Bank grassroots employee pressure sensing and behavior adjusting method

Abstract

The invention relates to the technical field of physiology and emotion monitoring, in particular to a bank grassroots employee pressure sensing and behavior adjusting method, which comprises the following steps: S1, monitoring the physiological data of employees, and performing data acquisition on the emotional states of the employees by using a facial expression recognition algorithm; s2, performing normalization processing to form a standardized data set; s3, constructing a pressure sensing model, and outputting a pressure state score; s4, the pressure states of the employees are divided into a low grade, a middle grade and a high grade; s5, making a personalized behavior adjustment strategy; s6, pushing the behavior adjustment strategy generated in the step S5 to the employee; s7, adjusting hyper-parameters of Gaussian process regression through an optimization algorithm; according to the method, by combining real-time physiological and emotion data monitoring, personalized pressure threshold adjustment and a Bayesian optimization algorithm, the accuracy of pressure sensing and the effectiveness of a personalized behavior adjustment strategy are improved, so that the working efficiency and mental health of employees are remarkably enhanced.

Description

Technical Field

[0001] The present invention relates to the field of physiological and emotional monitoring technology, and in particular to a method for stress perception and behavior adjustment of grassroots bank employees. Background Art

[0002] Grassroots bank employees often face high-intensity tasks, strict performance requirements and complex interpersonal relationships at work. These factors can cause employees to be under high mental stress and psychological burden for a long time. With the popularization of wearable devices and smart terminals, the technical means of monitoring employees' physiological data (such as heart rate, respiratory rate, skin electrical response) and emotional state have gradually matured, and can perceive employees' stress state in real time. However, existing technical means are mostly focused on simple physiological data monitoring, and do not combine individual differences of employees, such as historical behavior patterns, mental health status and personal preferences, resulting in the inability to formulate personalized stress management strategies for each employee.

[0003] Existing technologies have significant deficiencies in the accuracy and personalization of stress perception and behavior adjustment. First, many systems use fixed stress thresholds, ignore individual differences among employees, and cannot accurately identify different stress levels. Second, existing behavior adjustment strategies lack a dynamic feedback mechanism and fail to optimize based on employees' real-time status and feedback data, resulting in poor results of adjustment strategies. Therefore, a method for stress perception and behavior adjustment of grassroots bank employees is urgently needed to solve the above problems. Summary of the invention

[0004] Based on the above purpose, the present invention provides a method for stress perception and behavior adjustment of grassroots bank employees.

[0005] The stress perception and behavior adjustment method for grassroots bank employees includes the following steps:

[0006] S1: Deploy a sensor network in the working environment of grassroots bank employees to monitor their physiological data, including their heart rate, galvanic skin response, and respiratory rate, and use facial expression recognition algorithms to collect data on their emotional states;

[0007] S2: Normalize the physiological data and emotional data collected in S1, use the noise filtering algorithm to remove abnormal data and noise in the collection process, and extract features from various types of data to form a standardized data set;

[0008] S3: Based on the standardized data set generated in S2, a Gaussian process regression algorithm is used to construct a stress perception model, wherein the input of the stress perception model includes the employee's physiological data, emotional data, and work task load characteristics, and the output is a stress state score;

[0009] S4: Based on the output stress score of the stress perception model in S3, the employee's stress status is divided into three levels: low, medium, and high, and the classification results are adjusted in combination with the employee's personalized stress threshold;

[0010] S5: Based on the stress state level in S4, combined with the employee's historical behavior data, mental health records and personal preferences, formulate a personalized behavior adjustment strategy, which includes task adjustment, rest suggestions and emotional relief measures;

[0011] S6: Push the behavior adjustment strategy generated in S5 to employees through their wearable devices or office terminals, and provide execution reminders on the wearable devices or work interfaces;

[0012] S7: Monitor the physiological data and emotional state after the execution of the strategy in S6, transmit the collected feedback data back to the pressure perception model constructed in S3, and adjust the hyperparameters of Gaussian process regression through the optimization algorithm to achieve adaptive optimization of the model.

[0013] Optionally, the S1 specifically includes:

[0014] S11: Sensor network layout. In the working area of ​​the grassroots bank employees, a distributed sensor layout scheme is adopted to evenly install heart rate sensors, skin galvanic response sensors and respiratory rate sensors on the employees' work desks, chair backs and office equipment. The heart rate sensors are in contact with employees through wearable devices such as smart watches or chest straps; the skin galvanic response sensors are monitored through the contact surface of the work chair; the respiratory rate sensors obtain the employees' respiratory rate signals through infrared sensors installed near the employees' seats;

[0015] S12: Emotional state data collection: a facial expression recognition camera is set up in the work environment. The camera uses a convolutional neural network algorithm to perform real-time facial expression recognition to capture the facial expression changes of employees, including the expression feature points of eyebrows, mouth corners, and eye areas, and maps the feature points to emotional states, including pleasure, tension, and anxiety;

[0016] S13: Data integration and transmission: integrating the heart rate, galvanic skin response, respiratory rate data and emotional state data into the central data processing unit through the wireless sensor network to form the employee's physiological and emotional state data set.

[0017] Optionally, the S12 specifically includes:

[0018] S121: Feature point capture: The facial image of the employee is captured in real time through the facial expression recognition camera, and the image is input into the convolutional neural network for convolution operation to extract the facial feature points of the employee's eyebrows, mouth corners and eye areas;

[0019] S122: Feature map generation, downsampling the convolutional feature map through the pooling layer to retain feature information;

[0020] S123: Feature vector generation and classification, the pooled feature map is passed to the fully connected layer to generate the feature vector F of the facial expression;

[0021] S124: Emotional state mapping, mapping is performed by calculating the Euclidean distance between the facial expression feature vector F and the preset standard emotion feature vector E, and by comparing the Euclidean distance values, the standard emotion feature vector E corresponding to the smallest distance is selected and mapped to a specific emotional state, including pleasure, tension or anxiety.

[0022] Optionally, the S2 specifically includes:

[0023] S21: Data normalization processing: normalizing the collected physiological data and emotional data to ensure that data of different dimensions have consistent scales;

[0024] S22: Noise filtering: applying a noise filtering algorithm to the normalized data to remove abnormal data and noise. Specifically, a weighted moving average filter is used to smooth the data to remove sudden outliers.

[0025] S23: Abnormal data screening, using the triple standard deviation method to screen outliers in the data;

[0026] S24: Perform feature extraction on the normalized and noise-filtered data to extract the amplitude of heart rate changes, the amplitude of skin electrical response fluctuations, and the fluctuation period of respiratory rate. Perform dimensionality reduction on the multi-dimensional data through principal component analysis to extract features and form a standardized data set.

[0027] Optionally, the S3 specifically includes:

[0028] S31: Training data preparation, taking the standardized data set generated in S2 as input, the standardized data includes the physiological data and emotional data of the employee, and setting the input standardized data set to X=(x1, x2, ..., x n ), where x i is the feature vector at the i-th time point;

[0029] S32: The radial basis function is selected as the kernel function of the Gaussian process, and the expression is:

[0030] Among them, k(x i , x j ) is x i and x j The kernel function value between ∥x i -xj ∥ is the Euclidean distance between two eigenvectors, σ is the scale parameter of the kernel function;

[0031] S33: Based on the kernel function and training data, a Gaussian process regression model is constructed. It is assumed that there is a Gaussian process relationship between the features in the standardized data set and the pressure state, and the output f(X) of the model represents the pressure state; then the prior distribution of the Gaussian process is expressed as: Where f(X) is the predicted value of the pressure state, represents a Gaussian process, K(X, X) is the kernel matrix;

[0032] S34: Optimize the model through maximum likelihood estimation, optimize the model parameters, including the kernel function parameter σ, and use the model to predict the new pressure state value; the prediction formula is:

[0033] f * =K(X * ,X)K(X,X) -1 f(X), where f * For the new data point X * The predicted value of the pressure state, K(X * , X) is the kernel function value between the new data point and the training data, (X, X) -1 is the inverse of the training data kernel matrix;

[0034] S35: The final pressure perception model is constructed based on Gaussian process regression, and the complete expression is:

[0035] f(X * )=μ(X * )+K(X * ,X)K(K,X) -1 (y-μ(X)), where f(X * ) is the predicted pressure value of the new data point, μ(X * ) is the mean function, K(X * ,X) is the kernel function, and y is the actual pressure value of the training data.

[0036] Optionally, the pressure state score is output according to the pressure prediction value calculated by the pressure perception model in S35, including pressure prediction value normalization, by normalizing the output pressure prediction value f(X * ) to ensure that the pressure prediction value is within the preset range. The standardization formula is: Among them, f′(X * ) is the normalized pressure prediction value, f min and f max are the minimum and maximum values ​​in the historical pressure prediction values ​​respectively; the standardized value f′(X *) is limited to the range of 0 to 100.

[0037] Optionally, the S4 specifically includes:

[0038] S41: Classification of pressure state based on the normalized pressure prediction value f′(X * ), initially divide the stress status of employees into three levels: low, medium and high; the classification criteria are:

[0039]

[0040] Among them, L is the initial pressure level, f′(X * ) is the normalized pressure prediction value, ranging from 0 to 100, used to reflect different pressure levels;

[0041] S42: personalized stress threshold setting, combining each employee's personalized stress tolerance and setting personalized thresholds T1 and T2 according to the employee's historical stress data, where T1 represents the critical value of low stress and medium stress, and T2 represents the critical value of medium stress and high stress;

[0042] S43: Adjust the stress state by adjusting the stress state classification result according to the personalized stress thresholds T1 and T2; the adjustment rule expression is:

[0043]

[0044] Among them, L′ is the pressure level adjusted according to the personalized threshold;

[0045] S44: Output the pressure state, and output the final pressure state L′ as three levels: low, medium, and high.

[0046] Optionally, the S5 specifically includes:

[0047] S51: Historical behavior data analysis, obtaining employees’ historical behavior data, including employees’ work performance, rest frequency and emotional changes under different stress states, and using correlation analysis to calculate the relationship between employee behavior patterns and stress states;

[0048] S52: Analyze employees’ mental health records, assess their mental health status, including their history of anxiety and depression, score them using a scale, and determine their mental health threshold H;

[0049] S53: setting a personalized parameter set P according to the employee's personal preferences, including rest time, workload preference, and task type;

[0050] S54: Based on the analysis results of the above steps, generate a personalized behavior adjustment strategy, which includes task load adjustment, short-term rest arrangement and emotion regulation suggestions. Suppose the behavior adjustment strategy is S, then the specific formula is: S=αr+βH+γP, where α, β, γ are weight coefficients.

[0051] Optionally, the S6 specifically includes:

[0052] S61: According to the employee's usage habits and work scenarios, select the corresponding platform to push the behavior adjustment strategy. The specific platform includes the smart watch, smart bracelet or office terminal worn by the employee;

[0053] S62: Push the personalized behavior adjustment strategy generated in S5 to the employee's device via a wireless network, and the push content includes task adjustment, rest arrangement, and emotion regulation suggestions;

[0054] S63: When the employee's wearable device receives the pushed behavior adjustment policy, the device reminds the employee to execute the policy through vibration, sound or screen prompt;

[0055] S64: When an employee is in an office scene, a pop-up window or task bar notification will be generated on the office terminal interface to prompt the employee to make corresponding adjustments. The specific reminder information will be presented in the form of text, icons or animations.

[0056] Optionally, the S7 specifically includes:

[0057] S71: The physiological data and emotional data of the employees after they implement the behavior adjustment strategy recorded in S6 are transmitted to the stress perception model as feedback data, and the feedback data is normalized to ensure consistency with the previous standardized data;

[0058] S72: Compare the feedback data with the prediction result of the pressure perception model constructed in S3, and calculate the prediction residual ∈ of the model, the formula is: ∈=y true -f(X * ), where y true is the actual pressure value in the feedback data, f(X * ) is the predicted value of the model, ∈ is the residual;

[0059] S73: The Bayesian optimization algorithm is used to optimize the hyperparameters of the Gaussian process. The Bayesian optimization constructs a proxy model based on the Gaussian process on the objective function and selects the hyperparameters using the expected improvement criterion, so that the optimal value of the objective function is gradually approached.

[0060] S75: Retrain the Gaussian process regression model using the optimized hyperparameters.

[0061] Beneficial effects of the present invention:

[0062] The present invention effectively identifies and evaluates employees' stress levels by integrating real-time physiological data monitoring and emotional state analysis. It uses a Gaussian process regression model combined with personalized stress thresholds to accurately adjust the stress level classification according to the specific circumstances of each employee, thereby providing a more accurate personalized stress management strategy. This method not only improves the accuracy of stress state identification, but also enhances the practicality and effectiveness of stress management solutions.

[0063] The present invention realizes the real-time push and execution monitoring of behavior adjustment strategies through the dynamic feedback mechanism of smart terminals and wearable devices, and uses the Bayesian optimization algorithm to adaptively optimize the kernel function and hyperparameters of the stress perception model, ensuring that the model maintains optimal performance in a continuously changing working environment. This adaptive adjustment mechanism significantly improves the response speed and execution efficiency of stress management measures, helps promote the mental health and job satisfaction of employees, and ultimately achieves the goal of improving overall work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0065] Figure 1 A schematic diagram of a method for pressure perception and behavior adjustment according to an embodiment of the present invention;

[0066] Figure 2 The figure is a schematic diagram of a process of forming a standardized data set according to an embodiment of the present invention. DETAILED DESCRIPTION

[0067] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0068] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0069] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0070] like Figure 1-Figure 2 As shown in the figure, the stress perception and behavior adjustment method for grassroots bank employees includes the following steps:

[0071] S1: Deploy a sensor network in the working environment of grassroots bank employees to monitor their physiological data, including their heart rate, galvanic skin response, and respiratory rate, and use facial expression recognition algorithms to collect data on their emotional states;

[0072] S2: Normalize the physiological data and emotional data collected in S1, use the noise filtering algorithm to remove abnormal data and noise in the collection process, and extract features from various types of data to form a standardized data set;

[0073] S3: Based on the standardized data set generated in S2, the Gaussian process regression (GPR) algorithm is used to build a stress perception model. The input of the stress perception model includes the employee's physiological data, emotional data, and work task load characteristics, and the output is the stress state score. The Gaussian process regression algorithm quantifies the similarity between different features by defining a kernel function. The model can not only predict the stress score, but also provide uncertainty measurement for each prediction to help further optimize the decision.

[0074] S4: Based on the stress score output by the stress perception model in S3, the employee's stress status is divided into three levels: low, medium, and high. The classification results are adjusted in combination with the employee's personalized stress threshold to ensure the accuracy and personalization of stress status classification;

[0075] S5: Based on the stress state level in S4, combined with the employee's historical behavior data, mental health records and personal preferences, develop personalized behavior adjustment strategies, including task adjustment, rest suggestions and emotional relief measures;

[0076] S6: Push the behavior adjustment strategy generated in S5 to employees through their wearable devices or office terminals, and provide execution reminders on the wearable devices or work interfaces to ensure that employees make adjustments according to the strategy;

[0077] S7: Monitor the physiological data and emotional state after the execution of the strategy in S6, transmit the collected feedback data back to the pressure perception model constructed in S3, and adjust the hyperparameters of Gaussian process regression through the optimization algorithm to achieve adaptive optimization of the model.

[0078] S1 specifically includes:

[0079] S11: Sensor network layout. In the working area of ​​grassroots bank employees, a distributed sensor layout scheme is adopted. Heart rate sensors, galvanic skin response sensors and respiratory rate sensors are evenly installed on employees' work desks, chair backs and office equipment. Heart rate sensors contact employees through wearable devices such as smart watches or chest straps; galvanic skin response sensors monitor through the contact surface of work chairs; respiratory rate sensors obtain employees' respiratory rate signals through infrared sensors installed near employees' seats; sensor networks are used to collect employees' physiological data in real time. Heart rate sensors obtain employees' heart rate data through photoelectric volume pulse wave (PPG) technology, galvanic skin response sensors detect employees' psychological tension through changes in skin conductivity, and respiratory rate sensors capture changes in employees' chest or abdomen breathing rhythm through infrared ranging principles. All data are transmitted to the data processing center through wireless transmission modules;

[0080] S12: Emotional state data collection: a facial expression recognition camera is set up in the work environment. The camera uses a convolutional neural network (CNN) algorithm for real-time facial expression recognition to capture employees' facial expression changes, including the expression feature points of eyebrows, mouth corners, and eye areas, and maps the feature points to emotional states, including pleasure, tension, and anxiety;

[0081] S13: Data integration and transmission: integrating heart rate, galvanic skin response, respiratory rate data and emotional state data into the central data processing unit through the wireless sensor network (WSN) to form a complete employee physiological and emotional state data set for subsequent stress analysis and behavioral adjustment.

[0082] S12 specifically includes:

[0083] S121: Feature point capture: The facial image of the employee is captured in real time through the facial expression recognition camera, and the image is input into the convolutional neural network for convolution operation to extract the expression feature points of the employee's eyebrows, mouth corners and eye areas; the specific convolution operation is defined as:

[0084] Among them, g(x, y) is the convolution output, I(i, j) is the pixel value of the input image at position i, j, K(xi, yj) is the weight value of the convolution kernel at position xi, yj, m and n are the height and width of the convolution window respectively. Through this operation, the expression feature points of the employee's eyebrows, mouth corners and eye areas are extracted;

[0085] S122: Feature map generation, downsampling the convolutional feature map through the pooling layer, the pooling operation is: Among them, p(x, y) is the pooling output, I(i, j) is the pixel value in the pooling window, Represents average pooling, where m and n are the height and width of the pooling window, respectively. Pooling is used to reduce the size of the feature map and retain feature information;

[0086] S123: Feature vector generation and classification. The pooled feature map is passed to the fully connected layer to generate the feature vector F of the facial expression, which is expressed as: F = [f1, f2, ..., f k ], where f i represents the i-th feature point, k is the dimension of the feature vector;

[0087] S124: Emotional state mapping, mapping is performed by calculating the Euclidean distance between the facial expression feature vector f and the preset standard emotion feature vector E. The calculation formula of the Euclidean distance is:

[0088] Where D(F, E) is the Euclidean distance between the feature vector F and the standard emotion vector E, f i is the i-th feature point in the facial expression feature vector, e i is the ith feature point in the standard emotion vector, k is the dimension of the feature vector, and by comparing the Euclidean distance values, the standard emotion feature vector E corresponding to the smallest distance is selected and mapped to a specific emotional state, including pleasure, tension or anxiety. Through the Euclidean distance calculation, the feature points of facial expressions can be directly mapped to specific emotional states, providing accurate emotional input for the stress perception model.

[0089] S2 specifically includes:

[0090] S21: Data normalization processing, normalizing the collected physiological data (heart rate, skin galvanic response, breathing rate) and emotional data to ensure that data of different dimensions have consistent scales; the normalization process adopts Min-Max Normalization, and the normalization formula is: Among them, x is the original data value, x min is the minimum value of the data, x max is the maximum value of the data, and x′ is the normalized data. Through this normalization process, all data are converted to the [0,1] interval for easy subsequent processing;

[0091] S22: Noise filtering: Apply a noise filtering algorithm to the normalized data to remove abnormal data and noise. Specifically, a weighted moving average filter is used to smooth the data and remove sudden outliers. The formula is: Among them, y t is the smoothed data value at time t, x t-i is the original data, w is the size of the sliding window, α i is the weighting coefficient; the data smoothed by this algorithm can better reflect the real physiological and emotional state of employees;

[0092] S23: Abnormal data screening, using the triple standard deviation method (3σ method) to screen outliers in the data, specifically: Among them, z is the standard score, x is the data point, μ is the mean of the data, σ is the standard deviation of the data, and any value exceeding 3 times the standard deviation will be considered an outlier and filtered;

[0093] S24: Extract features from the normalized and noise-filtered data to extract the amplitude of heart rate changes, the amplitude of skin electrical response fluctuations, and the period of respiratory rate fluctuations. Reduce the dimension of the multi-dimensional data through principal component analysis (PCA) to extract representative features and form a standardized data set. The PCA dimensionality reduction formula is: Z=XW, where Z is the eigenvector after dimensionality reduction, X is the original data matrix, and W is the principal component matrix. This step ensures that the data set can accurately reflect the stress and emotional state of the employees. Through the above steps, the collected physiological and emotional data can be effectively normalized, noise can be removed, and key features can be extracted to provide high-quality input data for the stress perception model.

[0094] S3 uses Gaussian process regression algorithm to build a pressure perception model, which includes:

[0095] S31: Training data preparation. The standardized data set generated in S2 is used as input. The standardized data includes the employee's physiological data (heart rate, skin galvanic response, breathing rate) and emotional data. The input standardized data set is set to X = (x1, x2, ..., x n ), where x i is the feature vector at the i-th time point, containing the features of various types of data;

[0096] S32: The radial basis function is selected as the kernel function of the Gaussian process, and the expression is:

[0097] Among them, k(x i , x j ) is x i and x jThe kernel function value between ∥x i -x j ∥ is the Euclidean distance between two feature vectors, σ is the scale parameter of the kernel function, and this kernel function is used to quantify the similarity between different data points;

[0098] S33: Based on the kernel function and training data, a Gaussian process regression model is constructed. It is assumed that there is a Gaussian process relationship between the features in the standardized data set and the pressure state, and the output f(X) of the model represents the pressure state; then the prior distribution of the Gaussian process is expressed as: Where f(X) is the predicted value of the pressure state, represents a Gaussian process, K(X, X) is the kernel matrix, which is composed of the kernel function values ​​of all training data;

[0099] S34: Optimize the model through maximum likelihood estimation, optimize the model parameters, including the kernel function parameter σ, and use the model to predict the new pressure state value; the prediction formula is:

[0100] f * =K(X * ,X)K(X,X) -1 f(X), where f * For the new data point X * The predicted value of the pressure state, K(X * , X) is the kernel function value between the new data point and the training data, (X, X) -1 is the inverse of the training data kernel matrix;

[0101] S35: The final pressure perception model is constructed based on Gaussian process regression, and the complete expression is:

[0102] f(X * )=μ(X * )+K(X * ,X)k(X,X) -1 (y-μ(X)), where f(X * ) is the predicted pressure value of the new data point, μ(X * ) is the mean function, K(X * ,X) is the kernel function, y is the actual stress value of the training data; through the above steps, a model that can accurately predict the stress state of employees is constructed using the Gaussian process regression algorithm, providing technical support for stress perception and behavior adjustment.

[0103] In S35, the pressure state score is output according to the pressure prediction value calculated by the pressure perception model, including the pressure prediction value normalization, by outputting the pressure prediction value f(X * ) to ensure that the pressure prediction value is within the preset range. The standardization formula is: Among them, f′(X * ) is the normalized pressure prediction value, f min and f max are the minimum and maximum values ​​in the historical pressure prediction values ​​respectively; the standardized value f′(X * ) is limited to the range of 0 to 100 to facilitate the stress state scoring process in subsequent steps.

[0104] S4 specifically includes:

[0105] S41: Classification of pressure state based on the normalized pressure prediction value f′(X * ), initially divide the stress status of employees into three levels: low, medium and high; the classification criteria are:

[0106]

[0107] Among them, L is the initial pressure level, f′(X * ) is the normalized pressure prediction value, ranging from 0 to 100, used to reflect different pressure levels;

[0108] S42: Personalized stress threshold setting. Combined with each employee's personalized stress tolerance, personalized thresholds T1 and T2 are set according to the employee's historical stress data, where T1 represents the critical value of low stress and medium stress, and T2 represents the critical value of medium stress and high stress. The calculation formula of the personalized threshold is: T i =μ i +βσ i , where T i is the personalized stress threshold of the i-th employee, μ i is the employee's historical stress average, σ i is the standard deviation, β is the adjustment factor, which is set according to the individual needs of employees. This threshold is used to distinguish the stress tolerance of employees;

[0109] S43: Adjust the stress state by adjusting the stress state classification result according to the personalized stress thresholds T1 and T2; the adjustment rule expression is:

[0110]

[0111] Among them, L′ is the stress level adjusted according to the personalized threshold. Through this adjustment rule, it is ensured that the division of stress state can be more in line with the individual differences of employees;

[0112] S44: Output the stress state, and output the final stress state L′ as three levels: low, medium, and high, for use in the formulation of behavioral adjustment strategies in subsequent steps; through the above steps, combined with the dual adjustment of standardized stress prediction values ​​and personalized thresholds, accurate classification of employee stress states is achieved, ensuring the personalization and flexibility of stress perception.

[0113] S5 specifically includes:

[0114] S51: Historical behavior data analysis, obtain employees' historical behavior data, including employees' work performance, rest frequency and emotional changes under different stress states, and use correlation analysis to calculate the relationship between employee behavior patterns and stress states. The correlation calculation formula is: Among them, x i is the pressure state data, y i is the behavioral data, and r is the correlation coefficient between the two, which is used to determine the effectiveness of specific behaviors under different stress conditions;

[0115] S52: Analyze employees’ mental health records, assess their mental health status, including historical records of anxiety and depression, score them using scales (such as anxiety scale, depression scale), and determine their mental health threshold H. The mental health scoring formula is: Among them, H is the mental health score, S is the total score of the scale, and N is the number of scale items;

[0116] S53: According to the personal preferences of the employees, including rest time, workload preference and task type, a personalized parameter set P is set, which is expressed as: P = (p1, p2, ..., p n ), where p i Representing different preference dimensions, personal preference data is obtained through employee self-assessment or historical data extraction, and is used for the allocation of personalized tasks in behavior adjustment strategies;

[0117] S54: Based on the analysis results of the above steps, generate personalized behavior adjustment strategies, which include task load adjustment, short-term rest arrangements and emotion regulation suggestions. Suppose the behavior adjustment strategy is S, then the specific formula is: S = αr + βH + γP, where α, β, γ are weight coefficients, which are dynamically adjusted according to the employee's stress level L′ to generate personalized strategies that comprehensively consider the employee's behavior, mental health and preferences; through this process, combined with the stress state level, historical behavior data, mental health records and personal preferences, formulate personalized behavior adjustment strategies to ensure that employees can maintain mental health while working efficiently.

[0118] S6 specifically includes:

[0119] S61: According to the employee's usage habits and work scenarios, select the corresponding platform to push the behavior adjustment strategy. The specific platform includes the smart watch, smart bracelet or office terminal (such as computer, tablet) worn by the employee. The device is connected to the data processing center through a wireless network (Wi-Fi, Bluetooth, etc.);

[0120] S62: Push the personalized behavior adjustment strategy generated in S5 to the employee's device via the wireless network. The push content includes task adjustment, rest arrangement and emotion regulation suggestions. The push mechanism ensures timeliness. The employee status is monitored in real time through the background to ensure dynamic update and timely push of the strategy.

[0121] S63: When the employee's wearable device receives the pushed behavior adjustment strategy, the device reminds the employee to execute the strategy through vibration, sound or screen prompts. The device can also display detailed behavior adjustment suggestions through the interface, reminding the employee such as "take a 5-minute break" or "adjust the work rhythm".

[0122] S64: When an employee is in an office scene, a pop-up window or taskbar notification will be generated on the interface of the office terminal (such as a computer or tablet) to prompt the employee to make corresponding adjustments. The specific reminder information will be presented in the form of text, icons or animations to ensure that the employee can notice and execute the adjustments in time. Through the above steps, the behavior adjustment strategy can be effectively pushed through the employee's wearable device or office terminal, and through appropriate reminders, it can be ensured that the employee executes it in time, thereby achieving effective management of employee work pressure.

[0123] S7 specifically includes:

[0124] S71: The physiological data and emotional data of the employees after they implement the behavior adjustment strategy recorded in S6 are transmitted to the stress perception model as feedback data, and the feedback data is normalized to ensure consistency with the previous standardized data;

[0125] S72: Compare the feedback data with the prediction result of the pressure perception model constructed in S3, and calculate the prediction residual ∈ of the model, the formula is: ∈=y true -f(X * ), where y true is the actual pressure value in the feedback data, f(X * ) is the predicted value of the model, ∈ is the residual;

[0126] S73: The Bayesian optimization algorithm is used to optimize the hyperparameters of the Gaussian process. The Bayesian optimization constructs a proxy model based on the Gaussian process on the objective function and selects the hyperparameters using the expected improvement criterion so that the optimal value of the objective function is gradually approached. The formula for expected improvement is:

[0127] Where μ(x) is the mean prediction of the current proxy model, σ(x) is the uncertainty, Φ(Z) and are the cumulative distribution function and probability density function of the normal distribution, f(x best ) is the current optimal solution. By maximizing the expected improvement function, the best hyperparameters are selected for update;

[0128] S75: Use the optimized hyperparameters to retrain the Gaussian process regression model and update the model's predictive ability so that it can better adapt to new feedback data and improve the accuracy of pressure prediction.

[0129] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0130] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. The method of stress perception and behavior adjustment for grassroots bank employees is characterized by: The following steps are involved: S1: Deploy a sensor network in the working environment of grassroots bank employees to monitor their physiological data, including their heart rate, galvanic skin response, and respiratory rate, and use facial expression recognition algorithms to collect data on their emotional states; S2: Normalize the physiological data and emotional data collected in S1, use the noise filtering algorithm to remove abnormal data and noise in the collection process, and extract features from various types of data to form a standardized data set; S3: Based on the standardized data set generated in S2, a Gaussian process regression algorithm is used to construct a stress perception model, wherein the input of the stress perception model includes the employee's physiological data, emotional data, and work task load characteristics, and the output is a stress state score; S4: Based on the output stress score of the stress perception model in S3, the employee's stress status is divided into three levels: low, medium, and high, and the classification results are adjusted in combination with the employee's personalized stress threshold; S5: Based on the stress state level in S4, combined with the employee's historical behavior data, mental health records and personal preferences, formulate a personalized behavior adjustment strategy, which includes task adjustment, rest suggestions and emotional relief measures; S6: Push the behavior adjustment strategy generated in S5 to employees through their wearable devices or office terminals, and provide execution reminders on the wearable devices or work interfaces; S7: Monitor the physiological data and emotional state after the execution of the strategy in S6, transmit the collected feedback data back to the pressure perception model constructed in S3, and adjust the hyperparameters of Gaussian process regression through the optimization algorithm to achieve adaptive optimization of the model.

2. The method for stress perception and behavior adjustment of grassroots bank employees according to claim 1 is characterized in that: The S1 specifically includes: S11: Sensor network layout. In the working area of ​​the grassroots bank employees, a distributed sensor layout scheme is adopted to evenly install heart rate sensors, skin galvanic response sensors and respiratory rate sensors on the employees' work desks, chair backs and office equipment. The heart rate sensors are in contact with employees through wearable devices such as smart watches or chest straps; the skin galvanic response sensors are monitored through the contact surface of the work chair; the respiratory rate sensors obtain the employees' respiratory rate signals through infrared sensors installed near the employees' seats; S12: Emotional state data collection: a facial expression recognition camera is set up in the work environment. The camera uses a convolutional neural network algorithm to perform real-time facial expression recognition to capture the facial expression changes of employees, including the expression feature points of eyebrows, mouth corners, and eye areas, and maps the feature points to emotional states, including pleasure, tension, and anxiety; S13: Data integration and transmission: integrating the heart rate, galvanic skin response, respiratory rate data and emotional state data into the central data processing unit through the wireless sensor network to form the employee's physiological and emotional state data set.

3. The method for stress perception and behavior adjustment of grassroots bank employees according to claim 2 is characterized in that: The S12 specifically includes: S121: Feature point capture: The facial image of the employee is captured in real time through the facial expression recognition camera, and the image is input into the convolutional neural network for convolution operation to extract the facial feature points of the employee's eyebrows, mouth corners and eye areas; S122: Feature map generation, downsampling the convolutional feature map through the pooling layer to retain feature information; S123: Feature vector generation and classification, the pooled feature map is passed to the fully connected layer to generate the feature vector F of the facial expression; S124: Emotional state mapping, mapping is performed by calculating the Euclidean distance between the facial expression feature vector F and the preset standard emotion feature vector E, and by comparing the Euclidean distance values, the standard emotion feature vector E corresponding to the smallest distance is selected and mapped to a specific emotional state, including pleasure, tension or anxiety.

4. The method for stress perception and behavior adjustment of grassroots bank employees according to claim 1 is characterized in that: The S2 specifically includes: S21: Data normalization processing: normalizing the collected physiological data and emotional data to ensure that data of different dimensions have consistent scales; S22: Noise filtering: applying a noise filtering algorithm to the normalized data to remove abnormal data and noise. Specifically, a weighted moving average filter is used to smooth the data to remove sudden outliers. S23: Abnormal data screening, using the triple standard deviation method to screen outliers in the data; S24: Perform feature extraction on the normalized and noise-filtered data to extract the amplitude of heart rate changes, the amplitude of skin electrical response fluctuations, and the fluctuation period of respiratory rate. Perform dimensionality reduction on the multi-dimensional data through principal component analysis to extract features and form a standardized data set.

5. The method for stress perception and behavior adjustment of grassroots bank employees according to claim 1 is characterized in that: The S3 specifically includes: S31: Training data preparation, taking the standardized data set generated in S2 as input, the standardized data includes the physiological data and emotional data of the employee, and setting the input standardized data set to X=(x1, x2, ..., x n ), where x i is the feature vector at the i-th time point; S32: The radial basis function is selected as the kernel function of the Gaussian process, and the expression is: Among them, k(x i , x j ) is x i and x j The kernel function value between ∥x i -x j ∥ is the Euclidean distance between two eigenvectors, σ is the scale parameter of the kernel function; S33: Based on the kernel function and training data, a Gaussian process regression model is constructed. It is assumed that there is a Gaussian process relationship between the features in the standardized data set and the pressure state, and the output f(X) of the model represents the pressure state; then the prior distribution of the Gaussian process is expressed as: Where f(X) is the predicted value of the pressure state, represents a Gaussian process, K(X, X) is the kernel matrix; S34: Optimize the model through maximum likelihood estimation, optimize the model parameters, including the kernel function parameter σ, and use the model to predict the new pressure state value; the prediction formula is: f * =K(X * ,X)K(X,X) -1 f(X), where f * For the new data point X * The predicted value of the pressure state, K(X * , X) is the kernel function value between the new data point and the training data, (X, X) -1 is the inverse of the training data kernel matrix; S35: The final pressure perception model is constructed based on Gaussian process regression, and the complete expression is: f(X * )=μ(X * )+K(X * ,X)K(X,X) -1 (y-μ(X)), where f(X * ) is the predicted pressure value of the new data point, μ(X * ) is the mean function, K(X * ,X) is the kernel function, and y is the actual pressure value of the training data.

6. The method for stress perception and behavior adjustment of grassroots bank employees according to claim 5 is characterized in that: In the step S35, the pressure state score is output according to the pressure prediction value calculated by the pressure perception model, including the pressure prediction value normalization, by * ) to ensure that the pressure prediction value is within the preset range. The standardization formula is: Among them, f′(X * ) is the normalized pressure prediction value, f min and f max are the minimum and maximum values ​​in the historical pressure prediction values ​​respectively; the standardized value f′(X * ) is limited to the range of 0 to 100.

7. The method for stress perception and behavior adjustment of grassroots bank employees according to claim 6 is characterized in that: The S4 specifically includes: S41: Classification of pressure state based on the normalized pressure prediction value f′(X * ), initially divide the stress status of employees into three levels: low, medium and high; the classification criteria are: Among them, L is the initial pressure level, f′(X * ) is the normalized pressure prediction value, ranging from 0 to 100, used to reflect different pressure levels; S42: personalized stress threshold setting, combining each employee's personalized stress tolerance and setting personalized thresholds T1 and T2 according to the employee's historical stress data, where T1 represents the critical value of low stress and medium stress, and T2 represents the critical value of medium stress and high stress; S43: Adjust the stress state by adjusting the stress state classification result according to the personalized stress thresholds T1 and T2; the adjustment rule expression is: Among them, L′ is the pressure level adjusted according to the personalized threshold; S44: Output the pressure state, and output the final pressure state L′ as three levels: low, medium, and high.

8. The method for stress perception and behavior adjustment of grassroots bank employees according to claim 1 is characterized in that: The S5 specifically includes: S51: Historical behavior data analysis, obtaining employees’ historical behavior data, including employees’ work performance, rest frequency, and emotional changes under different stress states, and using correlation analysis to calculate the relationship between employee behavior patterns and stress states; S52: Analyze employees’ mental health records, assess their mental health status, including their history of anxiety and depression, score them using a scale, and determine their mental health threshold H; S53: setting a personalized parameter set P according to the employee's personal preferences, including rest time, workload preference, and task type; S54: Based on the analysis results of the above steps, generate a personalized behavior adjustment strategy, which includes task load adjustment, short-term rest arrangement and emotion regulation suggestions. Suppose the behavior adjustment strategy is S, then the specific formula is: ∈=αr+βH+γP, where α, β, γ are weight coefficients.

9. The method for stress perception and behavior adjustment of grassroots bank employees according to claim 1 is characterized in that: The S6 specifically includes: S61: According to the employee's usage habits and work scenarios, select the corresponding platform to push the behavior adjustment strategy. The specific platform includes the smart watch, smart bracelet or office terminal worn by the employee; S62: Push the personalized behavior adjustment strategy generated in S5 to the employee's device via a wireless network, and the push content includes task adjustment, rest arrangement, and emotion regulation suggestions; S63: When the employee's wearable device receives the pushed behavior adjustment policy, the device reminds the employee to execute the policy through vibration, sound or screen prompt; S64: When an employee is in an office scene, a pop-up window or task bar notification will be generated on the office terminal interface to prompt the employee to make corresponding adjustments. The specific reminder information will be presented in the form of text, icons or animations.

10. The method for stress perception and behavior adjustment of grassroots bank employees according to claim 1, characterized in that: The S7 specifically includes: S71: The physiological data and emotional data of the employees after they implement the behavior adjustment strategy recorded in S6 are transmitted to the stress perception model as feedback data, and the feedback data is normalized to ensure consistency with the previous standardized data; S72: Compare the feedback data with the prediction result of the pressure perception model constructed in S3, and calculate the prediction residual ∈ of the model, the formula is: ∈=y true -f(X * ), where y true is the actual pressure value in the feedback data, f(X * ) is the predicted value of the model, ∈ is the residual; S73: The Bayesian optimization algorithm is used to optimize the hyperparameters of the Gaussian process. The Bayesian optimization constructs a proxy model based on the Gaussian process on the objective function and selects the hyperparameters using the expected improvement criterion, so that the optimal value of the objective function is gradually approached. S75: Retrain the Gaussian process regression model using the optimized hyperparameters.

Images