Intelligent psychophysiological health detection method and system

Abstract

This application discloses an intelligent method and system for detecting psychological and physiological health, belonging to the field of health detection technology. The method achieves synchronous acquisition of dual-modal signals from the same source using a high frame rate camera, generates accurate evaluation coefficients through a multi-dimensional real-time verification algorithm, and adaptively calibrates them. It employs an attention mechanism dynamic correlation fusion algorithm to strengthen the intrinsic correlation between physiological and psychological indicators, combines an improved Bi-Mamba deep neural network to capture temporal features, and establishes a risk mapping model. Based on a three-dimensional evaluation system, it generates multi-scenario adapted detection reports. The system includes six modules: dual-modal acquisition, multi-indicator extraction, dynamic verification and calibration, deep correlation fusion, temporal risk modeling, and scenario-based reporting. This invention achieves synchronous acquisition, dynamic verification, deep fusion, and timely early warning of physiological and psychological signals, significantly improving detection accuracy and scenario adaptability, and providing precise data support for psychological and physiological health intervention.

Description

Technical Field

[0001] This application relates to the field of health detection technology, and in particular to an intelligent psychophysiological health detection method and system. Background Technology

[0002] With the increasing awareness of public health, integrated monitoring of psychological and physiological health has become a core need in fields such as public health, workplace health, and school health. Non-contact testing technologies, with their advantages of being non-invasive, convenient, and approachable, are gradually replacing traditional contact-based testing and subjective questionnaire assessments, becoming the mainstream approach for large-scale population health screening. Currently, most related technologies are based on principles such as computer vision and biosensing to collect single-dimensional health indicators. Some solutions attempt to improve the comprehensiveness of assessments through multimodal data fusion, but overall, they remain at the basic level of "data collection + simple analysis," and are widely used in scenarios such as health screening and psychological intervention assistance.

[0003] However, existing technologies face numerous insurmountable technical bottlenecks in practical applications. Firstly, data acquisition is asynchronous. Traditional multimodal solutions employ multiple devices to collect physiological and psychological signals separately, or a single device to collect two types of data sequentially, resulting in a time lag between the two types of signals and making it impossible to establish a real-time correlation between physiological fluctuations and changes in psychological state. Secondly, there is a lack of dynamic accuracy verification mechanisms. During detection, environmental interference such as uneven lighting and posture shifts can easily lead to the loss of feature point tracking and image frame rate shifts. However, existing systems do not conduct real-time assessments of data reliability, directly affecting the accuracy of detection results. Thirdly, feature fusion logic is simplistic, often employing data splicing or fixed-weight fusion methods without considering the inherent correlation between physiological and psychological indicators, leading to biased assessment results. Fourthly, risk warning capabilities are weak, only able to qualitatively describe the current health status and unable to achieve early risk quantification and warning of mental illness through temporal feature capture. Fifthly, scenario adaptability is insufficient; algorithm models and warning thresholds are fixed, making it difficult to meet personalized needs such as depression screening in schools and stress monitoring in enterprises.

[0004] Therefore, developing an intelligent psychological and physiological health detection method and system that can achieve simultaneous acquisition of physiological and psychological signals, dynamic verification of the detection process, deep feature correlation and fusion, clinical-level risk warning, and multi-scenario adaptation has become a key technical problem that urgently needs to be solved in this field. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this paper provides an intelligent method and system for detecting psychological and physiological health. This system achieves synchronous, non-intrusive acquisition of physiological and psychological signals, ensures the accuracy of the detection data through a dynamic verification mechanism, and utilizes an improved algorithm to achieve deep feature fusion and accurate early warning of mental illness risks. Simultaneously, it constructs a three-dimensional assessment system to complete comprehensive health detection and report output, solving the technical problems of signal asynchrony, lack of data verification, insufficient fusion, inaccurate early warning, and incomplete assessment in existing technologies. The technical solution is as follows: On the one hand, an intelligent method for detecting psychological and physiological health is provided, the method comprising: A high frame rate camera is used to capture facial video streams of the subject, and facial spectral reflectance signals and head and neck muscle micro-movement signals are extracted simultaneously. Hardware clock synchronization technology is used to align the timestamps of the two signals to ensure the consistency of the acquisition timing. Cortical blood flow reconstruction was performed on the spectral reflectance signal to restore changes in cortical blood flow, which was used to extract 7 physiological indicators. Feature point tracking and trajectory fitting are performed on muscle micromotor signals, and psychological health indicators are extracted by combining a psychological feature model library; The system monitors three core parameters in real time during the detection process: the number of feature point tracking loss, the degree of image frame rate shift, and the pixel-coordinate conversion efficiency. It calculates the detection accuracy evaluation coefficient through a normalization algorithm, forming a dynamic closed loop of "acquisition-evaluation-calibration". When the detection accuracy evaluation coefficient is lower than the preset judgment value, an adaptive calibration strategy is triggered: if the number of feature point tracking losses is abnormal, a voice prompt is issued to correct the subject's posture; if the image frame rate offset exceeds the standard, the camera exposure parameters are automatically adjusted; if the pixel-coordinate conversion efficiency is insufficient, the image acquisition area is optimized. An improved cross-modal feature fusion algorithm is adopted, which dynamically assigns weights based on the correlation strength between physiological and psychological indicators, strengthens the intrinsic correlation between cortical blood flow changes and muscle micromotor features, and generates a multi-dimensional fusion feature vector. The fused feature vectors are input into the improved Bi-Mamba deep neural network to simultaneously capture the long-term dependencies of 128 dynamic features, establish a real-time mapping model between physiological index fluctuations and mental illness risk, and output a five-level risk level including normal, borderline, and I-III level warnings. The test report is automatically generated based on a three-dimensional assessment system, which includes 7 physiological indicators, 14 mental health indicators and 27 psychological quality indicators. The report simultaneously outputs the risk level and corresponding intervention recommendations.

[0006] Optionally, the high frame rate camera has an acquisition frame rate of ≥120fps and a timestamp alignment accuracy controlled within 1ms to ensure the synchronization of the acquisition of spectral reflection signals and muscle micro-movement signals.

[0007] Optionally, the calculation logic of the normalization algorithm is as follows: map the three core parameters to the [0,1] interval respectively, calibrate the weight coefficients of each parameter through experimental data, and obtain the detection accuracy evaluation coefficient after weighted summation.

[0008] Optionally, the dynamic weight allocation rule is as follows: the correlation strength is determined by calculating the Pearson correlation coefficient between physiological indicators and psychological indicators. The higher the correlation strength, the greater the fusion weight is assigned, with the correlation weight between physiological indicators and mental illness indicators having the highest priority.

[0009] Optionally, the improved Bi-Mamba deep neural network introduces a temporal alignment unit to enhance the correlation of dynamic features within 30 seconds, thereby improving the real-time performance and accuracy of mental illness risk mapping.

[0010] On the other hand, an intelligent psychophysiological health detection system is provided for use in the above-described method, the system comprising: The dual-modal synchronous acquisition module includes a high frame rate camera and a hardware clock synchronization unit, which is used to synchronously acquire facial spectral reflectance signals and head and neck muscle micro-movement signals. The multi-dimensional indicator extraction module, connected to the dual-modal synchronous acquisition module, is used to extract 7 physiological indicators, 14 mental health indicators, and 27 psychological quality indicators. The dynamic calibration evaluation module, connected to the multi-dimensional index extraction module, is used to calculate the detection accuracy evaluation coefficient and trigger the adaptive calibration strategy. An improved feature fusion module, connected to the dynamic calibration and evaluation module, is used to assign dynamic weights based on the correlation strength of indicators and generate a fused feature vector. The temporal correlation modeling module is connected to the improved feature fusion module, which includes an improved Bi-Mamba deep neural network, used to establish a physiological-psychological risk mapping model and output the risk level; The 3D assessment report module, connected to the temporal correlation modeling module, is used to generate a test report containing intervention recommendations based on the 3D assessment system.

[0011] Optionally, the multi-dimensional indicator extraction module has a built-in physiological feature model library and a psychological feature model library. The physiological feature model library includes at least reconstruction algorithms for heart rate and blood oxygen indicators, and the psychological feature model library includes at least parsing algorithms for aggression and depression indicators.

[0012] Optionally, the dynamic calibration evaluation module has a built-in voice prompt unit and a parameter adjustment unit. The voice prompt unit is used to issue attitude correction prompts, and the parameter adjustment unit is used to optimize the camera exposure parameters.

[0013] Optionally, the improved feature fusion module supports real-time calculation of the correlation strength of indicators and connects with the cloud model library through an API interface to realize dynamic updates of weight allocation rules.

[0014] Optionally, the three-dimensional assessment report module supports batch export of reports and can use different color blocks to mark the report status of physiological abnormalities, psychological abnormalities, and both physiological and psychological abnormalities, for the purpose of quickly screening key targets.

[0015] Therefore, compared with the prior art, this application achieves at least the following technical effects.

[0016] This invention discloses an intelligent method and system for detecting psychological and physiological health, belonging to the field of health detection technology. It addresses the problems of signal asynchrony, lack of dynamic verification, simple fusion logic, and weak risk warning in existing technologies. The method achieves synchronous acquisition of co-source dual-modal signals using a high frame rate camera, generates accurate evaluation coefficients through a multi-dimensional real-time verification algorithm, and adaptively calibrates them. It employs an attention mechanism dynamic correlation fusion algorithm to strengthen the intrinsic correlation between physiological and psychological indicators, and combines an improved Bi-Mamba deep neural network to capture temporal features and establish a risk mapping model. Based on a three-dimensional evaluation system, it generates multi-scenario adapted detection reports. The system comprises six modules: co-source dual-modal acquisition, multi-indicator extraction, dynamic verification and calibration, deep correlation fusion, temporal risk modeling, and scenario-based reporting. This invention achieves synchronized acquisition, dynamic verification, deep fusion, and temporal warning of physiological and psychological signals, significantly improving detection accuracy and scenario adaptability, and providing precise data support for psychological and physiological health intervention. Attached Figure Description

[0017] Figure 1 The diagram illustrates a real-time interface for intelligent psychological and physiological health detection (frontal face detection prompt). Figure 2 The image illustrates a real-time interface for intelligent psychological and physiological health monitoring (with avatar size prompts). Figure 3 The diagram illustrates a real-time interface for intelligent psychological and physiological health monitoring (with face placement prompts). Figure 4 The diagram illustrates a real-time interface for intelligent psychological and physiological health monitoring (with lighting adjustment prompts). Figure 5 This diagram illustrates the results of a smart psychophysiological health test. Figure 6 This diagram illustrates the results of a mental health indicator test for intelligent psychophysiological health. Figure 7 This diagram illustrates the results of a smart psychophysiological health detection and mental illness risk assessment. Figure 8This diagram illustrates the results of an intelligent psychophysiological health assessment of occupational psychological indicators. Figure 9 This diagram illustrates an occupational health test result. Figure 10 This diagram illustrates the analysis results of an intelligent psychological and physiological health assessment. Figure 11 This diagram illustrates a detailed analysis of occupational psychological indicators in an intelligent psychophysiological health testing system. Figure 12 This diagram illustrates a detailed analysis of the psychological and physiological health assessment capabilities of an intelligent system. Figure 13 This diagram illustrates an interface for querying and batch exporting intelligent psychological and physiological health test reports. Figure 14 This diagram illustrates a list of files for an intelligent psychological and physiological health testing report. Figure 15 This diagram illustrates the management interface of an intelligent psychological and physiological health monitoring system. Figure 16 The diagram illustrates the data center interface of an intelligent psychological and physiological health monitoring system. Detailed Implementation

[0018] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0019] In this article, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0020] The purpose of this invention is to overcome the shortcomings of existing technologies and provide an intelligent psychological and physiological health detection method and system. This system enables synchronous and non-intrusive acquisition of physiological and psychological signals, ensures the accuracy of detection data through a dynamic verification mechanism, achieves deep feature fusion and accurate early warning of mental illness risk based on an improved algorithm, and constructs a three-dimensional assessment system to complete comprehensive health detection and report output. This solves the technical problems of signal asynchrony, lack of data verification, insufficient fusion, inaccurate early warning, and incomplete assessment in existing technologies.

[0021] To achieve the above objectives, this invention adopts the following technical solution: an intelligent psychological and physiological health detection method, comprising the steps of synchronous acquisition of homologous dual-modal signals, multi-dimensional indicator extraction, dynamic verification and adaptive calibration of the detection process, improved cross-modal feature fusion, deep neural network risk modeling, and generation of a three-dimensional assessment system report, which can achieve non-invasive, accurate, and comprehensive detection of psychological and physiological health; and correspondingly provides an intelligent psychological and physiological health detection system for implementing this method, the system consisting of functionally modular hardware and software units, with each module working together to complete the entire detection process, ensuring the effective implementation of the method.

[0022] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Example 1 This embodiment discloses an intelligent method for detecting psychological and physiological health, the method comprising the following steps: Step 1: Use a high frame rate camera to capture the facial video stream of the subject, and simultaneously extract facial spectral reflectance signals and head and neck muscle micro-movement signals from the captured video stream.

[0024] Hardware clock synchronization technology is used to align the timestamps of the two types of signals, thereby ensuring the consistency of the acquisition timing of the two types of signals and laying a timing foundation for subsequent correlation analysis of physiological and psychological indicators.

[0025] Step 2: Perform cortical blood flow reconstruction processing on the spectral reflectance signal extracted in Step 1). Reconstruct the pattern of facial cortical blood flow changes by analyzing the signal. Extract 7 physiological indicators based on the reconstructed cortical blood flow data to achieve quantitative characterization of physiological state. Step 3: Perform feature point tracking and trajectory fitting processing on the muscle micro-motion signals extracted in Step 1) to capture the micro-motion patterns of the head and neck muscles. At the same time, combine the preset psychological feature model library, and extract the corresponding mental health indicators based on the matching analysis between micro-motion features and psychological models to achieve quantitative representation of mental state. Step 4: During the above indicator extraction process, three core parameters in the entire detection process are monitored in real time: the number of feature point tracking loss, the degree of image frame rate offset, and the pixel-coordinate conversion efficiency. The three core parameters are processed by a normalization algorithm to calculate the detection accuracy evaluation coefficient. Based on the real-time feedback of this coefficient, a dynamic closed loop of "acquisition-evaluation-calibration" is formed to realize real-time monitoring and dynamic adjustment of the detection process.

[0026] Step 5: When the detection accuracy evaluation coefficient calculated in step 4) is lower than the preset judgment value, the adaptive calibration strategy is immediately triggered, and corresponding calibration measures are taken for different parameter abnormalities: If the number of feature point tracking loss is abnormal, the system automatically issues a voice prompt to guide the subject to correct facial posture; if the image frame rate offset exceeds the standard, the system sends a control command to the camera to automatically adjust the camera's exposure parameters; if the pixel-coordinate conversion efficiency is insufficient, the system automatically optimizes the acquisition area of ​​the facial image until the detection accuracy evaluation coefficient reaches the preset judgment value, thus restoring the validity of the detection data.

[0027] Step 6: After the data to be detected passes the accuracy verification, the extracted physiological and psychological health indicators are fused using an improved cross-modal feature fusion algorithm. The algorithm assigns dynamic weights to each indicator based on the correlation strength between the physiological and psychological indicators. By assigning weights, the intrinsic correlation between cortical blood flow changes and muscle micromotor features is strengthened, and finally a multi-dimensional fusion feature vector is generated to achieve deep fusion of physiological and psychological features.

[0028] Step 7: Input the multi-dimensional fusion feature vector generated in Step 6) into the improved Bi-Mamba deep neural network. The network synchronously captures the long-term temporal dependencies of 128 dynamic features. Based on the temporal correlation analysis of the features, a real-time mapping model between physiological index fluctuations and mental illness risk is established. The model outputs a five-level risk level, which includes normal, critical, Level I warning, Level II warning, and Level III warning, to achieve quantitative early warning of mental illness risk.

[0029] Step 8: Based on the preset three-dimensional assessment system, an automatic test report is generated. The three-dimensional assessment system integrates full-dimensional test data of 7 physiological indicators, 14 mental health indicators and 27 psychological quality indicators. The test report simultaneously outputs the risk level of the test subject and the corresponding health intervention suggestions, providing comprehensive and accurate data support for subsequent health interventions.

[0030] This embodiment achieves temporal unification of physiological and psychological signals through synchronous acquisition of homologous dual-modal signals, solving the core problem of signal asynchrony in existing technologies and laying the foundation for indicator correlation analysis. By constructing a dynamic closed loop of "acquisition-evaluation-calibration," real-time accuracy verification of the detection process is achieved, effectively avoiding interference from environmental factors such as lighting and posture, ensuring the validity and accuracy of the detection data. Relying on an improved cross-modal feature fusion algorithm, deep fusion of physiological and psychological features is achieved, breaking through the technical limitations of traditional simple splicing and fusion, and strengthening the intrinsic correlation between indicators. A risk mapping model is established through an improved Bi-Mamba deep neural network, enabling the capture of long-term dependencies of dynamic features and quantitative early warning of mental illness risks, improving the early warning capability of the detection. Finally, a three-dimensional evaluation system is used to generate a full-dimensional detection report, realizing a comprehensive assessment of mental and physiological health, providing comprehensive data support for health intervention, and overall improving the accuracy, real-time performance, and comprehensiveness of non-contact mental and physiological health detection.

[0031] Example 2 Based on the method of Embodiment 1, this embodiment further describes the high frame rate camera parameters in step 1).

[0032] The high frame rate camera has a capture frame rate of ≥120fps, and the timestamp alignment accuracy is controlled within 1ms through hardware clock synchronization technology. This parameter setting can ensure the synchronization of the acquisition of spectral reflection signals and muscle micro-movement signals, avoid signal feature loss due to insufficient frame rate, and at the same time, the microsecond-level timestamp alignment accuracy can eliminate the timing deviation of the two types of signals to the greatest extent, further ensuring the accuracy of subsequent physiological and psychological index correlation analysis.

[0033] This embodiment ensures the synchronization and integrity of dual-modal signals from the hardware acquisition level by limiting the frame rate and timestamp alignment accuracy of the high frame rate camera. The sufficiently high acquisition frame rate can clearly capture subtle changes in facial spectral reflectance and micro-movement features of head and neck muscles, avoiding the loss of feature information. The timestamp alignment accuracy within 1ms fundamentally eliminates the timing deviation of dual-modal signals, ensuring that fluctuations in physiological indicators and changes in psychological state can be correlated in real time, providing high-quality raw data for subsequent feature fusion and risk modeling, and improving the overall detection accuracy from the source of detection.

[0034] Example 3 Based on the method in Example 1, this example explicitly defines the calculation logic of the normalization algorithm. The calculation logic of the normalization algorithm is as follows: First, the three core parameters—the number of feature point tracking losses, the degree of image frame rate offset, and the pixel-coordinate conversion efficiency—are normalized and mapped to the [0,1] interval to eliminate the dimensional differences between different parameters. Then, the weight coefficients corresponding to each parameter are calibrated using experimental data. Based on the calibrated weight coefficients, the three normalized parameters are weighted and summed to finally obtain the detection accuracy evaluation coefficient.

[0035] This embodiment clarifies the calculation logic of the normalization algorithm, achieving standardized processing of core parameters with different dimensions, eliminating dimensional interference between parameters, and enabling effective fusion calculation of various parameters. Simultaneously, by calibrating weight coefficients using experimental data, the proportion of each parameter in the accuracy evaluation coefficient more closely matches the needs of actual detection scenarios, ensuring that the evaluation coefficient can truly and accurately reflect the accuracy of the detection process. This provides a scientific and reliable quantitative basis for the effective operation of the "acquisition-evaluation-calibration" dynamic closed loop, guaranteeing the rationality and effectiveness of the dynamic verification mechanism.

[0036] Example 4 According to the method in Example 1, this example defines the dynamic weight allocation rule of the improved cross-modal feature fusion algorithm: the dynamic weight allocation rule is to calculate the Pearson correlation coefficient between physiological indicators and psychological indicators, and determine the correlation strength between the two types of indicators based on the value of the Pearson correlation coefficient. The higher the correlation strength, the greater the fusion weight assigned to the corresponding indicator. At the same time, it is clarified that the correlation weight between physiological indicators and mental illness indicators has the highest priority.

[0037] This embodiment clarifies the dynamic weight allocation rules based on the Pearson correlation coefficient, providing a scientific quantitative basis for the weight allocation of cross-modal feature fusion. This allows the fusion process to accurately reflect the intrinsic correlation characteristics between physiological and psychological indicators, strengthening the contribution of strongly correlated features in the fusion result and weakening the interference of weakly correlated features. Simultaneously, by setting the priority of correlation weights between physiological and mental illness indicators, the fused feature vector can more accurately reflect the correlation between mental illness and physiological state, providing more targeted fused feature data for subsequent mental illness risk modeling and improving the modeling accuracy and early warning accuracy of the risk mapping model.

[0038] Example 5 According to the method of Example 1, this example defines the improvement points of the improved Bi-Mamba deep neural network: the improved Bi-Mamba deep neural network introduces a temporal alignment unit, which is used to enhance the correlation of dynamic features within 30 seconds. By focusing on capturing and analyzing the correlation of dynamic features within a short time window, the real-time performance and accuracy of the mental illness risk mapping model are improved.

[0039] This embodiment introduces a temporal alignment unit into the improved Bi-Mamba deep neural network, specifically enhancing the correlation of dynamic features within 30 seconds. This allows the network to more accurately capture the dynamic changes in physiological indicators and psychological characteristics within a short time window, aligning with the real-time changes in psychological and physiological states. Simultaneously, this improved design further enhances the network's ability to capture long-term dependencies of dynamic features, enabling the established risk mapping model to more quickly and accurately reflect the correlation between fluctuations in physiological indicators and the risk of mental illness. This effectively improves the real-time performance and early warning accuracy of the risk model, achieving a technological upgrade from static state description to dynamic risk warning.

[0040] Example 6 This embodiment discloses an intelligent psychological and physiological health detection system. The system is used to implement the intelligent psychological and physiological health detection method of any one of embodiments 1-5. The system includes the following functional modules.

[0041] The dual-modal synchronous acquisition module includes a high frame rate camera and a hardware clock synchronization unit. The high frame rate camera, as the core acquisition hardware, is used to acquire the facial video stream of the subject and simultaneously extract the facial spectral reflectance signal and the micro-movement signal of the head and neck muscles. The hardware clock synchronization unit works with the high frame rate camera to perform timestamp alignment processing on the two extracted signals to ensure the consistency of the acquisition timing of the dual-modal signals. The multi-dimensional indicator extraction module establishes a data connection with the dual-modal synchronous acquisition module, receives spectral reflectance signals and muscle micro-movement signals transmitted by the dual-modal synchronous acquisition module, and performs cortical blood flow reconstruction processing, feature point tracking and trajectory fitting processing based on the received signals. Combined with the psychological feature model library, it completes the extraction of 7 physiological indicators, 14 mental health indicators and 27 psychological quality indicators, and realizes the quantitative output of multi-dimensional indicators. The dynamic calibration and evaluation module establishes a data connection with the multi-dimensional index extraction module. It receives three core parameters in real time during the detection process: the number of feature point tracking loss, the degree of image frame rate offset, and the pixel-coordinate conversion efficiency. It calculates the detection accuracy evaluation coefficient through a normalization algorithm. Based on the value of the evaluation coefficient, it determines whether to trigger the adaptive calibration strategy and sends calibration instructions to the corresponding module to complete the dynamic verification and calibration of the detection process. An improved feature fusion module establishes a data connection with the dynamic calibration and evaluation module, receives physiological and psychological health indicators that have passed accuracy verification, and incorporates an improved cross-modal feature fusion algorithm. Based on the correlation strength between physiological and psychological indicators, dynamic weights are assigned to complete deep feature fusion and generate a multi-dimensional fusion feature vector. The temporal correlation modeling module establishes a data connection with the improved feature fusion module, receives multi-dimensional fused feature vectors, and has an improved Bi-Mamba deep neural network built in. The neural network captures the long-term temporal dependencies of dynamic features, establishes a real-time mapping model between physiological index fluctuations and mental illness risk, and outputs a five-level risk level through model calculation. The 3D assessment report module establishes a data connection with the time-series correlation modeling module and integrates the full-dimensional indicator data output by the multi-dimensional indicator extraction module. Based on the preset 3D assessment system, it integrates the risk level and full-dimensional indicator data to automatically generate a test report. The report synchronously outputs the risk level and corresponding intervention recommendations.

[0042] This embodiment, through a modular system design, enables the practical implementation of the detection methods described in Embodiments 1-5. Each module has a clear functional division and smooth data connection, allowing for the completion of the entire process from dual-modal signal acquisition to report generation. The modular design provides excellent scalability and maintainability, allowing for parameter adjustments or functional optimization of individual modules based on actual testing needs without requiring a complete system reconstruction. Simultaneously, the collaborative operation of each module, implementing hardware acquisition, signal processing, algorithm computation, and report generation in a layered manner, ensures efficient and accurate operation throughout the entire testing process. It stably achieves the technical effects of simultaneous acquisition, dynamic verification, deep fusion, risk warning, and comprehensive report generation of physiological and psychological signals, adapting to large-scale, multi-scenario non-contact psychological and physiological health testing needs.

[0043] Furthermore, the multi-dimensional indicator extraction module incorporates a physiological feature model library and a psychological feature model library. The physiological feature model library includes at least reconstruction algorithms for heart rate and blood oxygenation indicators. These algorithms are core algorithms for cortical blood flow reconstruction processing and are used to accurately extract core physiological indicators such as heart rate and blood oxygenation from spectral reflectance signals. The psychological feature model library includes at least analytical algorithms for aggression and depression indicators. These algorithms are core algorithms for analyzing muscle micromotor signals and are used to accurately extract core mental health indicators such as aggression and depression from muscle micromotor features.

[0044] Therefore, by limiting the internal model library and core algorithms of the multi-dimensional indicator extraction module, the accuracy of physiological and mental health indicator extraction is guaranteed at the algorithm level. The built-in dedicated model library and parsing algorithms can specifically meet the extraction needs of core indicators, effectively improving the efficiency and accuracy of indicator extraction. At the same time, the clear algorithm and model library settings make the module's function implementation more targeted, avoiding the extraction bias caused by general algorithms, providing high-quality indicator data for subsequent feature fusion and risk modeling, and further improving the overall detection accuracy of the system.

[0045] Furthermore, the dynamic calibration and evaluation module incorporates a voice prompt unit and a parameter adjustment unit. The voice prompt unit automatically issues a voice prompt when the number of feature point tracking losses is abnormal, guiding the subject to correct their facial posture. The parameter adjustment unit sends control commands to the camera when the image frame rate offset exceeds the standard, automatically optimizing the camera's exposure parameters and achieving adaptive adjustment of the acquisition parameters.

[0046] Therefore, by incorporating a voice prompt unit and a parameter adjustment unit into the dynamic calibration and evaluation module, the adaptive calibration strategy can be implemented quickly and efficiently. It can guide the subject's posture and adjust the collected parameters without human intervention, thereby improving the system's automation level and detection efficiency. At the same time, the dedicated functional unit makes the calibration operation more targeted, which can quickly resolve parameter anomalies during the detection process, effectively shorten the calibration time, ensure the smooth progress of the detection process, and further enhance the operational efficiency of the "acquisition-evaluation-calibration" dynamic closed loop.

[0047] Furthermore, the improved feature fusion module supports real-time calculation of the correlation strength of indicators. It can dynamically update the correlation strength value and adjust the fusion weight according to the real-time changes of physiological and psychological indicators. At the same time, the module reserves an API interface to connect with the cloud model library and realize the dynamic update of the weight allocation rules. This allows the fusion algorithm to keep up with the update needs of the cloud model library and continuously optimize the fusion effect.

[0048] Therefore, by enabling the improved feature fusion module to perform real-time calculation of the correlation strength of indicators and connect with the cloud model library, the cross-modal feature fusion process becomes dynamic and optimizable. It can adjust the fusion weights according to the real-time changes of indicators, making the fusion results more consistent with the real-time psychological and physiological state of the test subject. At the same time, the connection with the cloud model library allows the weight allocation rules to be updated remotely and dynamically. The optimization of the fusion algorithm can be completed without local algorithm reconstruction, which improves the system's iteration efficiency and scenario adaptability, enabling the system to continuously adapt to industry technology development and actual testing needs.

[0049] Furthermore, the 3D assessment report module supports batch export of test reports, enabling it to process test data from multiple test subjects simultaneously and generate and export test reports in batches, adapting to the needs of large-scale population screening. At the same time, the module has a report status annotation function, which can use different color blocks to annotate report statuses such as physiological abnormalities, psychological abnormalities, and both physiological and psychological abnormalities. Staff can quickly identify abnormal test reports through color blocks, enabling rapid screening of key individuals.

[0050] Therefore, by adding batch export and color block annotation functions to the 3D assessment report module, the actual application efficiency of the system has been greatly improved. The batch export function is suitable for the testing needs of large-scale groups such as campuses and enterprises, and can quickly generate and export multiple reports, reducing manual operation costs. The color block annotation function enables the visual and rapid identification of abnormal reports. Staff can quickly filter out key concerns without having to review the report content one by one, improving the response speed of health interventions and allowing test results to be transformed into intervention actions more quickly, further enhancing the practical application value of the system.

[0051] For the system provided in this application, there are corresponding system software interface diagrams for supplementary explanation, which are described below in conjunction with the appendix. Figures 1 to 16 Provide a software execution example.

[0052] Figure 1 The system's real-time detection interface was showcased. The left side of the interface displayed the status of physiological health indicator detection, while the right side showed real-time monitoring data of mental health indicators. The bottom prompted "Please face the camera directly." It is evident that the system guides the test subject to maintain the correct testing posture through text prompts in the early stages of detection, laying the foundation for accurate acquisition of subsequent dual-modal signals.

[0053] Figure 2 The system displays a real-time detection interface with a bottom prompt stating, "The image is too large or too small. Please ensure that the image size is basically consistent with the frame size." The interface shows that both physiological and psychological health indicators are in the detection process, indicating that the system monitors the image acquisition specifications in real time during the detection process and promptly issues adjustment prompts when abnormal image size occurs, ensuring the effectiveness of feature point tracking.

[0054] Figure 3 The system's real-time detection interface is shown, with a prompt at the bottom: "Please place your face in the frame, ensuring it matches the frame size, and keep your face facing the camera throughout the process." The interface also displays a respiratory curve. It is evident that the system ensures the subject's cooperation in completing the test and improves the accuracy of data collection through clear text guidance and visualization of physiological signals.

[0055] Figure 4The system displays a real-time detection interface with a bottom prompt that reads, "Uneven facial lighting, please adjust the light or angle." The interface shows partial detection data for physiological indicators (such as heart rate variability) and mental health indicators (such as aggression and stress). It is evident that the system monitors the lighting environment in real time during the detection process and promptly prompts adjustments when lighting interference occurs, reducing the impact of environmental factors on the detection results.

[0056] Figure 5 The table displays the results of physiological health tests, including seven physiological indicators: heart rate, blood oxygen, cardiovascular risk, heart rate variability, respiration, systolic blood pressure, and diastolic blood pressure. The minimum, average, maximum, and reference range of each indicator are clearly marked. The results indicate the abnormal conclusion of "bradycardia". It can be seen that the system generates quantitative results after extracting physiological indicators, which provides physiological data support for subsequent cross-modal fusion.

[0057] Figure 6 The system presents a table of mental health test results, covering 14 mental health indicators such as aggression, stress, and anxiety. It marks the minimum, average, and maximum values ​​of each indicator, and includes textual analysis of emotional state and sources of stress below. It can be seen that the system extracts psychological indicators based on muscle micro-movement signals and generates interpretation conclusions, realizing the quantitative and qualitative analysis of mental state.

[0058] Figure 7 The system displays the results of three risk assessments: mental and emotional health, anxiety, and depression. These assessments include risk coefficients (percentages) and risk levels (normal, borderline, Level II warning, etc.). For warning states, the system provides intervention suggestions such as "consulting a professional mental health professional." It is evident that the system outputs five levels of risk through a risk mapping model, achieving a leap from state description to risk warning.

[0059] Figure 8 The system presents the results of occupational psychological testing, including scores, levels, and assessment conclusions for 10 indicators such as self-confidence, self-control, and motivation. The bottom summarizes the total score and level, demonstrating that the system extracts occupation-related psychological quality indicators based on multi-dimensional psychological characteristics, providing data support for occupational health assessment.

[0060] Figure 9 The system displays occupational health test results, including scores, levels, and assessment conclusions for three indicators: physical fitness, vitality, and balance. It clearly states conclusions such as "being able to effectively manage and regulate physical and mental health, and paying attention to balancing work and life." This demonstrates that the system completes the detection and analysis of occupational health dimensions, enriching the occupational suitability of the test report.

[0061] Figure 10The system presents radar charts and tables for analyzing competencies, covering the scores and levels of 17 competency indicators, including communication skills, teamwork skills, and innovation skills. The tables provide detailed explanations of the evaluation conclusions for each indicator. It is evident that the system extracts competencies based on multimodal fusion data, achieving a comprehensive assessment of overall competencies.

[0062] Figure 11 The system provides a detailed analysis of the occupational psychological test results, including the scores, levels, assessment conclusions, and targeted suggestions for each indicator. For example, it states that the system can "manage and control emotions well and possess appropriate calmness and rationality." This demonstrates that the system's interpretation of psychological indicators is targeted, enhancing the practical value of the report.

[0063] Figure 12 The system presents detailed conclusions from the competency analysis, including indicators such as responsiveness and self-management ability. It uses scores to explain conclusions such as "able to basically understand and cope with change" and "able to effectively arrange and manage one's time and resources." This demonstrates that the system uses fused feature vectors to perform in-depth analysis of competencies, corresponding to the technical aspect of the "time-series correlation modeling module outputting multi-dimensional evaluation conclusions," providing users with a clear understanding of their capabilities.

[0064] Figure 13 The personal report query interface of the AI-powered intelligent dynamic emotion early warning platform is shown. It includes information such as file ID, name, affiliated unit, detection date, and report status. It supports the functions of "exporting the list" and "batch downloading". The interface marks the report generation status with different statuses, which shows that the system has the ability to manage reports and process them in batches, making it suitable for large-scale population detection scenarios.

[0065] Figure 14 The system presents a list of test reports, displaying multiple test report files named "Mr. Zhang," including file size and generation time. This shows that the system creates multiple test files for each test subject, corresponding to the "one file per person" setting in "File Management," enabling the traceability and historical comparative analysis of test data.

[0066] Figure 15 The system management interface is shown, which includes functional modules such as unit management, file management, activity management, report query, user management, and equipment management. It can be seen that the system achieves full-process management through modular design, corresponding to the system architecture of "six major modules working together, including dual-modal synchronous acquisition module and multi-dimensional indicator extraction module", which ensures the orderly execution of the entire testing process.

[0067] Figure 16The system's data center function is presented, supporting data queries by region, date, and unit, and displaying statistical information such as the number of measured units, the total number of measured persons, and the percentage of abnormalities. It can be seen that the system has the ability to summarize and analyze data, which corresponds to the functional requirements of "dynamic calibration and evaluation module to calculate the detection accuracy evaluation coefficient" and "data statistical analysis", providing macro data support for collective health monitoring.

[0068] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware, or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk. The above descriptions are merely optional embodiments of this application and are not intended to limit the application. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. An intelligent method for detecting psychological and physiological health, characterized in that, The method includes: A high frame rate camera is used to capture facial video streams of the subject, and facial spectral reflectance signals and head and neck muscle micro-movement signals are extracted simultaneously. Hardware clock synchronization technology is used to align the timestamps of the two signals to ensure the consistency of the acquisition timing. Cortical blood flow reconstruction was performed on the spectral reflectance signal to restore changes in cortical blood flow, which was used to extract 7 physiological indicators. Feature point tracking and trajectory fitting are performed on muscle micromotor signals, and psychological health indicators are extracted by combining a psychological feature model library; The system monitors three core parameters in real time during the detection process: the number of feature point tracking loss, the degree of image frame rate shift, and the pixel-coordinate transformation efficiency. It calculates the detection accuracy evaluation coefficient through a normalization algorithm, forming a dynamic closed loop of "acquisition-evaluation-calibration". When the detection accuracy evaluation coefficient is lower than the preset judgment value, an adaptive calibration strategy is triggered: if the number of feature point tracking losses is abnormal, a voice prompt is issued to correct the subject's posture; if the image frame rate offset exceeds the standard, the camera exposure parameters are automatically adjusted; if the pixel-coordinate conversion efficiency is insufficient, the image acquisition area is optimized. An improved cross-modal feature fusion algorithm is adopted, which dynamically assigns weights based on the correlation strength between physiological and psychological indicators, strengthens the intrinsic correlation between cortical blood flow changes and muscle micromotor features, and generates a multi-dimensional fusion feature vector. The fused feature vectors are input into the improved Bi-Mamba deep neural network to simultaneously capture the long-term dependencies of 128 dynamic features, establish a real-time mapping model between physiological index fluctuations and mental illness risk, and output a five-level risk level including normal, borderline, and I-III level warnings. The test report is automatically generated based on a three-dimensional assessment system, which includes 7 physiological indicators, 14 mental health indicators and 27 psychological quality indicators. The report simultaneously outputs the risk level and corresponding intervention recommendations.

2. The method according to claim 1, characterized in that, The high frame rate camera has a capture frame rate of ≥120fps and a timestamp alignment accuracy controlled within 1ms to ensure the synchronization of the acquisition of spectral reflection signals and muscle micro-movement signals.

3. The method according to claim 1, characterized in that, The calculation logic of the normalization algorithm is as follows: map the three core parameters to the [0,1] interval respectively, calibrate the weight coefficients of each parameter through experimental data, and obtain the detection accuracy evaluation coefficient after weighted summation.

4. The method according to claim 1, characterized in that, The dynamic weight allocation rule is as follows: the correlation strength is determined by calculating the Pearson correlation coefficient between physiological indicators and psychological indicators. The higher the correlation strength, the greater the fusion weight is assigned. Among them, the correlation weight between physiological indicators and mental illness indicators has the highest priority.

5. The method according to claim 1, characterized in that, The improved Bi-Mamba deep neural network introduces a temporal alignment unit to enhance the correlation of dynamic features within 30 seconds, thereby improving the real-time performance and accuracy of mental illness risk mapping.

6. An intelligent psychological and physiological health detection system, characterized in that, The system for implementing the method according to any one of claims 1-5, the system comprising: The dual-modal synchronous acquisition module includes a high frame rate camera and a hardware clock synchronization unit, which is used to synchronously acquire facial spectral reflectance signals and head and neck muscle micro-movement signals. The multi-dimensional indicator extraction module, connected to the dual-modal synchronous acquisition module, is used to extract 7 physiological indicators, 14 mental health indicators, and 27 psychological quality indicators. The dynamic calibration evaluation module, connected to the multi-dimensional index extraction module, is used to calculate the detection accuracy evaluation coefficient and trigger the adaptive calibration strategy. An improved feature fusion module, connected to the dynamic calibration and evaluation module, is used to assign dynamic weights based on the correlation strength of indicators and generate a fused feature vector. The temporal correlation modeling module is connected to the improved feature fusion module, which includes an improved Bi-Mamba deep neural network, used to establish a physiological-psychological risk mapping model and output the risk level; The 3D assessment report module, connected to the temporal correlation modeling module, is used to generate a test report containing intervention recommendations based on the 3D assessment system.

7. The system according to claim 6, characterized in that, The multi-dimensional indicator extraction module has a built-in physiological feature model library and a psychological feature model library. The physiological feature model library contains at least reconstruction algorithms for heart rate and blood oxygen indicators, and the psychological feature model library contains at least parsing algorithms for aggression and depression indicators.

8. The system according to claim 6, characterized in that, The dynamic calibration and evaluation module has a built-in voice prompt unit and a parameter adjustment unit. The voice prompt unit is used to issue attitude correction prompts, and the parameter adjustment unit is used to optimize the camera exposure parameters.

9. The system according to claim 6, characterized in that, The improved feature fusion module supports real-time calculation of the correlation strength of indicators and connects with the cloud model library through an API interface to realize dynamic updates of weight allocation rules.

10. The system according to claim 6, characterized in that, The 3D assessment report module supports batch export of reports and can use different color blocks to mark the report status of physiological abnormalities, psychological abnormalities, and both physiological and psychological abnormalities, which can be used to quickly screen key targets.

Images