Encrypted communication system for intervention in depressive adolescents with non-suicidal self-injury behavior

Abstract

The present application relates to the technical field of depression intervention of adolescent behavior, and discloses an encryption communication system for non-suicidal self-injury behavior intervention of depressed adolescents, which comprises a data acquisition module for obtaining communication behavior data, emotional state parameters and setting an encryption communication interval; a behavior analysis module capable of dividing behavior nodes and generating an interactive feature vector; a dynamic evaluation module capable of extracting high-risk indicators and establishing an encryption intervention rule; a channel optimization module responsible for identifying security levels and calculating data fluctuation density; a key management module capable of deriving an optimal key threshold and generating a key deviation sequence; and a strategy execution module for integrating and forming an intervention scheme. Through multi-dimensional data acquisition, dynamic behavior analysis and intelligent encryption intervention, the system realizes real-time monitoring and accurate intervention of self-injury behavior of depressed adolescents, improves the timeliness, individualization level and data security of intervention, and provides an efficient and secure technical solution for adolescent mental health intervention.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of behavior intervention for depressed adolescents, in particular to an encryption communication system for behavior intervention of non-suicidal self-injury of depressed adolescents. BACKGROUND

[0002] With increasing attention to mental health problems among adolescents, non-suicidal self-injury (NSSI) of depressed adolescents has become an important public health issue worldwide. Such behavior not only causes direct harm to the physical health of adolescents, but also may further exacerbate psychological problems and increase the risk of suicide, causing heavy burden to families and society. Currently, the intervention methods for depressed adolescents with NSSI mainly focus on traditional methods such as psychological counseling and drug treatment, but these methods generally have problems such as intervention lag, lack of individualization, and poor data security.

[0003] From the perspective of intervention timeliness, traditional intervention models rely on regular counseling or passive response after a crisis occurs, making it difficult to capture high-risk behavior signals exposed by adolescents in daily communication scenarios in real time. For example, the text content, interaction frequency, and other data of adolescents in social media and instant messaging often imply early signs of emotional fluctuations and self-injury tendencies, but existing technologies lack a system that can collect, analyze, and intervene in real time on these dynamic data.

[0004] In terms of individualized intervention, the behavior patterns and psychological states of different depressed adolescents differ significantly, and traditional "one-size-fits-all" intervention strategies cannot accurately match individual needs. For example, some adolescents may exhibit higher emotional fluctuations and self-injury risks during certain time periods (such as late at night), and existing systems cannot dynamically adjust intervention programs based on individual behavior characteristics.

[0005] Data security and privacy protection is another key challenge. The communication data and emotional state parameters of depressed adolescents are highly sensitive information, and traditional communication systems are vulnerable to leaks during data transmission and storage, leading to adolescents' reluctance to participate in interventions due to privacy concerns. In addition, existing encryption technologies are mostly general-purpose solutions, lacking dynamic encryption mechanisms deeply integrated with behavior intervention scenarios, making it difficult to balance data security while allowing flexible adjustment of intervention strategies.

[0006] In existing technologies, although some studies attempt to identify self-injury risks through data analysis, they generally have the following shortcomings: first, the data collection dimension is single, focusing only on emotional state or communication behavior in one aspect, lacking multi-dimensional data fusion analysis; second, encryption technology and intervention strategy are independent of each other, unable to dynamically optimize encryption strategies based on real-time behavior analysis results, resulting in a lack of synergy between intervention measures and data security protection; third, there is a lack of a closed-loop intervention execution mechanism, unable to automatically generate and execute individualized intervention programs based on real-time data fluctuations, resulting in insufficient intervention efficiency and precision.

[0007] Therefore, it is urgent to provide an integrated system capable of realizing multi-dimensional data real-time collection, dynamic behavior analysis and intelligent encryption intervention, so as to solve the problems of intervention lag, insufficient personalization and data security risks in the prior art, and to provide more effective technical support for precise intervention of NSSI of depressed adolescents. SUMMARY

[0008] The present application aims to provide an encryption communication system for intervention of NSSI of depressed adolescents, so as to solve the problems raised in the background art.

[0009] To achieve the above-mentioned purpose, the present application provides the following technical solution: an encryption communication system for intervention of NSSI of depressed adolescents, comprising:

[0010] A data collection module is configured to obtain communication behavior data and emotional state parameters of a target user authorized by the target user, set an encryption communication interval matched with the user behavior characteristics, and set the encryption communication interval as a time range for subsequent data processing.

[0011] A behavior analysis module is configured to divide a plurality of behavior nodes within the encryption communication interval, perform pattern recognition on the communication data of each behavior node, and generate an interaction feature vector corresponding to the behavior node.

[0012] A dynamic evaluation module is configured to extract a high-risk behavior indicator from the interaction feature vector, establish an encryption intervention rule associated with the behavior node, and obtain a key control parameter corresponding to the encryption intervention rule.

[0013] A channel optimization module is configured to identify a communication security level in the key control parameter, dynamically encrypt the high-risk behavior indicator according to the communication security level, and calculate a data fluctuation density of each behavior node under different encryption strategies.

[0014] A key management module is configured to derive an optimal key threshold value according to the data fluctuation density, compare the current communication data density with the optimal key threshold value, and generate a key deviation sequence.

[0015] A strategy execution module is configured to analyze the key deviation sequence, integrate the key deviation sequence into a communication intervention execution scheme based on the data fluctuation characteristics of the behavior node.

[0016] Preferably, the implementation of the behavior analysis module comprises: constructing a user feature library corresponding to the behavior node, wherein the user feature library contains a behavior parameter vector mapped from the communication behavior data and the emotional state parameters.

[0017] The behavior parameter vectors are subjected to similar behavior matching, and the behavior clusters of the behavior parameter vectors are divided according to the matching results; and the distribution gravity points of the communication data are extracted from the behavior cluster groups, and the distribution gravity points are set as the behavior nodes.

[0018] Preferably, the division of the behavior cluster groups of the behavior parameter vectors further comprises:

[0019] According to the communication frequency and the emotional fluctuation index in the behavior parameter vectors, the communication duration, the interaction interval and the emotional intensity parameters are extracted, and the behavior feature labels are generated based on the communication duration, the interaction interval and the emotional intensity parameters;

[0020] The behavior feature labels are associated with the behavior parameter vectors, the behavior similarity between the feature labels is calculated, and the parameter vectors with a similarity higher than a preset behavior threshold are screened to form the behavior cluster groups.

[0021] Preferably, the implementation manner of generating the interaction feature vectors corresponding to the behavior nodes comprises:

[0022] For each behavior node, the data fluctuation parameters of the behavior node in a preset period are obtained according to the time sequence distribution of the behavior node in the encrypted communication interval, and the communication fluctuation coefficient of the node is calculated;

[0023] When the communication fluctuation coefficient exceeds a first risk threshold, the node is marked as a high-risk node, and its communication data is extracted to form an interaction feature vector; when the communication fluctuation coefficient is lower than the first risk threshold, the node is marked as a safe node, and the communication data of adjacent nodes of the node is subjected to density superposition, and the superposed data is reconstructed into an interaction feature vector.

[0024] Preferably, the implementation manner of the dynamic evaluation module comprises:

[0025] The communication data proportion, the abnormal interaction proportion and the emotional fluctuation parameters are separated from the interaction feature vectors, and the encrypted intervention rules of the behavior nodes are generated based on the communication data proportion, the abnormal interaction proportion and the emotional fluctuation parameters;

[0026] If the number of behavior nodes covered by the current encrypted intervention rule is less than a preset behavior threshold, the interaction feature vectors of adjacent behavior nodes are traversed, and the communication indicators not included in the encrypted intervention rules of the adjacent nodes are added to the current rule.

[0027] Preferably, the implementation manner of the channel optimization module comprises: obtaining the time sequence parameters of the key validity period and the fluctuation parameters of the data encryption strength in the communication security level;

[0028] An encryption state network associated with the time sequence parameters and the fluctuation parameters is constructed, and the data fluctuation densities under different encryption strategies are determined according to the switching probabilities of various paths in the encryption state network.

[0029] Preferably, the encrypted state network is constructed further comprising:

[0030] identifying a cycle mode of the timing parameter, and setting the timing parameter as a starting point of the encrypted state network if the current cycle mode completely matches the preset communication cycle;

[0031] calculating a switching correlation degree of the timing parameter and the fluctuation parameter, and generating an intermediate point and a termination point of the encrypted state network in turn according to the switching correlation degree from high to low;

[0032] performing state reverse verification on the termination point, and outputting the termination point as a final link of the encrypted state network when the switching correlation degree of the termination point is lower than a preset switching threshold.

[0033] Preferably, the implementation manner of calculating the data fluctuation density comprises:

[0034] statistically calculating a mean value of the timing parameter and a range of the fluctuation parameter of each termination point in the encrypted state network, and calculating a global covariance of all node parameters;

[0035] differencing the mean value of the timing parameter of a single termination point and the mean value of the timing parameter of an adjacent node, dividing the difference value obtained by the global covariance to obtain a timing density coefficient; and simultaneously calculating a ratio of the range of the fluctuation parameter and the global covariance, and weighting and summing the ratio and the timing density coefficient as a data fluctuation density of the node.

[0036] Preferably, the implementation manner of deriving the optimal key threshold comprises:

[0037] extracting an encrypted mode in the historical data closest to the current data fluctuation density, and calculating an Euclidean distance of the data fluctuation density in the closest encrypted mode and the current data fluctuation density on a timing distribution as a first key reference value;

[0038] statistically calculating a difference in the number of peak points between the data fluctuation density in the closest encrypted mode and the current data fluctuation density, and taking the difference number as a second key reference value;

[0039] based on a linear combination of the first key reference value and the second key reference value, matching an optimal key threshold in a preset key threshold table.

[0040] Preferably, the implementation manner of the strategy execution module comprises: dividing a positive fluctuation domain and a reverse fluctuation domain according to a data fluctuation trend of each node in the key deviation sequence;

[0041] extracting a convergence frequency of the key deviation in the positive fluctuation domain and a diffusion frequency of the key deviation in the reverse fluctuation domain, dynamically reconciling the two according to a communication weight of a node, and generating a configuration parameter of a communication intervention execution scheme.

[0042] Compared with the prior art, the present application has the following advantages:

[0043] The data collection module captures real-time multi-dimensional dynamic data of adolescents by obtaining the communication behavior data and emotional state parameters of the target user and setting an encrypted communication interval matching the user behavior characteristics. This design breaks through the limitations of traditional intervention methods with single data collection and can more comprehensively reflect the real state of the user, providing a rich and accurate data basis for subsequent behavior analysis. For example, by integrating parameters such as communication frequency, interaction interval, and emotional intensity, the system can accurately identify differences in behavior patterns among different users, laying the foundation for personalized intervention.

[0044] The behavior analysis module realizes deep structured analysis of communication data by constructing a user feature library, dividing behavior clustering groups, and setting behavior nodes. This module not only classifies data with similar characteristics through similar behavior matching, but also extracts distribution gravity points as behavior nodes, allowing complex communication data to be converted into identifiable and analyzable structured units. This hierarchical analysis mechanism helps the system quickly locate key behavior nodes, such as behavior characteristic labels generated by communication duration and emotional fluctuation index, which can accurately identify high-risk behavior patterns that may pose self-harm risks, providing clear targets for subsequent dynamic assessment.

[0045] The dynamic assessment module extracts high-risk behavior indicators from the interaction feature vector and establishes encrypted intervention rules, while supplementing communication indicators by traversing adjacent nodes to ensure the comprehensiveness and dynamic adaptability of the intervention rules. This module can dynamically adjust intervention strategies based on real-time data, such as automatically integrating information from adjacent nodes when the number of behavior nodes covered by the current rule is insufficient, avoiding intervention loopholes caused by partial data. This mechanism enables the system to respond to behavior changes in a timely manner, improving the timeliness and relevance of intervention, such as quickly triggering high-risk node labeling and intervention rule generation when identifying abnormal communication fluctuation coefficients.

[0046] The channel optimization module realizes the deep integration of encryption strategies and behavior characteristics by constructing an encrypted state network and calculating data fluctuation density. This module not only considers the timing parameters of key validity period and the fluctuation parameters of data encryption strength, but also dynamically optimizes encryption strategies through switching probability and correlation analysis. For example, when detecting an increase in data fluctuation density, the system automatically increases encryption strength and adjusts the key validity period, ensuring that intervention information can be accurately and securely transmitted to the target user, avoiding security vulnerabilities or intervention delays caused by fixed encryption strategies.

[0047] The key management module realizes the dynamic matching of key parameters and data fluctuations by deriving the optimal key threshold and generating the key deviation sequence. Based on the comparative analysis of historical data and real-time data, the module can accurately calculate the optimal key threshold that adapts to the current behavior characteristics, such as matching the preset threshold table through the linear combination of Euclidean distance and peak point difference, so that the key parameters can be adjusted in real time with the data fluctuations. This dynamic key management mechanism not only improves the security of data encryption, but also provides precise control parameters for the policy execution module, ensuring the effectiveness of the intervention plan.

[0048] The policy execution module generates personalized communication intervention execution plans by dividing the fluctuation domain and adjusting the convergence / diffusion frequency. The module can dynamically adjust the configuration parameters of the intervention strategy according to the data fluctuation trend of each behavior node, such as adopting different intervention intensity and methods for positive and negative fluctuation domains. This differentiated intervention mechanism enables the system to provide precise intervention measures for different user behavior characteristics, such as strengthening real-time monitoring and psychological counseling for nodes with intense emotional fluctuations, improving the accuracy and effectiveness of intervention. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 The working principle diagram of the encryption communication system for the intervention of depressive adolescent non-suicidal self-injury behavior described in the present application;

[0050] Figure 2 The flowchart of the behavior node division method;

[0051] Figure 3 The design diagram of feature label association and cluster group screening;

[0052] Figure 4 The flowchart of high-risk node identification and feature vector reconstruction. DETAILED DESCRIPTION

[0053] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0054] Please refer to Figures 1-4 The encryption communication system for the intervention of depressive adolescent non-suicidal self-injury behavior described in the present application, which realizes the communication behavior monitoring, risk assessment and encryption intervention of target users through the cooperative work of multiple modules. Specifically, the following steps are included:

[0055] Data acquisition module: Obtain the communication behavior data (such as communication frequency, duration, interaction object, etc.) and emotional state parameters (such as heart rate variability, voice tone fluctuation collected through wearable devices or APP) of the target user authorized by the target user, set a matching encrypted communication interval based on the user behavior characteristics (such as historical communication rules, emotional fluctuation period), and the interval is the time range for subsequent data processing. The communication behavior data and emotional state parameters of the target user authorized by the target user refer to the communication behavior data and emotional state parameters involved in the present application, which have obtained the permission and authorization of the target user, and do not involve violating the requirements of personal data privacy protection.

[0056] Behavior analysis module: Divide multiple behavior nodes in the encrypted communication interval, and each node corresponds to the communication data of a specific time period. Perform pattern recognition on the communication data of each behavior node to generate an interaction feature vector corresponding to the node, which contains multi-dimensional features reflecting user communication patterns and emotional states.

[0057] Dynamic evaluation module: Extract high-risk behavior indicators (such as abnormal communication frequency, emotional drastic fluctuation, etc.) from the interaction feature vector, establish encrypted intervention rules associated with the behavior nodes, and obtain key control parameters (such as encryption strength, key validity period, etc.) corresponding to the rules.

[0058] Channel optimization module: Identify the communication security level (such as high, medium, and low risk levels) in the key control parameters, dynamically encrypt the high-risk behavior indicators according to the security level, and calculate the data fluctuation density of each behavior node under different encryption strategies. The density reflects the degree of data change.

[0059] Key management module: Derive the optimal key threshold value according to the data fluctuation density, compare the current communication data density with the optimal key threshold value, and generate a key deviation sequence that records the difference between the actual data and the ideal state.

[0060] Strategy execution module: Analyze the key deviation sequence, integrate the deviation sequence into a specific communication intervention execution scheme based on the data fluctuation characteristics of the behavior nodes, such as adjusting the communication content filtering rules, triggering the warning mechanism, etc.

[0061] The technical solutions of the present application will be further described in detail below in combination with specific embodiments.

[0062] Embodiment 1:

[0063] The embodiment relates to a specific implementation of a behavior analysis module. The behavior analysis module first constructs a user feature library corresponding to a behavior node, which contains a behavior parameter vector of communication behavior data and emotion state parameter mapping. The communication behavior data covers various interaction information of a target user in an encrypted communication interval, such as communication initiation time, communication duration, communication frequency with different interaction objects, and the number and types (text, voice, picture, etc.) of message sending and receiving; the emotion state parameters are collected through terminals such as wearable devices and mobile phone APPs, including but not limited to heart rate variability data, voice tone fluctuation amplitude, facial expression recognition emotion intensity index, etc. After standardization processing, these data are mapped into a multi-dimensional behavior parameter vector, each vector corresponding to the user behavior and emotion state in a specific time window, for example, a 15-minute time window, generating a vector composed of dimensions such as communication frequency, average communication duration, and heart rate variability coefficient in the window.

[0064] When performing similar behavior matching on the behavior parameter vector, a classic similarity calculation method in the data mining field is adopted, such as the cosine similarity algorithm or the Euclidean distance measurement method. By calculating the similarity values between different behavior parameter vectors, the vectors with a similarity higher than an initial set threshold are divided into the same behavior clustering group. In the process of dividing the behavior clustering group, more targeted feature parameters need to be extracted according to the communication frequency and emotion fluctuation index in the behavior parameter vector, specifically including: extracting the communication frequency in a unit time (i.e., communication frequency) from the communication frequency dimension, extracting the standard deviation of the emotion parameter in a set time period (i.e., emotion fluctuation intensity parameter) from the emotion fluctuation index dimension, and combining the communication initiation and end time in the communication behavior data to calculate the time interval between two adjacent communications (i.e., interaction interval parameter). Based on the extracted communication duration, interaction interval, and emotion intensity parameters, a behavior feature label that can intuitively reflect the user behavior mode and emotion characteristics is generated, for example: when the communication frequency of a certain behavior parameter vector is high (such as more than 5 times per hour), the interaction interval is short (such as an average interval of less than 10 minutes), and the emotion intensity parameter is large (such as a standard deviation exceeding a preset threshold), a "high-frequency short-interval anxious interaction" label is generated; if the communication frequency is low (such as less than 1 time per hour), the interaction interval is long (such as an average interval of more than 1 hour), and the emotion intensity parameter is small, a "low-frequency long-interval calm interaction" label is generated.

[0065] After associating the behavior feature labels with the corresponding behavior parameter vectors, the members of the clustering group are further screened by calculating the behavior similarity between the feature labels. The behavior similarity calculation here is based on the semantic correlation degree of the feature parameters contained in the labels. For example, the similarity between the labels of "high frequency short interval anxious interaction" and "high frequency long interval anxious interaction" is higher than that between the label of "low frequency long interval calm interaction", because both of them contain the semantics of "high frequency" and "anxious interaction". In specific implementation, the label can be converted into a vector form by using the word vector model (such as Word2Vec) in natural language processing, the cosine distance between the vectors is calculated to determine the similarity value, and the parameter vectors with a similarity higher than a preset behavior threshold are screened out, and finally several behavior clustering groups are formed, and the behavior parameter vectors in each clustering group have similar communication patterns and emotional characteristics.

[0066] After completing the division of the behavior clustering groups, the distribution center point of the communication data needs to be extracted from each clustering group, which is used as the representative feature of the clustering group to set the behavior node. The calculation method of the distribution center point is as follows: the arithmetic mean value of the data of each dimension of all behavior parameter vectors in the clustering group is taken to generate a new vector, which is used as the distribution center point of the clustering group. For example, a clustering group contains 100 behavior parameter vectors, each vector contains data of three dimensions of communication frequency, interaction interval and emotional intensity, and the communication frequency dimension value of the distribution center point is the average value of the communication frequency of 100 vectors, and the interaction interval and emotional intensity dimensions are the same. After setting the distribution center point as the behavior node, each behavior node corresponds to a time point or time period with typical behavior characteristics, which is used for subsequent dynamic analysis and risk assessment of the communication behavior of the target user. In this way, the behavior analysis module can convert complex user communication data and emotional state into structured behavior nodes and corresponding interaction feature vectors, providing basic data support for the subsequent dynamic assessment module to identify high-risk behavior indicators and establish encryption intervention rules. The whole process realizes the automatic classification and node setting of the behavior mode through the data-driven way, avoiding the subjectivity and randomness of manual intervention, and ensuring the objectivity and accuracy of the behavior analysis result.

[0067] Embodiment 2:

[0068] This embodiment further refines the process of generating the interaction feature vectors corresponding to behavioral nodes. For each behavioral node, it is first necessary to determine its temporal distribution within the encrypted communication interval, i.e., its temporal position (such as being in the beginning, middle, or end segment of the entire interval) and the time interval between adjacent nodes. Based on this temporal distribution, the data fluctuation parameters of the node within a preset period are obtained. The preset period can be flexibly set according to the actual application scenario; for example, it can be set to 1 hour for high-frequency communication users and 1 day for low-frequency communication users. The data fluctuation parameters include, but are not limited to, the rate of change of communication frequency within the preset period (i.e., the ratio of the difference between the current period's communication frequency and the previous period's communication frequency to the previous period's frequency), the standard deviation of the emotion parameter (reflecting the stability of the emotional state), and the coefficient of variation of the interaction interval (measuring the degree of fluctuation of the interaction interval). Using these parameters, the communication fluctuation coefficient of the node is calculated. This coefficient is obtained by comprehensively weighting each data fluctuation parameter. The weighting weights can be preset according to the importance of each parameter to risk assessment; for example, the communication frequency change rate accounts for 40%, the standard deviation of the emotion parameter accounts for 30%, and the coefficient of variation of the interaction interval accounts for 30%.

[0069] When the communication fluctuation coefficient exceeds the first risk threshold, the system determines that the node is a high-risk node. At this point, the communication data of that node is directly extracted to form an interaction feature vector. The scope of communication data extraction includes, but is not limited to: keywords in text communication, voiceprint features in voice communication (such as intonation and speech rate changes), identification information of the interaction object (such as phone number and social media account ID), and specific timestamps for initiation and termination of communication. After normalization, this data is combined into a multi-dimensional interaction feature vector according to a preset dimension order, such as a vector composed of dimensions like [keyword frequency, intonation fluctuation, unfamiliarity with the interaction object, and timestamp offset], which is used by the subsequent dynamic evaluation module to analyze high-risk behavior indicators.

[0070] When the communication fluctuation coefficient is below the first risk threshold, the node is marked as a safe node. Since the data fluctuation of a safe node is relatively small, analyzing its communication data alone may not effectively reflect user behavior characteristics. Therefore, it is necessary to perform density overlay processing on the communication data of its neighboring nodes. The range of neighboring nodes can be set to the nodes before and after the current safe node. 1 node The value is determined based on the length of the encrypted communication interval and the node density. For example, when the interval length is 24 hours and the node interval is 1 hour... It can be set to 1 or 2.

[0071] The specific method of density overlay is as follows: the communication data of the secure node and its neighboring nodes are concatenated in chronological order, and duplicate or conflicting data is deduplicated or weighted averaged. For example, if the secure node... The previous node is , the latter node is , the communication frequencies of the three nodes are , , , the communication frequency after superposition is:

[0072]

[0073] For the keyword frequency of the text content, the keyword lists of the three nodes can be combined, and the number of occurrences of each keyword is counted and normalized.

[0074] After completing the data superposition, the superimposed data needs to be reconstructed into an interaction feature vector. The reconstruction process needs to follow the same dimension definition as the high-risk node interaction feature vector to ensure consistency in subsequent analysis. For example, if the high-risk node interaction feature vector contains four dimensions, the data of the security node after density superposition also needs to be mapped to these four dimensions. In specific implementation, for data that cannot be directly mapped to the dimensions, feature engineering methods can be used for conversion, such as converting timestamp information into relative offset with the preset period, converting the social relationship chain depth of the interaction object into the stranger index, etc. In addition, to avoid the explosion of vector dimensions caused by the large amount of superimposed data, dimension reduction techniques such as principal component analysis (PCA) can be used to compress the data, retaining the principal components that best reflect the behavior characteristics.

[0075] When processing the security node, boundary conditions also need to be considered. For example, when the security node is located at the starting position of the encrypted communication interval, its forward adjacent node does not exist, and only the backward adjacent node is superimposed; if the security node is located at the end of the interval, only the forward adjacent node is superimposed. For the case of having only a single side adjacent node, asymmetric weighting can be used during superposition, such as the superimposed data of the starting node, the current node data weight accounts for 70%, and the backward node accounts for 30%, to highlight the basic role of the current node.

[0076] The interaction feature vector generated in the above manner can effectively capture the communication mode and emotional state characteristics of the target user at different behavior nodes, whether it is a high-risk node or a safe node. The vector of the high-risk node directly reflects the current possible abnormal behavior data, providing a basis for timely triggering encryption intervention rules; the vector of the safe node is reconstructed through adjacent data stacking, avoiding missing potential behavior trend changes due to single node data flatness, for example, the stacked data of multiple consecutive safe nodes may reveal slow accumulation of emotional fluctuations or communication mode changes, providing more comprehensive data support for long-term risk assessment of the system. The entire process realizes differentiated processing of different risk level behavior nodes through dynamic threshold judgment and data stacking strategy, ensuring timely response to high-risk behavior and enhancing trend analysis capability for safe behavior, so that the interaction feature vector can more accurately reflect the real behavior state of the user in the encryption communication interval.

[0077] Embodiment 3:

[0078] This embodiment describes the specific implementation of the dynamic evaluation module. The core function of the dynamic evaluation module is to extract key parameters from the interaction feature vector, establish encryption intervention rules associated with behavior nodes, and dynamically optimize rule content according to the rule coverage range. First, the module separates three types of key parameters from the interaction feature vector: communication data proportion, abnormal interaction proportion, and emotional fluctuation parameter. The communication data proportion reflects the proportion of different types of communication content in the overall data, such as the proportion of text message quantity in total communication quantity, the proportion of voice call duration in total communication duration, etc.; the abnormal interaction proportion is used to measure the frequency of interaction with irregular objects (such as unknown numbers, untagged contacts) or abnormal communication mode (such as high-frequency communication at night); the emotional fluctuation parameter is calculated based on physiological indicators (such as heart rate, skin conductance) collected by wearable devices or APP and communication content sentiment analysis results (such as text message sentiment polarity score), such as the ratio of low frequency component (LF) to high frequency component (HF) of heart rate variability (HRV), the fluctuation amplitude of mel frequency cepstral coefficient (MFCC) of voice signal, etc.

[0079] Based on the above three types of parameters, the module generates encryption intervention rules for the behavior nodes. The rule generation logic is based on preset risk judgment conditions, for example: when the abnormal interaction proportion of a certain behavior node exceeds 20% and the HRVLF / HF ratio in the emotion fluctuation parameter is greater than 3, generate the rule of "deep content encryption for the communication content of the node"; if the voice call duration in the communication data proportion exceeds 60% and the sentiment polarity score in the emotion fluctuation parameter is less than -0.5 (negative value represents negative emotion), generate the rule of "trigger real-time communication content monitoring and record logs". These rules convert abstract interaction features into executable encryption intervention actions through parameter threshold combination, and each rule corresponds to a unique behavior node and specific key control parameters (such as encryption algorithm type, key update frequency).

[0080] When the number of behavior nodes covered by the current encryption intervention rule is less than the preset behavior threshold (for example, the preset threshold is 10 nodes, and the current rule only covers 6 nodes), the module starts the rule expansion mechanism. The specific process is: traverse all adjacent nodes of the current behavior node (the range of adjacent nodes can be set to M nodes before and after, M is determined according to the node density of the encryption communication interval, such as one node per hour, M=2), extract the communication indicators in the adjacent node interaction feature vector that are not included in the current rule. For example, if the current rule mainly focuses on abnormal interaction proportion and emotion fluctuation parameter, and the adjacent node interaction feature vector contains indicators such as communication frequency mutation (such as an increase of 50% compared to the previous node) or interaction interval abnormally shortened (such as from an average of 30 minutes to 5 minutes), which are not included in the rule, these indicators will be added as new conditions to the current rule.

[0081] When expanding the rule, the effectiveness of the new communication indicators needs to be verified. The verification method is to analyze the correlation between the indicator and high-risk behavior in historical data, for example, by calculating the correlation coefficient between the communication frequency mutation indicator and the record of non-suicidal self-injury behavior, if the absolute value of the coefficient is greater than 0.4 (preset effective threshold), it is confirmed that the indicator has risk warning value and can be included in the rule; if the coefficient is below the threshold, the indicator is ignored to avoid rule redundancy. Through this data-driven verification mechanism, the scientificity and effectiveness of the new rule conditions are ensured.

[0082] After the rule expansion is completed, the newly generated rule needs to be checked for logical consistency to avoid conditions conflict or overlapping coverage. For example, if the original rule specifies "encrypt when abnormal interaction proportion > 20%", and the new rule specifies "encrypt when abnormal interaction proportion > 15% and communication frequency mutation", the execution priority or complementary relationship of the two rules needs to be clear, which can be achieved by setting rule levels (such as basic rules, extended rules) or condition merging (such as "abnormal interaction proportion > 15% and (communication frequency mutation or emotion fluctuation parameter > threshold)").

[0083] Another key aspect of the dynamic evaluation module is the acquisition of key control parameters. Each encryption intervention rule corresponds to a specific set of key control parameters, which are dynamically adjusted according to the risk level and intervention intensity of the rule. For example, for the "deep content encryption" rule, the key control parameters may include: using the AES-256 encryption algorithm (high security level algorithm), shortening the key validity period to 30 minutes (50% shorter than the regular validity period), setting the encrypted data fragment size to 1KB (smaller fragments improve transmission security); while for the "real-time communication monitoring" rule, the parameters may be: using the AES-128 algorithm, keeping the default key validity period of 1 hour, and setting the data fragment size to 4KB. The setting of parameters is based on the principles of cryptography and communication security policy, ensuring that behavior nodes of different risk levels receive matching encryption protection.

[0084] During rule execution, the module continuously monitors the parameter changes of the behavior nodes, and when the parameters of a certain node exceed the threshold set by the rule, the corresponding encryption intervention action is triggered immediately, and the rule execution log is recorded. The log content includes: rule triggering time, behavior node identification, triggering parameter value, executed encryption strategy and key control parameter details, which provide data support for subsequent system optimization and risk tracing.

[0085] Through the above process, the dynamic evaluation module realizes the automatic mapping from interaction features to intervention rules, as well as the dynamic expansion and optimization of rule coverage. The module not only can implement precise encryption intervention on the currently identified high-risk behavior nodes, but also can discover potential risk trends in advance through adjacent node feature analysis, expanding the foresight and comprehensiveness of the rules. The entire process is based on a data-driven rule generation and verification mechanism, avoiding the limitations of manually formulated rules, ensuring that the encryption intervention strategy can accurately match the target user's communication behavior and emotional state changes in real time, providing scientific and dynamic technical support for early intervention of depressive adolescents' non-suicidal self-injury behavior.

[0086] Embodiment 4:

[0087] This embodiment relates to the detailed process of the channel optimization module. The channel optimization module first acquires two types of core parameters in the communication security level: the timing parameter of the key validity period and the fluctuation parameter of the data encryption strength. The timing parameter is used to describe the distribution characteristics of the key validity period on the time axis, such as the time point of key replacement, the period length (such as replacing the key every hour), the matching degree of the validity period starting time and the user communication peak period, etc.; the fluctuation parameter reflects the changes of the data encryption strength over time, including the complexity of the encryption algorithm (such as switching from AES-128 to AES-256), the adjustment of the key length (such as expanding from 128 bits to 256 bits), the change of the data fragment size (such as reducing from 4KB to 1KB), etc.

[0088] In constructing the encrypted state network, the periodic pattern of the timing parameter is first identified. The periodic pattern analysis is achieved by extracting the time characteristics of the timing parameter, such as detecting the main frequency component of the key validity period replacement cycle by Fourier transform, or determining the cycle length by counting the number of key replacements in a fixed time period by sliding window statistics. If the current periodic pattern completely matches the preset communication period (such as the daily 9:00-22:00 as the high-frequency communication period, and the rest as the low-frequency period, which is statistically derived from the historical communication data of the target user), the current cycle start point of the timing parameter is set as the start point of the encrypted state network, serving as the time reference of the entire network.

[0089] The switching correlation degree of the timing parameter and the fluctuation parameter is calculated. The correlation degree is based on the co-occurrence frequency of the two types of parameters in the historical data, such as 80 times of key replacement accompanied by encryption strength enhancement in the past 100 times of key replacement, then the correlation degree of the timing parameter (key replacement) and the fluctuation parameter (encryption strength enhancement) is 80%. By traversing all possible parameter combinations, a correlation matrix is generated, and the correlation degrees are sorted from high to low to determine the intermediate point and the termination point of the encrypted state network in turn. The intermediate point represents the intermediate state of different parameter combinations, such as "key validity period shortened to 30 minutes and encryption algorithm replaced with AES-256"; the termination point represents the final stable state of the parameters, such as "key validity period 30 minutes, AES-256 encryption, and fragment size 1KB".

[0090] When performing state reverse verification on the termination point, starting from the termination point, the path is traced back to the start point along the parameter switching path, and it is checked whether each step of switching conforms to the parameter correlation logic in the historical data. For example, if the parameter combination corresponding to the termination point is "key validity period 15 minutes, ChaCha20 encryption", the backtracking path is "ChaCha20 encryption → key validity period shortened to 15 minutes", it needs to be verified whether there is a case of adjusting the encryption algorithm first and then shortening the key validity period in the historical data, if there is and the correlation degree is higher than the preset switching threshold (such as 50%), the termination point is confirmed to be valid; if there is not or the correlation degree is lower than the threshold, the termination point is excluded and the path is regenerated until the final link that meets the conditions is found.

[0091] When calculating the data fluctuation density, first, the parameters of each termination point in the encrypted state network are counted. The mean of the timing parameter is the average length of the key validity period corresponding to the termination point (such as the average of multiple samples), and the range of the fluctuation parameter is the difference between the maximum and minimum values of the data encryption strength parameter (such as the range of the encryption algorithm complexity score). The global covariance is obtained by calculating the covariance matrix of the timing parameter and the fluctuation parameter of all nodes, reflecting the overall correlation degree of the two types of parameters in the entire network.

[0092] For a single termination point, the time sequence density coefficient is calculated by subtracting the time sequence parameter mean value of the termination point from the time sequence parameter mean value of the adjacent node, taking the absolute value, and dividing by the corresponding dimension value of the global covariance. For example, the time sequence parameter mean value of termination point A is 30 minutes, the mean value of adjacent node B is 45 minutes, and the variance of the time sequence parameter in the global covariance is 225. The time sequence density coefficient is |30-45| / 225 = 15 / 15 = 1. The fluctuation parameter ratio is the fluctuation parameter range of the termination point divided by the square root of the variance of the fluctuation parameter in the global covariance. For example, the fluctuation parameter range is 10, and the variance of the fluctuation parameter in the global covariance is 25. The ratio is 10 / 5 = 2.

[0093] The time sequence density coefficient and the fluctuation parameter ratio are weighted and summed according to the preset weight to obtain the data fluctuation density. The weight setting needs to follow the communication security policy. For example, in a high-risk scenario, the time sequence density coefficient weight is set to 0.7, and the fluctuation parameter ratio weight is set to 0.3, to highlight the impact of key validity period changes on data stability. In a low-risk scenario, the weights of the two can be set to 0.5:0.5. The final value reflects the degree of data fluctuation under the encryption policy corresponding to the termination point. The larger the value, the more unstable the data, and the encryption policy needs to be adjusted to improve security.

[0094] In practical applications, the channel optimization module dynamically adjusts the encryption policy of each behavior node by updating the encryption state network and data fluctuation density in real time. For example, when a behavior node is marked as a high-risk node, the module determines the communication security level to be "high" according to the key control parameters corresponding to it, triggers the encryption state network path of the high security level, selects a termination point parameter combination with low data fluctuation density (such as a combination with short key validity period, high encryption strength, and small parameter fluctuation), and dynamically encrypts the high-risk behavior index to ensure the security of sensitive data during transmission.

[0095] The entire channel optimization process realizes the dynamic matching of encryption policy and user behavior risk through parameter correlation analysis, network modeling, and density calculation. The construction of the encryption state network is based on historical data and communication cycle characteristics, ensuring the logicality and effectiveness of parameter switching. The calculation of data fluctuation density quantifies the stability of the encryption policy from the time sequence and intensity dimensions, providing key evidence for the key management module to derive the optimal key threshold. This module not only improves the security of the communication channel, but also reduces communication delay or resource consumption caused by excessive encryption through a dynamic adjustment mechanism, achieving a balance between security and communication efficiency, and is suitable for the encryption protection needs of real-time communication data in the intervention scenario of depressive adolescents' non-suicidal self-injury behavior.

[0096] Embodiment 5:

[0097] The present embodiment illustrates the implementation details of the key management module and the policy execution module. When the key management module derives the optimal key threshold, it first selects the encryption mode closest to the current data fluctuation density from the historical data. The historical data is stored in the system database, and each encryption mode contains the data fluctuation density value, the corresponding key parameters and the timing distribution characteristics. The selection process selects the first K modes (K is a preset value, such as 5) with the smallest absolute difference between the current data fluctuation density and the historical record as the candidate set by calculating the absolute difference. For each candidate mode, calculate the Euclidean distance between its timing distribution and the current data, that is, compare the similarity of the key effective period change curve or the encryption strength fluctuation curve on the time axis. For example, if the encryption strength of the current data fluctuates greatly in the 14:00-16:00 period, and the fluctuation trend of a certain historical mode in the same period is similar, then the Euclidean distance between the two is small, and this value is used as the first key reference value.

[0098] The difference between the current data fluctuation density and the historical encryption mode in the number of peak points is counted. The peak point is defined as the point where the data fluctuation density exceeds the local threshold (such as the point where the density value of the adjacent two nodes exceeds), and the difference number is the absolute value of the difference between the number of peak points of the current data and the historical mode. For example, the current data has 4 peak points, and a certain historical mode has 6, then the difference number is 2, and this value is used as the second key reference value. The first reference value reflects the similarity of the timing distribution, and the second reference value reflects the difference in the fluctuation intensity, and the two together represent the overall matching degree of the current data and the historical mode.

[0099] Based on the linear combination (such as the sum or weighted sum, the weight can be set to 1:1) of the first key reference value and the second key reference value, the preset key threshold table is queried. The table is a two-dimensional mapping table, the horizontal axis is the first reference value interval, the vertical axis is the second reference value interval, and each cell stores the corresponding optimal key threshold (such as key effective period, encryption strength coefficient, etc.). For example, when the first reference value is 0.8 and the second reference value is 1, the optimal key threshold is determined to be "key effective period 45 minutes, encryption strength level 3" by looking up the table. The threshold table is generated by training historical data, and the correlation between parameters is fitted by clustering analysis or regression model to ensure the rationality of the threshold.

[0100] When the policy execution module analyzes the key deviation sequence, it first divides the positive fluctuation domain and the reverse fluctuation domain according to the data fluctuation trend (rising, falling or stable) of each node in the sequence. The positive fluctuation domain refers to the interval where the data fluctuation density of the continuous nodes shows an upward trend, and the reverse fluctuation domain refers to the interval where the data fluctuation density shows a downward trend. For example, if the density values of nodes i, i+1, i+2 in the sequence are 0.6, 0.7, 0.8 respectively, then nodes i to i+2 belong to the positive fluctuation domain; if the density values of nodes j, j+1, j+2 are 0.9, 0.7, 0.5, then they belong to the reverse fluctuation domain.

[0101] In the positive fluctuation domain, the convergence frequency of the key deviation is extracted, that is, the proportion of the number of nodes whose deviation value (the difference between the current data density and the optimal key threshold value) gradually decreases in the total number of nodes in the domain. For example, in the positive fluctuation domain, there are 5 nodes, 3 of which have a deviation value decreasing from 0.3 to 0.1, 0.2 to 0.1, and 0.4 to 0.2. The convergence frequency is 3 / 5 = 60%. In the negative fluctuation domain, the diffusion frequency of the key deviation is extracted, that is, the proportion of the number of nodes whose deviation value gradually increases. For example, in the negative fluctuation domain, there are 4 nodes, 2 of which have a deviation value increasing from 0.1 to 0.3 and from 0.2 to 0.5. The diffusion frequency is 2 / 4 = 50%.

[0102] The convergence frequency and diffusion frequency are dynamically reconciled according to the communication weight of the behavior node. The communication weight reflects the importance of the node and can be set according to factors such as the intimacy of the communication object (e.g., higher weight for family and friends), the sensitivity of the communication content (e.g., higher weight for self-harm topics), etc. The value range is 0-1. The reconciliation method is as follows: multiply the convergence frequency of the positive fluctuation domain by the average weight of the nodes in the domain, multiply the diffusion frequency of the negative fluctuation domain by the average weight of the nodes in the domain, and add the two values to obtain the final reconciliation value. For example, the average weight of the positive fluctuation domain is 0.8, the convergence frequency is 60%, and the contribution value is 0.8 x 0.6 = 0.48. The average weight of the negative fluctuation domain is 0.7, the diffusion frequency is 50%, and the contribution value is 0.7 x 0.5 = 0.35. The reconciliation value is 0.48 + 0.35 = 0.83.

[0103] The configuration parameters of the communication intervention execution scheme are generated according to the reconciliation value. The configuration parameters include encryption strength adjustment amplitude (e.g., increase the encryption strength by 1 level when the reconciliation value is higher than 0.8), key update frequency (e.g., extend the key validity period to 1.5 times the default value when the reconciliation value is lower than 0.5), warning trigger conditions (e.g., trigger intermediate warning and record logs when the reconciliation value is between 0.5 and 0.8), etc. These parameters are associated with the reconciliation value through pre-set mapping rules, such as piecewise functions or fuzzy logic rules, to ensure the flexibility and pertinence of the intervention strategy.

[0104] When executing the intervention scheme, the system adjusts the encryption strategy of each behavior node according to the configuration parameters. For example, for nodes with high weight in the positive fluctuation domain, if the reconciliation value shows that the deviation converges but there is still a high risk, the high encryption strength can be maintained and the key validity period can be shortened. For nodes with low weight in the negative fluctuation domain, if the deviation diffusion trend is gentle, the encryption strength can be appropriately reduced to reduce resource consumption. Through data fluctuation trend analysis and weight reconciliation, the dynamic optimization of the intervention strategy is realized, which not only ensures the strict protection of high-risk nodes, but also avoids excessive intervention on low-risk nodes, improving the overall operating efficiency of the system.

[0105] The key management module and the policy execution module form a closed-loop encryption policy adjustment mechanism through data-driven threshold derivation and deviation analysis. The former is based on historical pattern matching and parameter difference calculation to ensure the scientificity of the key threshold; the latter realizes the precision of the intervention scheme through fluctuation trend and weight reconciliation. The two work together to enable the system to dynamically adjust the encryption policy according to the real-time communication data of the target user, while ensuring the communication security, and to provide flexible and efficient technical support for the intervention of the non-suicidal self-injury behavior of depressed adolescents.

[0106] It should be noted that the relational terms herein such as first and second and the like are used solely to distinguish one entity or action from another, without necessarily requiring or implying any such actual relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus.

[0107] Although embodiments of the present application have been shown and described, it is to be understood that various modifications, substitutions, replacements and changes can be made to these embodiments without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A secure communication system for intervening in non-suicidal self-injurious behavior in depressed adolescents, characterized in that, Comprise: A data acquisition module for obtaining communication behavior data and emotional state parameters of a target user authorized by the target user, setting an encrypted communication interval matching the user behavior characteristics, the encrypted communication interval being a time range for subsequent data processing; A behavior analysis module for dividing a plurality of behavior nodes within the encrypted communication interval, performing pattern recognition on the communication data of each behavior node, and generating an interaction feature vector corresponding to the behavior node; A dynamic evaluation module for extracting high-risk behavior indicators from the interaction feature vector, establishing an encrypted intervention rule associated with the behavior node, and obtaining a key control parameter corresponding to the encrypted intervention rule; A channel optimization module for identifying a communication security level in the key control parameter, dynamically encrypting the high-risk behavior indicators according to the communication security level, and calculating the data fluctuation density of each behavior node under different encryption strategies; A key management module for deriving an optimal key threshold value according to the data fluctuation density, comparing the current communication data density with the optimal key threshold value, and generating a key deviation sequence; A strategy execution module for analyzing the key deviation sequence, integrating the key deviation sequence into a communication intervention execution scheme based on the data fluctuation characteristics of the behavior node; The implementation of the behavior analysis module includes: constructing a user feature library corresponding to the behavior node, the user feature library containing a behavior parameter vector mapped from the communication behavior data and the emotional state parameter; Similar behavior matching is performed on the behavior parameter vector, and the behavior parameter vector is divided into a behavior clustering group according to the matching result; the distribution center of the communication data is extracted from the behavior clustering group, and the distribution center is set as the behavior node; The division of the behavior clustering group of the behavior parameter vector also includes: According to the communication frequency and emotional fluctuation index in the behavior parameter vector, the communication duration, interaction interval and emotional intensity parameters are extracted, and the behavior feature label is generated based on the communication duration, interaction interval and emotional intensity parameters; The behavior feature label is associated with the behavior parameter vector, the behavior similarity between the feature labels is calculated, and the parameter vectors with a similarity higher than a preset behavior threshold are selected to form a behavior clustering group; The implementation of generating an interaction feature vector corresponding to the behavior node includes: For each behavior node, the data fluctuation parameters of the behavior node within a preset period are obtained according to the time sequence distribution of the behavior node in the encrypted communication interval, and the communication fluctuation coefficient of the node is calculated; When the communication fluctuation coefficient exceeds a first risk threshold, the node is marked as a high-risk node, and its communication data is extracted to form an interaction feature vector; when the communication fluctuation coefficient is lower than the first risk threshold, the node is marked as a safe node, and the communication data of adjacent nodes of the node is superimposed in density, and the superimposed data is reconstructed into an interaction feature vector; The implementation of the dynamic evaluation module includes: Separate the communication data proportion, abnormal interaction proportion and emotional fluctuation parameter from the interaction feature vector, and generate the encrypted intervention rule of the behavior node based on the above parameters; If the number of behavior nodes covered by the current encrypted intervention rule is less than a preset behavior threshold, the interaction feature vectors of adjacent behavior nodes are traversed, and the communication indicators not included in the encrypted intervention rules of the adjacent nodes are added to the current rule; The implementation of the channel optimization module comprises: obtaining a time sequence parameter of a key validity period in a communication security level and a fluctuation parameter of data encryption strength; constructing an encryption state network associated with the time sequence parameter and the fluctuation parameter, and determining data fluctuation density under different encryption strategies according to switching probabilities of various paths in the encryption state network; The construction of the encryption state network further comprises: identifying a periodic mode of the time sequence parameter, and setting the time sequence parameter as a starting point of the encryption state network if the current periodic mode completely matches a preset communication period; calculating switching correlation degrees of the time sequence parameter and the fluctuation parameter, and generating intermediate points and termination points of the encryption state network in turn according to the switching correlation degrees from high to low; performing state reverse verification on the termination points, and outputting a final link of the encryption state network when the switching correlation degree of the termination point is lower than a preset switching threshold; The implementation of the data fluctuation density calculation comprises: statistically obtaining a mean value of the time sequence parameter and a range of the fluctuation parameter of each termination point in the encryption state network, and calculating a global covariance of all node parameters; obtaining a difference between the mean value of the time sequence parameter of a single termination point and the mean value of the time sequence parameter of an adjacent node, dividing the difference by the global covariance to obtain a time sequence density coefficient; and simultaneously calculating a ratio of the range of the fluctuation parameter to the global covariance, and weighting and summing the ratio and the time sequence density coefficient as the data fluctuation density of the node; The implementation of deriving the optimal key threshold comprises: extracting an encryption mode in historical data that is closest to the current data fluctuation density, calculating an Euclidean distance between the data fluctuation density in the closest encryption mode and the current data fluctuation density in a time sequence distribution as a first key reference value; statistically obtaining a difference between the data fluctuation density in the closest encryption mode and the current data fluctuation density in the number of peak points, and taking the difference number as a second key reference value; linearly combining the first key reference value and the second key reference value to match an optimal key threshold in a preset key threshold table; The implementation of the strategy execution module comprises: dividing a forward fluctuation domain and a reverse fluctuation domain according to data fluctuation trends of each node in a key deviation sequence; extracting a convergence frequency of the key deviation in the forward fluctuation domain and a diffusion frequency of the key deviation in the reverse fluctuation domain, dynamically reconciling the two according to communication weights of nodes, and generating a configuration parameter of a communication intervention execution scheme.

Images