Estrus-sharing type health behavior intervention assisting dialogue robot based on self-consciousness emotion recognition

Abstract

The invention discloses an emotional assistant health behavior intervention dialogue robot based on self-consciousness emotion recognition. The emotional assistant health behavior intervention dialogue robot at least comprises a health behavior information acquisition module, a man-machine dialogue module, a self-consciousness emotion recognition module and a behavior intervention module. By combining a self-consciousness emotion recognition technology with a large model interaction technology and a behavior intervention psychological model, the system can improve the fine granularity of health behavior intervention in the aspect of emotion understanding, and further realizes more accurate and personalized intervention services. The system not only can promote the change of health behaviors, but also can provide a common intervention service, so that the health management ability of the user is effectively improved, and the durability and effect of behavior change are enhanced.

Description

Technical Field

[0001] The present invention belongs to the technical field of human-computer interaction, and is particularly suitable for assisting users in regulating their own health behaviors. It mainly relates to an empathetic health behavior intervention assistance robot based on self-awareness emotion recognition. Background Art

[0002] In recent years, with the continuous improvement of economic level, people's attention to health is no longer limited to clinical aspects such as disease diagnosis and treatment, but they strive to enhance health literacy and maintain a healthy lifestyle in daily life in order to achieve the purpose of strengthening their body and reducing the risk of disease.

[0003] Health behavior management aims to have a positive impact on personal health-related behaviors by designing effective intervention mechanisms. It is an important means to change users' health behaviors and cultivate a healthy lifestyle. A large number of studies have shown that unhealthy behaviors such as late sleep, lack of exercise, and improper diet will greatly increase the risk of various diseases such as cardiovascular disease, diabetes, liver disease, and even cause psychological diseases such as depression and autism. However, the current health literacy, healthy lifestyle and behavioral literacy of Chinese residents are at a low level, making it difficult to maintain healthy behaviors, which is not conducive to improving the physical fitness of the entire Chinese people. Therefore, effective health behavior intervention methods are of great significance to improving the health level of individuals and even the Chinese people.

[0004] Human-computer interaction is a discipline that studies the interaction between people and computers, aiming to design user-centered interactive technologies to improve people's quality of life. As the importance of health behavior management becomes known, more and more human-computer interaction research focuses on developing persuasive technologies for health behavior intervention. Most of the work in this field focuses on health behavior tracking and visualization. The system records health behaviors through sensors or relies on user self-reports, and visualizes user behaviors through text, numbers, sounds, gamification elements, and even 3D printed models. However, these methods rely too much on the user's self-discipline and health information processing capabilities, and ignore the user's personal emotions and initiative, that is, they cannot supplement health-related knowledge according to the user's individual situation, cannot adjust health recommendations based on the user's subjective emotions, and cannot support negotiations between users and intelligent systems.

[0005] In the field of psychology and behavioral science, previous studies have demonstrated the effectiveness of various intervention techniques in improving behavior. These interventions are generally divided into the following categories: 1) health behavior education, 2) cognitive behavioral therapy (CBT), and 3) relaxation, mindfulness, and hypnosis. However, these methods usually require personalized customization by health behavior experts, which is time-consuming and expensive, and difficult to be universal to the entire population. Therefore, the field is in urgent need of more accessible and cost-effective interventions suitable for a wider population. With the emergence of chatbots, many studies hope to design chatbots with the professional capabilities of health behavior experts and integrate persuasive strategies into chatbot dialogue design for effective behavioral intervention. However, most of the current work still uses pre-designed dialogues, which have great limitations in terms of freedom and empathy. Although empathetic dialogue has been a research hotspot in the field in recent years, most of the current work focuses on identifying and responding to basic emotions, such as happiness, sadness, or anger, but there is limited exploration of dealing with more complex, self-conscious emotions, such as shame, guilt, and pride. According to psychological research, these self-aware emotions are crucial in promoting behavioral change, but most existing chatbots cannot effectively understand and respond to them. Therefore, how to design a chatbot with high degree of freedom, high-level empathy, and effective health behavior intervention has become a hot topic for researchers at home and abroad. Summary of the invention

[0006] Technical problem: The present invention aims at the problems in the prior art and provides an empathic auxiliary health behavior intervention dialogue robot based on self-conscious emotion recognition. The robot supports user-defined behavior goals: such as the daily target sleep time. At the same time, the robot records the user's actual health behavior and mood to provide further support and intervention. The robot supports active and passive interactions. When the actual behavior of the user collected by the robot does not meet the target behavior, the robot initiates active interaction, understands the reasons for the user's inappropriate behavior and the emotions behind it in the communication with the user, provides empathic persuasion interaction, and corrects the user's behavior. When the user actively initiates a dialogue, the robot implements passive interaction, provides empathic responses based on the user's input dialogue and the self-conscious emotions reflected in the dialogue, and supports the user's behavior change. Whether in active or passive interaction, the robot can generate text that meets the guidelines of cognitive behavioral therapy to achieve effective behavioral intervention. At the same time, the robot can recognize and understand the user's basic emotions and complex self-conscious emotions, such as guilt, embarrassment, shame and pride, and effectively respond to these emotions in the intervention dialogue to improve the intervention effect.

[0007] In order to achieve the above-mentioned purpose, the technical solution adopted by the present invention is an empathy-assisted health behavior intervention dialogue robot based on self-awareness emotion recognition, which at least includes a health behavior information collection module, a human-computer dialogue module, a self-awareness emotion recognition module and a health behavior intervention module.

[0008] The health behavior information collection module is configured to support the user in setting behavior change goals, and record actual behaviors, and input the above results into the human-computer dialogue module;

[0009] The human-machine dialogue module is used to manage the communication between the user and the robot, receive and process the user's input, so as to perform emotion recognition and behavior intervention;

[0010] The self-awareness emotion recognition module: recognizes emotions based on the text input by the user in the human-computer dialogue, especially self-awareness emotions including guilt, shame, embarrassment and pride, and the recognition results are transmitted to the behavior intervention module;

[0011] The health behavior intervention module generates text of empathetic persuasive behavior intervention based on the user's behavior, user goals, and current emotions, and inputs it into the human-computer dialogue module. The text conforms to cognitive behavioral theory and the theory of self-awareness emotion formation in psychology, ensuring the correct understanding and effective response to the user's emotions, as well as effective intervention in behavior.

[0012] As an improvement of the present invention, the health behavior collection method includes one or more combinations of user uploading, mobile phone recording, wearable device collection, and motion capture device recording;

[0013] The health behavior information collection module records and cleans data based on multiple cloud services and records user data according to server time.

[0014] As an improvement of the present invention, the human-computer dialogue module at least includes text dialogue interaction, which can be combined with one or more combinations of audio, video, etc. The human-computer dialogue module includes two modes: actively initiating dialogue and passively answering dialogue.

[0015] As an improvement of the present invention, the self-awareness emotion recognition module realizes the recognition of self-awareness emotions, including guilt, embarrassment, shame, and pride, based on the user input text message, and transmits the emotion result label to the intermediate forwarding server.

[0016] The self-conscious emotion recognition module uses a two-stage recognition model for emotion recognition. The model includes the first stage of identifying whether the user input text contains self-conscious emotions; the second stage: further classifying the text that has been identified to contain self-conscious emotions to determine the specific type of self-conscious emotions, including guilt, shame, embarrassment or pride; compared with the method of directly identifying specific self-conscious emotions and general emotions, this method can effectively reduce the interference of category imbalance and insufficient data set on model training without reducing the recognition accuracy of specific self-conscious emotions, thereby improving the stability of the training model and reducing the difficulty of training. The self-conscious emotion recognition module is based on multiple data sources such as text, voice, image, and human posture, which are used alone or in combination to achieve emotion recognition.

[0017] As an improvement of the present invention, the health behavior intervention module uses a large language model to process the user's health behavior information, input text, recognized emotional results, current time and other inputs to generate personalized interventions and suggestions. The generated answer content is forwarded to the user end for display through the module; the health behavior intervention module integrates the user's health behavior information and current time context information, and applies cognitive behavioral therapy (CBT) theory to design health behavior intervention strategies, and provides behavior adjustment suggestions based on corresponding strategies for individual user health behavior problems.

[0018] As an improvement of the present invention, the health behavior intervention module can dynamically match the user's emotional state and select a more appropriate empathy response strategy by designing a more detailed empathy level adjustment mechanism, such as providing more comforting language when the user is frustrated and preferring action suggestions when the user is emotionally stable.

[0019] As an improvement of the present invention, the healthy behavior intervention module can add a functional module to assist users in behavior adjustment and emotion regulation based on the dialogue. For example, the system can help users regulate their emotions by providing real-time breathing training guidance and mindfulness exercises. At the same time, it provides clear goal decomposition, progress tracking and reward feedback mechanisms for behavior management, helping users gradually achieve healthy behavior goals and promote continuous behavior changes.

[0020] The empathetic health assistance intervention dialogue robot based on self-awareness emotion recognition includes the following steps:

[0021] S1: When a user uses a chatbot based on a mobile phone APP for the first time, the backend will send the device code of the user's device to the server. The server will query the user's device for the first time and get the result that the user's device code does not exist. After getting the result, it will send a message to the user to submit the user name and set the basic expected behavior change target. If the user has already registered and submitted the information, this step will be skipped.

[0022] S2: Users can submit relevant health behavior data for the day in the system, and then enter the dialogue interaction page.

[0023] S3: When used for the first time, the system will first greet the user, and then the user can interact with the software according to personal circumstances. The system will send the user's text input to the server for processing and generate corresponding replies. This is the active mode. After the first use, this mode mainly occurs when the user's actual behavior is inconsistent with the set goals. The system actively asks questions or suggestions and provides empathetic behavioral persuasion.

[0024] S4: When the user actively initiates a conversation, including asking for suggestions or expressing concerns, the system adopts a passive mode. In this mode, the system provides targeted intervention suggestions and support based on the specific content and emotions input by the user and based on the established cognitive behavioral therapy and motivational interviewing theories.

[0025] Beneficial effects: Compared with the prior art, the present invention has the following advantages:

[0026] (1) The invention combines computer technology with behavioral theory, integrating artificial intelligence technology and human-computer interaction technology, and is used in the field of health behavior intervention. By combining the self-conscious emotion generation mechanism, cognitive behavioral theory, etc., the invention provides multiple and scientific support in health behavior intervention. This interdisciplinary integration makes the intervention method more comprehensive and can help users achieve health behavior changes more effectively.

[0027] (2) The chatbot of the present invention not only provides rational information support, but also can identify and understand the user's self-conscious emotions, such as guilt, shame, embarrassment and pride, and generate appropriate persuasive feedback based on emotional empathy strategies. Compared with traditional interaction methods, the present invention is more empathetic while providing behavioral guidance, helping users feel supported and understood during the intervention process, thereby improving the intervention effect.

[0028] (3) Compared with traditional behavioral intervention programs that rely on health behavior experts, this invention is not limited by time, location, and human resources, and users can obtain personalized health behavior support anytime and anywhere. This autonomous design makes the intervention more flexible and convenient, and is particularly suitable for user groups that have difficulty obtaining traditional health behavior guidance, thus expanding the popularity of health behavior intervention.

[0029] (4) The present invention fully utilizes the advantages of human-computer interaction and realizes human-centered interaction, which can adjust the intervention strategy and feedback content in real time based on the user's health behavior and emotional state. Whether in active or passive interaction mode, the system can generate personalized text content that conforms to the principles of cognitive behavioral therapy based on the user's current state and goal progress, thereby effectively promoting the user's behavior change.

[0030] (5) The present invention is aimed at the health behavior management needs that are prevalent in modern society, and is particularly suitable for individuals who need to improve sleep, increase exercise, or adjust diet and other health behaviors. The system has strong applicability and a wide range of services. Regardless of age, gender or cultural background, it can provide appropriate health behavior support. It has significant social application value and helps to improve the health level of the entire social group. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0032] Figure 1 Workflow diagram for the intervention system

[0033] Figure 2 It is a workflow diagram of the intervention system in the field of sleep regulation;

[0034] Figure 3 It is the software user interface of the system (sample picture);

[0035] Figure 4 This is a flow chart of the interaction between the front-end and back-end of the system. DETAILED DESCRIPTION

[0036] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. Note that the following embodiments are merely illustrative in nature, and the present invention is not intended to limit its applicable objects or uses, and the present invention is not limited to the following embodiments.

[0037] To facilitate understanding, the nouns involved in the example are first explained.

[0038] Prompt Engineering: In the field of natural language processing (NLP), it is the process of designing and optimizing input text (i.e., prompts) to effectively guide AI models to generate the desired output. With the widespread application of large language models such as GPT, prompt engineering has become increasingly important. Its core goal is to influence the quality and accuracy of the model's response by designing accurate, concise, and instructive prompts. Prompt engineering has become increasingly important for leveraging existing large-scale pre-trained models (such as the GPT series of models) because these models do not need to be retrained, but their performance depends to a large extent on whether the input prompts are clear and accurate. By optimizing prompts, the effectiveness of the model can be improved and a lot of time and resources can be saved. The key elements of prompt engineering include (1) Designing effective prompts: The output of the language model is highly dependent on the input content. Prompt engineers need to find a suitable way to provide instructions to the model and clarify the expected answer type or context. (2) Prompt optimization and iteration: The initial prompt may not produce the expected results, so continuous iteration and optimization are required. (3) The importance of context and examples: Providing appropriate context information in the prompt, even including multiple examples, can help the model better understand the user's intentions.

[0039] Cognitive behavior therapy: It is a widely used psycho-social intervention therapy that can improve mental health. CBT focuses on the development of personal coping strategies, with the goal of solving the client's current problems, changing their incorrect cognitions (e.g., thoughts, beliefs, and attitudes), behaviors, and unhelpful patterns of emotion regulation, thereby relieving the client's distress and reducing their related symptoms. Cognition refers to how we think and understand what happens to us. CBT believes that people's emotional and behavioral responses are not only triggered by external events, but more by their cognition or interpretation of these events. Negative or irrational cognitions often lead to emotional distress. The way we behave can exacerbate or alleviate emotional problems. CBT emphasizes that by identifying and changing unfavorable behavior patterns, individuals can learn healthier coping mechanisms to help them better cope with stress and challenges in life. Cognitive behavioral therapy is used to treat depression and can also be used to treat many mental illnesses.

[0040] Embodiment 1:

[0041] Figure 1 A behavioral intervention system based on emotion recognition and coping strategies is described, which at least includes an input and interaction module, a self-awareness emotion recognition module, a situation analysis module, a strategy configuration module, a model reasoning and a psychological intervention module.

[0042] The input and interaction module is mainly designed according to the needs of software interaction, the concepts and types of behavioral intervention, theoretical guidance and needs, to input the user's output, such as text, voice, video, etc., into the software according to certain rules, and use the interactive performance on the user side as a carrier for the user to interact with the software to implement specific intervention measures for the user. At the same time, the input and interaction module will form a direct data flow and information exchange with the self-awareness emotion recognition module and the data processing module, analyze the data according to the corresponding strategies and intervention therapies, and integrate the appropriate information for processed output. This module is a module that users can interact directly with. It is a medium for users to submit input to the system according to the needs of behavioral intervention and obtain corresponding intervention measures according to the strategy of behavioral intervention.

[0043] The self-awareness emotion recognition module is mainly responsible for directly interacting with the user input content and passing the recognition results to the strategy configuration module and the data processing module, etc. According to the theoretical support of behavioral intervention, with appropriate strategy guidance, after data processing and analysis, it constitutes the output content that is beneficial to the needs of the psychological intervention module. The actual implementation of this module will be highly related to the media selection of behavioral intervention. Different information input methods will directly determine the model carried by the module and the processing logic of the input information. The results of the module output will determine the specific behavioral intervention strategy selected in the strategy configuration module, thereby affecting the output results of the final psychological intervention module.

[0044] The scenario analysis module mainly designs different corresponding methods for user input at different time and space nodes based on the needs and theories of behavioral intervention. For example, a module that analyzes past data is enabled before a fixed time node, and an intervention module related to possible scenarios is enabled before another time node. The design of this module can be related not only to the time and space characteristics of real life, but also to the user's own personality settings. For example, if some users want to obtain a more emotional and proactive intervention system in a certain period, the intervention components and intervention intensity can be increased according to the user's desired scenario. This module also includes analysis and processing of daily data, not just responding to a single user interaction, such as real-time personalized analysis of the basic information submitted by the user himself and the real-time situation when using the software. This module increases the content and richness of system interaction, and also provides users with the freedom to choose different styles of systems.

[0045] The strategy configuration module is mainly designed to design a set of strategy guidance corresponding to the processing results of different user outputs based on the needs and theories of behavioral intervention. The design form of these strategies is designed according to the form of psychological intervention modules. In a system with text as an interactive carrier, the strategy is usually designed as direct feedback based on the analysis results of user output. In a software system using a large model, the strategy can also be designed as a combination of the prompt words of the large model and direct feedback from other software systems, which increases the richness of the system interaction experience. In the emotion-related large model-based intervention system, the prompt words will be designed as corresponding guidance strategies based on the causes and coping methods of emotions.

[0046] The model reasoning and psychological intervention module, after obtaining a preliminary analysis of user data, such as sentiment analysis, and selecting corresponding strategies, mainly uses the big model as a reasoning module, uses text as an output interaction medium, takes user input and sentiment analysis results as direct inputs for reasoning, and uses strategy guidance as a requirement, so that the big model analyzes the user's emotional needs and gives corresponding psychological intervention measures based on the needs of strategy guidance, user input, and sentiment analysis results.

[0047] At the same time, in the psychological intervention module, in addition to the output results generated by the interactive module of the large model, the software system is also designed to have a variety of other interactive components. The system can use the user's basic input information to actively interact with the user around the user's basic needs, such as notifying the user to start the behavioral intervention system, and providing prompts and guidance to promote the user's mental health after obtaining the user's daily submitted information. The above components can be turned on or off by the user according to the user's usage habits.

[0048] This system will use sentiment analysis and daily interactions, utilize the reasoning capabilities of the big model, and combine big model reasoning with emotional needs, thereby helping cognitive behavioral therapy in specific areas, guiding changes in patients' thinking, cognition, and behavior, and ultimately achieving effective treatment of related psychological disorders.

[0049] Embodiment 2:

[0050] Figure 2 Describes the system flow of how current software uses emotion recognition and coping strategies to provide sleep assistance to users.

[0051] The following is a detailed description of the process:

[0052] 1. User input (input and interaction module)

[0053] The user enters a sleep-related problem or problem, such as staying up late or having difficulty falling asleep. This input can be provided in text form, which may describe the user's current state, mood, or sleep problem; it can also be entered in the form of filling in data in a dialog box, describing basic quantitative data related to sleep.

[0054] 2. Self-awareness emotion recognition model (self-awareness emotion recognition module)

[0055] The user's text input is first processed by the "Self-Awareness Emotion Recognition Model". The role of this model is to analyze the user's text and identify the user's emotional state, especially emotions related to self-awareness, such as shame, guilt, pride, and embarrassment. These emotions are usually related to the user's behavior patterns (such as staying up late or irregular work and rest), and the model helps identify the user's potential emotional drivers through analysis.

[0056] 3. Sleep goals and situation analysis (situation analysis module)

[0057] The large language model will analyze the user's target sleep time and the user. This includes: time factors, such as how long the current time is from the user's target sleep time. This user information will be compared and analyzed with the user's basic information and general information based on specific scenario requirements.

[0058] 4. Persuasion strategy and empathy strategy (strategy configuration module)

[0059] To help users achieve good sleep goals, the system uses a variety of strategies:

[0060] Persuasion strategies: These strategies are used to persuade users to adopt better sleep behaviors. They are basic prompt words used in system prompt engineering, including affirmation, persuasion, and reflection.

[0061] Empathy strategy: In order to better deal with users' self-conscious emotions, the system will adopt empathy strategies to convey understanding and support to users and help them alleviate possible negative emotions.

[0062] In the empathy strategy, for different emotions recognized in the user's input text, we will also design different strategy guidance as prompt word input.

[0063] The prompt words in strategy design are shown in the following table:

[0064]

[0065]

[0066] The strategy combinations used for different emotion recognition are shown in the following table:

[0067]

[0068] Based on the above two tables, clear strategic guidance is given in setting basic prompt words and determining corresponding prompt words according to recognized emotions.

[0069] 5. Large language model processing (model reasoning)

[0070] After the user's emotion is recognized, the input text and recognition results are passed to a large language model (the Qwen model is used in this example). Combined with the strategy guidance for the corresponding emotion as a prompt word, the model generates a personalized, sleep-related response based on the input. The model takes into account the user's goals and emotional state and provides adaptive suggestions or dialogues.

[0071] 6. ZzzMate App Generates Empathic Response (Sleep Intervention Module)

[0072] Based on the above analysis and strategies, the system generates an empathic response presented by the ZzzMate application on the user's mobile phone interface. The goal of these responses is to promote good sleep habits for users, relieve emotional stress, and guide them to better follow their sleep goals. The overall process combines self-aware emotion recognition, language model generation, and the use of persuasion and empathy strategies, combined with notification functions, information collection functions, and daily data analysis functions to help users adjust their behaviors and emotions to achieve better sleep results.

[0073] When users use the software, they will see several interfaces, such as Figure 3 As shown in the sample picture.

[0074] First, the user will enter the main interface, where they can see an interface picture that matches sleep, which is an image of a little girl sleeping, which determines the basic application purpose of the software, which is to help users sleep.

[0075] After clicking the start button, the user will see different subsequent pop-up windows based on their use of the software. If the user has not used the software, the software will initiate an information column requesting the user to submit the user name and set basic expected sleep information. If the user has already used and submitted the information, this step will be skipped. If the user opens the software for the first time that day, the server will obtain the result that the user has not logged in that day after querying. After obtaining the result, it will initiate an information column requesting the user to submit relevant sleep information. If the user has already logged in and submitted the information that day, this step will be skipped.

[0076] After submitting basic information, users can enter the interactive page. During the interactive phase, the system will first greet the user, and then the user can interact with the software based on their personal sleep conditions. The software will send the user's text input to the server for processing and generate corresponding replies. The system will also have different greetings and information extraction logic and reply logic at different times. The model with sentiment analysis and sentiment guidance can also analyze and reply based on the daily information entered by the user and the basic information set by the user, and provide analysis and guidance on daily sleep conditions.

[0077] The software uses the server to achieve interaction such as Figure 4 As shown in the figure, the user front-end interaction module involves direct user operations and interactions, and includes four main functions:

[0078] 1) Set a sleep time: The user sets a desired bedtime in the system. This goal will be used as a reference point for subsequent notifications and conversations.

[0079] 2) Daily Notifications: Based on the sleep goals set by the user, the system sends reminders or notifications to guide the user to prepare for sleep. The content of these notifications is personalized based on the user's emotional state or sleep progress.

[0080] 3) Submit daily information: When the user logs into the software for the first time every day, a dialog box will pop up asking the user to provide information related to bedtime, sleeping mood, etc.

[0081] 4) User input: The system provides a dialogue with the user. The user can enter the interactive content, which will be submitted to the backend together with the information provided in the previous steps.

[0082] In the backend processing, i.e. the data processing and emotion recognition module, the user input will go through a series of processes with the goal of understanding the user's emotional state, especially emotions related to self-awareness. First, the data will be subjected to basic text processing. This module includes the following steps:

[0083] 1) Data preprocessing: First, preprocess the user’s input data. Use the big model to optimize the user’s original input, such as translation and correcting sentence errors, to facilitate subsequent model analysis.

[0084] 2) Emotion Recognition Model: The preprocessed data is fed into the Self-Awareness Emotion Classification Model. The purpose of this model is to identify the user's emotions, especially those related to self-awareness, such as shame, guilt, pride, or embarrassment. Under different emotional states, the user's behavior may be driven by different emotions, which in turn affects their sleeping habits. The emotion classification model outputs the user's specific emotional state. This emotional information will serve as an important basis for generating personalized responses later.

[0085] The language model response generation module generates personalized responses based on user input and sentiment analysis results:

[0086] 1) User input information integration: This is a collection of information submitted by users, including the following:

[0087] ① User input: This is text information provided by the user, such as describing their current mood or sleeping problems.

[0088] ② Self-awareness emotion: This is the user’s emotional information generated by the emotion classification model, such as shame, guilt, etc., which reflects the user’s inner emotional state.

[0089] ③Designed prompts: These are prompts designed based on user input and self-aware emotions, with the goal of guiding the large language model to generate targeted suggestions or conversation content.

[0090] 2) LLM (Large Language Model) Reasoning: Next, the user’s input and the designed prompts are passed to the large language model. The model generates responses based on the context, which are personalized suggestions based on the user’s current emotional state and sleep goals.

[0091] 3) Empathy output: The responses generated by the large language model are passed back to the user. These responses may include encouraging words, helping users reflect on their emotions, or providing direct sleep advice.

[0092] Through the above steps and technical routes, we designed a software that can be installed on Android phones, can be used online at any time to obtain responses, and collect basic information permitted by users for analysis to generate responses.

[0093] This case is dedicated to developing an interactive system for psychological relaxation and counseling for people of all ages based on large models and behavioral intervention technology. It strives to attract people of all ages to actively receive consultations by virtue of the high convenience, fun and privacy of human-computer interaction of mobile phone software, and provide new methods for people of all ages to relieve anxiety, understand sleep information, promote sleep quality, and relax their bodies and minds, so as to effectively improve the mental health of the population. When the system matures, it is hoped that it can be expanded to more people with psychological treatment needs, and truly serve social and economic development and people's daily lives.

[0094] It should be noted that the above content only illustrates the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. For ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications all fall within the protection scope of the claims of the present invention.

Claims

1. An empathetic assisted health behavior intervention dialogue robot based on self-awareness emotion recognition, characterized in that: It at least includes a health behavior information collection module, a human-computer dialogue module, a self-awareness emotion recognition module and a health behavior intervention module. The health behavior information collection module is configured to support the user in setting behavior change goals, and record actual behaviors, and input the above results into the human-computer dialogue module; The human-machine dialogue module is used to manage the communication between the user and the robot, receive and process the user's input, so as to perform emotion recognition and behavior intervention; The self-awareness emotion recognition module: recognizes emotions based on the text input by the user in the human-computer dialogue, especially self-awareness emotions including guilt, shame, embarrassment and pride, and the recognition results are transmitted to the behavior intervention module; The health behavior intervention module generates text of empathetic persuasive behavior intervention based on the user's behavior, user goals, and current emotions, and inputs it into the human-computer dialogue module. The text conforms to cognitive behavioral theory and the theory of self-awareness emotion formation in psychology, ensuring the correct understanding and effective response to the user's emotions, as well as effective intervention in behavior.

2. The empathy-assisted health behavior intervention dialogue robot based on self-awareness emotion recognition according to claim 1 is characterized by: The health behavior collection methods include one or more combinations of user uploads, mobile phone records, wearable device collection, and motion capture device records; The health behavior information collection module records and cleans data based on multiple cloud services and records user data according to server time.

3. The empathy-assisted health behavior intervention dialogue robot based on self-awareness emotion recognition according to claim 1 is characterized by: The human-computer dialogue module at least includes text dialogue interaction, which can be combined with one or more combinations of audio, video, etc.

4. The empathy-assisted health behavior intervention dialogue robot based on self-awareness emotion recognition according to claim 1 is characterized by: The human-computer dialogue module includes two modes: actively initiating dialogue and passively answering dialogue.

5. An empathy-assisted health behavior intervention dialogue robot based on self-awareness emotion recognition according to claim 1, characterized in that: The self-awareness emotion recognition module realizes the recognition of self-awareness emotions, including guilt, embarrassment, shame, and pride, based on the user input text message, and transmits the emotion result label to the intermediate forwarding server.

6. An empathy-assisted health behavior intervention dialogue robot based on self-awareness emotion recognition according to claim 4, characterized in that: The self-conscious emotion recognition module uses a two-stage recognition model to perform emotion recognition. The model includes the first stage of identifying whether the user input text contains self-conscious emotions; the second stage: further classifying the text that has been identified to contain self-conscious emotions to determine the specific type of self-conscious emotions, including guilt, shame, embarrassment or pride. The self-awareness emotion recognition module is based on multiple data sources such as text, voice, image, and human posture, which are used alone or in combination to achieve emotion recognition.

7. The empathy-assisted health behavior intervention dialogue robot based on self-awareness emotion recognition according to claim 1 is characterized by: The health behavior intervention module uses a large language model to process the user's health behavior information, input text, recognized emotional results, and current time inputs to generate personalized interventions and suggestions. The generated answer content is forwarded to the user end through this module for display; The health behavior intervention module integrates the user's health behavior information and current time context information, applies cognitive behavioral therapy (CBT) theory to design health behavior intervention strategies, and provides behavior adjustment suggestions based on corresponding strategies for individual health behavior problems of users.

8. The empathy-assisted health behavior intervention dialogue robot based on self-awareness emotion recognition according to claim 6 is characterized by: The behavioral intervention module designs empathetic response strategies for each self-conscious emotion based on emotion formation theory, self-conscious emotion coping style theory, etc., so that the answers output by the robot have empathy capabilities.

9. The empathy-assisted health behavior intervention dialogue robot based on self-awareness emotion recognition according to claim 1, characterized in that: The method of use includes the following steps: S1: When a user uses a chatbot based on a mobile phone APP for the first time, the backend will send the device code of the user's device to the server. The server will query the user's device for the first time and get the result that the user's device code does not exist. After getting the result, it will send a message to the user to submit the user name and set the basic expected behavior change target. If the user has already registered and submitted the information, this step will be skipped. S2: Users can submit relevant health behavior data for the day in the system, and then enter the dialogue interaction page. S3: When used for the first time, the system will first greet the user, and then the user can interact with the software according to personal circumstances. The system will send the user's text input to the server for processing and generate corresponding replies. This is the active mode. After the first use, this mode mainly occurs when the user's actual behavior is inconsistent with the set goals. The system actively asks questions or suggestions and provides empathetic behavioral persuasion. S4: When the user actively initiates a conversation, including asking for suggestions or expressing concerns, the system adopts a passive mode. In this mode, the system provides targeted intervention suggestions and support based on the specific content and emotions input by the user and based on the established cognitive behavioral therapy and motivational interviewing theories.

Images