Psychological state assessment interaction system and method

Abstract

The embodiment of the invention belongs to the technical field of psychological state detection, and discloses a psychological state assessment interaction system and device, and the corresponding system comprises an AI interaction module which is used for interacting with a user and obtaining feedback information of the user in the interaction process, and the feedback information comprises voice reply data, text reply data and physiological data of a question; the deep learning model is used for extracting feature data of the feedback information; the support vector machine model is used for evaluating the psychological state of the user according to the feedback information; input data of a kernel function of the support vector machine model is feature data; and the VR video generation module is used for generating a corresponding VR video according to the psychological state and pushing the VR video to the user, and the VR video has an interaction function with the user so as to further determine a lower state of the psychological state of the user and autonomously modify the VR video according to the lower state. The emotion, cognition and behavior characteristics can be extracted from the dialogue of the user, and the psychological state is quantified.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of psychological state detection, in particular to the technical field of artificial intelligence assisted mental health assessment, and specifically to a psychological state assessment interaction system and method. BACKGROUND

[0002] Psychological state assessment is a process of evaluating, analyzing and monitoring the psychological state of individuals through various technical means. These psychological states include emotional, stress, anxiety, depression, mental health and other aspects. With the development of technology, the technology of psychological state detection is constantly innovating and progressing. In the prior art, the psychological state assessment method faces the following technical problems:

[0003] 1. Low efficiency of psychological assessment: traditional psychological assessment relies on questionnaires or manual interviews, which is time-consuming and subjective, and it is difficult to capture the emotional changes of users in real time.

[0004] 2. Fixed content of counseling: existing psychological counseling videos are mostly pre-made content, which cannot be dynamically adjusted according to the individual needs of users, and have poor adaptability.

[0005] 3. Lack of interactivity and immersion: there is a lack of technical solutions that combine AI real-time feedback and VR immersive experience, resulting in low user engagement.

[0006] 4. Security and privacy risks: generative AI may output harmful content, and existing systems lack multi-modal auditing mechanisms. SUMMARY

[0007] The purpose of the present application is to provide at least a psychological state assessment interaction system and method to overcome the above technical problems in the prior art psychological state assessment interaction process.

[0008] To solve the above technical problems, at least one embodiment of the present application provides a psychological state assessment interaction system, comprising:

[0009] An AI interaction module for interacting with a user and obtaining feedback information of the user during the interaction, the feedback information including voice reply data, text reply data and physiological data of the questions;

[0010] A deep learning model for extracting preliminary feature data of the feedback information and performing heterogeneous multi-modal fusion of the preliminary feature data under multiple scales to generate fusion feature data;

[0011] a support vector machine model configured to evaluate the psychological state of the user according to the feedback information; input data of a kernel function of the support vector machine model being the fusion feature data; a hyperplane of the support vector machine model being determined by a dimension of the feedback information; and a margin being determined by a support vector corresponding to the feedback information closest to the hyperplane;

[0012] a VR video generation module configured to generate a corresponding VR video according to the psychological state, and push the VR video to the user, wherein the VR video has an interactive function with the user, so as to further determine a sub-state of the psychological state of the user, and the VR video is autonomously modified according to the sub-state.

[0013] In some embodiments, the AI interaction module comprises a voice recognition feedback module and a wearable physiological detection device, the voice recognition feedback module is configured to convert the voice reply data into the text reply data and recognize the tone of the voice reply data, and the wearable physiological detection device is configured to acquire the physiological data, wherein the physiological data comprises facial expression, heart rate, blood pressure, body temperature, respiratory rate and eye movement tracking.

[0014] In some embodiments, a sensitivity of the kernel function of the support vector machine is determined by a quantity of the feedback information.

[0015] In some embodiments, a psychological state evaluation interaction system further comprises:

[0016] a keyword weight modification module configured to modify a weight of a preset keyword in the support vector machine model if the feedback information contains the preset keyword, wherein the weight is determined by a preset mapping relationship between the keyword and the weight.

[0017] In some embodiments, the VR video generation module is specifically configured to determine a selection interface scene, a style, a constraint condition and a visual parameter of the sub-state of a pre-generated VR video according to the psychological state.

[0018] In some embodiments, the VR video generation module is specifically configured to generate a VR video corresponding to the sub-state according to feedback information of the user on the selection interface scene.

[0019] In some embodiments, a psychological state evaluation interaction system further comprises:

[0020] a VR video review module configured to delete negative visual data, text data and voice data in the VR video.

[0021] At least one embodiment of the present application further provides a psychological state evaluation interaction method, comprising:

[0022] interact with the user and obtain feedback information of the user during the interaction, the feedback information including voice reply data, text reply data and physiological data of the user in response to the question;

[0023] extract preliminary feature data of the feedback information through a deep learning model and perform heterogeneous multi-modal fusion of the preliminary feature data under multiple scales to generate fusion feature data;

[0024] evaluate the psychological state of the user according to a support vector machine model and the feedback information, input data of a kernel function of the support vector machine model being the fusion feature data, a hyperplane of the support vector machine model being determined by a dimension of the feedback information, and a margin being determined by a support vector corresponding to the feedback information closest to the hyperplane;

[0025] generate a corresponding VR video according to the psychological state and push the VR video to the user, the VR video having an interactive function with the user to further determine a lower state of the psychological state of the user and autonomously modify the VR video according to the lower state.

[0026] In some embodiments, the interaction with the user and the obtaining of the feedback information of the user during the interaction include:

[0027] converting the voice reply data into the text reply data and recognizing a tone of the voice reply data;

[0028] obtaining the physiological data, the physiological data including facial expressions, heart rates, blood pressures, body temperatures, respiratory rates and eye movement tracking.

[0029] In some embodiments, a sensitivity of the kernel function of the support vector machine is determined by a quantity of the feedback information.

[0030] In some embodiments, the evaluation of the psychological state of the user according to the feedback information includes:

[0031] detecting keywords in the feedback information, modifying a weight of the keywords, and inputting the keywords with the modified weights into the support vector machine model.

[0032] In some embodiments, the generation of the corresponding VR video according to the psychological state includes:

[0033] determining a selection interface scene, a style, a constraint condition and a visual parameter of the VR video according to the psychological state.

[0034] In some embodiments, the autonomous modification of the VR video according to the lower state includes:

[0035] According to the feedback information of the user for the selection interface scene, a VR video corresponding to the lower state is generated.

[0036] In some embodiments, a mental state evaluation interaction method further comprises:

[0037] Delete the negative visual data, text data and voice data in the VR video.

[0038] At least one embodiment of the present application also provides an electronic device, comprising: at least one processor; and a memory in communication connection with the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the mental state evaluation interaction method described above.

[0039] At least one embodiment of the present application also provides a computer readable storage medium storing a computer program, which is executed by a processor to implement the mental state evaluation interaction method described above.

[0040] The mental state evaluation interaction system and method provided by the embodiments of the present application correspond to a system comprising: an AI interaction module for interacting with a user and obtaining feedback information of the user during the interaction, the feedback information including voice reply data, text reply data and physiological data of the question; a deep learning model for extracting preliminary feature data of the feedback information and performing heterogeneous multi-modal fusion of the preliminary feature data under multiple scales to generate fusion feature data; a support vector machine model for evaluating the mental state of the user according to the feedback information; the input data of the kernel function of the support vector machine model is the fusion feature data; the hyperplane of the support vector machine model is determined by the dimension of the feedback information, and the interval is determined by the support vector corresponding to the nearest negative feedback information of the hyperplane; a VR video generation module for generating a corresponding VR video according to the mental state and pushing it to the user, and the VR video has an interactive function with the user to further determine the lower state of the mental state of the user and modify the VR video according to the lower state.

[0041] In summary, the present application evaluates the mental state of the user through AI dialogue analysis, automatically generates an adaptive VR video, and supports real-time intervention by a tutor, achieving the following technical effects:

[0042] 1. Accurate evaluation: extract emotional, cognitive and behavioral characteristics from the dialogue to quantify the mental state.

[0043] 2. Dynamic generation: generate personalized prompts by combining expert databases and user data to render high-definition VR videos.

[0044] 3. Safe and controllable: multi-modal review ensures that the content is harmless, and the tutor can adjust the generation logic at any time.

[0045] 4. Data-driven optimization: record user feedback and optimize long-term tutoring plans. BRIEF DESCRIPTION OF DRAWINGS

[0046] One or more embodiments are illustrated by way of example in the figures that form a part of this patent specification, and in which like referenced numbers indicate corresponding elements in the various figures.

[0047] Figure 1 is a block diagram of a psychological state evaluation interaction system provided by an embodiment of the present application;

[0048] Figure 2 is a flowchart of a psychological state evaluation interaction method provided by a specific embodiment of the present application;

[0049] Figure 3 is a mind map of a psychological state evaluation interaction method provided by a specific embodiment of the present application;

[0050] Figure 4 is a schematic diagram of the working principle of a psychological state evaluation interaction device provided by a specific embodiment of the present application;

[0051] Figure 5 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0052] To make the objectives, technical solutions, and advantages of the embodiments of the present application clearer, the embodiments of the present application will be described in detail below with reference to the drawings. However, those skilled in the art can understand that, in the embodiments of the present application, many technical details are presented in order to make the readers better understand the present application. However, the technical solutions claimed by the present application can be implemented even without these technical details and based on various changes and modifications of the following embodiments. The division of the following embodiments is for the convenience of description, and should not constitute any limitation on the specific implementation of the present application, and the embodiments can be combined and referenced with each other on the premise of not contradicting.

[0053] Hereinafter, the term "include" or "may include" used in various embodiments of the present application indicates existence of a certain function, operation, or element, and does not limit one or more functions, operations, or elements to be added. Also, the term "include" or "have" and its conjugates used in various embodiments of the present application are used to indicate a certain characteristic, number, step, operation, element, component or a combination thereof, but not to exclude the existence or possibility of adding one or more other characteristics, numbers, steps, operations, elements, components or combinations thereof.

[0054] In various embodiments of the present application, the expression "or" or "at least one of A, B, and C" includes any combination of the listed items A, B, and C, or any combination of A, B, and C. For example, the expression "A or B" or "at least one of A or B" can include A, B, or both A and B.

[0055] The expression used in various embodiments of the present application, such as "first", "second", etc., can modify various constituent elements in various embodiments, but can not limit the corresponding constituent elements. For example, the above expressions do not limit the order and / or importance of the elements. The above expressions are used only for the purpose of distinguishing one element from another element. For example, a first user device and a second user device indicate different user devices, although both are user devices. For example, a first element can be called a second element, and likewise, a second element can be called a first element, without departing from the scope of various embodiments of the present application.

[0056] It should be noted that if a certain constituent element is "connected" to or "linked" to another constituent element, the first constituent element can be directly connected to the second constituent element, and a third constituent element can be "connected" between the first constituent element and the second constituent element. Conversely, when a certain constituent element is "directly connected" to or "directly linked" to another constituent element, it can be understood that there is no third constituent element between the first constituent element and the second constituent element.

[0057] The terminology used in the description of the various embodiments of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the various embodiments of the application. As used herein, the singular forms "a," "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the various embodiments of the application belong. The terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

[0058] In order to make the purposes, technical solutions and advantages of the present application clearer, further specific description will be given below in combination with embodiments and drawings. The specific embodiments and their descriptions are only used to explain the present application and do not limit the present application.

[0059] At present, the existing psychological state evaluation interaction system has the following technical problems: AI psychological chat robot: only provides text conversation, no video generation and VR interaction function. VR exposure therapy system: relies on preset scene, cannot dynamically generate content. AI video generation tool: does not integrate psychological evaluation and harmful information filtering.

[0060] Embodiment one:

[0061] Based on this, the psychological state evaluation interaction system of the embodiment can be applied to electronic devices with communication, calculation and data storage capabilities. As shown in Figure 1 , the system can include:

[0062] The AI interaction module 801 is used for interaction with the user and obtaining feedback information of the user in the interaction process, and the feedback information includes voice reply data, text reply data and physiological data of the question;

[0063] The deep learning model 802 is used for extracting preliminary feature data of the feedback information and performing heterogeneous multi-modal fusion under multiple scales on the preliminary feature data to generate fusion feature data;

[0064] The support vector machine model 803 is used for evaluating the psychological state of the user according to the feedback information; the input data of the kernel function of the support vector machine model is the fusion feature data; the hyperplane of the support vector machine model is determined by the dimension of the feedback information, and the interval is determined by the support vector corresponding to the nearest negative feedback information of the hyperplane;

[0065] The VR video generation module 804 is configured to generate a corresponding VR video according to the psychological state, and push the VR video to the user, and the VR video has an interaction function with the user, so as to further determine a lower state of the psychological state of the user, and autonomously modify the VR video according to the lower state.

[0066] The embodiment of the present application provides a psychological state evaluation interaction system, which comprises an AI interaction module, a deep learning model, a support vector machine model and a VR video generation module.

[0067] In summary, the present application evaluates the psychological state of the user through AI dialogue analysis, automatically generates adaptive VR videos, and supports real-time intervention by instructors, achieving the following technical effects:

[0068] 1. Precise evaluation: extract emotional, cognitive and behavioral characteristics from dialogue to quantify psychological state.

[0069] 2. Dynamic generation: generate personalized prompts by combining expert database and user data to render high-definition VR videos.

[0070] 3. Safe and controllable: multi-modal review ensures harmless content, and instructors can adjust the generation logic at any time.

[0071] 4. Data-driven optimization: record user feedback to optimize long-term counseling plans.

[0072] The AI interaction module specifically comprises the following core modules:

[0073] Voice recognition feedback module: integrated with Google Speech-to-Text or Whisper API to support real-time text conversion and noise suppression.

[0074] Text input interface: provides API endpoints or UI components (such as chat windows) to receive user text.

[0075] Physiological data collection module: Obtain heart rate, galvanic skin response (GSR), etc. through Bluetooth / Wi-Fi connection with wearable devices (such as Apple Watch, Empatica E4). Use analysis cameras to capture facial expressions / eye movements.

[0076] Standardize timestamp, user ID, data type (JSON format).

[0077] After obtaining physiological data, use a time series database (such as InfluxDB) to store real-time physiological data, and a relational database to store interaction logs.

[0078] It should be noted that before obtaining the user's feedback information, the user's consent needs to be obtained and the user's privacy needs to be protected.

[0079] The workflow of the AI interaction module is as follows: Ask the user questions through voice / text, the user wears the device simultaneously to synchronize physiological data, timestamp aligns multi-source data, filters invalid signals (such as motion artifacts). After the user receives the AI reply, a new round of data collection starts, forming a continuous optimization cycle.

[0080] Deep learning models are used to process complex feedback information (such as user reviews, sensor data, log files, image / video clips, etc.) from different sources (modalities) and different granularities (scales), with the goal of generating a more comprehensive and robust feature representation for downstream tasks (such as sentiment analysis, question classification, prediction, recommendation, etc.).

[0081] Specifically, structured or numerical representations that represent the key attributes of the original feedback information (text reviews, ratings, voice recordings, video clips, clickstream logs, physiological signals (such as heart rate), facial expression images) are extracted. Each modality of data has its specific feature extraction method. Feedback information comes from different types (heterogeneous) sources (modalities), such as text + image + audio + numerical sensor data. The data structure and semantic meaning of each modality are different.

[0082] Data can be observed and analyzed at different granularities or time / space scales. For example: Text: character level, word level, phrase level, sentence level, paragraph level, document level. Image: pixel level, local block level, object level, scene level. Time series data (audio, sensor): millisecond level sampling point, short time frame (tens of milliseconds), segment level (several seconds), event level, session level.

[0083] Finally, the feature data from different modalities and different scales are integrated into a unified and more informative representation. The final output, a multi-modal and multi-scale information fused feature vector or representation, is used for subsequent analysis or model input.

[0084] For heterogeneous multi-modal fusion, we design one feature encoder / mapper (preferably a neural network) for each modality (or each modality-scale combination) to map their respective preliminary features into a shared semantic subspace or common representation space. In this space, features of different modalities / scales are comparable, and then fused.

[0085] During fusion, we first start from the finest-grained scale and gradually fuse to coarser-grained scales. For example: first fuse word-level features of all modalities -> sentence-level features -> paragraph-level features. Then, fuse features of all modalities at the same scale (e.g., fuse "segment-level" features of all modalities) to get the fused representation at this scale, and then fuse the fused representations of different scales. Finally, use an attention module to learn the importance weights of features of different scales (even different modality-scale combinations), and then perform weighted summation to generate the final fused representation. This allows the model to dynamically focus on the most relevant scale and modality information.

[0086] Deep learning models and support vector machine models are used specifically for:

[0087] 1. Dialog information processing (deep learning module).

[0088] Semantic features are extracted using a pre-trained language model (such as BERT): E text = BERT(X text ) (get sentence embedding vector, dimension d text )

[0089] Here BERT(X text ) represents the output representation obtained after inputting the text X text into the BERT model. E text represents the original text input, which can be a sentence, a paragraph, or a word sequence. d text represents the dimension of the sentence embedding vector, which depends on the configuration of the BERT model used:

[0090] BERT-base: d text = 768

[0091] BERT-large: d text = 1024

[0092] 2. Physiological parameter feature engineering.

[0093] Numerical parameters (heart rate, blood pressure, etc.) are directly standardized:

[0094] X phys_num = [X hr , X bp,sys , X bp,dia , Xtemp ,X rr ]∈R 5 ;

[0095] (standardized, mean vector μ, standard deviation vector σ);

[0096] Here Xphys_num is the original numerical physiological parameter vector, containing multiple physiological indicators (e.g. heart rate, blood pressure, etc.), with dimension R 5 (i.e. 5-dimensional vector).

[0097] Specific physiological parameters:

[0098] X hr : Heart Rate, usually in beats per minute (bpm).

[0099] X bp,sys : Systolic Blood Pressure, in mmHg.

[0100] X bp,dia : Diastolic Blood Pressure, in mmHg.

[0101] Note: Systolic and Diastolic Blood Pressure can be combined into one "Blood Pressure" indicator (e.g. [X bp,sys ,X bp,dia ]), occupying 2 dimensions.

[0102] X temp : Temperature, possibly in °C or °F.

[0103] X rr : Respiratory Rate, in beats per minute.

[0104] X^phys_num is the standardized physiological parameter vector, X phys_num is the unprocessed physiological parameter vector.

[0105] Facial expression and eye movement tracking features are extracted through a CNN (Convolutional Neural Network):

[0106] H face = CNN face (X face ) ∈ R d1 (Facial expression feature extraction);

[0107] H eye = CNN eye (X eye ) ∈ R d2(Eye movement feature extraction)

[0108] Here X face means the original facial image data, usually a multi-channel (such as RGB) two-dimensional or three-dimensional matrix. CNN face means a convolutional neural network model (such as VGG, ResNet or a custom lightweight network) specifically designed for processing facial expressions. H face means the facial expression feature vector extracted by CNN, with dimension R d1 .

[0109] X eye means the original eye movement data. CNN eye means a convolutional neural network for processing eye movement data. H eye means the eye movement feature vector extracted by CNN, with dimension R d2 . The dimensions d1, d2: the length of the feature vector, determined by the model architecture.

[0110] 3. Feature fusion (Deep learning module)

[0111] After concatenating the dialogue, physiological, and adjustment instruction features, they are fused through a fully connected layer:

[0112] Here H raw means the original joint feature vector after horizontally concatenating all features. C means the adjustment instruction feature (such as user-set preferences or control instructions). The total dimension is d text +5+d1+d2+m.

[0113] H dl = σ (W1·H raw +b1) (activation function such as ReLU rectified linear unit, to get deep features);

[0114] Here W1 means the weight matrix of the fully connected layer. b1 means the bias vector of the fully connected layer. σ means the activation function, which introduces nonlinearity. H dl means the fused deep feature representation, used for downstream tasks (such as classification, regression). Among them, is the weight matrix, d raw is the total dimension of the original features, d h is the output dimension of the fully connected layer hidden layer.

[0115] 4. Mental state classification (Support Vector Machine model)

[0116] The features output by the deep learning are input into the support vector machine model for classification:

[0117]

[0118] SVM kernel Support Vector Machine model, H dl Deep fusion feature vector output by the fully connected layer. n denotes the number of Support Vectors, i.e., the number of samples in the training set that play a key role in the classification boundary. a i Lagrange multiplier (weight) of the i-th Support Vector, which is optimized by training the Support Vector Machine model. y i True label of the i-th Support Vector. Kernel Function, used to calculate the similarity between the input feature H dl and the i-th Support Vector. Bias term of the classifier, obtained by training. f(H dl ) denotes the decision function output of the Support Vector Machine model, representing the original classification score (not normalized) of the input feature H dl . The final output is the psychological state evaluation result (probability or category):

[0119] p(y|X) = softmax(f(H dl )) or y = argmax(f(H dl )).

[0120] Here p(y|X) denotes the probability distribution of predicting the psychological state under the condition of inputting the multi-modal feature X (text, physiology, face, eye movement, etc.). f(H dl ) denotes the original decision value of the Support Vector Machine model, which can be a scalar (binary classification) or a vector (multi-classification). Softmax(·) serves to convert the original decision value into a probability distribution. y denotes the directly predicted psychological state category label, which takes the category with the maximum decision value. argmax(·) serves to return the index of the maximum value in the vector (i.e., the corresponding category).

[0121] The purpose of the support vector machine model is to find an optimal hyperplane in a high-dimensional space so that different classes of samples are separated as much as possible. The process includes: finding a hyperplane that separates different classes of data points, then improving the generalization ability of the model by maximizing the margin. The margin refers to the distance between the hyperplane and the nearest data point (support vector) to it. Support vectors are those data points closest to the hyperplane, and the goal of the support vector machine model is to maximize the distance of these support vectors to the hyperplane. The larger the margin, the stronger the generalization ability of the model, that is, its performance will be better when facing unknown data. It should be noted that support vectors are points located between two classes, closest to the separation hyperplane. They are crucial to the establishment of the decision boundary, and the classification effect of the support vector machine model depends only on these support vectors, not on other data points.

[0122] The psychological state evaluation module is specifically used to first construct a feature vector from the multi-dimensional features of user feedback information (voice, text, physiological data), and then learn a hyperplane determined by the dimensions of the feedback information through a support vector machine model to maximize the positive and negative sample margin. It can be understood that this way pays more attention to the recognition of negative psychological states (such as anxiety, anger), and is suitable for scenarios that require timely intervention such as customer service and education.

[0123] For the VR video generation module, first, an initial VR scene is generated according to the psychological state, and an interactive trigger is embedded, then user behavior (action, voice, physiological data) is captured in real time. Analyze the interaction data to determine the lower psychological state (such as mild anxiety → social avoidance), and finally modify the VR content based on the lower state to form a closed-loop feedback.

[0124] Embodiment Two:

[0125] In some embodiments, the AI interaction module includes a voice recognition feedback module and a wearable physiological detection device. The voice recognition feedback module is used to convert the voice reply data into text reply data and identify the tone of the voice reply data. The wearable physiological detection device is used to obtain the physiological data, which includes facial expressions, heart rate, blood pressure, body temperature, respiratory rate, and eye movement tracking.

[0126] The main task of the voice recognition feedback module is to convert the user's voice reply into text reply data, and identify the tone (such as emotion, speech rate, voice intensity, etc.) in the voice to further assess the user's emotional state.

[0127] The speech recognition feedback module uses automatic speech recognition technology to convert the user's speech input into corresponding text, which involves acoustic models, language models, and processing of speech signals. In addition to text conversion, the speech recognition feedback module also needs to analyze the tone of the speech (such as pitch, volume, speed, tone, etc.) to identify emotional states. For example, a rapid tone may indicate anxiety, while a low voice may indicate depression. Speech emotion analysis techniques include emotion-based speech recognition models, preferably using deep learning methods to analyze emotional features in speech.

[0128] The wearable physiological detection device not only provides an immersive virtual reality experience, but also monitors and collects physiological data in real time. These physiological data provide important feedback on the user's physiological state, helping to more accurately assess their psychological state.

[0129] The camera equipped on the wearable physiological detection device can capture the user's facial expressions through computer vision algorithms (such as OpenCV, deep learning models, etc.). Facial expressions can reflect the user's emotional state, such as a smile indicating happiness and a frown indicating anxiety or unease.

[0130] The wearable physiological detection device can also be equipped with physiological sensors (such as heart rate monitors, skin temperature sensors, etc.) to monitor the user's physiological indicators (heart rate, blood pressure, body temperature, respiratory rate):

[0131] Heart rate: Reflects the user's tension or relaxation state. For example, heart rate increases when anxious and decreases when relaxed.

[0132] Blood pressure: High blood pressure may indicate that the user is under high pressure or anxious, and low blood pressure may be related to emotions such as depression.

[0133] Body temperature: Changes in body temperature may reflect the user's physiological or emotional fluctuations.

[0134] Respiratory rate: Frequent shallow breathing indicates anxiety or stress, while deep and regular breathing may indicate relaxation.

[0135] Eye tracking: Eye tracking can detect the user's eye movements, fixation points, and eye fixation positions. Through eye movement data, the user's attention concentration, anxiety or confusion state can be analyzed. For example, rapid eye movements or long-term fixation on a point may indicate anxiety or distress.

[0136] Based on the text reply and tone analysis, emotional features (such as anxiety, relaxation, happiness, depression, etc.) are extracted. Combined with facial expressions and physiological data (such as heart rate, respiratory rate, etc.), the user's physiological response is further determined to help determine whether the user is in a state of stress, anxiety, or relaxation. These information is analyzed comprehensively through support vector machine model algorithm to obtain the final psychological state evaluation result.

[0137] In some embodiments, the sensitivity of the kernel function of the support vector machine is determined by the amount of feedback information.

[0138] The "sensitivity" of a kernel function refers to the degree to which the model responds to changes in data or the size of the training data. In a support vector machine model, the choice and tuning of the kernel function is crucial to the performance of the model, especially when the amount of feedback information (such as user's physiological data, speech data, etc.) changes, the sensitivity of the kernel function also changes.

[0139] When the amount of data is small: If the amount of feedback information is small, the model is easily influenced by each data point. At this time, it is very important to choose the appropriate kernel function and adjust its hyperparameters (such as the gamma parameter of the RBF kernel). If the kernel function is too sensitive (i.e., the gamma value is too large), the model may overfit, resulting in poor generalization ability on new data. Conversely, if the kernel function is not sensitive enough (i.e., the gamma value is too small), it may not be able to capture the complex patterns in the data, resulting in underfitting.

[0140] When the amount of data is large: When the amount of feedback information increases, the performance of the support vector machine improves, because more data can help the model better capture the global structure of the data. In this case, the sensitivity of the kernel function needs to be relatively low (i.e., reduce the value of gamma) to avoid overfitting, because more samples can help the model identify the overall trend of the data rather than being affected by individual outliers.

[0141] Diversity of training data: If the types and ranges of feedback information are very extensive (for example, there are physiological data, speech data and facial expression data), the kernel function of the support vector machine model may need higher sensitivity to handle different data patterns. At this time, the hyperparameters of the kernel function need to be adjusted according to the distribution of the data to balance the response to local data patterns and the learning of global data structure.

[0142] In a support vector machine model, adjusting the sensitivity of the kernel function can be achieved through hyperparameter optimization. The main hyperparameters include:

[0143] Gamma value of RBF kernel: Gamma controls the sensitivity of the RBF kernel, the larger the value, the more sensitive the response of the kernel function, prone to overfitting; the smaller the value, the less sensitive the response, may lead to underfitting. Adjusting the size of the gamma value can be achieved through cross-validation.

[0144] Penalty parameter C: The C value controls the degree of punishment for errors. A larger C value makes the model pay more attention to the accurate classification of training data, which may lead to overfitting; a smaller C value makes the model tolerate more errors, thereby improving the generalization ability.

[0145] When there is a small amount of feedback information: Careful selection of kernel function parameters is needed, and cross-validation is needed to select appropriate gamma and C values. Due to the small amount of training data, the model is more susceptible to each data point, so the gamma value can be appropriately reduced to make the model less sensitive to noise.

[0146] When there is a large amount of feedback information: With the increase in the amount of data, the gamma value can be appropriately increased to better capture the details in the data. However, at the same time, overfitting needs to be avoided, so cross-validation or other regularization methods (such as adjusting the C value) can be used to ensure the generalization ability of the model.

[0147] Specifically, grid search or random search can be used to automatically adjust the hyperparameters of the kernel function to find the most appropriate settings under the current amount of feedback data. If the system can monitor the number and distribution of feedback information in real time, the sensitivity of the kernel function can be dynamically adjusted to ensure that the model always works in the optimal state.

[0148] In some embodiments, one embodiment provides a psychological state assessment interaction system, further comprising:

[0149] A keyword weight modification module is configured to modify the weight of the keyword in the support vector machine model if the feedback information contains a preset keyword; the weight is determined by a preset mapping relationship between the keyword and the weight.

[0150] Specifically, first, the keywords need to be extracted from the feedback information. These keywords can be significant words in the text data (such as important terms in user feedback), or features generated by external perception devices (such as speech recognition, facial expression analysis, etc.).

[0151] If the feedback information is in text form, keywords can be extracted through natural language processing methods. If the feedback information is voice data, it needs to be converted into text through speech recognition technology, and then keywords are extracted. The emotional information in the voice (such as speech speed, tone, etc.) can also be part of the keyword features. For physiological data (such as heart rate, body temperature, etc.), data points with significant impact can be extracted by defining certain thresholds or rules as keywords.

[0152] After extracting the keywords, the importance of the keywords to the model's prediction results can be reflected by modifying the weights of the keywords.

[0153] Specifically:

[0154] Frequency or importance-based weight adjustment: If a keyword appears frequently in feedback information or is assessed as important through certain means (such as sentiment analysis, TF-IDF scoring, etc.), its weight can be increased, considering it more important for classification or prediction tasks.

[0155] Sentiment analysis weighting: In sentiment analysis, the intensity of keywords' sentiment (e.g., obtained through sentiment dictionaries) can be used to adjust their weights. For example, certain negative sentiment keywords (such as "anxiety") can be given higher weights as they have a greater impact on mental state prediction.

[0156] Feedback pattern-based weight adjustment: If the feedback patterns corresponding to keywords are considered to have higher predictive power in past learning, the weights of these keywords can be adjusted through historical data.

[0157] Adaptive adjustment during training: The weights can be dynamically adjusted during training through methods such as gradient descent, using the model's prediction errors to guide weight updates.

[0158] Here is an example flow: Suppose a user expresses "I recently feel very anxious, heart racing, breathing rapidly" through voice feedback.

[0159] Keyword extraction: Convert it to text through speech recognition: "I recently feel very anxious, heart racing, breathing rapidly". Extract keywords: "anxiety", "heart racing", "breathing rapidly".

[0160] Keyword weight adjustment: According to sentiment analysis, the keyword "anxiety" can be given a higher weight. "Heart racing" and "breathing rapidly" can be weighted according to the intensity of their physiological responses. Construct feature vector: Numericalize the weight values of these keywords and combine them into a feature vector, for example:

[0161] ["anxiety": 0.8, "heart racing": 0.5, "breathing rapidly": 0.6], input into a support vector machine model to make an emotional state prediction.

[0162] In some embodiments, the VR video generation module is specifically configured to determine the selection interface scene, style, constraint condition, and visual parameter of the lower state of the pre-generated VR video according to the mental state.

[0163] Taking the "anxiety" superstate as an example, the lower states and corresponding VR parameters are defined as shown in Table 1:

[0164] Table 1

[0165]

[0166]

[0167] A color space conversion method for determining a lower state of a pre-generated VR video according to a psychological state is provided herein, and a color space conversion formula is provided:

[0168] H' = H + 120 x anxiety index

[0169] S' = max(0.7, 1-0.3 x respiratory rate)

[0170] V' = 0.5 + 0.5 x sin(2π x HRV_normalized)

[0171] In the above formula, the HSV color space is dynamically adjusted, H: hue, S: saturation, and V: lightness.

[0172] In the above manner, the system can achieve a millisecond-level response closed loop from psychological state recognition to VR scene parameter adjustment, while ensuring user experience comfort and accurately implementing psychological intervention. Actual deployment needs to be combined with local cultural characteristics (such as differences in red sensitivity of Eastern users) for parameter fine-tuning.

[0173] In some embodiments, the VR video generation module is specifically configured to generate a VR video corresponding to the lower state according to feedback information of the user for the selection interface scene.

[0174] The embodiment of the present application provides a psychological state evaluation interaction system, which comprises: an AI interaction module, which is used for interacting with a user and acquiring feedback information of the user in an interaction process, the feedback information comprising voice reply data, text reply data and physiological data of a question; a deep learning model, which is used for extracting feature data of the feedback information; a support vector machine model, which is used for evaluating a psychological state of the user according to the feedback information; input data of a kernel function of the support vector machine model is the feature data; a hyperplane of the support vector machine model is determined by a dimension of the feedback information, and a gap is determined by support vectors corresponding to negative feedback information closest to the hyperplane; a VR video generation module, which is used for generating a corresponding VR video according to the psychological state and pushing the VR video to the user, and the VR video has an interaction function with the user, so as to further determine a lower state of the psychological state of the user and autonomously modify the VR video according to the lower state.

[0175] The embodiment of the present application provides a psychological state evaluation interaction system, which comprises: an AI interaction module, which is used for interacting with a user and acquiring feedback information of the user in an interaction process, the feedback information comprising voice reply data, text reply data and physiological data of a question; a deep learning model, which is used for extracting feature data of the feedback information; a support vector machine model, which is used for evaluating a psychological state of the user according to the feedback information; input data of a kernel function of the support vector machine model is the feature data; a hyperplane of the support vector machine model is determined by a dimension of the feedback information, and a gap is determined by support vectors corresponding to negative feedback information closest to the hyperplane; a VR video generation module, which is used for generating a corresponding VR video according to the psychological state and pushing the VR video to the user, and the VR video has an interaction function with the user, so as to further determine a lower state of the psychological state of the user and autonomously modify the VR video according to the lower state.

[0176] Embodiment three:

[0177] To further illustrate the scheme, the present application also provides an embodiment of a mental state assessment interaction method, see Figure 2 and Figure 3 , specifically including the following.

[0178] First, see Figure 4 , the present application provides a mental state assessment interaction device, which comprises a user interaction module, a psychological assessment engine, a video Prompt generator, a video generation and review module, a VR playback and intervention module, and a data management and reporting module, specifically:

[0179] User interaction module: the user inputs the psychological state description (such as "I have been insomnia, always worry about work mistakes") through voice / text dialogue with AI large model. If the dialogue information is insufficient, the model can actively ask questions (such as: "What perspective do you need? What are the characteristics of the role?") to supplement the insufficient information. Support multi-modal input (such as expression recognition, voice tone analysis).

[0180] The psychological assessment engine includes:

[0181] Analysis layer: use large language model to extract keywords (such as "insomnia" → sleep disorder, "worry about mistakes" → anxiety tendency), and analyze the intention of the dialogue.

[0182] Emotion quantification layer: output emotion score (anxiety value: 0.82 / 1.0) through emotion analysis model (such as VADER). Determine the style (realistic, cartoon, cyberpunk) or emotion (tense, warm) implied in the dialogue.

[0183] Entity extraction: extract key entities (such as roles, objects, scenes) and attributes (colors, actions, times).

[0184] Expert library matching: according to the expert preset psychological knowledge base, video feature point scanning, and according to EMDR, DSM-5 or ICD-11 standard matching counseling strategy (such as "cognitive behavioral therapy-sleep restriction method").

[0185] Video Prompt generator: convert the extracted information into Prompt format supported by video generation tools.

[0186] Video generation and review module: call AI video generation class large model tools.

[0187] VR playback and intervention module: Users watch videos through VR headsets (such as PICO 4) using streaming technology, and the system can monitor physiological indicators (such as heart rate, blood pressure, and variability) in real time. The VR motion capture device can be used to obtain user feature actions, so that the system can make corresponding adjustments. The tutor console: can pause the playback, modify the Prompt parameters (such as adding "deep breathing guidance"), and real-time regenerate the VR video. The tutor console includes at least one of pausing the playback, modifying the Prompt parameters, and inserting interactive instructions.

[0188] Data management and reporting module: stores all session data and generates visual reports. Includes: emotional change trend chart, high-frequency question keyword cloud, and video effectiveness score (based on user feedback).

[0189] Based on the above-mentioned psychological state evaluation interaction device, the specific implementation of the psychological state evaluation interaction method provided by the application includes the following steps:

[0190] S1: Interact with the user and obtain feedback information of the user during the interaction, and the feedback information includes voice reply data, text reply data and physiological data of the question;

[0191] The user inputs the psychological state description (such as "I have insomnia recently and always worry about making mistakes at work") through voice / text dialogue with the AI large model. If the dialogue information is insufficient, the model can actively ask questions (such as: "What perspective do you need? What are the characteristics of the role?") to supplement the insufficient information. In addition, it also supports multi-modal input (such as expression recognition, voice tone analysis).

[0192] S2: Extract feature data of the feedback information through a deep learning model;

[0193] S3: Evaluate the psychological state of the user according to the support vector machine model and the feedback information; the input data of the kernel function of the support vector machine model is the feature data; the hyperplane of the support vector machine model is determined by the dimension of the feedback information, and the interval is determined by the support vector corresponding to the nearest negative feedback information of the hyperplane distance;

[0194] Use a large language model to extract keywords (such as "insomnia" → sleep disorder, "worry about making mistakes" → anxiety tendency) and analyze the intent of the dialogue. Output the emotional score (anxiety value: 0.82 / 1.0) through the sentiment analysis model (such as VADER). Determine the style (realistic, cartoon, cyberpunk) or emotion (tense, warm) implied in the dialogue. Extract key entities (such as characters, objects, and scenes) and attributes (color, action, and time). According to the expert pre-set psychological knowledge base, video feature point scanning, and matching the counseling strategy (such as "cognitive behavioral therapy-sleep restriction method") according to the EMDR, DSM-5 or ICD-11 standard.

[0195] S4: Generate a corresponding VR video according to the psychological state and using the video generation large model, and push it to the user, and the VR video has an interactive function with the user to further determine the lower state of the user's psychological state, and modify the VR video according to the lower state.

[0196] Convert the extracted information into a Prompt format supported by the video generation tool, which needs to include:

[0197] Scene description (background, subject, action) sets the environmental tone, provides spatial sense and atmosphere. Determine the focal point of the picture, carry the narrative or visual impact. Give dynamic performance, avoid stiff pictures.

[0198] Visual parameters (resolution, frame rate, lens) can automatically supplement default parameters (resolution, frame rate) or recommended options.

[0199] Style reference (artist, film style) captures keywords to determine style, which can also be manually or automatically set to default style.

[0200] Constraints (avoid blur, flicker, etc.) can set and modify default constraints.

[0201] Input: psychological assessment results (such as "high anxiety + sleep problems") + user preferences (such as "natural scenes").

[0202] Output: Structured Prompt example:

[0203]

[0204] S5: Review VR video.

[0205] Specifically, the following reviews are required: visual review: detect violent and bloody content. Voice review: filter negative suggestive statements (such as "give up"). Key word blacklist can be set for quick comparison and exclusion. The output is in VR compatible format (such as 360° panoramic video).

[0206] S6: Play and intervene in VR video.

[0207] Using streaming technology, users watch videos through VR headsets (such as PICO 4), and the system can monitor physiological indicators (such as heart rate, blood pressure, and variability) in real time. Combined with VR motion capture equipment to obtain user characteristic actions, the system can make corresponding adjustments. The video can be paused, the Prompt parameters can be modified (such as adding "deep breathing guidance"), and the VR video can be regenerated in real time.

[0208] S7: Generate a psychological state assessment report.

[0209] Store all conversation data, generate visual reports, including: emotional change trend chart, high-frequency question keyword cloud, and video effectiveness score (based on user feedback).

[0210] Specific embodiment: social anxiety counseling

[0211]

[0212] Embodiment 2: Depression auxiliary treatment

[0213]

[0214]

[0215] The specific embodiment of the present application provides a psychological state evaluation interaction method, which comprises the following steps: first, interacting with the user and obtaining feedback information of the user during the interaction, the feedback information including voice reply data, text reply data and physiological data of the question; extracting feature data of the feedback information through a deep learning model; then, evaluating the psychological state of the user according to a support vector machine model and the feedback information; the input data of the kernel function of the support vector machine model is the feature data; the hyperplane of the support vector machine model is determined by the dimension of the feedback information, and the interval is determined by the support vector corresponding to the nearest negative feedback information of the hyperplane; finally, generating a corresponding VR video according to the psychological state and pushing it to the user, and the VR video has an interaction function with the user to further determine the lower state of the psychological state of the user, and the VR video is modified automatically according to the lower state. The present application has the following beneficial effects:

[0216] 1. Evaluation efficiency is improved: 10 minutes of conversation can complete the evaluation of traditional 1 hour interview.

[0217] 2. Content personalization: the video is generated to match the psychological characteristics of the user (accuracy rate 92%, based on 1000 test data).

[0218] 3. Real-time intervention capability: the counselor adjusts the video to be regenerated in less than 3 seconds (RTX 4090 graphics card).

[0219] 4. Safety guarantee: harmful content interception rate is 99.3% (test data set: 100,000 negative prompts).

[0220] In summary, the specific embodiment of the present application provides a psychological state evaluation interaction method, which first integrates AI conversation analysis, dynamic video generation and VR real-time intervention. It conforms to the principles of psychological treatment (such as EMDR, exposure therapy, mindfulness training, desensitization training, etc.). It can be applied to psychological counseling agencies, telemedicine platforms and personal health management APPs.

[0221] Embodiment four:

[0222] Another embodiment of the present application relates to an electronic device, such as Figure 5 As shown, the electronic device specifically includes the following:

[0223] The processor (processor) 1201, memory (memory) 1202, communication interface (communications interface) 1203 and bus 1204;

[0224] The processor 1201, memory 1202, communication interface 1203 communicate with each other through the bus 1204; the communication interface 1203 is used to realize the information transmission between the server side device and the user side device and other related devices;

[0225] The processor 1201 is used to call the computer program in the memory 1202, and the processor executes the computer program to realize all steps in the psychological state evaluation interaction method in the above embodiment, for example, the processor executes the computer program to realize the following steps:

[0226] Interact with the user and obtain the feedback information of the user in the interaction process, the feedback information including voice reply data, text reply data and physiological data of the question;

[0227] Extract the feature data of the feedback information through the deep learning model;

[0228] According to the support vector machine model and the feedback information, the psychological state of the user is evaluated; the input data of the kernel function of the support vector machine model is the feature data; the hyperplane of the support vector machine model is determined by the dimension of the feedback information, and the interval is determined by the support vector corresponding to the nearest negative feedback information of the hyperplane distance;

[0229] According to the psychological state, the corresponding VR video is generated and pushed to the user, and the VR video has an interactive function with the user to further determine the lower state of the psychological state of the user, and the VR video is modified according to the lower state.

[0230] The memory and the processor are connected in a bus manner, the bus can include any number of interconnected buses and bridges, the bus connects various circuits of one or more processors and memories together. The bus can also connect various other circuits such as peripheral devices, voltage stabilizers and power management circuits together, which are well known in the art, therefore, further description will not be made herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be one element or multiple elements such as multiple receivers and transmitters, which provide units for communicating with various other devices on the transmission medium. The data processed by the processor is transmitted on the wireless medium through the antenna, further, the antenna also receives data and transmits the data to the processor.

[0231] The processor is responsible for managing the bus and general processing, and can also provide various functions, including timing, peripheral interface, voltage regulation, power management and other control functions. And the memory can be used to store the data used by the processor in the execution operation.

[0232] Embodiment five:

[0233] Another embodiment of the application relates to a computer readable storage medium, which stores a computer program. The computer program is executed by the processor to realize the steps in the above-mentioned embodiment of the mental state evaluation interaction method, and the steps include:

[0234] Interact with the user and obtain feedback information of the user during the interaction, the feedback information including voice reply data, text reply data and physiological data of the question;

[0235] Extract feature data of the feedback information through a deep learning model;

[0236] Evaluate the mental state of the user according to a support vector machine model and the feedback information; input data of a kernel function of the support vector machine model is the feature data; a hyperplane of the support vector machine model is determined by a dimension of the feedback information, and a margin is determined by a support vector corresponding to the feedback information closest to the hyperplane;

[0237] Generate a corresponding VR video according to the mental state and push it to the user, and the VR video has an interaction function with the user to further determine a lower state of the mental state of the user, and modify the VR video according to the lower state.

[0238] The various embodiments described in this specification are intended to be illustrative only and in no way limit the scope of the application. Those skilled in the art will be able to devise numerous alternative arrangements without departing from the scope of the application. Numerous specific details are described in this specification to provide a thorough understanding of the application. However, embodiments of the application can be practiced without these specific details. In other instances, well known methods have not been described in detail in order to avoid obscuring the application. References to a method comprising two or more steps, are intended to mean that the method can comprise these two or more steps in any suitable order or simultaneously where possible, or that the method can comprise additional steps which are not disclosed.

[0239] The above description discloses only a few variations. Many other variations are possible, as will be apparent to those with ordinary skill in the art having the benefit of this disclosure. The scope of the application is not to be limited by the specific descriptions and specific examples provided above.

[0240] Although the method operations were described in a specific order, this was done for convenience only. In other embodiments, the order of the operations can be different, the specific operations can be different, or various operations can be omitted or added. In some embodiments, the order of the operations can be changed, or two or more operations can be performed at the same time, or at different times. In some embodiments, the operations can be performed in an order different than that described in the examples or flow charts.

[0241] The present application is described in terms of exemplary embodiments, devices (systems), and computer programs, in flowcharts and / or block diagrams. It will be understood that each block of the flowcharts and / or block diagrams, and combinations of blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processing system, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowcharts and / or block diagrams block or blocks. Figure 1 The flowcharts and / or block diagrams have been simplified for ease of discussion, and so should not be construed as requiring a particular order of implementation in accordance with any aspect disclosed herein. Figure 1 The flowcharts and / or block diagrams have been simplified for ease of discussion, and so should not be construed as requiring a particular order of implementation in accordance with any aspect disclosed herein.

[0242] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the flowcharts and / or block diagrams block or blocks. Figure 1 The flowcharts and / or block diagrams have been simplified for ease of discussion, and so should not be construed as requiring a particular order of implementation in accordance with any aspect disclosed herein. Figure 1 The flowcharts and / or block diagrams have been simplified for ease of discussion, and so should not be construed as requiring a particular order of implementation in accordance with any aspect disclosed herein.

[0243] These computer program instructions can also be loaded into a computer or other programmable data processing devices, so that a series of operational steps are generated to realize the computer-implemented processes, and the instructions executed on the computer or other programmable devices provide a process for implementing the functions specified in the flowchart Figure 1 one flow or multiple flows and / or the functions specified in the block Figure 1 one flow or multiple flows and / or the functions specified in the block

[0244] The principles and implementation manners of the present application are described in the specific embodiments. The above embodiment descriptions are only used to help understand the method and core idea of the present application. Meanwhile, for those skilled in the art, according to the idea of the present application, the specific implementation manners and application ranges will be changed. In summary, the content of the present application should not be understood as a limitation.

Claims

1. A psychological state assessment interactive system, characterized in that, include: The AI ​​interaction module is used to interact with users and obtain user feedback information during the interaction process. The feedback information includes voice response data, text response data, and physiological data. A deep learning model is used to extract preliminary feature data from the feedback information and to perform heterogeneous multimodal fusion on the preliminary feature data at multiple scales to generate fused feature data. A support vector machine model is used to evaluate the user's psychological state based on the feedback information; the input data of the kernel function of the support vector machine model is the fused feature data. The hyperplane of the support vector machine model is determined by the dimension of the feedback information, and the interval is determined by the support vector corresponding to the nearest negative feedback information to the hyperplane; The VR video generation module is used to generate corresponding VR videos based on the psychological state and push them to the user. The VR videos have interactive functions with the user to further determine the lower-level state of the user's psychological state and to autonomously modify the VR videos based on the lower-level state.

2. The psychological state assessment interactive system according to claim 1, characterized in that, The AI ​​interaction module includes a voice recognition feedback module and a wearable physiological detection device. The voice recognition feedback module is used to convert the voice response data into text response data and to recognize the tone of the voice response data. The wearable physiological detection device is used to acquire the physiological data, which includes facial expressions, heart rate, blood pressure, body temperature, respiratory rate, and eye tracking.

3. The psychological state assessment interactive system according to claim 1, characterized in that, The sensitivity of the kernel function of the support vector machine is determined by the amount of feedback information.

4. The psychological state assessment interactive system according to claim 1, characterized in that, Also includes: The keyword weight modification module is used to modify the weight of the keyword in the support vector machine model if the feedback information contains a preset keyword. The weights are determined by a preset mapping relationship between keywords and weights.

5. The psychological state assessment interactive system according to claim 1, characterized in that, The VR video generation module is specifically used to determine the selection interface scene, style, constraints, and visual parameters of the lower-level state of the pre-generated VR video based on the psychological state.

6. The psychological state assessment interactive system according to claim 5, characterized in that, The VR video generation module is specifically used to generate a VR video corresponding to the lower-level state based on the user's feedback information on the selected interface scenario.

7. The psychological state assessment interactive system according to claim 1, characterized in that, Also includes: The VR video review module is used to remove negative visual data, text data, and audio data from the VR video.

8. A psychological state assessment interaction method applied to the psychological state assessment interaction system according to any one of claims 1 to 7, characterized in that, include: Interact with users and obtain user feedback information during the interaction process. The feedback information includes voice response data, text response data, and physiological data. Preliminary feature data of the feedback information is extracted through a deep learning model, and heterogeneous multimodal fusion is performed on the preliminary feature data at multiple scales to generate fused feature data. The user's psychological state is assessed based on the support vector machine model and the feedback information. The input data of the kernel function of the support vector machine model is the fused feature data; The hyperplane of the support vector machine model is determined by the dimension of the feedback information, and the interval is determined by the support vector corresponding to the nearest negative feedback information to the hyperplane; A corresponding VR video is generated based on the user's psychological state and pushed to the user. The VR video has interactive functions to further determine the user's lower-level psychological state and to autonomously modify the VR video based on the lower-level psychological state.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the interactive method for assessing mental states as described in claim 1.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the interactive method for assessing mental states as described in claim 1.

Images