An execution module used to bind the verified data with the optimized seat allocation data and complete the test interface unlocking and task loading.

Images