A psychological counseling level bidirectional psychological buffer method based on multi-scene AI voice analysis and intelligent earphone

Abstract

This invention relates to the field of artificial intelligence psychological intervention technology, and in particular to a psychological counseling-level two-way psychological buffering method and smart earphone based on multi-scenario AI voice analysis. The method includes the following steps: collecting the speaker's voice and converting it into text and emotional information; training a neural network model using a dataset composed of emotional information to obtain an AI emotion model; analyzing and processing the two-way psychological buffering based on the wearer's personal personality traits, the level of harm in the dialogue, and text information, and outputting personalized psychological buffering scripts; recognizing text and emotional information from the speaker's voice, and extracting personal personality traits by combining the wearer's historical voice data; accurately calculating the dialogue harm score using a dialogue harm scoring formula, and determining whether to activate the two-way psychological buffering mechanism based on the harm level, thereby achieving timely emotional intervention and protection, fundamentally eliminating self-denial, strengthening psychological boundaries, and preventing the intergenerational transmission and cross-scenario spread of emotions.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence psychological intervention technology, specifically to a psychological counseling-level two-way psychological buffering method and a smart headset based on multi-scenario AI voice analysis. Background Technology

[0002] In the current social environment, language-based psychological stress is characterized by its universality, cross-scenario nature, and bidirectional transmission across all population groups. It is not limited to parent-child scenarios; various groups face intense emotional impacts and psychological strain from language-based stress. Adolescents bear multiple pressures, including academic competition, high-pressure parental supervision, teacher criticism, and negative comparisons and exclusion among classmates. Parents' oppressive lecturing and comparative negativity often stem from their own anxieties about parenting, workplace stress, and life burdens, leading to a psychological imbalance that unconsciously transfers negative emotions to their children, creating a vicious cycle of parent-child psychological distress. Meanwhile, working professionals, as the core bearers of social pressure... Individuals often face multiple forms of verbal abuse, including verbal suppression from superiors, psychological pressure, performance pressure, negative comparisons from colleagues, malicious communication, difficult clients, and incessant lecturing in the workplace. Prolonged exposure to accusations, complaints, denials, and psychological pressure can easily trigger anxiety, depression, self-doubt, and burnout, and may even lead to the transfer of workplace stress to the family, becoming a source of negative emotions within the family. In addition, adults also need to cope with the repeated venting of negative experiences by elders, verbal trauma from their family of origin, and the comparative lecturing from relatives and friends. Various high-pressure professionals, freelancers, and students all face different sources of verbal psychological stress.

[0003] Such negative language is not simply a communication barrier; it is essentially an external projection of the speaker's own stress, anxiety, past trauma, and psychological imbalance. Listeners who passively endure this for extended periods are highly susceptible to developing various psychological problems such as depression and anxiety, and it can even create a chain of psychological stress transmission across generations and different scenarios, resulting in a lasting negative impact on themselves.

[0004] While some psychological counseling devices exist on the market, most are limited to passive intervention after the fact, making it difficult to intervene and protect during conversations in real time, and easily missing the best opportunity for emotional intervention. At the same time, these devices cannot accurately quantify the damage caused by conversations based on the individual personality characteristics of the user, and it is also difficult to carry out personalized and targeted psychological adjustment. Therefore, in response to the above pain points, this paper proposes a psychological counseling-level two-way psychological buffering method based on multi-scenario AI voice analysis and a matching smart headset. Summary of the Invention

[0005] The purpose of this invention is to provide a psychological counseling-level bidirectional psychological buffering method and a smart earphone based on multi-scenario AI voice parsing, so as to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: As an optional solution to the psychological counseling-level bidirectional psychological buffering method and smart earphone based on multi-scene AI voice parsing described in this invention, the psychological counseling-level bidirectional psychological buffering method and smart earphone based on multi-scene AI voice parsing includes the following steps: S1: Collect the speaker's voice and convert it into text and emotional information. Analyze the text information to obtain the speaker's identity information and dialogue scenario attribute information. Collect and convert the wearer's historical voice data to determine the wearer's personal personality traits; S2: Train the neural network model using a dataset composed of emotional information to obtain an AI emotion model. Input the obtained text information and emotional information into the AI ​​emotion model and output the emotion type. S3: Determine the level of harm in the conversation by combining emotion type and individual personality traits; when the level of harm reaches the negative threshold, activate two-way psychological buffer; the level of harm includes positive and negative thresholds; S4: Two-way psychological buffer analyzes and processes the wearer's personal characteristics, the level of harm in the conversation, and the text information, and outputs personalized psychological buffering messages through smart headphones.

[0007] As an optional solution to the psychological counseling-level bidirectional psychological buffering method and smart earphone based on multi-scene AI voice parsing described in this invention, wherein: a scenario-based statement classification database is constructed based on scenario attribute features; Identity information includes parents, teachers, leaders, colleagues, clients, elders, classmates, etc., and dialogue scenario attribute information includes parent-child communication, campus interaction, workplace office, workplace interpersonal relationships, family elder communication, etc.

[0008] As an optional solution for the psychological counseling-level bidirectional psychological buffering method and smart earphone based on multi-scenario AI voice parsing described in this invention, the identity information and text information are processed by text vectorization. Quantification is performed based on the emotion type in the emotional information: The base score for calmness is 0. Slight dissatisfaction has a base score of 2. Irritability has a base score of 4. Anger has a base score of 6. Anger / blame has a base score of 8. Abusive language / aggressive attacks have a base score of 10. By combining the base score of emotion type, volume and speech rate features are extracted from the emotion type to calculate the final quantitative value of emotional harm:

[0009] Where b is the final quantification value of emotional harm based on emotion type and tone intensity. The damage value is the value of the emotion itself. This is the volume deviation coefficient. This is the speech rate deviation coefficient, ranging from 0.8 to 1.2. The louder the volume and the faster the speech, the more intense the emotions, and the closer the emotional damage coefficient is to 1.2. The volume is even, the speaking speed is stable, and the emotional harm coefficient is close to 0.8; The identity information and text information are processed into text vectors, and combined with the final emotional harm quantification value to construct a formula for calculating the dialogue harm score, as follows:

[0010] Among them, the dialogue damage score is D, the wearer's identity information is α, the text information is a, the final emotional damage quantification value is b, K1 and K2 are the weighting coefficients of the corresponding information dimensions, and k1+k2=1; The dialogue damage rating D ranges from 0 to 10. Determine if the dialogue damage rating D is greater than 3; if yes, classify the corresponding dialogue damage level as abnormal; otherwise, classify the corresponding dialogue damage level as normal.

[0011] As an optional solution of the psychological counseling-level two-way psychological buffering method and smart earphone based on multi-scenario AI voice analysis described in this invention, the degree of harm of the dialogue is determined according to the dialogue harm score. The negative threshold is set based on the maximum and minimum values ​​of the degree of harm in historical data; The severity of damage is divided into two dialogue damage levels: normal and abnormal. The abnormal level is further divided into mild, moderate, and severe. When the dialogue damage level is judged to be normal, two-way psychological buffer is not enabled; When the damage level assessment during a conversation is abnormal, activate the two-way psychological buffer.

[0012] As an optional solution to the psychological counseling-level bidirectional psychological buffering method and smart earphone based on multi-scenario AI voice parsing described in this invention, the wearer's personal personality characteristics are quantified to construct a personality assessment calculation formula, defined as follows:

[0013] Where S is the extroversion index. The percentage of spoken words to total words. The percentage of times a topic is initiated proactively. The percentage of positive emotion words Standardized average speech rate , , and These correspond to the weighting coefficients of each indicator.

[0014] As an optional solution to the psychological counseling-level bidirectional psychological buffering method and smart earphone based on multi-scenario AI voice analysis described in this invention, the following is included: determining the emotional type of the speaker through text and emotional information, and determining the degree of harm of the conversation by combining the wearer's personal personality characteristics, including: A1: Based on the wearer's personal personality traits, which include introversion and extroversion; A2: Determine if the wearer's personality trait is introverted; if yes, proceed to A3; if no, the dialogue damage level remains the same. A3: Determine if the dialogue damage level is normal; if yes, the dialogue damage level remains unchanged; otherwise, the dialogue damage level increases by one level.

[0015] As an optional solution of the psychological counseling-level bidirectional psychological buffering method and smart earphone based on multi-scenario AI voice analysis described in this invention, it includes: earphone body, sound receiving module, micro air pump and earplug structure; A miniature air pump is installed inside one side of the earphone body. The output end of the miniature air pump is connected to an earplug structure. The earplug structure is installed on the outside of the earphone body. A sound receiving module for sound reception is also installed on one side of the earphone body. The earbud structure includes a connecting sleeve that connects to the earphone body and an inflation tube that connects to the output end of a miniature air pump. The connecting sleeve has a partition plate that divides the inside of the connecting sleeve into two chambers. An earbud support frame is also fixedly connected to one side of the connecting sleeve, and an inflation airbag is installed on the outside of the earbud support frame. The other end of the inflation tube is connected to the airbag inflation tube and the drying tube, and the other end of the airbag inflation tube is connected to the airbag, while the other end of the drying tube is connected to the earplug support frame. A humidity detection unit is installed on one side of the earbud support frame, and humidity adjustment vent and humidity adjustment air inlet are respectively opened on both sides of the earbud support frame.

[0016] As an optional solution of the psychological counseling-level bidirectional psychological buffering method and smart earphone based on multi-scenario AI voice analysis described in this invention, a discharge hole is also provided on one side of the connecting sleeve, and the chamber of the connecting sleeve where the discharge hole is located is connected to the humidity regulating air inlet.

[0017] As an optional solution of the psychological counseling-level bidirectional psychological buffering method and smart earphone based on multi-scenario AI voice analysis described in this invention, a first solenoid valve is installed on the outside of the airbag inflation tube, and a second solenoid valve is installed on the outside of the drying tube.

[0018] Compared with the prior art, the beneficial effects of the present invention are: By recognizing textual and emotional information from the speaker's voice, the system simultaneously obtains the speaker's identity information and the attributes of the dialogue scenario. Combined with the wearer's historical voice data, it extracts personal personality traits. Then, relying on the dialogue damage scoring formula, it can accurately calculate the dialogue damage score and determine whether to activate the two-way psychological buffer mechanism based on the damage level. This enables timely emotional intervention and protection, fundamentally eliminating self-denial, strengthening psychological boundaries, and blocking the intergenerational transmission and cross-scenario spread of emotions. These earphones automatically regulate the internal environment of the ear canal when worn, using a humidity detection unit to monitor the humidity level in the ear canal in real time. Given that a consistently damp ear canal is prone to the growth of fungi and bacteria, which can lead to problems such as otitis externa and fungal otitis externa, these earphones utilize a built-in micro-air pump to achieve slow airflow within the ear canal, maintaining a dry environment and providing ear canal protection. Attached Figure Description

[0019] Figure 1 A flowchart of a two-way psychological buffering method for psychological counseling based on multi-scenario AI voice parsing; Figure 2 A schematic diagram of the overall structure of a smart earphone; Figure 3 This is a schematic diagram of the earbud structure of a smart earphone; Figure 4 This is a schematic diagram of the discharge port structure of a smart earphone.

[0020] In the diagram: 1-Earphone body, 2-Radio module, 3-Miniature air pump, 4-Earplug structure, 401-Connecting sleeve, 402-Separator plate, 403-Earplug support frame, 404-Inflatable airbag, 405-Airbag inflation tube, 406-Drying tube, 407-First solenoid valve, 408-Second solenoid valve, 409-Inflation tube, 410-Emission hole, 411-Humidity adjustment exhaust hole, 412-Humidity adjustment air inlet, 413-Humidity detection unit. Detailed Implementation

[0021] Example 1: Please refer to Figure 1 The present invention provides a technical solution: A psychological counseling-level bidirectional psychological buffering method based on multi-scenario AI voice parsing is applied to a smart earphone device. The smart earphone has a built-in voice acquisition module, an AI processing module, and an audio output module. The AI ​​processing module preloads an AI emotion model and a scenario-based sentence classification database. The specific implementation steps are as follows: S1: Data Acquisition and Database Construction a. Voice Acquisition and Basic Information Conversion: The smart earphone uses a built-in high-sensitivity microphone to acquire the voice signals of both parties in a conversation (wearer and speaker) in real time. The sampling rate is set to 16kHz and the sampling precision is 16bit. The voice signal is converted into text information through an ASR algorithm. At the same time, the voice emotion recognition algorithm (based on Mel-frequency cepstral coefficients MFCC to extract voice features) is used to extract the speaker's emotional information (including preliminary determination of emotion intensity and emotion category). b. Identity Information and Scene Attribute Analysis: Keyword extraction (using the TF-IDF algorithm) is performed on the converted text information to identify the identity information of the speakers. This identity information includes parents, teachers, leaders, colleagues, clients, elders, classmates, etc. For example, when keywords such as "dad" and "mom" appear in the text, the speaker is identified as a parent; when keywords such as "General Manager Wang" and "Manager Li" appear, the speaker is identified as a leader; when keywords such as "Teacher Zhang" and "classmate" appear, the speaker is identified as a teacher and a classmate, respectively. Simultaneously, scene keyword analysis is performed on the text (e.g., "homework" and "exam" correspond to campus interaction; "work report" and "project coordination" correspond to workplace communication; "housework" and "caring for health" correspond to communication with family elders; "parent-child games" and "tutoring homework" correspond to parent-child communication) to obtain the dialogue scene attribute information. c. Collection of historical voice data and determination of personality traits of wearers: The smart earphones collect historical voice data of the wearer over the past 30 days (collection time of no less than 1 hour per day, covering voice interaction in different scenarios), convert it into text data through ASR algorithm, and extract indicators such as the number of spoken words, total number of words, number of times the wearer actively initiated topics, total number of topics, number of positive emotion words (such as "happy", "smooth", "keep going"), total number of words, number of questions / interaction sentences (such as "what do you think?" "is this okay?"), and average speech rate, thereby obtaining basic data for determining the wearer's personal personality traits; d. Construction of a scenario-based statement classification database: Based on scenario attribute characteristics (parent-child communication, school interaction, workplace, workplace interpersonal relationships, family elder communication, etc.), a scenario-based statement classification database is constructed. Each scenario corresponds to a sub-database, which contains common positive, neutral, and negative statements in that scenario. For example, negative statements in parent-child communication scenarios include "How can you be so stupid?" and "You can't even do this simple thing right." Negative statements in workplace scenarios include "Your efficiency is too low" and "This plan is completely unacceptable." Negative statements in school interaction scenarios include "Your grades are so bad, don't play with me." At the same time, the database is updated in real time, and every time 100 scenario-related statements are added, the database is automatically classified and archived.

[0022] S2: AI Emotion Model Pre-training and Inference a. Model Pre-training: The AI ​​emotion model adopts a CNN-LSTM hybrid network structure. The training dataset uses publicly available emotional speech datasets (such as the IEMOCAP dataset) combined with a custom-collected multi-scene speech dataset (covering scenarios such as parent-child, school, and workplace, with a total of 100,000 speech samples, each labeled with emotion type and speech type). During training, the text information and emotional information converted from speech are used as input, and the emotion type (happy, calm, angry, wronged, furious, depressed, etc.) is used as output. The model parameters are optimized by gradient descent. The training iterations are 50 rounds, the learning rate is set to 0.001, and the pre-training is completed when the model accuracy reaches more than 92%. b. Model Inference: Input the text and emotion information of the interlocutor obtained in S1 into the pre-trained AI emotion model and output the corresponding emotion type; for example, when the interlocutor's text is "How could you make a mistake again? You've disappointed me so much", the emotion information is "anger (intensity 8.5)". S3: Emotion Assessment and Harm Level Evaluation a. Identity and Scene Matching: Based on the identity information of the interlocutor identified in S1, the corresponding dialogue scene attribute information is matched. For example, if the interlocutor's identity is "parent", then the "parent-child communication" scene is matched; if the identity is "leader", then the "workplace office" scene is matched; if the identity is "teacher", then the "school interaction" scene is matched, ensuring the relevance of the subsequent harm assessment. b. Determination of the degree of harm: Combining the textual and emotional information of the interlocutor, as well as the wearer's personal personality traits obtained from S1, a score is calculated using the dialogue harm scoring formula to determine the degree of harm in the dialogue; the relevant calculation formula for dialogue harm scoring is as follows:

[0023] Among them, the dialogue harm score is D (ranging from 0 to 10), and the wearer's identity information is α (assigned a value based on the closeness of the relationship between the wearer and the identity, the degree of authority, etc., for parents). =0.8, Leader =0.9, teacher =0.85, colleague =0.6, customer =0.7, elders =0.8, classmate =0.5), text information is a (assigned based on the proportion of negative words in the text; a=0 when the proportion of negative words is 0%, and a=10 when the proportion is 100%), emotion information is b (calculated based on the final emotional harm quantification score of emotion type and tone intensity; when the score exceeds 10, it is calculated as 10), K1 and K2 are the weighting coefficients of the corresponding information dimensions, where K1=0.4 and K2=0.6 (obtained by fitting historical data to ensure the accuracy of the score); Harm severity level classification: The dialogue harm score D is calculated using the relevant formula, and corresponds to two dialogue harm levels: normal and abnormal. The abnormal level is further divided into mild, moderate, and severe. The specific classification criteria are as follows: Normal grade: D < 3 points, in which case two-way psychological buffering is not activated; Abnormal Level - Mild: 3 points ≤ D < 5 points; Abnormal grade - moderate: 5 points ≤ D < 8 points; Abnormal grade - Severe: D ≥ 8 points; When the level of harm reaches an abnormal level (i.e., the negative threshold is 3 points, which is set based on the maximum harm level of 10 points and the minimum harm level of 0 points in historical data, and takes 30% of the difference between the two as the negative threshold to ensure coverage of all possible negative dialogue scenarios), two-way psychological buffer is activated. c. Adaptive adjustment of damage level based on personality traits: The wearer's personal personality traits are quantified through a personality assessment, the calculation formula of which is as follows:

[0024] Where S is the extroversion index (ranging from 0 to 1). The percentage of spoken words to total words. The percentage of times a topic is initiated proactively. The percentage of positive emotion words Standardized average speech rate (normalizing the average speech rate to the 0-1 range). , , and These correspond to the weighting coefficients of each indicator, where, =0.25、 =0.35、 =0.25、 =0.15 (determined through fitting a large number of personality samples); It should be added that the ratio of spoken words to total words can be obtained from text information, while the ratio of the number of times a topic is initiated and the ratio of positive emotion words can be obtained through semantic analysis or semantic models.

[0025] The wearer's personality is determined based on the extroversion index S: when S ≥ 0.5, the wearer is considered extroverted; when S < 0.5, the wearer is considered introverted. The dialogue damage level is adaptively adjusted based on the wearer's individual personality traits, as follows: A1: If the wearer is an introverted person, and the dialogue damage level is normal (D < 3 points), then the damage level remains unchanged; If the dialogue damage level is abnormal (mild, moderate, severe), the damage level increases by one level (mild → moderate, moderate → severe, severe remains unchanged); for example, if an introverted wearer's dialogue damage score is 4 (mild), it will be adaptively adjusted to moderate; if the score is 6 (moderate), it will be adjusted to severe.

[0026] A2: When the wearer is an extroverted personality, the dialogue damage level will remain unchanged and will not be adjusted. For example, if the dialogue damage score of an extroverted wearer is 4 (mild), the mild level will remain unchanged.

[0027] S4: Two-way psychological buffer speech output and playback a. Two-way psychological buffer analysis and processing: After enabling two-way psychological buffer, the AI ​​processing module performs comprehensive analysis and processing based on the identity information of the interlocutor, the severity rating of the dialogue harm, and the text information. Among them, identity information determines the tone of the speech (e.g., a gentle and empathetic tone is output when applied to parents, a rational and guiding tone is output when applied to leaders, and a relaxed and comforting tone is output when applied to classmates), the severity rating determines the depth of the speech (simple comforting speech is used for mild harm, empathetic and guiding speech is used for moderate harm, and professional psychological counseling-level speech is used for severe harm), and text information determines the targeting of the speech (combining the specific criticisms and accusations in the text to avoid empty speech). b. Personalized Psychological Buffering Script Generation: Based on the analysis results, script templates corresponding to the scenario and damage level are retrieved from the scenario-based statement classification database and then personalized by combining the wearer's individual personality traits; for example: Scenario 1: Parent-child communication, identity: parent, harm level: mild (before adaptive adjustment for introverted wearers), text feature: "How can you be so stupid? You can't even do this simple thing well", generated dialogue: "Dad / Mom is just too anxious. It's not that Dad really thinks you're stupid. You've already tried very hard. Just take your time." Scenario 2: Workplace, Identity: Leader, Damage Level: Moderate (Extroverted Person), Text Feature: "Your efficiency is too low, this plan is completely unacceptable", Generated Response: "The leader's criticism is to encourage you to do better. We can first sort out the problems with the plan and optimize it step by step. You can definitely do it well." Scenario 3: Campus interaction, Identity: classmate, Harm level: severe (adaptive adjustment for introverted wearers), Text feature: "Your grades are so bad, don't play with me," Generated dialogue: "His words are too hurtful. Grades don't define you. You have many strengths. Don't let other people's negativity negate you. Take it slow, and you will become more and more excellent."

[0028] a. Script Output: The generated personalized psychological buffer script is played through the audio output module of the smart headset. The playback volume is set to a comfortable level for the wearer (default 60%, which can be manually adjusted), and the playback duration is controlled at 5-10 seconds to avoid affecting the normal communication between the two parties. At the same time, if the damage level is severe, the script will be played again after a 30-second interval to enhance the psychological buffer effect.

[0029] Specific application examples Taking a 15-year-old student (extroverted personality, S=0.635) as an example, with the wearer being a parent and the dialogue scenario being parent-child communication, the specific implementation process is as follows: Collected audio dialogue: Parent's voice: "Why did you do so poorly on this exam again? I tutor you every day, but you don't care at all. I'm so disappointed in you." Wearer's voice: "I didn't want to do poorly on the exam either. I've already tried my best." Information Conversion and Analysis: The parent's voice is converted into text information, and the emotional information is identified as "anger (final emotional harm quantification value of 8.5)"; the identity of the speaker is identified as "parent", and the dialogue scenario attribute is "parent-child communication". In this embodiment, negative words account for 80% of the text information. Collect the wearer's historical voice data, process the data, and calculate... =0.6, =0.7, =0.6, =0.6, and the personality assessment calculation S=0.25×0.6+ 0.35×0.7+0.25×0.6+0.15×0.6=0.635, which is judged as extroverted personality; Reasoning: Input the parents' text and emotional information into the AI ​​emotion model, and output the emotion type "anger"; Harm severity assessment: α=0.8 (parental role), a=8 (negative vocabulary percentage 80%), b=8.5 (final emotional harm quantification value calculated to be 8.5). Based on the dialogue harm score, D=0.8(0.4*8+0.6*8.5)=6.64 points, the original harm level is moderate; because the wearer is an extroverted personality, the harm level remains unchanged and reaches the negative threshold (3 points), so two-way psychological buffer is activated; Buffered dialogue output: Combining the parent's identity, moderate level of harm, and text information, a personalized dialogue is generated: "Mom is just too anxious about your grades. It's not that I'm denying your efforts. One bad test doesn't mean anything. Let's analyze the problem together, and we'll definitely improve next time." The dialogue is played through smart headphones at 60% volume for 7 seconds.

[0030] Emotional damage coefficient and The calculation formula is as follows: Extract the short-time average loudness from the speech, denoted as... Based on the average volume of the wearer's daily conversation. , Calculate the volume deviation coefficient: , in, The truncation function restricts the result to the interval [min, max], i.e., between [0.8, 1.2]. Extract the raw speech rate value from the speech, denoted as... Based on normal daily speaking speed, it is recorded as , Calculate the speech rate deviation coefficient: , in, The truncation function restricts the result to the interval [min, max], i.e., between [0.8, 1.2]. When the average loudness extracted from the speech Average volume of daily conversations with the wearer When they are the same, The value is 1, when the average loudness extracted from the speech is... Compared to the average volume of the wearer's daily conversation When it is large, then its Greater than 1; Extracting raw speech rate values ​​from speech Based on normal daily speaking speed When they are the same, The value is 1, representing the original speech rate extracted from the speech. Based on a normal speaking speed When it is large, then its Greater than 1; The louder and faster the speaker's voice, the greater the calculated final emotional damage quantification value 'b' will be compared to the base value. This results in a higher calculated dialogue damage score and a higher level of dialogue damage. Conversely, when the speaker's voice is soft and their speech is slow, the calculated emotional damage coefficient will be lower. and When the value is less than 1, the final emotional damage quantification value b is based on the base score, and there will be no effect on the wearer by lowering the volume or slowing down the speaking speed of the speaker.

[0031] Through the above embodiments, accurate identification of dialogue-related harm and personalized two-way psychological buffering can be achieved in multiple scenarios, effectively alleviating negative emotions brought about by dialogue, achieving a buffering effect at the level of psychological counseling, and adapting to the needs of wearers with different identities and personalities.

[0032] The smart earphones include the earphone body 1, the microphone module 2, the miniature air pump 3, and the earbud structure 4; A miniature air pump 3 is installed inside one side of the headphone body 1. The output end of the miniature air pump 3 is connected to an earplug structure 4. The earplug structure 4 is installed on the outside of the headphone body 1. A radio module 2 for radio reception is also installed on one side of the headphone body 1. The earplug structure 4 includes a connecting sleeve 401 connected to the earphone body 1 and an inflation tube 409 connected to the output end of the miniature air pump 3. The connecting sleeve 401 is fixedly connected to a partition plate 402 for dividing the interior of the connecting sleeve 401 into two chambers. An earplug support frame 403 is also fixedly connected to one side of the connecting sleeve 401, and an inflation airbag 404 is installed on the outside of the earplug support frame 403. The other end of the inflation tube 409 is connected to the airbag inflation tube 405 and the drying tube 406, and the other end of the airbag inflation tube 405 is connected to the inflatable airbag 404, and the other end of the drying tube 406 is connected to the earplug support frame 403. A humidity detection unit 413 is installed on one side of the earbud support frame 403. A humidity adjustment vent 411 and a humidity adjustment vent 412 are respectively opened on both sides of the earbud support frame 403.

[0033] A discharge hole 410 is also provided on one side of the connecting sleeve 401, and the chamber of the connecting sleeve 401 where the discharge hole 410 is located is connected to the humidity regulating air inlet 412.

[0034] A first solenoid valve 407 is installed on the outside of the airbag inflation tube 405, and a second solenoid valve 408 is installed on the outside of the drying tube 406.

[0035] When the wearer puts on the headphones immediately after showering, the ear canal is humid. The humidity detection unit 413 collects and monitors the ear canal humidity data in real time. When the ear canal is detected to be humid, the control unit automatically starts the micro air pump 3 and opens the second solenoid valve 408. Airflow is slowly delivered through the inflation tube 409, flows along the drying tube 406, and sequentially passes through the connecting sleeve 401, earplug support 403, humidity regulating air inlet 412 and humidity regulating exhaust 411 to form a circulating ventilation path, and finally is discharged through the exhaust hole 410, thereby gently drying and dehumidifying the inside of the ear canal.

[0036] A constantly damp ear canal is prone to the growth of fungi and bacteria, easily leading to ear diseases such as otitis externa and fungal otitis media. This earphone utilizes a built-in micro-pump to create a slow airflow system in the ear canal, continuously maintaining a dry environment, effectively reducing the risk of ear inflammation and providing long-lasting ear canal protection.

[0037] Meanwhile, when the external environment is noisy and interferes with the normal listening of the headphones, the control terminal can drive the micro air pump 3 to work and open the first solenoid valve 407. The airflow is introduced into the airbag 404 through the airbag inflation tube 405. The airbag 404 inflates at a uniform speed and fits tightly against the skin inside the ear, strengthening the ear's sealing and blocking of external noise through physical sealing, effectively achieving physical noise reduction and improving the quietness of wearing and listening.

[0038] This article uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only for the purpose of helping to understand the method and core ideas of the present invention. The above descriptions are only preferred embodiments of the present invention. It should be noted that due to the limitations of textual expression, while there are objectively infinite specific structures, those skilled in the art can make several improvements, modifications, or changes without departing from the principles of the present invention, and can also combine the above technical features in an appropriate manner. These improvements, modifications, changes, or combinations, or the direct application of the inventive concept and technical solution to other situations without modification, should all be considered within the scope of protection of the present invention.

Claims

1. A two-way psychological buffering method for psychological counseling based on multi-scenario AI voice parsing, characterized in that: Includes the following steps: S1: Collect the speaker's voice and convert it into text and emotional information. Analyze the text information to obtain the speaker's identity information and dialogue scenario attribute information. Collect and convert the wearer's historical voice data to determine the wearer's personal personality traits; S2: Train the neural network model using a dataset composed of emotional information to obtain an AI emotion model. Input the obtained text information and emotional information into the AI ​​emotion model and output the emotion type. S3: Determine the level of harm in the conversation by combining emotion type and individual personality traits; when the level of harm reaches the negative threshold, activate two-way psychological buffer; the level of harm includes positive and negative thresholds; S4: Two-way psychological buffer analyzes and processes the wearer's personal characteristics, the level of harm in the conversation, and the text information, and outputs personalized psychological buffering messages through smart headphones.

2. The bidirectional psychological buffering method for psychological counseling based on multi-scenario AI voice parsing as described in claim 1, characterized in that: Construct a scenario-based statement classification database based on scenario attribute features; Identity information includes parents, teachers, leaders, colleagues, clients, elders, classmates, etc., and dialogue scenario attribute information includes parent-child communication, campus interaction, workplace office, workplace interpersonal relationships, family elder communication, etc.

3. The bidirectional psychological buffering method for psychological counseling based on multi-scenario AI voice parsing as described in claim 1, characterized in that: Perform text vectorization processing on identity information and text information; Quantification is performed based on the emotion type in the emotional information: The base score for calmness is 0. A slight dissatisfaction score is 2. The base score for irritability is 4. Anger has a base score of 6. Anger / blame has a base score of 8. Abusive language / aggressive attacks have a base score of 10. By combining the base score of emotion type, volume and speech rate features are extracted from the emotion type to calculate the final quantitative value of emotional harm: ， Where b is the final quantification value of emotional harm based on emotion type and tone intensity. The damage value is the value of the emotion itself. This is the volume deviation coefficient. This is the speech rate deviation coefficient, ranging from 0.8 to 1.

2. The louder the volume and the faster the speech, the more intense the emotions, and the closer the emotional damage coefficient is to 1.

2. The volume is even, the speaking speed is stable, and the emotional harm coefficient is close to 0.8; The identity information and text information are processed into text vectors, and combined with the final emotional harm quantification value to construct a formula for calculating the dialogue harm score, as follows: ， Among them, the dialogue damage score is D, the wearer's identity information is α, the text information is a, the final emotional damage quantification value is b, K1 and K2 are the weighting coefficients of the corresponding information dimensions, and k1+k2=1; The dialogue damage rating D ranges from 0 to 10. Determine if the dialogue damage rating D is greater than 3; if yes, classify the corresponding dialogue damage level as abnormal; otherwise, classify the corresponding dialogue damage level as normal.

4. The bidirectional psychological buffering method for psychological counseling based on multi-scenario AI voice parsing as described in claim 3, characterized in that: The degree of harm caused by the dialogue is determined based on the dialogue harm score. The negative threshold is set based on the maximum and minimum values ​​of the degree of harm in historical data; The severity of damage is divided into two dialogue damage levels: normal and abnormal. The abnormal level is further divided into mild, moderate, and severe. When the dialogue damage level is judged to be normal, two-way psychological buffer is not enabled; When the damage level assessment during a conversation is abnormal, activate the two-way psychological buffer.

5. The bidirectional psychological buffering method for psychological counseling based on multi-scenario AI voice parsing according to claim 1, characterized in that: The wearer's personal personality traits are quantified to construct a personality assessment formula, defined as follows: ， Where S is the extroversion index. The percentage of spoken words to total words. The percentage of times a topic is initiated proactively. The percentage of positive emotion words Standardized average speech rate (words per minute) , , and These correspond to the weighting coefficients of each indicator.

6. The bidirectional psychological buffering method for psychological counseling based on multi-scenario AI voice parsing as described in claim 1, characterized in that: The emotional type of the interlocutor is determined by analyzing textual and emotional information, and the degree of harm caused by the conversation is assessed by combining this with the wearer's personal personality traits. A1: Based on the wearer's personal personality traits, which include introversion and extroversion; A2: Determine if the wearer's personality trait is introverted; if yes, proceed to A3; if no, the dialogue damage level remains the same. A3: Determine if the dialogue damage level is normal; if yes, the dialogue damage level remains unchanged; otherwise, the dialogue damage level increases by one level.

7. A smart earphone, characterized in that: It includes the headphone body (1), the radio module (2), the miniature air pump (3), and the earplug structure (4). A miniature air pump (3) is installed inside one side of the headphone body (1). The output end of the miniature air pump (3) is connected to an earplug structure (4). The earplug structure (4) is installed on the outside of the headphone body (1). A radio module (2) for radio reception is also installed on one side of the headphone body (1). The earplug structure (4) includes a connecting sleeve (401) connected to the headphone body (1) and an inflation tube (409) connected to the output end of the micro air pump (3). The connecting sleeve (401) is fixedly connected to a partition plate (402) for dividing the interior of the connecting sleeve (401) into two chambers. The connecting sleeve (401) is also fixedly connected to one side of an earplug support frame (403), and an inflation airbag (404) is installed on the outside of the earplug support frame (403). The other end of the inflation tube (409) is connected to the airbag inflation tube (405) and the drying tube (406), and the other end of the airbag inflation tube (405) is connected to the inflatable airbag (404), and the other end of the drying tube (406) is connected to the earplug support frame (403); A humidity detection unit (413) is installed on one side of the earbud support frame (403). Humidity adjustment exhaust port (411) and humidity adjustment air inlet port (412) are respectively opened on both sides of the earbud support frame (403).

8. The smart earphone according to claim 7, characterized in that: A discharge hole (410) is also provided on one side of the connecting sleeve (401), and the chamber of the connecting sleeve (401) where the discharge hole (410) is located is connected to the humidity regulating air inlet (412).

9. A smart earphone according to claim 7, characterized in that: A first solenoid valve (407) is installed on the outside of the airbag inflation tube (405), and a second solenoid valve (408) is installed on the outside of the drying tube (406).

Images