A behavior evaluation system and method based on cloud computing

Abstract

The present invention discloses a behavior evaluation system and method based on cloud computing, which relates to the technical field of behavior evaluation data analysis. The system includes: a data monitoring and processing module, a behavior data analysis module, and a behavior evaluation and feedback module. The data monitoring and processing module performs real-time monitoring on an evaluation object by setting a non-contact acquisition sensor in the current activity environment of the evaluation object, and is also used to collect and obtain the situation information of the evaluation object. The behavior data analysis module trains a situation-behavior reference deep learning model, calculates the reference deviation of the evaluation parameters of each behavior state of the evaluation object, and is also used to construct a behavior state evaluation knowledge graph and update and iterate the historical data of the reference deviation of the evaluation parameters of the behavior state. The behavior evaluation and feedback module performs real-time update and feedback of the evaluation results on the evaluation object according to the behavior state evaluation knowledge graph data.

Description

Technical Field

[0001] The present invention relates to the technical field of behavioral assessment data analysis, and particularly to a cloud computing-based behavioral assessment system and method. Background Art

[0002] In the field of children's behavioral state assessment, accurate and comprehensive assessment is crucial for formulating personalized education and training programs. However, the following technical defects exist in the commonly used behavioral state assessment methods nowadays: on the one hand, when collecting data on the assessment object, methods such as questionnaire interviews and specific situation observations are generally used, which are greatly affected by the subjective factors of the assessor and the assessment results have large errors; on the other hand, relying on the traditional interactive assessment method has a relatively direct contact with the assessment object, which has a subjective interference on the behavioral state of the assessment object and cannot reflect the most real behavioral state feedback of the assessment object; in addition, using periodic assessment on the assessment object cannot intuitively show the changes in the behavioral state of the assessment object at different times.

[0003] Therefore, a cloud computing-based behavioral assessment system and method are needed to solve the above technical problems, realize real-time and non-contact collection of the behavioral state of the assessment object, and combine intelligent analysis technology to provide more objective and accurate assessment results. Summary of the Invention

[0004] The purpose of the present invention is to provide a cloud computing-based behavioral assessment system and method to solve the problems raised in the prior art.

[0005] To achieve the above purpose, the present invention provides the following technical solutions:

[0006] A cloud computing-based behavioral assessment method, the method is divided into the following steps:

[0007] Step S100: Obtain the behavioral state assessment parameters and context parameter information of the assessment object, mark the reference values for the behavioral state assessment parameters of the assessment object in different context parameters, and train the context-behavior reference deep learning model;

[0008] Step S200: Obtain the behavioral analysis assessment dimension information of the current assessment object, mark the weight coefficients of each behavioral state parameter in different assessment dimensions, and construct a behavioral state assessment knowledge graph for the current assessment object;

[0009] Step S300: Set non-contact collection sensors in the activity environment of the current assessment object, monitor the assessment object in real time, obtain the real-time behavioral state assessment parameter data of the assessment object, and calculate the reference deviation of each behavioral state assessment parameter of the assessment object;

[0010] Step S400: Analyze the deviation synchronization coefficient among the behavior state evaluation parameters through cloud computing, and then update the node data of the behavior state evaluation knowledge graph;

[0011] Step S500: Conduct real-time behavior evaluation on the evaluation object according to the behavior state evaluation knowledge graph, and draw a radar chart based on the historical evaluation result data divided by each time period as the evaluation result of the current evaluation object for each time period for feedback.

[0012] In the above technical solution, the step S100 includes the following steps:

[0013] Step S101: Obtain the behavior state evaluation parameters and context parameter information of the evaluation object; the behavior state evaluation parameters refer to the parameters that measure the behavior characteristics of the evaluation object and affect the behavior evaluation result of the evaluation object; the context parameters refer to the parameters that reflect the context characteristics of the evaluation object;

[0014] Step S102: Obtain the historical context data of the evaluation object, and perform reference value annotation on the behavior state evaluation parameters in the context through manual annotation;

[0015] For any combination of context parameters, store the annotation results as a two-dimensional array composed of a context parameter vector and a reference value vector of behavior state evaluation parameters;

[0016] Step S103: Use the annotated data to train the context-behavior reference deep learning model; the input of the context-behavior reference deep learning model is the context parameter vector, and the output of the context-behavior reference deep learning model is the reference value vector of behavior state evaluation parameters under the input context parameter vector;

[0017] Construct a dataset through manual annotation to train the deep learning model, construct a context-behavior relationship chain, reduce the dependence on experts in complex contexts, improve the standardization degree of behavior state evaluation, and ensure the generalization ability of the algorithm.

[0018] In the above technical solution, the step S200 includes the following steps:

[0019] Step S201: Obtain the behavior analysis evaluation dimension information of the current evaluation object, and annotate the weight coefficients of each behavior state parameter in different evaluation dimensions;

[0020] Step S202: Construct a behavior state evaluation knowledge graph for the current evaluation object; the behavior state evaluation knowledge graph includes behavior state nodes and evaluation dimension nodes, and the edge weight between the behavior state nodes and the evaluation dimension nodes is the weight coefficient of each behavior state evaluation parameter in the corresponding evaluation dimension of each evaluation dimension node; among them, each behavior state node corresponds to each behavior state evaluation parameter one by one, and each evaluation dimension node corresponds to each evaluation dimension one by one;

[0021] Construct a behavior state evaluation knowledge graph, analyze the implicit associations among different behavior state evaluation parameters, avoid the one-sidedness of traditional evaluation methods, and effectively enhance the hierarchy and systematicness of the system evaluation results.

[0022] In the above technical solution, the step S300 includes the following steps:

[0023] Step S301: Set a contactless acquisition sensor in the current evaluation object's activity environment to monitor the evaluation object and its surrounding situation in real time;

[0024] Step S302: Perform feature extraction, data dimensionality reduction, and normalization on the data of each contactless acquisition sensor, convert the acquired data into one-dimensional data, and use it as the measured value of the behavior state evaluation parameter of the evaluation object;

[0025] Step S303: Calculate the reference deviation of each behavior state evaluation parameter of the evaluation object; use the context-behavior reference deep learning model, take the context parameters of the evaluation object's surrounding situation as input, obtain the reference value of the behavior state evaluation parameter of the evaluation object, and calculate the difference between the measured value and the reference value of each behavior state evaluation parameter of the evaluation object as the reference deviation of each behavior state evaluation parameter of the evaluation object;

[0026] Using contactless acquisition sensor data to monitor the evaluation object in real time effectively reduces the interference of data acquisition on the behavior state of the evaluation object, makes the acquired data more authentic and scientific, and further ensures the accuracy of the system evaluation results.

[0027] In the above technical solution, the step S400 includes the following content:

[0028] Upload the historical data of the reference deviation of each behavior state evaluation parameter of the evaluation object to the cloud for storage, and simultaneously analyze the deviation synchronization coefficient among each behavior state evaluation parameter through cloud computing;

[0029] For any behavior state evaluation parameters x and y, the calculation formula for the deviation synchronization coefficient of y with respect to x is as follows:

[0030] k_(y→x)=(1 / Nd)×sum(δx / (δy+ε));

[0031] Among them, \(k_{(y\rightarrow x)}\) is the deviation synchronization coefficient of behavior state evaluation parameter \(y\) with respect to \(x\), \(N_d\) is the number of reference deviation historical data of each behavior state evaluation parameter of the evaluation object stored in the cloud, \(\delta x\) is the reference deviation of behavior state evaluation parameter \(x\), \(\delta y\) is the reference deviation of behavior state evaluation parameter \(y\), \(\varepsilon\) is a constant, and \(\sum(\delta x / (\delta y + \varepsilon))\) is the numerical sum of the ratio of reference deviations \(\delta x / (\delta y + \varepsilon)\) of behavior state evaluation parameters \(x\) and \(y\) in all historical data stored in the cloud;

[0032] Update the deviation synchronization coefficient between each behavior state transition parameter to the edge weight between the corresponding behavior state nodes in the behavior state evaluation knowledge graph, and perform real-time update iteration on the weights of each behavior state node in the behavior state evaluation knowledge graph;

[0033] For any behavior state evaluation parameter \(x\), the iteration formula of the corresponding behavior state node in the behavior state evaluation knowledge graph is as follows:

[0034] \(x_{update}=(1 - \alpha - \beta)\times x_{ori}+\alpha\times\delta x+\beta / (N_p)\times\sum(k_{(p\rightarrow x)}\times\delta p)\);

[0035] Among them, \(x_{update}\) is the iteration data of the behavior state node corresponding to behavior state evaluation parameter \(x\), \(x_{ori}\) is the original data of the behavior state node corresponding to behavior state evaluation parameter \(x\), \(\delta x\) is the reference deviation of behavior state evaluation parameter \(x\), \(N_p\) is the number of other behavior state evaluation parameters except behavior state evaluation parameter \(x\), \(\sum(k_{(p\rightarrow x)}\times\delta p)\) is the numerical sum of the product of the deviation synchronization coefficient of each behavior state evaluation parameter other than behavior state evaluation parameter \(x\) with respect to \(x\) and the reference deviation, \(\alpha\) is the reference deviation compensation coefficient, \(\beta\) is the deviation synchronization compensation coefficient, \(\alpha>0\), \(\beta>0\) and \(\alpha+\beta<1\);

[0036] Through deviation synchronization analysis, the mutual influence between parameters is measured, breaking through the one-sidedness of the method of independently calculating the scores of each behavior parameter in the isolated evaluation mode, and improving the dynamic adaptability of the system; and by analyzing the deviation synchronization coefficient, compensation calculation is performed on each behavior state evaluation parameter, and then the behavior state evaluation knowledge graph is updated in real time, further avoiding the defect that the traditional method using static reports cannot effectively reflect the change trend of the behavior state of the evaluation object.

[0037] In the above technical solution, the step S500 includes the following content:

[0038] Obtain the behavior state evaluation knowledge graph data, extract the latest iteration data of each behavior state node adjacent to each evaluation dimension node, and calculate the node weights of each evaluation dimension node by weighting with the edge weights between the evaluation dimension nodes and the adjacent behavior state nodes as coefficients. Use the calculated node weights as the evaluation results of each evaluation dimension and upload them to the cloud for storage.

[0039] Set the behavior state evaluation period T, divide time periods by T, divide the historical data of the node weights of each evaluation dimension node, extract the minimum and maximum values of the node weights of each evaluation dimension node in the historical data within each time period through cloud computing, and draw a radar chart based on the extracted data at the end time point of each time period. Use the radar chart as the behavior state evaluation results of the current evaluation object for each time period and provide feedback.

[0040] By using the radar chart to provide feedback on the evaluation results within each time period, the user interaction experience is further improved. Through data screening, the evaluation result intervals within each time period are used to replace the single evaluation results of the static report in the traditional method, which more intuitively reflects the fluctuations and changes in the behavior state of the evaluation object within each time period and ensures the accuracy of subsequent education and training strategies.

[0041] Apply a behavior evaluation system based on cloud computing in the above technical solution, which includes: a data monitoring and processing module, a behavior data analysis module, and a behavior evaluation and feedback module.

[0042] The data monitoring and processing module sets non-contact acquisition sensors in the current evaluation object's activity environment to monitor the evaluation object in real time and is also used to collect and obtain the situation information of the evaluation object. The behavior data analysis module trains a situation-behavior reference deep learning model, calculates the reference deviations of the evaluation parameters of each behavior state of the evaluation object, and is also used to construct a behavior state evaluation knowledge graph and update and iterate the data according to the historical data of the reference deviations of the behavior state evaluation parameters. The behavior evaluation and feedback module updates and provides feedback on the evaluation results of the evaluation object in real time according to the behavior state evaluation knowledge graph data.

[0043] In the above technical solution, the data monitoring and processing module includes: a behavior state monitoring unit and a situation monitoring unit.

[0044] The behavior state monitoring unit sets non-contact acquisition sensors in the current evaluation object's activity environment to monitor the behavior state of the evaluation object in real time. The situation monitoring unit collects and obtains the situation information of the evaluation object through real-time monitoring.

[0045] In the above technical solution, the behavior data analysis module includes: a reference behavior analysis unit, a behavior deviation analysis unit, and a knowledge graph update unit.

[0046] The reference behavior analysis unit labels reference values for the behavior state evaluation parameters of the evaluation object in different context parameters, trains a context-behavior reference deep learning model, and then analyzes the reference values of the behavior state evaluation parameters of the evaluation object in various contexts; the behavior deviation analysis unit calculates the deviation between the monitored value and the reference value of the behavior state evaluation parameters of the evaluation object in real time, and then analyzes the deviation synchronization coefficient between the behavior state evaluation parameters; the knowledge graph update unit updates and iterates the data in the constructed behavior state evaluation knowledge graph according to the deviation synchronization coefficient between the behavior state evaluation parameters and in combination with the monitored data between the behavior state evaluation parameters.

[0047] In the above technical solution, the behavior evaluation feedback module includes: a behavior state evaluation unit and an evaluation result feedback unit;

[0048] The behavior state evaluation unit evaluates the behavior state of the evaluation object according to the data in the behavior state evaluation knowledge graph and updates the evaluation result in real time; the evaluation result feedback unit screens the evaluation results of each evaluation dimension in each time period and then draws a radar chart, which is fed back as the behavior state evaluation result of the current evaluation object in each time period.

[0049] Compared with the prior art, the beneficial effects of the present invention are:

[0050] In the present invention, a non-contact data acquisition method is adopted to monitor the evaluation object in real time, avoiding the interference of traditional wearable devices and contact access on the behavior state of the evaluation object, effectively improving the accuracy of the evaluation result, and also significantly enhancing the practicability of the system in various types of environments;

[0051] In the present invention, by constructing a behavior state evaluation knowledge graph to comprehensively analyze the behavior state of the evaluation object, and using the method of cloud computing to update and iterate the behavior state node data in the knowledge graph in real time, and then more truly and effectively feedback the evaluation results of the evaluation object in each evaluation dimension;

[0052] In the present invention, by analyzing the deviation synchronization coefficient between the behavior state evaluation parameters, the data update of each behavior state evaluation parameter is synchronized, reducing the bias of the data update of different behavior state evaluation parameters in the context where the evaluation object is located, effectively improving the comprehensiveness of the behavior state evaluation of the evaluation object, and further ensuring the accuracy and rationality of the evaluation result. Brief Description of the Drawings

[0053] Figure 1 It is a flowchart of a behavior evaluation method based on cloud computing according to the present invention;

[0054] Figure 2 It is an organizational structure diagram of a behavior evaluation system based on cloud computing according to the present invention. Detailed implementation manners

[0055] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0056] Embodiment: Please refer to Figure 1 - Figure 2 , the present invention provides the following technical solutions:

[0057] A behavior evaluation method based on cloud computing, the method is divided into the following steps:

[0058] Step S100: Obtain the behavior state evaluation parameters and context parameter information of the evaluation object, mark the reference values for the behavior state evaluation parameters of the evaluation object in different context parameters, and train the context-behavior reference deep learning model;

[0059] Step S200: Obtain the behavior analysis evaluation dimension information of the current evaluation object, mark the weight coefficients of each behavior state parameter in different evaluation dimensions, and construct a behavior state evaluation knowledge graph for the current evaluation object;

[0060] Step S300: Set a contactless acquisition sensor in the activity environment of the current evaluation object, monitor the evaluation object in real time, obtain the real-time behavior state evaluation parameter data of the evaluation object, and calculate the reference deviation of each behavior state evaluation parameter of the evaluation object;

[0061] Step S400: Analyze the deviation synchronization coefficient between each behavior state evaluation parameter through cloud computing, and then update the node data of the behavior state evaluation knowledge graph;

[0062] Step S500: Perform real-time behavior evaluation on the evaluation object according to the behavior state evaluation knowledge graph, and draw a radar chart based on the historical data of the evaluation results divided by each time period as the evaluation results of each time period of the current evaluation object for feedback.

[0063] The step S100 includes the following steps:

[0064] Step S101: Obtain the behavior state evaluation parameters and context parameter information of the evaluation object; the behavior state evaluation parameters refer to the parameters that measure the behavior characteristics of the evaluation object and affect the behavior evaluation results of the evaluation object; the context parameters refer to the parameters that reflect the context characteristics of the evaluation object;

[0065] Step S102: Obtain the historical data of the situation where the evaluation object is located, and label the reference values of the behavior state evaluation parameters in the situation through manual annotation;

[0066] For any combination of situation parameters, store the annotation results in a two-dimensional array composed of a situation parameter vector and a reference value vector of the behavior state evaluation parameter;

[0067] Step S103: Train a situation-behavior reference deep learning model using the annotated data; the input of the situation-behavior reference deep learning model is a situation parameter vector, and the output of the situation-behavior reference deep learning model is a reference value vector of the behavior state evaluation parameter under the input situation parameter vector;

[0068] In specific implementation, taking the autism evaluation of children as an example, based on the existing autism evaluation criteria, behavior state evaluation parameters such as eye contact, facial expression, language communication, body movements, and physiological signs can be selected; at the same time, situation parameters such as environmental temperature and humidity, personnel density, environmental noise intensity, situation type, and communication type can be selected as the situation parameters of the environment where the evaluation object is located;

[0069] When collecting data, cameras, WIFI_Sensing, millimeter-wave radars, and audio collectors can be used to collect raw signals, and then feature extraction and data dimensionality reduction are performed on the collected data to convert the raw signals into one-dimensional data with the required features in the raw signals; for eye contact and facial expressions, image collection is performed through a camera, and a facial feature recognition algorithm is used to analyze the facial features of the evaluation object and identify the eye orientation data of the evaluation object; for body movements, according to the collected image information, the current action posture of the evaluation object is analyzed through pose estimation; for physiological signs, etc., WIFI_Sensing is used to monitor the breathing and heart rate data of the evaluation object, and then the multi-dimensional physiological sign data is converted into one-dimensional data through time-frequency analysis and data dimensionality reduction;

[0070] Furthermore, use the processed data to construct a data set, and use a data-driven method to construct a deep learning model to achieve the adaptive mapping of multi-to-multi parameters of situation-behavior.

[0071] The step S200 includes the following steps:

[0072] Step S201: Obtain the behavior analysis evaluation dimension information of the current evaluation object, and label the weight coefficients of each behavior state parameter in different evaluation dimensions;

[0073] Step S202: Construct a behavior state evaluation knowledge graph for the current evaluation object; the behavior state evaluation knowledge graph includes behavior state nodes and evaluation dimension nodes, and the edge weights between the behavior state nodes and the evaluation dimension nodes are the weight coefficients of each behavior state evaluation parameter in the evaluation dimension corresponding to each evaluation dimension node; wherein, each behavior state node corresponds to each behavior state evaluation parameter one by one, and each evaluation dimension node corresponds to each evaluation dimension one by one;

[0074] In specific implementation, when considering the behavior state evaluation of autistic children, the combinations of relevant behavior state evaluation parameters are different in different situations. For example, in the classroom teaching situation, in the teaching situation, relevant situation parameters include personnel density, environmental temperature, situation type, communication type, etc., and it is necessary to consider the eye contact, body movements, physiological signs, etc. of the evaluation object. While in the communication situation in the family environment, the personnel density of the situation where the evaluation object is located does not need to be considered, and the facial expression of the evaluation object needs to be considered additionally.

[0075] The step S300 includes the following steps:

[0076] Step S301: Set non-contact acquisition sensors in the activity environment of the current evaluation object to monitor the evaluation object and its surrounding situation in real time;

[0077] Step S302: Perform feature extraction, data dimensionality reduction and normalization processing on the data of each non-contact acquisition sensor, and convert the acquired data into one-dimensional data as the measured values of the behavior state evaluation parameters of the evaluation object;

[0078] Step S303: Calculate the reference deviation of each behavior state evaluation parameter of the evaluation object; use the situation-behavior reference deep learning model, take the situation parameters of the situation where the evaluation object is located as the input, obtain the reference values of the behavior state evaluation parameters of the evaluation object, and calculate the difference between the measured values and the reference values of each behavior state evaluation parameter of the evaluation object as the reference deviation of each behavior state evaluation parameter of the evaluation object.

[0079] The step S400 includes the following content:

[0080] Upload the historical data of the reference deviation of each behavior state evaluation parameter of the evaluation object to the cloud for storage, and synchronously analyze the deviation synchronization coefficient between each behavior state evaluation parameter through cloud computing;

[0081] For any behavior state evaluation parameters x and y, the calculation formula of the deviation synchronization coefficient of y with respect to x is as follows:

[0082] k_(y→x)=(1 / Nd)×sum(δx / (δy+ε));

[0083] Among them, \(k_{(y\rightarrow x)}\) is the deviation synchronization coefficient of behavior state evaluation parameter \(y\) with respect to \(x\), \(N_d\) is the number of reference deviation historical data of each behavior state evaluation parameter of the evaluation object stored in the cloud, \(\delta x\) is the reference deviation of behavior state evaluation parameter \(x\), \(\delta y\) is the reference deviation of behavior state evaluation parameter \(y\), \(\varepsilon\) is a constant, and \(\sum(\delta x / (\delta y + \varepsilon))\) is the numerical sum of the ratio of reference deviations \(\delta x / (\delta y + \varepsilon)\) of behavior state evaluation parameters \(x\) and \(y\) in all historical data stored in the cloud;

[0084] In specific implementation, all behavior state evaluation parameters are initialized. Assume \(\varepsilon = 0.00001\). At the same time, each behavior state parameter is retained to two decimal places for calculation to avoid the denominator value being 0. Assume the reference deviations of behavior state evaluation parameters \(x\) and \(y\) in historical data are respectively: Data 1: \(\delta x = 0.1\), \(\delta y = 0.1\); Data 2: \(\delta x = 0.2\), \(\delta y=-0.1\); Data 3: \(\delta x=-0.1\), \(\delta y = -0.25\); Then, according to the formula, the deviation synchronization coefficient \(k_{(y\rightarrow x)}=-0.2\) of \(y\) with respect to \(x\) can be calculated;

[0085] Update the deviation synchronization coefficients between each behavior state transition parameter to the edge weights between corresponding behavior state nodes in the behavior state evaluation knowledge graph, and perform real-time update and iteration on the weights of each behavior state node in the behavior state evaluation knowledge graph;

[0086] For any behavior state evaluation parameter \(x\), the iteration formula of the corresponding behavior state node in the behavior state evaluation knowledge graph is as follows:

[0087] \(x\_update=(1 - \alpha-\beta)\times x\_ori+\alpha\times\delta x+\beta / (N_p)\times\sum(k_{(p\rightarrow x)}\times\delta p)\);

[0088] Among them, \(x\_update\) is the iteration data of the behavior state node corresponding to behavior state evaluation parameter \(x\), \(x\_ori\) is the original data of the behavior state node corresponding to behavior state evaluation parameter \(x\), \(\delta x\) is the reference deviation of behavior state evaluation parameter \(x\), \(N_p\) is the number of other behavior state evaluation parameters except behavior state evaluation parameter \(x\), \(\sum(k_{(p\rightarrow x)}\times\delta p)\) is the numerical sum of the product of the deviation synchronization coefficient of each behavior state evaluation parameter except behavior state evaluation parameter \(x\) with respect to \(x\) and the reference deviation, \(\alpha\) is the reference deviation compensation coefficient, \(\beta\) is the deviation synchronization compensation coefficient, \(\alpha>0\), \(\beta>0\) and \(\alpha+\beta<1\);

[0089] In specific implementation, assume α = β = 0.05. There are 4 behavior state evaluation parameters {a, x, y, z} involved in the current scenario. Among them, the value of a is 0.6, δa = 0.1, δx = 0.2, δy = -0.2, δz = -0.25. Then, scale the deviation synchronization coefficients between the behavior state evaluation parameters to the range of [0, 1] through numerical scaling. Assume the deviation synchronization coefficients of x, y, z with respect to a are: k_(x→a) = 0.3, k_(y→a) = -0.4, k_(z→a) = 0.8. According to the formula calculation, a_update = 0.544.

[0090] The step S500 includes the following content:

[0091] Obtain the behavior state evaluation knowledge graph data, extract the latest iteration data of each behavior state node adjacent to the nodes of each evaluation dimension, and calculate the node weights of each evaluation dimension node with the edge weights between the evaluation dimension node and the adjacent behavior state nodes as coefficients, and upload the evaluation results of each evaluation dimension to the cloud for storage;

[0092] Set the behavior state evaluation period T, divide the time period by T, divide the historical data of the node weights of each evaluation dimension node, extract the minimum and maximum values of the node weights of each evaluation dimension node in the historical data within each time period through cloud computing, and draw a radar chart based on the extracted data at the end time point of each time period as the behavior state evaluation result of the current evaluation object for each time period for feedback;

[0093] In specific implementation, update the behavior state evaluation parameters at equal time intervals or when the evaluation object's scenario switches, and simultaneously update the values of each evaluation dimension, and then calculate and update the evaluation results of each evaluation dimension node at the same time;

[0094] Furthermore, classify and store the data divided within each time period, screen out the maximum and minimum values of the evaluation results of each evaluation dimension in the cloud, and synchronously mark the maximum and minimum values of each evaluation dimension in the radar chart, so as to intuitively reflect the fluctuations of the evaluation results of the evaluation object in each evaluation dimension within each time period. Compared with the traditional method of timed access evaluation, the above scheme more intuitively and popularly reflects the behavior state changes of the evaluation object within each time period, and thus more effectively assists in education and training treatment.

[0095] A behavior assessment system based on cloud computing, the system includes: a data monitoring and processing module, a behavior data analysis module, and a behavior assessment feedback module;

[0096] The data monitoring and processing module monitors the evaluation object in real time by setting non-contact acquisition sensors in the current activity environment of the evaluation object, and is also used to collect the situation information of the evaluation object; the behavior data analysis module trains a situation-behavior reference deep learning model, calculates the reference deviation of the evaluation parameters of each behavior state of the evaluation object, and is also used to construct a behavior state evaluation knowledge graph, and updates and iterates the data according to the historical data of the reference deviation of the evaluation parameters of the behavior state; the behavior evaluation and feedback module updates and feedbacks the evaluation results of the evaluation object in real time according to the data of the behavior state evaluation knowledge graph.

[0097] The data monitoring and processing module includes: a behavior state monitoring unit and a situation monitoring unit;

[0098] The behavior state monitoring unit monitors the behavior state of the evaluation object in real time by setting non-contact acquisition sensors in the current activity environment of the evaluation object; the situation monitoring unit collects the situation information of the evaluation object through real-time monitoring.

[0099] The behavior data analysis module includes: a reference behavior analysis unit, a behavior deviation analysis unit, and a knowledge graph update unit;

[0100] The reference behavior analysis unit annotates the reference values of the evaluation parameters of the behavior state of the evaluation object in different situation parameters, trains a situation-behavior reference deep learning model, and then analyzes the reference values of the evaluation parameters of the behavior state of the evaluation object in various situations; the behavior deviation analysis unit calculates the deviation between the monitored value and the reference value of the evaluation parameters of the behavior state of the evaluation object in real time, and then analyzes the deviation synchronization coefficient between the evaluation parameters of each behavior state; the knowledge graph update unit updates and iterates the data in the constructed behavior state evaluation knowledge graph according to the deviation synchronization coefficient between the evaluation parameters of each behavior state, combined with the monitoring data between the evaluation parameters of the behavior state.

[0101] The behavior evaluation and feedback module includes: a behavior state evaluation unit and an evaluation result feedback unit;

[0102] The behavior state evaluation unit evaluates the behavior state of the evaluation object according to the data in the behavior state evaluation knowledge graph, and updates the evaluation results in real time; the evaluation result feedback unit screens the evaluation results of each evaluation dimension in each time period and then draws a radar chart, which is used as the evaluation result of the behavior state of the current evaluation object in each time period for feedback.

[0103] It is obvious to those skilled in the art that the present invention is not limited to the details of the above-described exemplary embodiments, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention. Therefore, in any aspect, the embodiments should be regarded as exemplary and non-limiting. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be embraced within the present invention. Any reference signs in the claims should not be construed as limiting the claims involved.

Claims

1. A behavior evaluation method based on cloud computing, characterized in that , The method is divided into the following steps: Step S100: Obtain the behavior state evaluation parameters and context parameter information of the evaluation object, mark the reference values of the behavior state evaluation parameters of the evaluation object in different context parameters, store the marked results in a two-dimensional array composed of a context parameter vector and a behavior state evaluation parameter reference value vector, and train the context-behavior reference deep learning model; The input of the context-behavior reference deep learning model is the context parameter vector, and the output of the context-behavior reference deep learning model is the behavior state evaluation parameter reference value vector under the input context parameter vector; Step S200: Obtain the behavior analysis evaluation dimension information of the current evaluation object, mark the weight coefficients of each behavior state parameter in different evaluation dimensions, and construct a behavior state evaluation knowledge graph for the current evaluation object; Step S300: Set a contactless acquisition sensor in the activity environment of the current evaluation object, monitor the evaluation object in real time, obtain the real-time behavior state evaluation parameter data of the evaluation object, and calculate the reference deviation of each behavior state evaluation parameter of the evaluation object; Step S400: Analyze the deviation synchronization coefficient between each behavior state evaluation parameter through cloud computing, and update and iterate the weight of each behavior state node in the behavior state evaluation knowledge graph in real time, and then update the node data of the behavior state evaluation knowledge graph; Step S500: Conduct a real-time behavior evaluation of the evaluation object according to the behavior state evaluation knowledge graph, and extract the minimum and maximum values of the weights of each evaluation dimension node in the historical data of each time period through cloud computing according to the evaluation result historical data divided by each time period, and draw a radar chart as the evaluation result of each time period of the current evaluation object for feedback.

2. The method for behavior evaluation based on cloud computing according to claim 1, wherein The following steps are included in the step S100: Step S101: Obtain the behavior state evaluation parameters and context parameter information of the evaluation object; the behavior state evaluation parameters refer to the parameters that measure the behavior characteristics of the evaluation object and affect the behavior evaluation result of the evaluation object; the context parameters refer to the parameters that reflect the context characteristics of the evaluation object; Step S102: Obtain the historical context data of the evaluation object, mark the reference values of the behavior state evaluation parameters in the context through manual marking, and use the marked data to train the context-behavior reference deep learning model.

3. The behavioral evaluation method based on cloud computing according to claim 1, characterized in that The following steps are included in the step S200: Step S201: Obtain the behavior analysis evaluation dimension information of the current evaluation object, and mark the weight coefficients of each behavior state parameter in different evaluation dimensions; Step S202: Construct a behavior state evaluation knowledge graph for the current evaluation object; the behavior state evaluation knowledge graph includes behavior state nodes and evaluation dimension nodes, and the edge weight between the behavior state nodes and the evaluation dimension nodes is the weight coefficient of each behavior state evaluation parameter in the corresponding evaluation dimension of each evaluation dimension node; among them, each behavior state node corresponds to each behavior state evaluation parameter one by one, and each evaluation dimension node corresponds to each evaluation dimension one by one.

4. A behavior evaluation method based on cloud computing according to claim 1, characterized in that, The step S300 includes the following steps: Step S301: Set up a contactless acquisition sensor in the current activity environment of the evaluation object to monitor the evaluation object and its surrounding situation in real time; Step S302: Extract features, reduce data dimensions, and normalize the data of each contactless acquisition sensor, and convert the acquired data into one-dimensional data as the measured value of the behavior state evaluation parameter of the evaluation object; Step S303: Calculate the reference deviation of each behavior state evaluation parameter of the evaluation object; use the context-behavior reference deep learning model, take the context parameters of the situation where the evaluation object is located as input, obtain the reference value of the behavior state evaluation parameter of the evaluation object, and calculate the difference between the measured value and the reference value of each behavior state evaluation parameter of the evaluation object as the reference deviation of each behavior state evaluation parameter of the evaluation object.

5. A behavior evaluation method based on cloud computing according to claim 1, characterized in that, The said step S400 includes the following contents: Upload the historical data of the reference deviation of each behavior state evaluation parameter of the evaluation object to the cloud for storage, and synchronously analyze the deviation synchronization coefficient between each behavior state evaluation parameter through cloud computing; For any behavior state evaluation parameters x and y, the calculation formula of the deviation synchronization coefficient of y with respect to x is as follows: k_(y→x)=(1 / Nd)×sum(δx / (δy+ε)); Where, k_(y→x) is the deviation synchronization coefficient of behavior state evaluation parameter y with respect to x, Nd is the number of historical data of the reference deviation of each behavior state evaluation parameter of the evaluation object stored in the cloud, δx is the reference deviation of behavior state evaluation parameter x, δy is the reference deviation of behavior state evaluation parameter y, ε is a constant, and sum(δx / (δy+ε)) is the numerical sum of the ratio δx / (δy+ε) of the reference deviations of behavior state evaluation parameters x and y in all historical data stored in the cloud; Update the deviation synchronization coefficient between each behavior state transition parameter to the edge weight between the corresponding behavior state nodes in the behavior state evaluation knowledge graph, and perform real-time update and iteration on the weights of each behavior state node in the behavior state evaluation knowledge graph; For any behavior state evaluation parameter x, the iteration formula of the corresponding behavior state node in the behavior state evaluation knowledge graph is as follows: x_update=(1-α-β)×x_ori+α×δx+β / (N_p)×sum(k_(p→x)×δp); Where, x_update is the iterative data of the behavior state node corresponding to behavior state evaluation parameter x, x_ori is the original data of the behavior state node corresponding to behavior state evaluation parameter x, δx is the reference deviation of behavior state evaluation parameter x, N_p is the number of other behavior state evaluation parameters except behavior state evaluation parameter x, sum(k_(p→x)×δp) is the numerical sum of the product of the deviation synchronization coefficient of each behavior state evaluation parameter other than behavior state evaluation parameter x with respect to x and the reference deviation, α is the reference deviation compensation coefficient, β is the deviation synchronization compensation coefficient, α>0, β>0 and α+β<1.

6. A behavior evaluation method based on cloud computing according to claim 1, characterized in that The said step S500 includes the following contents: Obtain the knowledge graph data for behavior state evaluation, extract the latest iterative data of each behavior state node adjacent to the evaluation dimension node, calculate the node weight of each evaluation dimension node by weighting with the edge weight between the evaluation dimension node and the adjacent behavior state node as the coefficient, and use it as the evaluation result of each evaluation dimension and upload it to the cloud for storage; Set the behavior state evaluation period T, divide the time period by T, divide the historical data of the node weights of each evaluation dimension node, extract the minimum and maximum values of the node weights of each evaluation dimension node in the historical data within each time period through cloud computing, and draw a radar chart based on the extracted data at the end time point of each time period, and use it as the behavior state evaluation result of the current evaluation object for each time period for feedback.

7. A cloud computing-based behavior evaluation system applying the cloud computing-based behavior evaluation method according to any one of claims 1-6, characterized in that, The system includes: a data monitoring and processing module, a behavior data analysis module, and a behavior evaluation and feedback module; The data monitoring and processing module performs real-time monitoring of the evaluation object by setting a non-contact acquisition sensor in the activity environment of the current evaluation object, and is also used to collect and obtain the situation information of the evaluation object; the behavior data analysis module trains a situation-behavior reference deep learning model, calculates the reference deviation of each behavior state evaluation parameter of the evaluation object, and is also used to construct a behavior state evaluation knowledge graph and update and iterate the data according to the historical data of the reference deviation of the behavior state evaluation parameters; the behavior evaluation and feedback module updates and feeds back the evaluation result of the evaluation object in real time according to the behavior state evaluation knowledge graph data.

8. A behavior evaluation system based on cloud computing according to claim 7, characterized in that, The data monitoring and processing module includes: a behavior state monitoring unit and a situation monitoring unit; The behavior state monitoring unit performs real-time monitoring of the behavior state of the evaluation object by setting a non-contact acquisition sensor in the activity environment of the current evaluation object; the situation monitoring unit collects and obtains the situation information of the evaluation object through real-time monitoring.

9. The behavior evaluation system based on cloud computing according to claim 7, wherein, The behavior data analysis module includes: a reference behavior analysis unit, a behavior deviation analysis unit, and a knowledge graph update unit; The reference behavior analysis unit annotates the reference values of the behavior state evaluation parameters of the evaluation object in different situation parameters, trains a situation-behavior reference deep learning model, and then analyzes the reference values of the behavior state evaluation parameters of the evaluation object in various situations; the behavior deviation analysis unit calculates the deviation between the monitored value and the reference value of the behavior state evaluation parameter of the evaluation object in real time, and then analyzes the deviation synchronization coefficient between the behavior state evaluation parameters; the knowledge graph update unit updates and iterates the data in the constructed behavior state evaluation knowledge graph according to the deviation synchronization coefficient between the behavior state evaluation parameters and combines the monitored data between the behavior state evaluation parameters.

10. A behavior evaluation system based on cloud computing according to claim 7, characterized in that, The behavior evaluation and feedback module includes: a behavior state evaluation unit and an evaluation result feedback unit; The behavior state evaluation unit evaluates the behavior state of the evaluation object according to the data in the behavior state evaluation knowledge graph and updates the evaluation result in real time; the evaluation result feedback unit filters the evaluation results of each evaluation dimension within each time period and then draws a radar chart, which is used as the behavior state evaluation result of the current evaluation object for each time period for feedback.

Images