Employee status assessment and intervention methods, devices, electronic equipment, and storage media

Abstract

This application discloses a method, device, electronic device, and storage medium for assessing and intervening in employee status, relating to the field of enterprise management technology. The method includes: real-time collection of multimodal data related to employee work status; feature extraction and encryption of the multimodal data to generate encrypted feature vectors; psychological health risk assessment based on the encrypted feature vectors using a preset data fusion model to generate a psychological risk index; and automatic triggering and execution of corresponding pre-configured intervention measures according to the risk level of the psychological risk index. This addresses the problems of delayed and passive assessment, isolated and one-sided data sources, lack of precision in intervention measures, and privacy and compliance risks associated with data collection in related technologies. It achieves the technical effects of early proactive prevention of employee mental health problems, comprehensive and objective assessment of employee status, accurate matching of personalized intervention needs, and protection of data privacy and compliance.

Description

Technical Field

[0001] This application relates to the field of enterprise management technology, and in particular to a method, apparatus, electronic device and storage medium for assessing and intervening in employee status. Background Technology

[0002] Currently, corporate employee mental health management mainly relies on questionnaires, employee assistance programs, and offline lectures. Some companies try to use wearable devices to generate rough early warnings based on single data points, while work status assessment relies on managers' subjective observations and lagging performance data.

[0003] These methods have significant shortcomings: assessments are delayed and passive, mostly involving post-event interventions, making it difficult to identify potential problems early; data sources are isolated and one-sided, with psychological assessments and work performance data being disconnected, and a single data source cannot comprehensively and objectively reflect the complex psychological state of employees; intervention measures lack precision, mostly being general solutions that are difficult to match the individual needs of employees; at the same time, the data collection process is prone to privacy concerns and compliance risks. Summary of the Invention

[0004] This application provides a method, apparatus, electronic device, and storage medium for assessing and intervening in employee status. It addresses the problems in related technologies, such as delayed and passive assessments, isolated and incomplete data sources, lack of precision in intervention measures, and privacy and compliance risks associated with data collection.

[0005] According to a first aspect of this application, a method for assessing and intervening in employee status is provided, comprising:

[0006] Real-time collection of multimodal data related to employees' work status; The multimodal data is subjected to feature extraction and encryption to generate an encrypted feature vector; Based on the encrypted feature vector, a psychological health risk assessment is performed using a preset data fusion model to generate a psychological risk index. Based on the risk level of the psychological risk index, the corresponding pre-configured intervention measures are automatically triggered and executed.

[0007] According to a second aspect of this application, an employee status assessment and intervention device is provided, comprising: The data acquisition module is configured to collect multimodal data related to employees' work status in real time. The first generation module is configured to extract features from the multimodal data and encrypt them to generate encrypted feature vectors. The second generation module is configured to perform a mental health risk assessment based on the encrypted feature vector and a preset data fusion model to generate a mental health risk index. The execution module is configured to automatically trigger and execute corresponding pre-configured intervention measures based on the risk level to which the psychological risk index belongs.

[0008] According to a third aspect of this application, an electronic device is provided, comprising: At least one processor; and memory that is communicatively connected to at least one processor; The memory stores instructions that can be executed by at least one processor, which enables the at least one processor to perform the employee status assessment and intervention method described in the first aspect above.

[0009] According to a fourth aspect of this application, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause a computer to execute the employee status assessment and intervention method described in the first aspect above.

[0010] According to a fifth aspect of this application, a computer program product is provided, including a computer program that, when executed by a processor, implements the employee status assessment and intervention method as described in the first aspect above.

[0011] This application enables the real-time collection of multimodal data related to employee work status, feature extraction, and encryption to generate encrypted feature vectors. Based on these encrypted feature vectors, a pre-defined data fusion model is used to conduct a mental health risk assessment to generate a mental health risk index. Finally, based on the risk level of the mental health risk index, corresponding pre-configured intervention measures are automatically triggered and executed. Therefore, this application can solve the problems of delayed and passive assessment, isolated and one-sided data sources, lack of precision in intervention measures, and privacy and compliance risks in data collection in related technologies. It achieves the technical effects of early and proactive prevention of employee mental health problems, comprehensive and objective assessment of employee status, accurate matching of personalized intervention needs, and protection of data privacy and compliance.

[0012] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0013] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0014] Figure 1A flowchart illustrating an employee status assessment and intervention method provided in an embodiment of this application; Figure 2 A flowchart illustrating another method for assessing and intervening in employee status provided in this application embodiment; Figure 3 A flowchart illustrating another method for assessing and intervening in employee status provided in this application embodiment; Figure 4 This is a schematic diagram of the structure of an employee status assessment and intervention device provided in an embodiment of this application. Detailed Implementation

[0015] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0016] The following description, with reference to the accompanying drawings, describes a method, apparatus, electronic device, and storage medium for assessing and intervening in employee status according to embodiments of this application.

[0017] Figure 1 This is a flowchart illustrating an employee status assessment and intervention method provided in an embodiment of this application.

[0018] like Figure 1 As shown, the method includes the following steps: Step 101: Collect multimodal data related to employees' work status in real time.

[0019] In some embodiments, multimodal data encompasses physiological data, behavioral data, and task load data related to employee work status. Real-time acquisition means that the data acquisition process is uninterrupted and synchronized with the employee's work progress, ensuring that status information at every key time point is captured.

[0020] All devices and sensors used for data collection are authorized and compliant. Wearable devices must meet ISO / IEC safety standards and are commonly in the form of smart badges or wristbands. They can accurately collect physiological data such as heart rate variability, skin conductance, body temperature, movement trajectory, and activity level. This data is stably transmitted to a local gateway or edge computing unit via Bluetooth technology. Heart rate variability can reflect the regulatory state of the human autonomic nervous system, skin conductance can sensitively capture changes in skin conductivity caused by emotional fluctuations, and body temperature and movement data can help determine the employee's fatigue level and daily activity status.

[0021] The workstation is equipped with non-invasive sensors including a keyboard kinetic sensor and a camera. The keyboard kinetic sensor can record detailed parameters such as keystroke frequency, key pressure, key interval time, and input error rate. The camera focuses on real-time posture analysis, collecting visual data such as sitting posture, limb movement frequency, and facial orientation. To protect privacy, the data collected by the camera undergoes strict anonymization processing. Before analysis, the video frames are blurred in real time, and sensitive information in the analysis results is further processed. At the same time, the keyboard kinetic data and the data collected by the camera are synchronously fused through a timestamp alignment mechanism to ensure the time consistency of the data.

[0022] The task load data comes from commonly used enterprise OA, CRM and other systems. The fields collected include the number of tasks, task deadlines, collaboration frequency and so on. This data is anonymized through techniques such as desensitization, aggregation and adding differential privacy noise to ensure that the data cannot be used to identify specific employees. The comprehensive and multi-dimensional real-time data collection provides a comprehensive, objective and compliant basis for subsequent employee status assessment and effectively avoids the limitations of a single data source.

[0023] Step 102: Extract features from the multimodal data and encrypt it to generate an encrypted feature vector.

[0024] In some embodiments, the feature extraction process for the collected multimodal data is completed on employee terminals or departmental gateways. Lightweight AI models are deployed here, which balance data processing efficiency and resource consumption control, and are adaptable to the operating environment of the terminals or gateways. The feature extraction stage performs targeted transformations for different types of multimodal data, refining the scattered raw data into high-dimensional feature vectors with clear evaluation significance.

[0025] Among them, the continuous heart rate signal in physiological data will be converted into a stress index that reflects the physical and mental stress state; the keystroke frequency, key pressure, interval time and error rate captured by the keyboard dynamics sensor will be converted into a focus fluctuation curve that reflects the changes in work focus; the posture data after desensitization will be converted into a body tension index that reflects the body's tense state; and various task load data will also be processed to form characteristic indicators that meet the assessment needs.

[0026] After feature extraction, all generated feature vectors are encrypted locally. Reliable encryption ensures data security during subsequent transmission, preventing the leakage of sensitive information. The final encrypted feature vectors are then uploaded to the cloud as core data for subsequent evaluation. This process effectively refines the raw data, enhancing its evaluative value, while local encryption safeguards privacy and security, laying a solid data foundation and security barrier for subsequent mental health risk assessments.

[0027] Step 103: Based on the encrypted feature vector, conduct a mental health risk assessment through a preset data fusion model to generate a mental health risk index.

[0028] In some embodiments, a pre-defined data fusion model is deployed in the cloud and is specifically designed to integrate and analyze encrypted feature vectors from multiple dimensions such as employee behavior, physiological state, and task load. Its core advantage lies in its ability to fully explore the intrinsic correlation between different modalities of data and avoid the one-sidedness of evaluation based on a single data dimension.

[0029] The model employs an attention mechanism that dynamically adjusts the weighting of different modal features based on the actual relevance of the data and the evaluation scenario. For example, during high-intensity project phases, the weight of task-load-related features will be appropriately increased, while in daily work scenarios, the weight of physiological state and behavioral features will be more prominent. This dynamic weighting method makes the evaluation process more in line with the actual situation.

[0030] During the evaluation process, after receiving the encrypted feature vector, the model will deeply integrate and quantitatively analyze the multi-source data through a preset algorithm. It will fully consider the synergistic effects of various factors, such as the stress index derived from heart rate variability, the focus fluctuation curve derived from keyboard dynamics, the body tension index corresponding to posture data, and task load-related indicators. Finally, it will output a psychological risk index ranging from 0 to 100. The higher the index value, the greater the psychological health risk currently faced by the employee.

[0031] This assessment process does not rely on subjective human judgment and is based entirely on objective data, ensuring the fairness and accuracy of the results. Its beneficial effect is that it can provide a precise and quantitative basis for subsequent risk level classification, allowing mental health risk assessment to break free from the subjective limitations of the traditional model and become more comprehensive and scientific.

[0032] Step 104: Based on the risk level of the psychological risk index, automatically trigger and execute the corresponding pre-configured intervention measures.

[0033] In some embodiments, the system has a comprehensive, pre-defined tiered intervention rule base. Risk level determination is strictly based on the psychological risk index and related state characteristics. Level 1 risk corresponds to a psychological risk index exceeding 70 for at least 2 hours; Level 2 risk is a psychological risk index exceeding 85 accompanied by a sharp drop in work efficiency; and Level 3 risk is the detection of extreme physiological signals such as a sudden change in heart rate. Differentiated intervention measures are pre-configured for different risk levels and can be executed automatically without manual triggering. For Level 1 risk, the system will immediately push mild stress-reduction resources, such as a 5-minute guided mindfulness breathing exercise, to help employees quickly relieve mild stress. For Level 2 risk, the system will anonymously notify the HRBP, suggesting they conduct a light-weight conversation with the employee and provide targeted support while protecting employee privacy. If Level 3 risk is triggered, the system will immediately activate emergency protocols, including issuing system alerts and arranging for professional intervention, to promptly address serious mental health crises.

[0034] To ensure the efficient implementation of intervention measures, the system supports seamless integration with the company's internal HR system and instant messaging tools, ensuring that intervention notices are accurately delivered to relevant personnel and intervention resources are smoothly pushed to employees.

[0035] The beneficial effect of this tiered automatic intervention model is that it achieves precise matching between intervention measures and risk levels, avoids the blindness of general interventions, improves the timeliness and effectiveness of interventions, and truly achieves proactive response and scientific management of employees' mental health risks.

[0036] Compared with related technologies, this embodiment collects multimodal data related to employees' work status in real time; extracts features from the multimodal data and encrypts it to generate encrypted feature vectors; based on the encrypted feature vectors, it conducts a mental health risk assessment through a preset data fusion model to generate a mental health risk index; and automatically triggers and executes corresponding pre-configured intervention measures according to the risk level of the mental health risk index. This solves the problems of delayed and passive assessment, isolated and one-sided data sources, lack of precision in intervention measures, and privacy and compliance risks associated with data collection in related technologies. It achieves the technical effects of early proactive prevention of employee mental health problems, comprehensive and objective assessment of employee status, accurate matching of personalized intervention needs, and protection of data privacy and compliance.

[0037] Figure 2 A flowchart illustrating another method for assessing and intervening in employee status provided in this application embodiment includes the following steps: Step 201: Collect physiological time-series data through the wearable device worn by the employee.

[0038] In some embodiments, the wearable devices worn by employees must be officially authorized and comply with ISO / IEC safety standards. Common forms include smart name tags or wristbands. These devices are small, easy to wear, and do not interfere with employees' normal work. The device is specifically designed to collect employees' physiological time-series data, which refers to physiological indicators continuously recorded over time. Core data include heart rate variability, skin conductance, body temperature, movement trajectory, and activity level.

[0039] Among these metrics, heart rate variability reflects the regulatory function of the human autonomic nervous system and is an important basis for assessing physical and mental stress; skin conductance changes with emotional fluctuations, accurately capturing the degree of emotional arousal; body temperature, movement trajectory, and activity level can intuitively reflect the employee's physical fatigue state and daily activity intensity. The collected physiological time-series data is stably transmitted to the local gateway or edge computing unit via Bluetooth technology, ensuring the real-time performance and security of data transmission and avoiding data delays or loss.

[0040] The beneficial effect of this data collection method is that it can continuously and accurately obtain dynamic changes in employees' physiological levels, providing real and reliable physiological data for subsequent mental health assessments.

[0041] Step 202: Collect anonymized behavioral time-series data using sensors deployed on employee workstations, including but not limited to keyboard kinetic sensors and image acquisition devices.

[0042] In some embodiments, the sensors deployed on employee workstations are all non-invasive designs that will not interfere with employees' work operations and strictly comply with privacy protection requirements. The sensor types include, but are not limited to, keyboard kinetic sensors and image acquisition devices.

[0043] Keyboard dynamics sensors capture various behavioral parameters of employees in real time during keyboard operation, generating anonymized behavioral time-series data, including keystroke frequency, key pressure, key interval time, and input error rate. This data can dynamically reflect changes in employees' operating habits and work focus.

[0044] Image acquisition devices (such as cameras) focus on real-time posture analysis, continuously recording visual data such as employees' sitting posture, frequency of limb movements, and facial orientation. To protect privacy, the acquisition process undergoes double desensitization processing. First, the video frames are blurred in real time before analysis, and then sensitive information in the analysis results is further optimized. At the same time, the data collected by the keyboard kinetic sensor and the data acquired by the image acquisition device are synchronously fused through a timestamp alignment mechanism to ensure the temporal consistency of behavioral time series data.

[0045] The advantage of this data collection method is that it can comprehensively capture the dynamic behavior during the work process while protecting employee privacy, providing rich behavioral dimensions to support employee status assessment.

[0046] Step 203: Obtain anonymized work task load data related to employees by communicating with the enterprise's internal information system.

[0047] In some embodiments, a data interaction channel is established with commonly used enterprise information systems such as OA and CRM through secure and compliant communication connections to ensure the security and stability of data transmission. During the interaction, the focus is on acquiring anonymized work load data related to employees. This data is updated in real time according to a time series, accurately reflecting the dynamic changes in employee work tasks. Core collection fields include the number of tasks, task deadlines, and collaboration frequency. To prevent data leakage or reverse identification of specific employees, all acquired work load data undergoes strict anonymization processing, specifically employing techniques such as desensitization, data aggregation, and the addition of differential privacy noise. This ensures the irreversible identifiability of the data from the outset, satisfying data collection requirements while fully respecting employee privacy. This data acquisition method supplements core objective data in employees' work scenarios, making multimodal data collection more closely aligned with actual work, effectively compensating for the lack of correlation between physiological and behavioral data in work scenarios, and making subsequent evaluations more targeted.

[0048] Step 204: At the edge of data acquisition, perform local processing and feature extraction on physiological time-series data, anonymized behavioral time-series data, and anonymized workload data to generate feature vectors.

[0049] In some embodiments, the edge side of data acquisition specifically refers to employee terminals or departmental gateways, where lightweight AI models are pre-deployed. These models can efficiently process data with limited hardware resources, ensuring processing speed without consuming excessive system resources.

[0050] For the collected physiological time-series data, anonymized behavioral time-series data, and anonymized work workload data, the edge computing platform will perform targeted local processing and feature extraction. For physiological time-series data, continuous heart rate signals will be converted into a stress index that directly reflects physical and mental stress levels, and skin conductance data will be processed into a feature indicator reflecting the degree of emotional arousal. At the same time, relevant features reflecting fatigue status will be extracted by combining body temperature, movement trajectory, and activity level data. For anonymized behavioral time-series data, parameters such as keystroke frequency, key pressure, interval time, and error rate related to keyboard dynamics will be converted into attention fluctuation curves, and desensitized posture data will be converted into a body tension index reflecting the state of physical tension. For anonymized work workload data, feature indicators that reflect the intensity and rhythm of work pressure will be extracted from fields such as the number of tasks, deadlines, and collaboration frequency.

[0051] The entire processing and extraction process is completed locally, avoiding long-distance transmission of raw data. This improves processing efficiency and reduces the risk of data leakage. The resulting high-dimensional feature vectors can accurately carry the core evaluation value of various types of data, laying a solid foundation for subsequent encryption and evaluation work. Its beneficial effect is that it realizes efficient transformation and value extraction of raw data, while ensuring the privacy and security of the data processing process.

[0052] Step 205: Encrypt the feature vector to generate the encrypted feature vector and transmit it to the cloud, while keeping the multimodal data locally.

[0053] In some embodiments, after feature extraction and feature vector generation are completed, the feature vector is encrypted locally. The encryption method used has high security and reliability, which can effectively resist the risk of theft and tampering during data transmission and ensure the confidentiality of the feature vector.

[0054] After encryption, the generated encrypted feature vector is transmitted to the cloud via a secure communication channel for subsequent mental health risk assessment. Meanwhile, all unprocessed raw data is completely preserved at the data collection edge—employee terminals or departmental gateways—and is not uploaded to the cloud or other external systems, thus preventing the leakage of raw sensitive data at its source.

[0055] This model of "locally encrypted upload of feature vectors and local retention of raw data" not only meets the data requirements for centralized cloud-based assessments but also maximizes the protection of employee privacy, complying with data compliance requirements. Its beneficial effects lie in achieving dual protection of data privacy and transmission security while ensuring the smooth progress of subsequent assessments, effectively alleviating employee concerns about data leaks, and enhancing the system's compliance and credibility.

[0056] Step 206: Using a federated learning framework, aggregate encrypted model parameters from multiple edge nodes in the cloud to update the data fusion model.

[0057] In some embodiments, the data fusion model is updated using a horizontal federated learning framework, which enables collaborative optimization of the model without aggregating the original data. The edge nodes participating in the collaborative update include multiple data acquisition terminals such as wearable devices worn by employees and workstations. After each edge node completes data processing and feature extraction locally, it only uploads encrypted local model parameters or gradients to the cloud, without involving the transmission of any original sensitive data, thus avoiding the risk of data leakage at the source. The cloud server uses secure multi-party computation technology to aggregate the encrypted model parameters from all edge nodes, ensuring that parameter information is not leaked or tampered with during the aggregation process through encryption algorithms. The updated data fusion model can employ an LSTM-based temporal anomaly detection model, which has excellent temporal data processing capabilities and can accurately capture the dynamic changes in employee status. The model supports asynchronous update mode to adapt to the data processing rhythm of different edge nodes. During communication, it uses encrypted transmission protocols based on homomorphic encryption or secure multi-party computation and is equipped with a strict key management system, further strengthening the security of data transmission and processing. Its beneficial effects are that it enables continuous optimization of the model with the support of multi-node data, improving the accuracy and adaptability of model evaluation, and also ensures the compliance of data processing and alleviates employees' privacy concerns through the privacy protection features of federated learning.

[0058] Step 207: Using the updated data fusion model, perform multimodal fusion analysis on the received encrypted feature vector to calculate the psychological risk index.

[0059] In some embodiments, the updated data fusion model is deployed in the cloud and has the ability to efficiently process multimodal encrypted feature vectors. Upon receiving encrypted feature vectors from edge nodes, the model initiates a multimodal fusion analysis process. These encrypted feature vectors cover core information across multiple dimensions, including the employee's physiological state, work behavior, and workload. The model incorporates an attention mechanism that dynamically adjusts the weight of each modality feature based on different assessment scenarios, job characteristics, and project stages. For example, during critical project phases, the weight of workload-related features is appropriately increased, while in daily work scenarios, the influence of physiological and behavioral features is more prominent, ensuring that the fusion analysis better aligns with actual assessment needs. By deeply mining the intrinsic correlations between various modalities, the model performs quantitative calculations and comprehensive judgments on the encrypted feature vectors, ultimately outputting a psychological risk index ranging from 0 to 100. The index value directly corresponds to the severity of the employee's current mental health risk. This process is entirely based on objective data calculations, eliminating the limitations of subjective human judgment. Its beneficial effect lies in achieving a comprehensive and accurate assessment of employee mental health risks, providing a scientific and quantitative core basis for subsequent risk level classification and targeted intervention.

[0060] Step 208: Compare the psychological risk index with the risk threshold dynamically adjusted based on the employee's historical baseline and job attributes to determine the risk level.

[0061] In some embodiments, the risk threshold is not fixed but dynamically adjusted based on multi-dimensional references. The core criteria include an individual historical baseline established for each employee and a group baseline adapted to different job attributes. The individual historical baseline is a personalized reference standard formed by integrating an employee's psychological state data and work performance data over a period of time, accurately reflecting the employee's normal psychological level. Job attributes fully consider the differences in work pressure, work pace, and responsibilities among different positions. For example, the risk threshold benchmark for high-intensity R&D positions will differ from that of regular administrative positions. Furthermore, it is flexibly fine-tuned based on different stages of the project cycle. When determining the risk level, the system accurately compares the real-time generated psychological risk index with this dynamically adjusted risk threshold, classifying the corresponding risk level according to preset rules. The first level corresponds to a psychological risk index exceeding 70 for 2 consecutive hours; the second level is a psychological risk index exceeding 85 accompanied by a sudden drop in work efficiency; and the third level is the detection of extreme physiological signals such as abnormal heart rate changes. This dynamic threshold-based comparison method can fully adapt to individual differences among employees and the characteristics of work scenarios. Its beneficial effect is to significantly reduce the probability of false alarms in risk level determination, making risk identification more targeted and accurate.

[0062] Step 209: If the risk level is Level 1, then push preset psychological adjustment resources to the employee.

[0063] In some embodiments, Level 1 risk corresponds to an employee's psychological risk index exceeding 70 for two consecutive hours. At this point, the employee is experiencing mild psychological stress, which has not yet significantly negatively impacted work performance. The system automatically pushes preset psychological adjustment resources to the employee. These resources focus on mild stress reduction and commonly include 5-minute guided mindfulness breathing exercises, short relaxation meditation audio clips, and mood regulation tips. Resource delivery is completed through convenient channels such as internal instant messaging tools, ensuring that employees can quickly access and utilize fragmented time for adjustment. These preset resources are selectively chosen to avoid taking up too much work time while effectively helping employees alleviate current mild stress and guiding them to quickly adjust their mental state. Its beneficial effect lies in achieving early, mild intervention for mental health issues, timely relieving mild stress, preventing further accumulation and escalation of risk, and maintaining the convenience and efficiency of intervention without the need for manual intervention.

[0064] Step 210: If the risk level is Level 2, then anonymously notify the relevant human resources personnel to initiate the proactive care process.

[0065] In some embodiments, the criteria for determining Level 2 risk is a psychological risk index exceeding 85 accompanied by a sharp drop in work performance. At this point, the employee's psychological stress has significantly impacted their work, requiring professional care and support. The system will send an anonymous notification to relevant human resources personnel (primarily HRBPs). The notification content only includes risk warnings and necessary work-related information, without disclosing employee personal privacy details, effectively avoiding the stigma associated with mental health issues. Upon receiving the notification, HR personnel will initiate a proactive care process, establishing contact with the employee through lightweight communication to gently understand their sources of stress and work difficulties, providing targeted support and advice, such as coordinating work tasks and offering resource assistance. Its beneficial effect lies in achieving warm and precise care while fully protecting employee privacy. It avoids the blindness of generic interventions while allowing employees to feel organizational support, helping to alleviate psychological stress and restore work performance.

[0066] Step 211: If the risk level is Level 3, then trigger an emergency response protocol that includes human intervention.

[0067] In some embodiments, Level 3 risk represents a severe emergency situation, determined by criteria including the detection of extreme physiological signals such as sudden changes in heart rate, potentially accompanied by an extremely high psychological risk index. In this case, the employee's mental and physical health faces a significant threat, requiring urgent intervention. The system immediately triggers an emergency response protocol including human intervention. First, it sends an instant alert to pre-set relevant personnel, including HR managers, department heads, and professional psychological intervention personnel. The alert information clearly indicates the risk level and key warning signals to ensure a rapid response from relevant personnel. The human intervention phase involves multi-faceted collaborative action. Professional psychological intervention personnel contact the employee immediately to provide emergency psychological counseling and assessment. Department heads coordinate work arrangements and provide necessary support based on the actual situation. If the employee experiences physical discomfort, the system also assists in contacting medical resources for timely treatment. Its beneficial effects include rapid response to severe mental health crises, minimizing emergency intervention response time, providing strong protection for the employee's mental and physical health, effectively reducing the probability of extreme situations, and demonstrating the organization's responsibility and care for its employees.

[0068] Figure 3 A flowchart illustrating another method for assessing and intervening in employee status provided in this application embodiment includes the following steps: Step 301: Collect multimodal data related to the employee's work status in real time.

[0069] Step 302: Extract features from the multimodal data and encrypt it to generate an encrypted feature vector.

[0070] Step 303: Based on the encrypted feature vector, a mental health risk assessment is performed using a preset data fusion model to generate a mental health risk index.

[0071] Step 304: Based on the risk level of the psychological risk index, automatically trigger and execute the corresponding pre-configured intervention measures.

[0072] For a description of steps 301-304, please refer to the description of steps 101-104 in the above embodiment. This embodiment will not repeat them in detail.

[0073] Step 305: Record the intervention measures implemented each time and the changes in the employee's psychological risk index after the intervention.

[0074] In some embodiments, a comprehensive data recording process is automatically initiated each time the system triggers and executes an intervention. The recorded content covers both complete details of the intervention, including the specific type and channel of the psychological adjustment resources pushed, the identity of the HR personnel notified anonymously and the notification time, the initiation point of the emergency response protocol, information on personnel involved in manual intervention, and the action process; and the complete trajectory of changes in the employee's psychological risk index before and after the intervention, specifically including the initial psychological risk index at the moment of intervention execution, real-time risk indices at key time points such as 1 hour, 3 hours, and 24 hours after intervention, as well as core data such as the magnitude and trend of risk index increases and decreases. All recorded data is stored using compliant encryption methods and is strictly managed separately from employee personal identification information, raw physiological and behavioral data, establishing independent encrypted data archives to ensure data security, integrity, and traceability, while also complying with data privacy protection regulations to prevent information leakage.

[0075] Step 306: Based on the changed data, the intervention strategy rules are iteratively optimized using a reinforcement learning algorithm.

[0076] In some embodiments, based on the intervention measures and risk index change data recorded in step 305, the system introduces a reinforcement learning algorithm to continuously iteratively optimize the intervention strategy rules. This algorithm, with "intervention effectiveness" as its core objective, constructs a closed-loop learning mechanism of "intervention implementation - risk feedback." A significant decrease in the psychological risk index and a gradual recovery of employee work status after intervention are set as positive reward signals, while no improvement, an increase in the risk index, or poor employee feedback are set as negative feedback signals. The algorithm deeply analyzes the actual adaptability of various intervention measures for different employees under different job characteristics, project stages, and risk levels. For example, for R&D employees at level one risk, if data shows that stress-reducing audio related to professional skills is more effective than general mindfulness guidance, the priority of intervention resource delivery for this type of employee at the corresponding risk level will be adjusted. If it is found that for some employees at level two risk, initiating a proactive care process outside of peak work hours is more effective, the timing rules for initiating the care process will be optimized. The entire optimization process is conducted in a cloud-based encrypted environment, only accessing encrypted change data and avoiding access to employees' sensitive personal information. The iteratively updated intervention strategies and rules are automatically synchronized to the system's pre-set tiered intervention rule library, ensuring that subsequent interventions more accurately match employees' individual needs and real-world scenarios. Its beneficial effect lies in enabling intervention strategies to have dynamic adaptive capabilities, continuously improving the targeting and effectiveness of interventions, and driving the continuous upgrading of the entire mental health management system's efficiency.

[0077] Figure 4 This is a schematic diagram of the structure of an employee status assessment and intervention device provided in an embodiment of this application, as shown below. Figure 4As shown, it includes: acquisition module 401, first generation module 402, second generation module 403, and execution module 404.

[0078] The data acquisition module 401 is configured to collect multimodal data related to the employee's work status in real time; The first generation module 402 is configured to extract features from the multimodal data and encrypt them to generate an encrypted feature vector. The second generation module 403 is configured to perform a mental health risk assessment based on the encrypted feature vector and a preset data fusion model to generate a mental health risk index. The execution module 404 is configured to automatically trigger and execute the corresponding pre-configured intervention measures based on the risk level to which the psychological risk index belongs.

[0079] In some examples of this embodiment, the acquisition module 401 is specifically configured to acquire physiological time-series data through a wearable device worn by the employee; acquire anonymized behavioral time-series data through sensors deployed on the employee's workstation, including but not limited to keyboard kinetic sensors and image acquisition devices; and acquire anonymized work task load data related to the employee through a communication connection with the enterprise's internal information system.

[0080] In some examples of this embodiment, the first generation module 402 is specifically configured to perform local processing and feature extraction on physiological time-series data, anonymized behavioral time-series data, and anonymized work task load data at the edge of data acquisition to generate feature vectors; encrypt the feature vectors to generate encrypted feature vectors and transmit them to the cloud, while keeping the original data locally.

[0081] In some examples of this embodiment, the second generation module 403 is specifically configured to aggregate encrypted model parameters from multiple edge nodes in the cloud through a federated learning framework to update the data fusion model; and to use the updated data fusion model to perform multimodal fusion analysis on the received encrypted feature vectors to calculate the psychological risk index.

[0082] In some examples of this embodiment, the execution module 404 is specifically configured to compare the psychological risk index with a risk threshold dynamically adjusted based on the employee's historical baseline and job attributes to determine the risk level; if the risk level is the first level, then push preset psychological adjustment resources to the employee; if the risk level is the second level, then anonymously notify human resources personnel to initiate the proactive care process; if the risk level is the third level, then trigger an emergency response protocol that includes human intervention.

[0083] It should be noted that other corresponding descriptions of the functional units involved in the employee status assessment and intervention device provided in this embodiment can be found in [reference]. Figure 1 , Figure 2 and Figure 3 The corresponding description in [the document] will not be repeated here.

[0084] Based on the above, Figure 1 , Figure 2 and Figure 3 The embodiment illustrates a method for assessing and intervening in employee status. Correspondingly, this embodiment also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned method. Figure 1 , Figure 2 and Figure 3 This illustrates a method for assessing and intervening in employee status.

[0085] Based on the above, Figure 1 , Figure 2 and Figure 3 The embodiment illustrates a method for assessing and intervening in employee status. Correspondingly, this embodiment also provides a computer program product on which a computer program is stored. When executed by a processor, this computer program implements the above-described method. Figure 1 , Figure 2 and Figure 3 This illustrates a method for assessing and intervening in employee status.

[0086] Based on this understanding, the technical solution of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause a computer device (such as personal computer, server, or network device, etc.) to execute the methods of various implementation scenarios of this application.

[0087] Based on the above, Figure 1 , Figure 2 and Figure 3 This illustrates a method for assessing and intervening in employee status, and Figure 4 To achieve the above objectives, the present application also provides an electronic device, such as a personal computer or a server, in the illustrated virtual device embodiment. This device includes a storage medium and a processor; the storage medium stores a computer program; the processor executes the computer program to implement the above-described virtual device. Figure 1 , Figure 2 and Figure 3 This illustrates a method for assessing and intervening in employee status.

[0088] In some embodiments, the aforementioned physical device may further include a user interface, a network interface, a camera, radio frequency (RF) circuitry, sensors, audio circuitry, a Wi-Fi module, etc. The user interface may include a display screen, an input unit such as a keyboard, etc., and optionally, a USB interface, a card reader interface, etc. In some embodiments, the network interface may include a standard wired interface, a wireless interface (such as a Wi-Fi interface), etc.

[0089] The storage medium may also include an operating system and a network communication module. The operating system is a program that manages the hardware and software resources of the aforementioned physical device, supporting the operation of information processing programs and other software and / or programs. The network communication module is used to enable communication between the various components within the storage medium, as well as communication with other hardware and software in the information processing physical device.

[0090] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0091] The above are merely specific embodiments of this application, enabling those skilled in the art to understand or implement this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to these embodiments, but is to be accorded the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A method for assessing and intervening in employee status, characterized in that, include: Real-time collection of multimodal data related to employees' work status; The multimodal data is subjected to feature extraction and encryption to generate an encrypted feature vector; Based on the encrypted feature vector, a psychological health risk assessment is performed using a preset data fusion model to generate a psychological risk index. Based on the risk level of the psychological risk index, the corresponding pre-configured intervention measures are automatically triggered and executed.

2. The method for assessing and intervening in employee status according to claim 1, characterized in that, The real-time acquisition of multimodal data related to the working status includes: Physiological time-series data are collected through wearable devices worn by the employees; Anonymized behavioral time-series data is collected by sensors deployed on employee workstations, including but not limited to keyboard kinetic sensors and image acquisition devices; By communicating with the enterprise's internal information system, anonymized work workload data related to employees can be obtained.

3. The method for assessing and intervening in employee status according to claim 1, characterized in that, The step of extracting features from the multimodal data and encrypting it to generate an encrypted feature vector includes: At the edge of data acquisition, physiological time-series data, anonymized behavioral time-series data, and anonymized workload data are processed locally and feature extracted to generate feature vectors. The feature vector is encrypted to generate the encrypted feature vector and transmitted to the cloud, while the multimodal data is stored locally.

4. The method for assessing and intervening in employee status according to claim 1, characterized in that, The step of conducting a mental health risk assessment based on the feature vector using a preset data fusion model to generate a mental health risk index includes: The data fusion model is updated by aggregating encrypted model parameters from multiple edge nodes in the cloud using a federated learning framework. Using the updated data fusion model, multimodal fusion analysis is performed on the received encrypted feature vector to calculate the psychological risk index.

5. The method for assessing and intervening in employee status according to claim 1, characterized in that, The automatic triggering and execution of pre-configured intervention measures based on the risk level of the psychological risk index includes: The risk level is determined by comparing the psychological risk index with a risk threshold that is dynamically adjusted based on the employee's historical baseline and job attributes. If the risk level is Level 1, then preset psychological adjustment resources will be pushed to the employee; If the risk level is Level 2, then the relevant human resources personnel will be anonymously notified to initiate the proactive care process; If the risk level is Level 3, an emergency response protocol involving human intervention will be triggered.

6. The method for assessing and intervening in employee status according to claim 1, characterized in that, Also includes: Record each intervention measure implemented and the changes in the employee's psychological risk index after the intervention; Based on the aforementioned change data, the intervention strategy rules are iteratively optimized using reinforcement learning algorithms.

7. A device for assessing and intervening in employee status, characterized in that, include: The data acquisition module is configured to collect multimodal data related to employees' work status in real time. The first generation module is configured to extract features from the multimodal data and encrypt them to generate encrypted feature vectors. The second generation module is configured to perform a mental health risk assessment based on the encrypted feature vector and a preset data fusion model to generate a mental health risk index. The execution module is configured to automatically trigger and execute corresponding pre-configured intervention measures based on the risk level to which the psychological risk index belongs.

8. An electronic device, characterized in that, include: At least one processor; and a memory communicatively connected to the at least one processor; The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the employee status assessment and intervention method according to any one of claims 1-6.

9. A non-transitory computer-readable storage medium storing computer instructions, characterized in that, The computer instructions are used to cause the computer to perform the employee status assessment and intervention method according to any one of claims 1-6.

10. A computer program product, characterized in that, It includes a computer program that, when executed by a processor, implements the method for assessing and intervening in the employee status according to any one of claims 1-6.

Images