Radar human-computer interface evaluation method and system based on multi-modal data fusion
By using multimodal data fusion technology, the problems of insufficient evaluation standardization and environmental adaptability in radar human-machine interface design have been solved, enabling scientific quantitative evaluation and optimization of the interface and improving the efficiency and security of interface design.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- THE 724TH RESEARCH INSTITUTE OF CHINA STATE SHIPBUILDING CORP LTD
- Filing Date
- 2025-12-27
- Publication Date
- 2026-04-17
AI Technical Summary
Existing radar human-machine interface designs lack standardized evaluation, have insufficient environmental adaptability, and fragmented multimodal interactions. They are difficult to dynamically quantify user cognitive load and operational efficiency in complex environments, leading to interface design flaws that increase the probability of misoperation and affect combat effectiveness.
A multimodal data fusion method is adopted. By constructing a multimodal synchronous evaluation environment, multi-stage testing is performed to collect operational performance, eye movement, EEG and subjective evaluation data, perform spatiotemporal alignment and fusion, calculate the radar interface health index, generate interface bottleneck analysis diagram and output optimization suggestions, and conduct A/B testing for verification.
It enables scientific, objective, and accurate evaluation of radar human-machine interfaces, can capture fluctuations in operator cognitive state in real time, improves interface design iteration efficiency, reduces the probability of misoperation, and improves operator performance in complex environments.
Smart Images
Figure CN121880786A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to human-computer interaction evaluation technology, specifically to a radar human-computer interface evaluation method and system based on multimodal data fusion. Background Technology
[0002] Current radar human-machine interface (HMI) design faces core challenges such as a lack of standardized evaluation, insufficient environmental adaptability, and fragmented multimodal interaction. Traditional radar interface evaluation relies primarily on expert review or static task performance analysis, lacking dynamic quantitative assessment of user cognitive load and operational efficiency in complex environments. This is especially critical in high-risk domains, where the interface serves as the bridge between the operator and the complex radar system. Inefficient, confusing, or difficult-to-understand interfaces severely slow down target identification, tracking, and decision-making, which is undoubtedly fatal in time-sensitive applications. Furthermore, in today's complex environments, operators endure immense physiological and psychological stress; a flawed interface significantly increases the probability of misoperation, potentially leading to serious consequences such as accidental touches and missed detections. The purpose of radar HMI evaluation is to improve operational effectiveness and reduce human error; establishing intuitive and user-friendly interfaces can reduce training costs and time. As radar HMI evaluation adapts to technological advancements and software-defined radar becomes a trend, radar HMIs require frequent updates and iterations. Rigorous evaluation is essential to ensure that each update brings positive improvements. Summary of the Invention
[0003] The purpose of this invention is to provide a radar human-machine interface evaluation method and system based on multimodal data fusion.
[0004] The technical solution to achieve the purpose of this invention is: a radar human-machine interface evaluation method based on multimodal data fusion, comprising the following steps:
[0005] S1. Construct a multimodal synchronous evaluation environment, execute multi-stage standardized test protocols, and control the tested operator to perform multi-stage test tasks, including static interface evaluation tasks, dynamic operation tasks, and stress environment task tests.
[0006] S2. While the tested operator performs a multi-stage test task, multimodal data is collected. The multimodal data includes operational performance data, eye-tracking data, electroencephalogram (EEG) data, and subjective evaluation data. Operational performance data includes task completion time and accuracy rate; eye-tracking data includes theta waves, pupil diameter, fixation point dispersion, and eye-tracking fixation hotspots; EEG data includes periods of high EEG load and areas of high error operation; subjective evaluation data is obtained based on a multi-dimensional subjective load score, which includes interface comprehensibility, operational satisfaction, task confidence, and perceived usability.
[0007] S3. Perform spatiotemporal alignment and fusion on the collected multimodal data, calculate the operational performance calibration score, physiological load calibration score, and subjective experience calibration score, and calculate the radar interface health index through a scenario-adaptive dynamic weighting algorithm to complete the quantitative assessment;
[0008] S4. Based on eye-tracking fixation hotspots, periods of high EEG load, and areas of high error operation, overlay rendering is performed on the interface screenshot to generate an interface bottleneck analysis diagram and output interface optimization suggestions. The optimization suggestions include at least one of layout reorganization, color coding optimization, interaction process simplification, and information display enhancement.
[0009] S5. After modifying the prototype interface based on the optimization suggestions, repeat steps S1 to S4 to conduct A / B testing and verification until the radar interface health index increases beyond the preset threshold.
[0010] Furthermore, the multimodal synchronous evaluation environment includes a radar human-machine interface system to be evaluated, a radar signal generator, an environment simulator, an eye tracker, an EEG acquisition device, and a synchronization controller. The synchronization controller is used to receive event marker signals injected by the radar signal generator and uniformly send synchronization trigger signals to the eye tracker and the EEG acquisition device.
[0011] Furthermore, the test tasks performed on the operator under test in step S1 include:
[0012] Static interface assessment task: Present an overview interface of radar health status, record the operator's accuracy in understanding key information and reaction time, including the health status and threat level of each subsystem;
[0013] Dynamic operational tasks: Simulate typical combat or operational processes, including search, identification, tracking, and decision-making, and record task completion efficiency and error rate;
[0014] Stress Environment Task: Simulate a complex electromagnetic environment by injecting electromagnetic interference signals to test the interface robustness and operator stress resistance. The electromagnetic interference signals include noise and deceptive interference.
[0015] Furthermore, in step S3, the spatiotemporal alignment is based on the synchronous trigger mark, which aligns the operation performance data, eye-tracking data, and EEG data on a unified time axis to form a multimodal data stream with a timestamp error of less than 10ms; the spatiotemporal alignment is implemented using the Precise Time Protocol (PTP).
[0016] Furthermore, in step S3, the operation performance calibration score is... ,Depend on Current user operation performance score User historical average performance The performance score of expert-level operators is calculated using the following criteria:
[0017] ;
[0018] Among them, each level , , The actual operational performance score is given by the operational performance calculation method P, which is calculated from the task completion time and accuracy rate.
[0019] ;
[0020] ;
[0021] ;
[0022] in, Here, T_actual is the actual time the user takes to complete the task, and T_benchmark is the benchmark completion time, representing the efficiency weighting coefficient. To ensure accurate performance weighting, A_task represents task-level accuracy, and A_action represents operation-level accuracy.
[0023] Furthermore, in step S3, the physiological load calibration score... The calculation formula is:
[0024] ;
[0025] Where B represents the actual score of physiological load:
[0026] ;
[0027] Where ENI is the environmental noise index, α is the adjustment coefficient, S_θ, S_p, and S_d are the standardized scores of eye movement theta wave, pupil diameter, and fixation point dispersion, respectively, and β_θ, β_p, and β_d are the weights of the corresponding indicators.
[0028] Furthermore, in step S3, the subjective experience calibration score... It is a weighted fusion of multi-dimensional subjective scores, and the calculation formula is:
[0029] ;
[0030] in and These are the standard deviation and mean of the subjective ratings given by all participants on this interface, respectively. This is the adjustment coefficient, where S is the actual subjective experience score. The formula is:
[0031] ;
[0032] R_i is the original score of the i-th dimension, W_i is its weight, C_i is the score consistency coefficient of this dimension, CF is the credibility factor, and δ is the adjustment coefficient.
[0033] Furthermore, in step S3, the radar interface health index incorporates a dynamic function that varies with the assessment task stage and environmental state to address the index weighting issue under different scenarios. The specific method is as follows:
[0034] Define task phase vector and environment state vector :
[0035]
[0036]
[0037] The weight function distribution of each indicator is as follows , and for:
[0038] ;
[0039] in, , , These are the basic weights for operational performance (P), physiological workload (B), and subjective experience (S). , and These are the various offsets for each task stage, including performance adjustment (P), physiological load (B), and subjective experience (S). , These are the environmental state offsets of operational performance correction (P), physiological load (B), and subjective experience (S);
[0040] The Radar Interface Health Index (R-CHCI) is:
[0041] ;
[0042] in, This is a consistency penalty term used to handle inconsistencies between metrics.
[0043] ;
[0044] in, , , These are the normalized operational performance calibration score, physiological load calibration score, and subjective experience calibration score. , It is the magnification factor.
[0045] Furthermore, in step S4, the layout reorganization includes:
[0046] Following the three-zone rule, the interface is divided into a status zone, an operation zone, and an early warning zone. The status zone is used to display core parameters and health status, the operation zone is used to arrange commonly used controls, and the early warning zone is used to display faults and alarms.
[0047] The most frequently used operation control key is placed in the golden line of sight area, which is defined as the display range of ±15° horizontal line of sight and 0°-30° vertical line of sight.
[0048] By compressing the menu depth, high-frequency functions in the second and third level menus are promoted to the first level menu or set as shortcut operations, and information is managed using a card-style design that can be collapsed or expanded, with only core information displayed by default.
[0049] A radar human-machine interface evaluation system based on multimodal data fusion, used to implement the aforementioned radar human-machine interface evaluation method based on multimodal data fusion, specifically includes:
[0050] The task control module is used to control the operator under test to execute multi-stage test tasks;
[0051] The multimodal data acquisition module is used to simultaneously collect operational performance, eye movement, electroencephalogram (EEG), and subjective evaluation data.
[0052] The data fusion and computation module is used for spatiotemporal alignment, calculation of calibration scores, and interface health index.
[0053] The bottleneck analysis module is used to generate heatmaps of bottleneck areas on the interface and output optimization suggestions.
[0054] The closed-loop verification module is used to perform A / B testing and evaluate the optimization effect.
[0055] Compared with existing technologies, the significant advantages of this invention are: 1) By introducing environmental interference factors, personal baseline calibration, and dynamic weight design, it can adaptively calculate and perceive the unique scenarios, environment, user status, and dynamic interface of radar operation. 2) Utilizing multimodal data overlay analysis, it can not only discover defects but also explain the causes of defects, providing a direct basis for optimization. 3) It can capture the fluctuations in the operator's cognitive state under complex and high-pressure tasks in real time, making the evaluation results closer to real application scenarios. 4) It forms a complete closed loop of evaluation-analysis-optimization-verification, significantly improving the efficiency and effectiveness of interface design iteration. Attached Figure Description
[0056] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation
[0057] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0058] A radar human-machine interface evaluation method based on multimodal data fusion is proposed, and the scheme is as follows:
[0059] S1. Construct a multimodal synchronous evaluation environment;
[0060] The multimodal synchronous testing environment includes: the radar human-machine interface system to be tested, a radar signal generator, an eye tracker, an EEG acquisition device, and a synchronization controller. The synchronization controller is used to receive the event marker signal injected by the radar signal generator and uniformly send synchronization trigger signals to the eye tracker and the EEG acquisition device to ensure that the timestamp error of all data sources is less than 10ms. Specific requirements are shown in Table 1.
[0061] Table 1: Components of the synchronous data acquisition environment deployment in this invention
[0062]
[0063] S2. Execute a multi-stage standardized test protocol to control the operator under test to perform the following test tasks and simultaneously collect multimodal data;
[0064] Perform the following test tasks:
[0065] S2a: Static Interface Assessment Task: Present an overview of radar health status and record the operator's accuracy and reaction time in understanding key information (such as the health status of each subsystem and threat level).
[0066] S2b: Dynamic Operation Tasks: Simulate typical task stages or operation processes (such as search, identification, tracking, and decision-making), and record task completion efficiency, error rate, etc.
[0067] S2c: Stress Environment Task: Simulate a complex electromagnetic environment by injecting electromagnetic interference signals (such as noise and deceptive interference) through a radar signal generator, and test the interface robustness and operator stress resistance.
[0068] During the above tasks, the following multimodal data streams will be collected simultaneously:
[0069] Operational performance data: task completion time, accuracy, etc.;
[0070] Eye movement data: theta waves, pupil diameter, fixation point dispersion, and fixation hotspots;
[0071] EEG data: periods of high EEG load and areas of high error rate;
[0072] Subjective evaluation data: Based on subjective workload ratings obtained from multiple dimensions, which generally include interface understandability, operational satisfaction, task confidence, and perceived usability.
[0073] S3, Multimodal data fusion and quantitative evaluation;
[0074] S3a: Spatiotemporal alignment: Based on synchronous trigger tags, operational performance data, eye-tracking data, and EEG data are aligned on a unified time axis to form a multimodal data stream.
[0075] S3b: Based on multimodal data results, construct a dynamic scenario-weighted operational performance calibration index.
[0076] Define the task phase vector as Phase = {search, identification, tracking, decision-making, ...};
[0077] Define the environment state vector as State = {normal, disturbance, emergency, ...};
[0078] Build This indicates the operational performance index after calibration, which is... Current user operation performance score User historical average performance The performance score of expert-level operators is calculated using the following criteria:
[0079]
[0080] in, , , The actual operational performance score is given by the operational performance calculation method P, which is calculated from the task completion time and accuracy rate.
[0081]
[0082]
[0083]
[0084] In the formula, This is the efficiency weighting coefficient, which is based on an efficiency weighting coefficient between 0 and 1. It can be adjusted according to the task type. For example, it can be set to 0.6 for "tracking tasks" and 0.4 for "search tasks". The actual time it takes for the user to complete the task. Based on the completion time, For task-level accuracy, For operational-level accuracy; To accurately measure performance weights, which are used to balance task success and operational accuracy, the weights are usually determined based on the current task type. They are typically set to 0.7, indicating that more importance is placed on whether the current task is ultimately completed.
[0085] S3c: Based on multimodal data results, construct a dynamic scenario-weighted physiological load calibration index. Indicates the physiological load index after calibration:
[0086]
[0087]
[0088] Where ENI is the environmental noise index, calculated from the interference signal-to-noise ratio currently output by the radar signal generator (e.g., ENI = (reference SNR - current SNR) / reference SNR); α is the adjustment coefficient, obtained by performing linear regression analysis on physiological load data collected at different environmental noise index levels (0, 0.5, 1.0, 2.0) to fit the environmental interference influence coefficient. ;
[0089] In the formula, , , , where represents the weight of each indicator. The weights are dynamically adjusted based on the task type to reflect the different cognitive needs and priorities of individuals in terms of cognitive resources. Typical task weight selections are shown below:
[0090]
[0091] It is the average relative power proportion of the theta wave (4-8Hz) in the current mission segment; This represents the absolute value of the average amplitude of the pupil diameter change during the current task segment; The gaze dispersion rate of the current task segment;
[0092] S3d: Based on multimodal data results, construct a dynamic scenario-weighted subjective experience calibration index. The subjective experience score index after calibration is a weighted fusion of subjective scores from multiple dimensions (including interface comprehensibility, operation satisfaction, task confidence, and perceived usability).
[0093]
[0094] in and These are the standard deviation and mean of the subjective ratings given by all participants on this interface, respectively. is the adjustment coefficient, and S is the subjective experience score.
[0095]
[0096] R_i is the raw score for the i-th dimension (including four dimensions: interface comprehensibility, operation satisfaction, task confidence, and perceived usability).
[0097] W_i represents the weight of the i-th dimension. The weights are determined using the analytic hierarchy process (AHP) based on the results of a survey by domain experts. The weight matrix is then normalized and averaged row-wise to obtain the eigenvector (i.e., the weight vector) as (0.30, 0.25, 0.25, 0.2).
[0098] C_i is the rating consistency coefficient for this dimension, used to automatically identify and reduce the interference of "internal inconsistency among raters" or "mediocrity bias" on the results. , For the predicted score of the i-th dimension, The rating scale ranges from 100.
[0099] For credibility factors; ; Give the user a corrected score across all dimensions (i.e.) The standard deviation of ). This is the magnification factor (e.g., 0.5). It is a non-positive value if users rate it highly consistently across all dimensions. (small), then Approaching 0 has little impact on the total score S; if the evaluations of each dimension are extremely contradictory ( If the value is large, then CF will be a significantly negative value, directly affecting the total score.
[0100] δ is an adjustment coefficient used to adjust the influence weight of the penalty item, determining the strength of the credibility penalty. To avoid highly contradictory ratings and user feedback dominating the overall evaluation, a maximum tolerable penalty is set for the system. This maximum tolerable penalty has been demonstrated to be appropriate. When the value is >25, the feedback should be significantly suppressed, and the theoretical value δ boundary is set to 2.5. Analysis of 50 valid subjective rating data collected in the preliminary experiment shows that the calculated δ data verification value is 0.93. To balance system safety and sensitivity to routine assessment, δ = 1.0 is ultimately chosen.
[0101] 4. S3e: Construct a dynamic scenario-weighted interface health index, with weights using a function that dynamically changes with the task stage and environmental state of the assessment.
[0102]
[0103] in,
[0104]
[0105]
[0106] In the formula, , , The basic weights for operational performance (P), physiological load (B), and subjective experience (S) are determined using the Analytic Hierarchy Process (AHP). A judgment matrix was constructed using this matrix (ten experts in radar human-computer interaction were invited to perform pairwise comparisons of the three weights based on scaling methods, and matrix A was obtained by synthesizing the questionnaire results). Matrix A is as follows:
[0107]
[0108] After normalizing matrix A, the average of the rows is calculated to obtain the eigenvector (i.e., the weight vector) w = [0.633, 0261, 0.106]; then the basic weights of operational performance P, physiological load B, and subjective experience S are [0.633, 0261, 0.106].
[0109] In the formula , and These are the offsets of various tasks in the task stage, including performance adjustment (P), physiological load (B), and subjective experience (S), and their values are determined by the characteristics of the current task stage. For example, in the "decision-making stage," mental demands are high, and physiological load is high. Set to positive; during the "parameter setting" stage, operational efficiency is crucial. Set to positive;
[0110] In the formula , and It refers to the environmental state offset of operational performance calibration (P), physiological load (B), and subjective experience (S); it is determined based on the operational environment during the current task execution. For example, in an emergency task, reliability is the top priority. Subjective experience is set to negative. Operational performance Physiological load is set to positive;
[0111] In the formula As a consistency penalty item, , It is the score after calibration of operational performance (P), physiological load (B), and subjective experience (S) and normalized to [0,1].
[0112] ;
[0113] This is the amplification factor, which is usually set to a negative number to make this item behave as a penalty.
[0114] S3f: Interface Bottleneck Unit Analysis:
[0115] a) Generate a multimodal data overlay heatmap, overlaying eye-tracking fixation hotspots, periods of high EEG load (theta wave power exceeding the resting baseline), and high error operation areas onto the interface screenshot to visually identify interface elements that cause high cognitive load and operational confusion.
[0116] b) Analyze the physiological load changes corresponding to each operation process in the dynamic task to locate the high-load bottleneck links in the task flow.
[0117] S4. Generate optimization suggestions and closed-loop verification;
[0118] Based on the bottleneck identification results of S3f, specific interface optimization suggestions are output, including but not limited to: layout reorganization, color coding optimization, interaction flow simplification, and information display enhancement. After modifying the prototype interface based on the optimization suggestions, repeat steps S1-S3 to conduct A / B testing and verification until the R-CHCI index increases by more than 10% above the preset threshold.
[0119] As a specific implementation method, layout reorganization generally includes spatial layout optimization, visual encoding, and noise reduction techniques, among which spatial layout optimization includes:
[0120] a) Following the three-zone rule, the interface is divided into a status zone (core parameters, health status), an operation zone (common controls), and an alert zone (faults / alarms) to reduce the visual search range.
[0121] b) Place the most frequently used operation keys in the golden line of sight area (horizontal line of sight ±15°, vertical display range 0°-30°) to reduce head and eye movement.
[0122] c) Reduce menu depth by elevating frequently used functions from second- and third-level menus to the first-level menu or setting them as shortcuts. Use a collapsible / expandable card design to manage information, displaying only core information by default, with details expanded as needed.
[0123] This invention also proposes a radar human-machine interface evaluation system based on multimodal data fusion, used to implement the aforementioned radar human-machine interface evaluation method based on multimodal data fusion, specifically including:
[0124] The multimodal data acquisition module is used to collect operational performance, eye-tracking, electroencephalogram (EEG), and subjective data.
[0125] The signal injection and synchronization control module is used to simulate target echo and electromagnetic interference environment and synchronize multi-source data;
[0126] The data fusion and analysis module is used to achieve spatiotemporal alignment of multimodal data and calculate the R-HCI index;
[0127] The interface optimization suggestion module is used to output optimization solutions.
[0128] In summary, this invention, by simultaneously collecting and deeply integrating operational performance, eye-tracking, electroencephalogram (EEG) and subjective evaluation data, constructs a closed-loop system of design-simulation-testing-optimization, thereby achieving scientific, objective, and accurate evaluation and optimization guidance for the design quality of radar human-machine interfaces.
[0129] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0130] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these modifications and improvements all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A radar human-machine interface evaluation method based on multimodal data fusion, characterized in that, Includes the following steps: S1. Construct a multimodal synchronous evaluation environment, execute multi-stage standardized test protocols, and control the tested operator to perform multi-stage test tasks, including static interface evaluation tasks, dynamic operation tasks, and stress environment task tests. S2. While the tested operator performs a multi-stage test task, multimodal data is collected. The multimodal data includes operational performance data, eye-tracking data, electroencephalogram (EEG) data, and subjective evaluation data. Operational performance data includes task completion time and accuracy rate; eye-tracking data includes theta waves, pupil diameter, fixation point dispersion, and eye-tracking fixation hotspots; EEG data includes periods of high EEG load and areas of high error operation; subjective evaluation data is obtained based on a multi-dimensional subjective load score, which includes interface comprehensibility, operational satisfaction, task confidence, and perceived usability. S3. Perform spatiotemporal alignment and fusion on the collected multimodal data, calculate the operational performance calibration score, physiological load calibration score, and subjective experience calibration score, and calculate the radar interface health index through a scenario-adaptive dynamic weighting algorithm to complete the quantitative assessment; S4. Based on eye-tracking fixation hotspots, periods of high EEG load, and areas of high error operation, overlay rendering is performed on the interface screenshot to generate an interface bottleneck analysis diagram and output interface optimization suggestions. The optimization suggestions include at least one of layout reorganization, color coding optimization, interaction process simplification, and information display enhancement. S5. After modifying the prototype interface based on the optimization suggestions, repeat steps S1 to S4 to conduct A / B testing and verification until the radar interface health index increases beyond the preset threshold.
2. The radar human-machine interface evaluation method based on multimodal data fusion according to claim 1, characterized in that, The multimodal synchronous evaluation environment includes a radar human-machine interface system to be evaluated, a radar signal generator, an environment simulator, an eye tracker, an EEG acquisition device, and a synchronization controller. The synchronization controller is used to receive event marker signals injected by the radar signal generator and uniformly send synchronization trigger signals to the eye tracker and the EEG acquisition device.
3. The radar human-machine interface evaluation method based on multimodal data fusion according to claim 1, characterized in that, The test tasks performed on the operator under test in step S1 include: Static interface assessment task: Present an overview interface of radar health status, record the operator's accuracy in understanding key information and reaction time, including the health status and threat level of each subsystem; Dynamic operational tasks: Simulate typical combat or operational processes, including search, identification, tracking, and decision-making, and record task completion efficiency and error rate; Stress Environment Task: Simulate a complex electromagnetic environment by injecting electromagnetic interference signals to test the interface robustness and operator stress resistance. The electromagnetic interference signals include noise and deceptive interference.
4. The radar human-machine interface evaluation method based on multimodal data fusion according to claim 1, characterized in that, In step S3, the spatiotemporal alignment is based on the synchronous trigger mark, which aligns the operation performance data, eye-tracking data, and EEG data on a unified time axis to form a multimodal data stream with a timestamp error of less than 10ms; the spatiotemporal alignment is implemented using the Precise Time Protocol (PTP).
5. The radar human-machine interface evaluation method based on multimodal data fusion according to claim 1, characterized in that, In step S3, the operation performance calibration score is... ,Depend on Current user operation performance score User historical average performance The performance score of expert-level operators is calculated using the following criteria: ； Among them, each level , , The actual operational performance score is given by the operational performance calculation method P, which is calculated from the task completion time and accuracy rate. ； ； ； in, Here, T_actual is the actual time the user takes to complete the task, and T_benchmark is the benchmark completion time, representing the efficiency weighting coefficient. To ensure accurate performance weighting, A_task represents task-level accuracy, and A_action represents operation-level accuracy.
6. The radar human-machine interface evaluation method based on multimodal data fusion according to claim 1, characterized in that, In step S3, the physiological load calibration score The calculation formula is: ； Where B represents the actual score of physiological load: ； Where ENI is the environmental noise index, α is the adjustment coefficient, S_θ, S_p, and S_d are the standardized scores of eye movement theta wave, pupil diameter, and fixation point dispersion, respectively, and β_θ, β_p, and β_d are the weights of the corresponding indicators.
7. The radar human-machine interface evaluation method based on multimodal data fusion according to claim 1, characterized in that, In step S3, the subjective experience calibration score is... It is a weighted fusion of multi-dimensional subjective scores, and the calculation formula is: ； in and These are the standard deviation and mean of the subjective ratings given by all participants on this interface, respectively. This is the adjustment coefficient, where S is the actual subjective experience score. The formula is: ； R_i is the original score of the i-th dimension, W_i is its weight, C_i is the score consistency coefficient of this dimension, CF is the credibility factor, and δ is the adjustment coefficient.
8. The radar human-machine interface evaluation method based on multimodal data fusion according to claim 1, characterized in that, In step S3, the radar interface health index incorporates a dynamic function that varies with the assessment task stage and environmental conditions to address the weighting issue of indicators under different scenarios. The specific method is as follows: Define task phase vector and environment state vector : The weight function distribution of each indicator is as follows , and for: ； in, , , These are the basic weights for operational performance (P), physiological workload (B), and subjective experience (S). , and These are the various offsets for each task stage, including performance adjustment (P), physiological load (B), and subjective experience (S). , These are the environmental state offsets of operational performance correction (P), physiological load (B), and subjective experience (S); The Radar Interface Health Index (R-CHCI) is: ； in, This is a consistency penalty term used to handle inconsistencies between indicators. ； in, , , These are the normalized operational performance calibration score, physiological load calibration score, and subjective experience calibration score. , It is the magnification factor.
9. The radar human-machine interface evaluation method based on multimodal data fusion according to claim 1, characterized in that, In step S4, the layout reorganization includes: Following the three-zone rule, the interface is divided into a status zone, an operation zone, and an early warning zone. The status zone is used to display core parameters and health status, the operation zone is used to arrange commonly used controls, and the early warning zone is used to display faults and alarms. The most frequently used operation control key is placed in the golden line of sight area, which is defined as the display range of ±15° horizontal line of sight and 0°-30° vertical line of sight. By compressing the menu depth, high-frequency functions in the second and third level menus are promoted to the first level menu or set as shortcut operations, and information is managed using a card-style design that can be collapsed or expanded, with only core information displayed by default.
10. A radar human-machine interface evaluation system based on multimodal data fusion, characterized in that, The radar human-machine interface evaluation method based on multimodal data fusion as described in any one of claims 1-9 specifically includes: The task control module is used to control the operator under test to execute multi-stage test tasks; The multimodal data acquisition module is used to simultaneously collect operational performance, eye movement, electroencephalogram (EEG), and subjective evaluation data. The data fusion and computation module is used for spatiotemporal alignment, calculation of calibration scores, and interface health index. The bottleneck analysis module is used to generate heatmaps of bottleneck areas on the interface and output optimization suggestions. The closed-loop verification module is used to perform A / B testing and evaluate the optimization effect.