Psychological tutoring system based on digital human technology

Abstract

The invention provides a psychological tutoring system based on a digital human technology, and relates to the technical field of psychological health. The system captures the emotional state of a user in real time in a multi-modal interaction mode, generates a real-time emotional state group of the user by combining multi-dimensional data fusion of facial micro-expressions, voice spectrums and dialogue texts, and maps the real-time emotional state group to a three-dimensional emotional space for accurate emotion recognition. Based on an emotion-behavior mapping rule, the system dynamically adjusts the digital person and provides an intervention scheme. And when the emotion fluctuation of the user is detected to exceed a preset threshold value, the system automatically triggers a digital human hiding mechanism, gradually weakens visual, auditory and action stimulation, and provides psychological pacification and guidance. According to the method, the accuracy of psychological state recognition and the pertinence of intervention are remarkably improved, real-time and continuous psychological health support is provided for the user, and the method is suitable for psychological health assistance and emotional state analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of mental health, and specifically to a psychological counseling system based on digital human technology. Background Art

[0002] In the field of mental health, it mainly relies on face-to-face consultations or remote communication means such as phone calls and videos. Although these methods can help users relieve psychological pressure to a certain extent, they are limited by time and space and it is difficult to achieve real-time and continuous emotional monitoring and intervention. In addition, traditional psychological counseling usually relies on the experience and subjective judgment of psychological counselors and lacks objective emotional data support, resulting in the difficulty of quantifying and evaluating the counseling effect.

[0003] However, existing psychological counseling systems often can only perform emotional analysis in a single way and it is difficult to comprehensively capture the true emotional state of users. The accuracy between the emotional recognition result and the intervention plan is insufficient, resulting in poor mental health counseling effects and the inability to provide timely psychological comfort and guidance. Summary of the Invention

[0004] To achieve the above purposes, the present invention is realized through the following technical solutions: A psychological counseling system based on digital human technology, including,

[0005] User Interaction Module: Based on the digital human, it conducts real-time interaction with the user through multi-modal interaction methods, collects the user's data, and generates a real-time emotional state group of the user according to the data;

[0006] Analysis Module: Maps the real-time emotional state group of the user to a three-dimensional emotional space constructed by pleasure, activation, and dominance, real-time identifies the emotional state of the user, and generates an emotional analysis report;

[0007] Psychological Counseling Module: Based on the emotional analysis report and the coordinates of the three-dimensional emotional space, dynamically adjusts the output and intervention plan of the digital human through emotional-behavior mapping rules;

[0008] Emergency Response Module: Real-time monitors the data generated during the interaction of the user, detects the user's emotional fluctuations by analyzing the user's emotional state group, and sets thresholds according to the change ranges of pleasure, activation, and dominance. When it detects that the user's emotional fluctuations exceed the threshold, it triggers the digital human hiding mechanism and immediately provides psychological comfort and guidance to the user through the digital human;

[0009] Effect Evaluation Module: Quantitatively evaluates the output and intervention plan of the digital human through a standardized psychological scale, the user's feedback data, and the number of times the digital human hiding mechanism is triggered.

[0010] The user interaction module captures the movement trajectory of the user's face through a high-precision camera, then analyzes the micro-expressions in real time, extracts the movement amplitude and frequency characteristics of the facial area, and forms a micro-expression vector; it collects the user's voice signal through a microphone, analyzes the fundamental frequency fluctuation range and energy distribution density using speech processing technology, and constructs an acoustic feature vector; it conducts sentiment analysis on the user's dialogue text through natural language processing technology, extracts sentiment term frequencies and sentence complexity, and forms a text feature vector; it performs multi-dimensional alignment on the micro-expression vector, acoustic feature vector, and text feature vector through a weighted fusion algorithm, dynamically adjusts the weights of each feature, generates the user's real-time emotional state group, and converts it into the multi-modal output parameters of the digital human; for example, when the user's voice features fluctuate greatly, the system automatically increases the acoustic weight; when strong emotional words appear in the text, the system enhances the semantic weight; when the facial micro-expressions are continuously abnormal, the system increases the visual weight. For example, when the movement amplitude of the corners of the mouth exceeds ±2 times the standard deviation of the facial micro-expression vector value in the calm state of the face and lasts for more than 5 seconds, it is considered that the facial micro-expressions are continuously abnormal.

[0011] According to the coordinate distribution of the user's real-time emotional state group in three-dimensional space, automatically calculate the micro-expression intensity, limb movement amplitude, and speech synthesis parameters that the digital human should present; the user's real-time emotional state group automatically calculates the micro-expression intensity, limb movement amplitude, and speech synthesis parameters of the digital human through emotion-behavior mapping rules; by mapping the coordinate values of pleasure, activation, and dominance in three-dimensional space to the corresponding behavior parameter ranges respectively.

[0012] Pleasure is a dimension in the three-dimensional emotional space, indicating the positive or negative degree of the user's emotion; the system maps the pleasure coordinate value to the micro-expression parameters of the digital human through a linear function, specifically manifested as the changes in the curvature of the corners of the mouth and the position of the eyebrows; let the pleasure coordinate be V, then the calculation formulas for the curvature of the corners of the mouth S and the position of the eyebrows B are: S = k1V + c1, B = k2V + c2, where k1 and k2 are slope coefficients, and c1 and c2 are intercept constants; for example, set k1 = 1° / unit, c1 = 0°, k2 = 0.55mm / unit, c2 = 0mm; when the pleasure V = 5, the curvature of the corners of the mouth S = 5°, and the position of the eyebrows B = 2.5mm, presenting an obvious smiling expression.

[0013] Activation indicates the excitement level of the user's emotion. The system maps the activation coordinate value to the limb movement amplitude of the digital human through a piecewise function; let the activation coordinate be A, then the calculation formula for the head swing amplitude H is: where A 阈值 and A 上限 are preset threshold and upper limit values. For example, A 阈值 = 3, A 上限= 7; When the activation level exceeds the upper limit, the system triggers a forward lean action of the upper body, and the forward lean angle T = 0.8A. For example, when the activation level A = 8, the head swing amplitude H = 20°, and the forward lean angle of the upper body T = 6.4°.

[0014] The calculation of the speech synthesis parameters uses the weighted sum of the pleasure level and the activation level as the input, adjusts the fundamental frequency and the speech rate through an exponential function, and adjusts the speech intensity according to the dominance coordinate value; let the pleasure level be V and the activation level be A, then the weighted sum W = αV + βA, where α and β are weight coefficients, and α = 0.6 and β = 0.4 are set to reflect the different effects of the pleasure level and the activation level on speech synthesis.

[0015] The calculation formulas for the fundamental frequency F and the speech rate R are as follows: where F0 and R0 are the initial fundamental frequency and speech rate, and k3 and k4 are adjustment coefficients; for example, when W = 13, the fundamental frequency increases by 65 Hz and the speech rate increases by 130%.

[0016] The adjustment of the speech intensity is: the speech intensity I = I0 + k5D, where D is the dominance coordinate value, I0 is the initial speech intensity, and k5 is the adjustment coefficient; for example, when the dominance D = 5, the speech intensity increases by 15 dB.

[0017] The psychological counseling module analyzes the changing trends of the pleasure level, activation level, and dominance level of the user in the three-dimensional emotion space and performs similarity matching with historical data; for example, if the pleasure level of the user continuously drops below the set threshold in three consecutive conversations, the system determines that the user has persistent low mood and marks it as an abnormal state; the system verifies the accuracy of the diagnosis by clustering and analyzing the distribution law of the three-dimensional coordinates and comparing it with the standard threshold of the psychological scale; for example, when the pleasure level coordinate value of the user is lower than the depression threshold of the psychological scale, the system will generate a high-confidence diagnosis result; based on the real-time emotion state group, the system dynamically designs an intervention plan: if the pleasure level is low, the system calls the positive emotion guidance content library, adjusts the speech intonation of the digital human to a gentle mode, and increases encouraging body movements; if the activation level rises abnormally and is accompanied by fluctuations in the dominance level, the system starts a cognitive behavioral therapy segment, guides the user to conduct emotion attribution analysis through the dialogue text, and generates an emotion analysis report to record the user's emotion changes and intervention effects.

[0018] The digital human hiding mechanism is combined with the emotion-behavior mapping rules and the user's real-time emotional state group. When the user's emotional fluctuation exceeds the preset threshold, according to the coordinate distribution of the user's real-time emotional state group in the three-dimensional emotional space, the output parameters of the digital human are adjusted reversely according to the emotion-behavior mapping rules. In the initial stage, the edges of the digital human image are blurred, and the blurred range expands towards the center at a ratio of 10% of the display area per second. At the same time, the overall contrast of the image is reduced, with the contrast value decreasing by 15% per second, gradually weakening the visual stimulation of the digital human; then, the voice output of the digital human is adjusted gradually. First, the voice volume is reduced by 3 decibels per second, and at the same time, the reverberation effect of the voice is increased to simulate the feeling of the sound gradually moving away. Then, the speech rate is gradually slowed down, with the speech rate decreasing by 10% per second, gradually weakening the voice stimulation; finally, the action presentation of the digital human is faded out. First, the amplitude of the limb movements is reduced by 20% per second, and then the execution frequency of the actions is reduced by 30% per second until the digital human completely stops the action output. The whole process is completed within 10 to 15 seconds. During this period, the user's emotional fluctuation is continuously monitored. If the fluctuation value drops back to the safe range, the digital human's various modality outputs are gradually restored in the reverse order until the normal interaction state is completely restored.

[0019] (I) Beneficial effects

[0020] The present invention provides a psychological counseling system based on digital human technology, having the following beneficial effects:

[0021] 1. By combining multi-modal data of facial micro-expressions, voice spectra, and dialogue texts, the present invention constructs a multi-dimensional emotional feature vector and maps it to a three-dimensional emotional space, achieving a comprehensive and accurate recognition of the user's emotional state; based on the emotion-behavior mapping rules, the system can dynamically adjust the micro-expressions, limb movements, and voice parameters of the digital human, providing a personalized psychological intervention plan, significantly improving the accuracy of psychological state recognition and the pertinence of intervention.

[0022] 2. The present invention can monitor the user's emotional fluctuation in real time. When it detects that the emotional fluctuation exceeds the preset threshold, it automatically triggers the digital human hiding mechanism and provides psychological comfort and guidance; effectively relieving the user's psychological pressure during sudden emotional fluctuations and enhancing the system's emergency handling ability.

[0023] 3. By standardizing psychological scales, user feedback data, and the triggering times of the digital human hiding mechanism, the system can quantitatively evaluate the effect of psychological intervention, further improving the scientificity and effectiveness of mental health. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a schematic flowchart of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0025] The technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0026] This system captures the user's emotional state in real time through a multimodal interaction method, combines multi-dimensional data fusion of facial micro-expressions, voice spectra, and dialogue texts, generates a real-time emotional state group of the user, and maps it to a three-dimensional emotional space for accurate emotion recognition. The system includes a user interaction module, an analysis module, a psychological counseling module, an emergency response module, and an effect evaluation module.

[0027] The user interaction module captures the movement trajectory of the user's face through a high-precision camera, analyzes the changes in micro-expressions in real time, extracts the movement amplitude and frequency characteristics of the facial area, and forms a micro-expression vector. For example, when the movement trajectory of the user's facial key points shows that the upward amplitude of the corners of the mouth is 5 millimeters and the position of the eyebrows rises by 2 millimeters, the system generates a corresponding micro-expression vector. At the same time, the system collects the user's voice signal through a microphone, uses speech processing to analyze the fundamental frequency fluctuation range and energy distribution density, and constructs an acoustic feature vector. For example, when the fundamental frequency fluctuation range of the user's voice is 100 Hz to 200 Hz and the energy distribution density is 0.5, the system generates a corresponding acoustic feature vector. In addition, through natural language processing technology, the system conducts sentiment analysis on the user's dialogue text, extracts the sentiment word frequency and sentence complexity, and forms a text feature vector. For example, when the frequency of positive emotion words such as happy and joyful in the user's dialogue text is 10 times per minute and the sentence complexity is medium, the system generates a corresponding text feature vector. Through a weighted fusion algorithm, the micro-expression vector, acoustic feature vector, and text feature vector are multi-dimensionally aligned, and the weights of each feature are dynamically adjusted to generate a real-time emotional state group of the user and convert it into multi-modal output parameters of the digital human. For example, when the voice feature fluctuation of the user is large, the system automatically increases the acoustic weight; when strong emotion words appear in the text, the system enhances the semantic weight; when the facial micro-expression continues to be abnormal, the system increases the visual weight.

[0028] The analysis module maps the user's real-time emotional state group to a three-dimensional emotional space constructed by pleasure, activation, and dominance. For example, pleasure can represent the positive or negative degree of the user's emotion, activation represents the excitement degree of the user's emotion, and dominance represents the degree of control over the emotion by the user. The system maps the pleasure coordinate value to the micro-expression parameters of the digital human through a linear function, specifically manifested as the changes in the mouth corner arc and eyebrow position. Let the pleasure coordinate be V, then the calculation formulas for the mouth corner arc S and eyebrow position B are: S = k1V + c1, B = k2V + c2, where k1 and k2 are slope coefficients, and c1 and c2 are intercept constants. For example, k1 = 0.5, c1 = 0, k2 = 0.3, c2 = 0 are set. When the pleasure V = 5, the mouth corner arc S = 2.5°, and the eyebrow position B = 1.5mm, presenting an obvious smiling expression. The mapping of activation uses a piecewise function. Let the activation coordinate be A, then the calculation formula for the head swing amplitude H is: where A 阈值 and A 上限 are preset threshold and upper limit values. For example, A 阈值 = 3, A 上限 = 7; when the activation exceeds the upper limit, the system triggers the upper body forward tilt action, and the forward tilt angle T = 0.8A. For example, when the activation A = 8, the head swing amplitude H = 20°, and the upper body forward tilt angle T = 6.4°. The calculation of the speech synthesis parameters uses the weighted sum of pleasure and activation as the input, adjusts the fundamental frequency and speech rate through an exponential function, and adjusts the speech intensity according to the dominance coordinate value at the same time. Let the pleasure be V and the activation be A, then the weighted sum W = αV + βA, where α and β are weight coefficients, and α = 0.6, β = 0.4 are set. The calculation formulas for the fundamental frequency F and speech rate R are where F0 and R0 are the initial fundamental frequency and speech rate, and k3 and k4 are adjustment coefficients. For example, when W = 13, the fundamental frequency is increased by 65Hz, and the speech rate is increased by 130%. The adjustment of the speech intensity is I = I0 + k5D, where D is the dominance coordinate value, I0 is the initial speech intensity, and k5 is the adjustment coefficient. For example, when the dominance D = 5, the speech intensity is increased by 15dB. For example, when the coordinates of the user in the three-dimensional emotional space show that the pleasure is 7, the activation is 5, and the dominance is 6, the system determines that the user is in a pleasant emotional state and the intensity is at a medium level; if the pleasure is 2, the activation is 8, and the dominance is 3, it is determined that the user is in an angry emotional state and the intensity is strong. The system determines the user's current emotional type and intensity through the emotional behavior conversion rule, and organizes the identified emotional state information into an emotional analysis report.

[0029] The psychological counseling module dynamically adjusts the micro - expressions, body movements, voice parameters and intervention plans of the digital human based on the sentiment analysis report through sentiment - behavior mapping rules. For example, when the user's pleasure level is low, the system calls up positive emotion - guiding content, adjusts the digital human's voice intonation to a gentle mode, and increases encouraging body movements. If the activation level rises abnormally and is accompanied by fluctuations in the dominance level, the system starts a cognitive - behavioral therapy segment, guides the user to conduct emotion attribution analysis through dialogue text, and generates a sentiment analysis report. The system calculates the change trends of the user's pleasure, activation and dominance levels in the three - dimensional sentiment space through time - series analysis methods, and performs similarity matching with historical data. For example, if the user's pleasure level continuously drops below the set threshold in three consecutive conversations, the system determines that the user has persistent low mood and marks it as an abnormal state. The system verifies the accuracy of the diagnosis by comparing the distribution rules of the three - dimensional coordinates through cluster analysis with the standard thresholds of psychological scales. For example, when the coordinate value of the user's pleasure level is lower than the depression threshold of the psychological scale, the system generates a diagnosis result with high confidence.

[0030] The emergency response module sets dynamic thresholds based on the pleasure, activation and dominance levels in the three - dimensional sentiment space. The safe range of pleasure is [4, 8], and when it exceeds the range, it is determined that the user's emotion is negative or overly excited. The activation range is [3, 7], and when it is abnormal, it reflects emotional depression or excitement. The dominance range is [3, 6], and when it exceeds the range, it indicates emotional out - of - control or excessive repression. When any dimension exceeds the threshold or, the system triggers the digital human hiding mechanism and immediately provides psychological comfort and guidance to the user through the digital human. For example, when the user's emotion fluctuation value exceeds the safe range of pleasure of 8, the system reversely adjusts the output parameters of the digital human according to the coordinate distribution of the user's real - time emotion state group in the three - dimensional sentiment space according to the sentiment - behavior mapping rules. In the initial stage, the edges of the digital human image are blurred, and the blurred range expands towards the center at a ratio of 10% of the display area per second. At the same time, the overall contrast of the image is reduced, with the contrast value decreasing by 15% per second, so that the visual stimulation of the digital human gradually weakens. Then, the voice output of the digital human is adjusted gradually. First, the voice volume is reduced by 3 decibels per second, and at the same time, the reverberation effect of the voice is increased to simulate the feeling of the sound gradually moving away. Then, the speech speed is gradually slowed down, with the speech speed ratio decreasing by 10% per second, so that the voice stimulation gradually weakens. Finally, the action presentation of the digital human is faded out. First, the amplitude of the body movements is reduced by 20% per second, and then the execution frequency of the movements is reduced, with the number of movements decreasing by 30% per second until the digital human completely stops the action output. The whole process is completed within 10 to 15 seconds. During this period, the user's emotion fluctuation is continuously monitored. If the fluctuation value drops back to the safe range, the digital human's various modality outputs are gradually restored in the reverse order until the normal interaction state is completely restored.

[0031] The effect evaluation module quantitatively evaluates the intervention effect through standardized psychological scales, the feedback data of users, and the number of times the digital human hiding mechanism is triggered. For example, the system evaluates the effect of psychological intervention by comparing the scores of psychological scales of users before and after using the system. If the user feedback data shows that the emotional state has been significantly improved and the number of times the digital human hiding mechanism is triggered has decreased, the significance of the mental health effect is improved.

[0032] The present invention combines multi-modal data of facial micro-expressions, voice spectra, and dialogue texts to construct multi-dimensional emotional feature vectors and map them to a three-dimensional emotional space, realizing a comprehensive and accurate recognition of the user's emotional state. Based on the emotion-behavior mapping rules, the system can dynamically adjust the micro-expressions, body movements, and voice parameters of the digital human, providing personalized psychological intervention programs, significantly improving the accuracy of emotion recognition and the pertinence of intervention. The system can real-time monitor the emotional fluctuations of users. When it detects that the emotional fluctuations exceed the preset threshold, it automatically triggers the digital human hiding mechanism and provides psychological comfort and guidance, effectively relieving the psychological pressure of users during sudden emotional fluctuations, enhancing the emergency handling ability of the system during the psychological counseling process of users, and further improving the scientificity and effectiveness of mental health.

[0033] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirits of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A psychological counseling system based on digital human technology, characterized in that, Including: User Interaction Module: Conducts real-time interaction with users based on digital humans through multi-modal interaction methods, collects user data, and generates a real-time emotional state group of users according to the data. Analysis Module: Maps the real-time emotional state group of users to a three-dimensional emotional space constructed by pleasure, activation, and dominance, real-time identifies the emotional state of users, and generates an emotional analysis report. Psychological Counseling Module: Based on the emotional analysis report and the coordinates of the three-dimensional emotional space, dynamically adjusts the output and intervention plan of the digital human through emotion-behavior mapping rules. Emergency Response Module: Real-time monitors the data generated by users during interaction, detects user mood fluctuations by analyzing the emotional state group of users, and sets thresholds according to the change ranges of pleasure, activation, and dominance. When it detects that the user mood fluctuation exceeds the threshold, it triggers the digital human hiding mechanism and immediately provides psychological comfort and guidance to users through the digital human. Effect Evaluation Module: Quantitatively evaluates the output and intervention plan of the digital human through standardized psychological scales, user feedback data, and the number of times the digital human hiding mechanism is triggered.

2. The psychological counseling system based on digital human technology according to claim 1, wherein: The User Interaction Module generates a micro-expression vector by capturing the movement trajectories of the facial key points of users; forms a multi-dimensional emotional feature vector through the acoustic feature vector generated by the fundamental frequency fluctuation and energy distribution characteristics in the speech spectrum; generates a real-time emotional state group of users through weighted fusion of the emotional tendency word frequencies and multi-dimensional emotional feature vectors in the dialogue text.

3. The psychological counseling system based on digital human technology according to claim 2, characterized in that: The real-time emotional state group of users normalizes the captured movement trajectories of facial key points, extracts the movement amplitude and frequency characteristics of the facial area to form a micro-expression vector; simultaneously analyzes the fundamental frequency fluctuation range and energy distribution density in the speech spectrum to construct an acoustic feature vector; aligns the above features with the emotional tendency word frequencies and sentence complexity extracted from the dialogue text in multiple dimensions and dynamically adjusts the weights of each feature; maps the weighted fusion feature vector to the three-dimensional emotional space to generate a real-time emotional state group of users; automatically calculates the micro-expression intensity, limb movement amplitude, and speech synthesis parameters that the digital human should present according to the coordinate distribution of the real-time emotional state group of users in the three-dimensional space.

4. The psychological counseling system based on digital human technology according to claim 3, characterized in that: The real-time emotional state group of the user automatically calculates the micro-expression intensity, body movement amplitude, and voice synthesis parameters of the digital human through emotion-behavior mapping rules; by mapping the coordinate values of pleasure, activation, and dominance in three-dimensional space to the corresponding behavior parameter ranges respectively; for the micro-expression intensity, the pleasure coordinate value is converted into the mouth corner arc and eyebrow position through a linear function: for each unit increase in pleasure, the mouth corner arc and eyebrows move; the calculation of the body movement amplitude is based on a piecewise function of the activation coordinate value. When the activation is lower than the threshold and the head swing amplitude is between the threshold and the upper limit, the head swing amplitude increases; when it exceeds the upper limit, the amplitude increases again and triggers the upper body forward tilt action; the calculation of the voice synthesis parameters uses the weighted sum of pleasure and activation as the input, and adjusts the fundamental frequency and speech rate through an exponential function: for each unit increase in the weighted sum, the fundamental frequency increases and the speech rate speeds up; for each unit increase in dominance, the voice intensity increases.

5. The psychological counseling system based on digital human technology according to claim 1, characterized in that: In the analysis module, emotion-behavior conversion rules are used to determine the current emotion type and intensity of the user; if the coordinates of the user in the three-dimensional emotion space show high pleasure, medium activation, and high dominance, it is determined that the user is in a pleasant emotional state and the intensity is at a medium level; if the pleasure is low, the activation is high, and the dominance is low, the user is in an angry emotional state and the intensity is strong; the identified emotion state information is sorted into an emotion analysis report.

6. The psychological counseling system based on digital human technology according to claim 1, characterized in that: The psychological counseling module compares the three-dimensional coordinates of the current emotion state group with historical data to identify the persistent abnormalities or mutation characteristics of the emotion state; by analyzing the distribution law of the three-dimensional coordinates and comparing with the standard values of the psychological scale, the accuracy of the diagnosis is verified; an intervention plan is designed based on the real-time emotion state group of the user. If the pleasure is on the low side, positive emotion guidance content is called, the voice intonation of the digital human is adjusted to a gentle mode, and encouraging body movements are increased; if the activation abnormally increases and is accompanied by fluctuations in dominance, a cognitive behavioral therapy segment is started.

7. A psychological counseling system based on digital human technology according to claim 1, characterized in that: The digital human hiding mechanism is combined with the emotion-behavior mapping rules and the real-time emotion state group of the user. When the user's emotion fluctuates beyond the preset threshold, according to the coordinate distribution of the real-time emotion state group of the user in the three-dimensional emotion space, the output parameters of the digital human are adjusted in reverse according to the emotion-behavior mapping rules. In the initial stage, the edges of the digital human image are blurred, and at the same time, the overall contrast of the image is reduced, so that the visual stimulation of the digital human is gradually weakened; the voice output of the digital human is adjusted gradually, the voice volume is reduced, the reverberation effect of the voice is increased, simulating the feeling of the sound gradually moving away, and then the speech rate is gradually slowed down, so that the voice stimulation is gradually weakened; the action presentation of the digital human is processed in a fade-out manner, the amplitude of the body movement is reduced, and then the execution frequency of the action is reduced until the digital human completely stops the action output. During this period, the user's emotion fluctuation is continuously monitored. If the fluctuation value falls back to the safe range, the digital human output is gradually restored in the reverse order until it completely returns to the normal interaction state.

Images