Intelligent health information consultation system and method based on interactive voice response

Abstract

The invention relates to the field of intelligent consultation of health information, and particularly discloses an intelligent health information consultation system and method based on interactive voice response, and the method comprises the steps: firstly obtaining a psychological consultation voice signal of a consultant and psychological health information data of a plurality of different emotion labels in a database, and enabling the psychological consultation voice signal and the psychological health information data to serve as input data; then, the deep learning technology is used for conducting feature extraction and analysis on the input data, and then feature information is comprehensively analyzed to obtain a classification result used for representing psychological emotion labels of consultants. In this way, the system can generate personalized psychological health voice responses based on the psychological emotion labels. Therefore, timely, convenient and confidential psychological support can be effectively provided for the consultant, resonance of the consultant is further caused, and the consultant is helped to relieve pressure and dispatch mind of the consultant.

Description

Technical Field

[0001] The present application relates to the field of intelligent consultation of health information, and more specifically, to a health information intelligent consultation system and method based on interactive voice response. Background Art

[0002] With the development of society and the economy, college students face increasing temptations and challenges. College is a crucial period for students to develop a sound worldview, values, and social outlook. Because of their higher education, college students are more attentive to their inner selves. Therefore, universities must prioritize the mental health of their students, as it is crucial to their growth.

[0003] Mental health education for college students refers to the extent to which college counselors address the mental health of their students. Mental health education encompasses aspects of personal emotions, interests, values, and attitudes. A healthy mental state is crucial for college students to develop positive values. A positive mindset promotes their growth and maturity, helping them connect with others and establish healthy interpersonal relationships. Furthermore, mental health education can cultivate a positive outlook on life and values, giving them the courage to face difficulties and challenges.

[0004] Psychological counseling provides face-to-face interaction with mental health educators, making it the most direct, effective, and memorable way to strengthen mental health education for college students. After all, college students are relatively inexperienced and often lack a deep understanding of mental health. By discussing their concerns with mental health counselors, students can develop a deeper understanding and resolve their psychological issues.

[0005] However, the number of psychological counselors on campus is limited, and psychological counseling usually requires appointments, which is limited in time and may not be able to meet the needs of college students in a timely manner. In addition, some college students may feel ashamed or embarrassed about seeking psychological counseling and give up seeking counseling.

[0006] Therefore, we look forward to a health information intelligent consultation system and method based on interactive voice response, which can analyze the psychological counseling problems of college students through voice response, provide timely, convenient and confidential psychological support, and then resonate with the consultants, helping them relieve stress and dispel their inner distress. Summary of the Invention

[0007] In order to solve the above-mentioned technical problems, the present application is proposed. The embodiment of the present application provides a health information intelligent consultation system and method based on interactive voice response, which first obtains the consultant's psychological consultation voice signal and the psychological health information data of multiple different emotional labels in the database as input data, and then uses deep learning technology to extract and analyze the features of these input data respectively, and then comprehensively analyzes these feature information to obtain classification results for representing the consultant's psychological emotional labels. In this way, the system can generate personalized psychological health voice responses based on the psychological emotional labels. In this way, it can effectively provide consultants with timely, convenient and confidential psychological support, thereby arousing the consultant's resonance and helping them relieve stress and dispel their inner depression.

[0008] According to a first aspect of the present application, there is provided a health information intelligent consultation system based on interactive voice response, comprising:

[0009] The health information consultation data acquisition module is used to obtain the psychological consultation voice signal of the consultant and obtain psychological health information data with multiple different emotional labels from the database;

[0010] A health information consultation feature acquisition module is used to extract and analyze the psychological consultation voice signal of the consultant and the psychological health information data with multiple different emotion labels obtained from the database to obtain the consultant's voice feature vector and the psychological health information feature matrix;

[0011] A health information consultation feature fusion module is used to perform feature fusion on the consultant's voice feature vector and the mental health information feature matrix to obtain a mental health consultation information matching feature vector;

[0012] The health information consultation feature classification result generation module is used to match feature vectors based on the mental health consultation information to generate personalized mental health voice responses.

[0013] In combination with the first aspect of the present application, in a health information intelligent consultation system based on interactive voice response in the first aspect of the present application, the health information consultation feature acquisition module includes: a consultant consultation information acquisition unit, used to feature encode the consultant's psychological consultation voice signal to obtain the consultant's voice feature vector; a psychological health information acquisition unit, used to feature encode the psychological health information data with multiple different emotional labels obtained from the database to obtain the psychological health information feature matrix.

[0014] According to a second aspect of the present application, a health information intelligent consultation method based on interactive voice response is provided, which includes:

[0015] Obtain the counselor's psychological consultation voice signal and obtain psychological health information data with multiple different emotional labels from the database;

[0016] Performing feature extraction and analysis on the counselor's psychological counseling voice signal and the mental health information data with multiple different emotion labels obtained from the database to obtain a counselor's voice feature vector and a mental health information feature matrix;

[0017] Performing feature fusion on the consultant's voice feature vector and the mental health information feature matrix to obtain a mental health consultation information matching feature vector;

[0018] Feature vectors are matched based on the mental health consultation information to generate a personalized mental health voice response.

[0019] Compared with the existing technology, the present application provides an intelligent health information consultation system and method based on interactive voice response. It first obtains the consultant's psychological consultation voice signal and the psychological health information data of multiple different emotional labels in the database as input data, and then uses deep learning technology to extract and analyze the features of these input data respectively, and then comprehensively analyzes these feature information to obtain classification results for representing the consultant's psychological emotional labels. In this way, the system can generate personalized psychological health voice responses based on the psychological emotional labels. In this way, it can effectively provide consultants with timely, convenient, and confidential psychological support, thereby resonating with the consultants and helping them relieve stress and dispel their inner distress. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The above and other purposes, features, and advantages of the present application will become more apparent through a more detailed description of the embodiments of the present application in conjunction with the accompanying drawings. The accompanying drawings are intended to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the drawings, the same reference numerals generally represent the same components or steps.

[0021] Figure 1 The figure shows a schematic block diagram of a health information intelligent consultation system based on interactive voice response according to an embodiment of the present application.

[0022] Figure 2 The figure illustrates a schematic block diagram of a health information consultation feature acquisition module in a health information intelligent consultation system based on interactive voice response according to an embodiment of the present application.

[0023] Figure 3 The figure illustrates a schematic block diagram of a consultant's consultation information acquisition unit in a health information consultation feature acquisition module in a health information intelligent consultation system based on interactive voice response according to an embodiment of the present application.

[0024] Figure 4 The figure illustrates a schematic block diagram of a mental health information acquisition unit in a health information consultation feature acquisition module in a health information intelligent consultation system based on interactive voice response according to an embodiment of the present application.

[0025] Figure 5 The figure illustrates a schematic block diagram of a health information consultation feature classification result generation module in a health information intelligent consultation system based on interactive voice response according to an embodiment of the present application.

[0026] Figure 6 The figure illustrates a flow chart of a health information intelligent consultation method based on interactive voice response according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described herein.

[0028] Exemplary Systems

[0029] Figure 1 FIG2 shows a schematic block diagram of a health information intelligent consultation system based on interactive voice response according to an embodiment of the present application. Figure 1 As shown, the health information intelligent consultation system 100 based on interactive voice response according to the embodiment of the present application includes: a health information consultation data acquisition module 110, which is used to obtain the consultant's psychological consultation voice signal and obtain multiple psychological health information data with different emotional labels from the database; a health information consultation feature acquisition module 120, which is used to extract and analyze the consultant's psychological consultation voice signal and the psychological health information data with multiple different emotional labels obtained from the database to obtain the consultant's voice feature vector and the psychological health information feature matrix; a health information consultation feature fusion module 130, which is used to fuse the consultant's voice feature vector and the psychological health information feature matrix to obtain a psychological health consultation information matching feature vector; a health information consultation feature classification result generation module 140, which is used to generate a personalized psychological health voice response based on the psychological health consultation information matching feature vector.

[0030] School-based mental health counseling is a vital service that provides mental health supports and services to students. It promotes student psychological health and well-being, identifies and addresses mental health issues, supports academic achievement, fosters social-emotional learning, responds to crises and trauma, promotes inclusion and belonging, and fosters school success. Investing in school-based mental health counseling is essential to ensuring all students succeed in school and in life.

[0031] However, the number of psychological counselors on campus is limited, and counseling typically requires appointments, which limits availability and may not be able to meet the needs of college students in a timely manner. Furthermore, some college students may feel stigmatized or embarrassed about seeking counseling and may forgo it. Therefore, a health information intelligent consultation system and method based on interactive voice response is desired. This system analyzes college students' psychological counseling issues through voice response, provides timely, convenient, and confidential psychological support, and thereby resonates with counselors, helping them relieve stress and resolve inner distress.

[0032] Deep learning is a subfield of artificial intelligence (AI) that mimics the neural network structure of the human brain to learn and understand complex data representations. The core of deep learning is to learn the characteristics and patterns of data through multi-layer neural networks, thereby automating the processing and prediction of various tasks.

[0033] Deep learning technology plays a key role in the interactive voice response-based health information intelligent consultation system of the present application. For example, deep learning technology helps the system analyze the emotions and feelings in the user's voice, thereby better understanding the user's inner state and needs. Therefore, in the specific embodiment of the present application, deep learning technology is used to extract and analyze features of the input data.

[0034] In this embodiment of the present application, the health information consultation data acquisition module 110 is used to acquire the consultant's psychological consultation voice signal and retrieve mental health information data with multiple different emotional tags from a database. It should be understood that the consultant's voice signal can provide important information about their mental health status, such as mood, tone, and speaking style. This information can be used to analyze the consultant's emotions. The database is used to acquire mental health information data with multiple different emotional tags, such as "How to cope with anxiety and stress?", "Common symptoms and diagnosis of depression," "How to build a positive self-image?", "Exploring methods of self-awareness and self-regulation," "The importance and skills of stress management," and "Learning strategies for relaxation and coping with stress." This enables the system to provide consultants with comprehensive information and resources and retrieve database information most relevant to the consultant's specific needs. Therefore, in order to more effectively provide consultants with mental health advice and support, the consultant's psychological consultation voice signal and the acquisition of mental health information data with multiple different emotional tags from the database are used.

[0035] Specifically, when obtaining the counselor's psychological consultation voice signal, voice recognition software or API can be used to convert the voice signal into text. When obtaining psychological health information data with multiple different emotional labels in the database, natural language processing (NLP) technology can be used.

[0036] In this embodiment of the present application, the health information consultation feature acquisition module 120 is configured to perform feature extraction and analysis on the counselor's psychological consultation voice signal and the psychological health information data with multiple different emotion labels obtained from the database to obtain a counselor's voice feature vector and a psychological health information feature matrix. It should be understood that after collecting these input data, feature extraction and analysis tasks are further performed on these input data.

[0037] Specifically, Figure 2 FIG2 is a schematic block diagram of a health information consultation feature acquisition module in a health information intelligent consultation system based on interactive voice response according to an embodiment of the present application. Figure 2 As shown, the health information consultation feature acquisition module 120 includes: a consultant consultation information acquisition unit 121, used to feature encode the consultant's psychological consultation voice signal to obtain the consultant's voice feature vector; a psychological health information acquisition unit 122, used to feature encode the psychological health information data with multiple different emotional labels obtained from the database to obtain the psychological health information feature matrix.

[0038] First, the psychological consultation voice signal of the consultant is subjected to feature extraction and analysis. Specifically, Figure 3 The figure shows a schematic block diagram of the consultant information acquisition unit in the health information consultation feature acquisition module of the health information intelligent consultation system based on interactive voice response according to an embodiment of the present application. Figure 3 As shown, the consultant consultation information acquisition unit 121 includes: a speech denoising subunit 121-1, which is used to pass the consultant's psychological consultation speech signal through a speech denoising module based on an autoencoder to obtain a consultant's speech signal after denoising; a consultant's speech feature acquisition subunit 121-2, which is used to pass the waveform of the consultant's speech signal after denoising through a consultant's speech feature acquisition module based on a three-dimensional convolutional neural network model to obtain a consultant's speech feature graph; and a speech feature graph pooling subunit 121-3, which is used to perform a pooling operation on the consultant's speech feature graph to obtain the consultant's speech feature vector.

[0039] It's understandable that speech signals are often contaminated by background noise, such as ambient noise, breathing, and keyboard tapping. This noise can interfere with speech recognition and analysis, leading to distortion or misunderstanding of the consultation content. Therefore, to improve the accuracy of speech signal feature extraction, a speech noise reduction module based on an autoencoder is first used to reduce the noise of the client's psychological consultation speech signal. The autoencoder learns speech-related features from the speech signal while removing noise and other unwanted components.

[0040] In a specific embodiment of the present application, the speech noise reduction subunit 121-1 is used to: use the feature encoder of the speech noise reduction module to extract speech features from the consultant's psychological consultation speech signal; and use the feature decoder of the speech noise reduction module to decode the speech features to obtain the consultant's speech signal after noise reduction.

[0041] Next, it should be understood that a speech signal is a time series data. That is, the noise-reduced client's speech signal includes not only frequency-domain features but also time-series features. Given that the 3D convolutional neural network model can simultaneously process the time, frequency, and channel dimensions of a speech signal, a client speech feature acquisition module based on the 3D convolutional neural network model is used to extract features from the noise-reduced client's speech signal. This allows for the extraction of more comprehensive and robust speech features.

[0042] In a specific embodiment of the present application, the consultant voice feature acquisition subunit 121-2 is used to: use each layer of the three-dimensional convolutional neural network model to perform the following operations on the input data in the forward pass of the layer: use the convolution units of each layer of the three-dimensional convolutional neural network model to perform convolution processing based on the three-dimensional convolution kernel on the input data to obtain a convolution feature map; use the pooling units of each layer of the three-dimensional convolutional neural network model to perform pooling processing based on the local feature matrix on the convolution feature map to obtain a pooled feature map; and use the activation units of each layer of the three-dimensional convolutional neural network model to perform nonlinear activation on the feature values ​​of each position in the pooled feature map to obtain an activation feature map; wherein, the output of the last layer of the three-dimensional convolutional neural network model is the consultant voice feature map.

[0043] Furthermore, given the large size of the feature graph, processing it may require a large amount of memory and more complex computations. Therefore, a pooling operation is further performed on the consultant's speech feature graph. This can reduce the size of the speech feature graph, thereby reducing the computational complexity and memory usage of the model.

[0044] Then, feature extraction and analysis are performed on the mental health information data with multiple different emotion tags obtained from the database. Specifically, Figure 4The figure shows a schematic block diagram of a mental health information acquisition unit in a health information consultation feature acquisition module in a health information intelligent consultation system based on interactive voice response according to an embodiment of the present application. Figure 4 As shown, the mental health information acquisition unit 122 includes: an embedding coding subunit 122-1, configured to embed and code the mental health information data with multiple different emotion labels obtained from the database to obtain multiple mental health information embedding vectors; and a mental health information extraction subunit 122-2, configured to arrange the multiple mental health information embedding vectors into a mental health information embedding matrix according to the emotion label dimension and then pass the matrix through a mental health information feature acquisition module based on a feature encoder to obtain the mental health information feature matrix. The feature encoder includes a two-dimensional convolution kernel.

[0045] It should be understood that the mental health information data with multiple different emotional labels obtained from the database is text data, and it is difficult for the machine to directly recognize this data. Therefore, the text data is first converted into a form that the machine can recognize. A common method is to use word embedding. Word embedding is a technology that maps words to vector space. Each word is represented by a vector that captures the semantic and syntactic information of the word. Further, considering that the Clip model can learn cross-modal representations, this means that it can map text and image data to a common vector space. Therefore, in order to convert the textual mental health information data into a machine-recognizable form, the Clip model is used to embed the mental health information data with multiple different emotional labels obtained from the database to obtain multiple mental health information embedding vectors.

[0046] In a specific embodiment of the present application, the embedding coding subunit 122-1 is used to: perform word segmentation processing on each mental health information data of the multiple different emotional labels in turn to obtain a word sequence; use the embedding layer of the sequence encoder of the Clip model to map each word in the word sequence into a word embedding vector to obtain a sequence of word embedding vectors; use the converter-based Bert model of the sequence encoder of the Clip model to perform global context semantic encoding on the sequence of embedding vectors to obtain multiple feature vectors; and cascade the multiple feature vectors to obtain each mental health information embedding vector in the multiple mental health information embedding vectors.

[0047] Next, consider that while embedding coding converts raw data into a computer-processable form, it fails to extract the most representative and useful features. Therefore, it's necessary to filter and refine the important information within the embedding vectors. Furthermore, it's understandable that, compared to directly extracting features from multiple vectors, preprocessing the data and arranging it into a matrix format may help improve the data's expressiveness and model performance. Specifically, arranging the embedding vectors into a matrix format based on the emotional label dimension can provide a structured representation for the data, allowing information under different emotional labels to be clearly organized and distinguished. Therefore, the multiple mental health information embedding vectors are first arranged into a mental health information embedding matrix based on the emotional label dimension. A mental health information feature acquisition module based on a feature encoder is then used to perform feature extraction on the mental health information embedding matrix. The feature encoder includes a two-dimensional convolution kernel. Two-dimensional convolution operations are translation-invariant, meaning they identify the same features regardless of their position within the matrix. This property helps the system better understand the emotional content and semantic information within the text, thereby improving the effectiveness and performance of the model.

[0048] In a specific embodiment of the present application, the mental health information extraction subunit 122-2 is used to: use the layers of the feature encoder to perform the following on the input data in the forward pass of the layer: use the convolution units of the layers of the feature encoder to perform convolution processing based on a two-dimensional convolution kernel on the input data to obtain a convolution feature map; use the pooling units of the layers of the feature encoder to perform pooling processing on the convolution feature map along the channel dimension to obtain a pooling feature map; and use the activation units of the layers of the feature encoder to perform nonlinear activation on the eigenvalues ​​of each position in the pooling feature map to obtain an activation feature map; wherein the output of the last layer of the feature encoder is the mental health information feature matrix.

[0049] In an embodiment of the present application, the health information consultation feature fusion module 130 is used to perform feature fusion on the consultant's voice feature vector and the mental health information feature matrix to obtain a mental health consultation information matching feature vector. It should be understood that voice features may contain information such as emotion and intonation, while mental health information features contain content such as emotion and themes in the text. Combining the two can comprehensively consider information from different modalities and provide a more comprehensive information basis. Therefore, the consultant's voice feature vector and the mental health information feature matrix are further subjected to feature fusion. This can better understand the consultant's needs and emotional state.

[0050] In the embodiment of the present application, the health information consultation feature classification result generation module 140 is used to generate a personalized mental health voice response based on the mental health consultation information matching feature vector. Specifically, Figure 5 FIG2 is a schematic block diagram of a health information consultation feature classification result generation module in a health information intelligent consultation system based on interactive voice response according to an embodiment of the present application. Figure 5 As shown, the health information consultation feature classification result generation module 140 includes: a health consultation information acquisition unit 141, which is used to pass the mental health consultation information matching feature vector through a mental health consultation matching feature extraction module based on a convolutional neural network model to obtain a mental health consultation classification feature vector; a spatial mapping compensation unit 142, which is used to perform core space mapping adjustment based on feature basis regression on the mental health consultation classification feature vector to obtain a compensated mental health consultation classification feature vector; a health consultation information classification unit 143, which is used to obtain a classification result from the compensated mental health consultation classification feature vector, and the classification result is used to represent the consultant's psychological emotional label; a health consultation information response unit 144, which is used to generate a personalized mental health voice response based on the psychological emotional label.

[0051] It should be understood that while fusing the consultant's voice feature vector and the mental health information feature matrix provides a more comprehensive feature representation, it is difficult to capture the most representative features. Therefore, to improve the system's ability to understand and process data, a mental health consultation matching feature extraction module based on a convolutional neural network model is also needed to perform deeper feature extraction on the mental health consultation information matching feature vector. Feature extraction can help the system better distinguish different types of data, improving the accuracy and efficiency of classification or recognition tasks.

[0052] In a specific embodiment of the present application, the health consultation information acquisition unit 141 is used to: use each layer of the convolutional neural network model to perform the following on the input data in the forward pass of the layer: convolution processing on the input data based on the convolution kernel to generate a convolution feature map; global mean pooling processing based on the feature matrix on the convolution feature map to generate a pooled feature map; and nonlinear activation of the eigenvalues ​​at each position in the pooled feature map to generate an activation feature map; wherein the output of the last layer of the convolutional neural network model is the mental health consultation classification feature vector, the input from the second layer to the last layer of the convolutional neural network model is the output of the previous layer, and the input of the convolutional neural network model is the mental health consultation information matching feature vector.

[0053] In particular, when performing speech denoising through an autoencoder, although noise can be effectively removed, there is also a chance that some subtle speech features that are valuable for sentiment analysis will be removed. When extracting speech feature maps, the three-dimensional convolutional neural network model can capture some feature changes in spatial and temporal dimensions, but it may not be able to fully cover some complex, nonlinear feature relationships. Subsequently, during the mean pooling operation, the average value calculation method adopted to simplify the feature representation may lead to the loss of feature details, especially those local feature differences that are crucial for distinguishing different emotional states. In the embedding coding and feature fusion stages, if the deep correlation between mental health information and speech features is not accurately captured, the final generated mental health counseling classification feature vector may fail to fully represent all the useful information in the original data, thereby affecting the accuracy of the classification results. Based on this, before passing the mental health counseling information matching feature vector through the classifier, the mental health counseling classification feature vector is first adjusted based on the core space mapping of the feature basis regression to obtain the compensated mental health counseling classification feature vector.

[0054] Specifically, the spatial mapping compensation unit 142 is used to: first, construct a pixel-level fine-grained correlation matrix of the mental health consultation classification feature vector, which is expressed as follows:

[0055]

[0056] Wherein, V represents the mental health consultation classification feature vector, v i and v j Respectively represent the eigenvalues ​​of the ith and jth positions of the mental health consultation classification feature vector, d(v i ,v j ) indicates calculating the Euclidean distance, D i,j Represents the eigenvalue of the (i, j) position of the pixel-level fine-grained correlation matrix.

[0057] That is, by establishing a pixel-level fine-grained association matrix of micro-units in the feature space, the system can dynamically identify the mapping relationship between key emotional triggers in the speech signal and the mental health knowledge base. For example, silent intervals in speech features may have a high-weight association with the "social avoidance" label in the database, while high-frequency word repetition may activate the emotional classification path of "anxiety tendency." This directional association mechanism not only strengthens the semantic coherence between the mental health consultation classification feature vectors. The generated pixel-level fine-grained association matrix significantly improves the system's accuracy in parsing ambiguous psychological demands.

[0058] Secondly, the core features of the pixel-level fine-grained correlation matrix are extracted based on the convolution layer to obtain the core spatial nonlinear excitation matrix of the mental health consultation classification, which is expressed as follows:

[0059] M=Conv(D)

[0060] Wherein, D represents the pixel-level fine-grained association matrix, Conv represents the convolutional layer, and M represents the core spatial nonlinear excitation matrix of the mental health counseling classification.

[0061] That is, through sliding learning across association domains of the convolution kernel, the core spatial nonlinear excitation pattern of mental health consultation classification is dynamically constructed. Specifically, at this stage, the convolution kernel does not focus on spatial continuity in traditional image processing, but instead performs feature abstraction on specific association patterns in the pixel-level fine-grained association matrix, adaptively identifying multi-hop association paths between labels in speech features, while suppressing irrelevant noise associations, thereby forming a core spatial nonlinear excitation matrix for mental health consultation classification that can characterize the essence of complex psychological states. This significantly enhances the system's penetration of implicit psychological demands, enabling the system to accurately locate key association clusters that affect psychological state classification while maintaining the original temporal integrity of the speech signal, providing decision support for generating personalized responses with clinical intervention value.

[0062] Then, the pixel-level fine-grained correlation matrix is ​​subjected to spectral feature decomposition to obtain a set of mental health consultation classification feature primitive encoding vectors, which is expressed as follows:

[0063]

[0064] Where T represents the transpose of the vector, Λ represents the diagonal matrix, λ1 and λ m Represent the first and mth eigenvalues ​​of the diagonal matrix respectively, U represents the set of primitive encoding vectors of mental health consultation classification features, x1, x2, x m Represent the first, second and mth mental health consultation classification feature primitive encoding vectors respectively.

[0065] That is, by constructing an orthogonal basis vector space, the chaotic pixel-level fine-grained correlation matrix is ​​decoupled into interpretable basic structural units of psychological states. Specifically, spectral decomposition establishes a structured representation capability for the dynamics of psychological state evolution. When multiple contradictory associations coexist in the consultant's voice, the set of psychological health consultation classification feature primitive encoding vectors can be separated into dominant primitives (such as achievement anxiety) and interfering primitives (such as short-term fatigue response) through orthogonal projection. This enables the system to dynamically adjust its response strategy based on the strength of the primitive combination, significantly improving the clinical rationality and individual adaptability of psychological health consultation interventions.

[0066] Next, each mental health consultation classification feature primitive encoding vector in the set of mental health consultation classification feature primitive encoding vectors is input into the feature prominence adjustment module based on the self-attention mechanism to obtain a set of enhanced mental health consultation classification feature primitive encoding vectors, which is expressed as follows:

[0067] Y=Transformer{[x1,x2,…,x m ]}=[y1,y2,…,y m ]

[0068] Among them, Transformer represents the self-attention mechanism, Y represents the set of enhanced mental health consultation classification feature primitive encoding vectors, y1, y2, y m Represent the first, second and mth enhanced mental health consultation classification feature primitive encoding vectors respectively.

[0069] Specifically, a dynamic cross-primitive weight transfer network is constructed, enabling the system to automatically strengthen key primitives related to the current psychological crisis level based on the real-time consultation context. Specifically, the self-attention mechanism accurately captures the semantic penetration of potential crisis signals through cross-primitive correlation analysis, avoiding the risk of misjudgment caused by equalizing primitives and achieving context-sensitive reconstruction of psychological state primitive combinations. The resulting enhanced mental health consultation classification feature primitive encoding vector, through the attention mechanism, highlights the synergistic effect of the "cognitive-behavioral dissociation" primitive and the "somatization symptom" primitive, thereby guiding the system to generate a complex response strategy that combines cognitive correction with physiological relaxation guidance.

[0070] Then, each enhanced mental health consultation classification feature primitive coding vector in the set of the enhanced mental health consultation classification feature primitive coding vectors is mapped to the mental health consultation classification core space nonlinear excitation matrix to obtain a set of mental health consultation classification feature primitive core mask coding vectors, which is expressed as follows:

[0071]

[0072] in, represents matrix multiplication, S represents the characteristic scale of the nonlinear activation matrix in the core space of mental health consultation classification, and y i represents the i-th enhanced mental health consultation classification feature primitive encoding vector, L represents the length of the enhanced mental health consultation classification feature primitive encoding vector, z i Represents the core mask encoding vector of the i-th mental health consultation classification feature primitive.

[0073] That is, a holographic fusion mechanism for cross-scale psychological features is constructed, and nonlinear projection is used to conduct adversarial interactions between global psychological primitives and local association patterns. Specifically, when contradictory expressions appear in voice interaction, the projection process forces the surface primitives to collide with the emotion masking association clusters in the nonlinear excitation matrix of the mental health counseling classification core space, stimulating the core mask encoding vector of the mental health counseling classification feature primitives that represents the intensity of psychological disguise. This enables the system to penetrate interference at the level of linguistic meaning and capture cognitive conflict signals at the user's subconscious level, forming a multi-dimensional penetrating representation capability of psychological states, significantly improving the targeting and timeliness of psychological counseling.

[0074] Finally, the set of the mental health consultation classification feature primitive core mask code vectors is fused to obtain the compensated mental health consultation classification feature vector, which is expressed as follows:

[0075] V′=Concat{z1,z2,…,z m}

[0076] Among them, Concat represents the cascade function, z1, z2, z m Represent the first, second and mth mental health consultation classification feature primitive core mask encoding vectors respectively, and V' represents the compensated mental health consultation classification feature vector.

[0077] In other words, the dynamic weights of cross-element interactions are reconstructed through a nonlinear fusion mechanism. For example, when both "descriptions of procrastination" and "expressions of perfectionism" are detected during voice interaction, the fused post-compensation mental health consultation classification feature vector can trigger a dual-track response mechanism involving cognitive reconstruction and meaning exploration through quantitative modeling of the nonlinear interactions between elements.

[0078] Specifically, in the technical solution of this application, considering that a classifier can help map abstract feature vectors to specific emotion labels, such as happiness, sadness, anxiety, etc., thereby more intuitively understanding the emotional state of the consultant. Therefore, the classifier is further used to perform feature classification on the post-compensation mental health consultation classification feature vector. This helps the system more accurately identify and understand the consultant's emotional needs, providing more targeted support and advice.

[0079] The system then generates personalized voice responses based on these psychological and emotional tags. This allows the system to provide customized support and advice based on the client's emotional state and needs. This personalized feedback better meets the client's needs, enhancing their sense of engagement and empathy, thereby improving the system's effectiveness and user experience.

[0080] It's worth noting that in addition to using a classifier to classify the post-compensation mental health consultation classification feature vectors, clustering algorithms can also be used to perform cluster analysis on the counselor's mental health information. Cluster analysis can group similar data points into categories without pre-defining the categories, which helps uncover hidden patterns and structures in the data. The following are the steps for implementing this method: 1. Data Preparation: Collect psychological counseling problem data from college students and extract corresponding feature vectors to represent each consultation problem. Ensure that the data format is uniform and the feature vectors have a certain dimensionality and information content. 2. Feature Vector Standardization: Standardize the extracted mental health consultation feature vectors to ensure that the scales of each feature are consistent, so that the clustering algorithm can accurately identify similarities between data points. 3. Clustering Algorithm Selection: Select an appropriate clustering algorithm, such as K-means clustering, hierarchical clustering, or density clustering. Choose the most appropriate algorithm based on the data volume and characteristics. 4. Cluster Analysis: Input the standardized feature vectors into the selected clustering algorithm to cluster the data points based on a similarity metric. The algorithm automatically identifies patterns between the data points and generates clustering results. 5. Determine the number of clusters: Before applying the clustering algorithm, you need to determine the number of clusters to generate. Methods such as the elbow rule and silhouette coefficient can be used to select the optimal number of clusters. 6. Generate psychological and emotional labels: Based on the clustering results, assign a psychological and emotional label to each cluster, representing the emotional characteristics of that type of consultation question. These labels can reflect the psychological states and emotional needs of different categories. 7. Personalized voice response: Based on the psychological and emotional label of each cluster, generate corresponding personalized mental health voice responses. Ensure that the voice response can provide appropriate support and advice for different psychological and emotional labels.

[0081] In summary, the intelligent health information consultation system based on interactive voice response according to the embodiment of the present application is explained, which first obtains the psychological consultation voice signal of the consultant and the psychological health information data of multiple different emotional labels in the database as input data, and then uses deep learning technology to extract and analyze the features of these input data respectively, and then comprehensively analyzes these feature information to obtain classification results for representing the consultant's psychological emotional labels. In this way, the system can generate personalized psychological health voice responses based on the psychological emotional labels. In this way, it can effectively provide consultants with timely, convenient, and confidential psychological support, thereby arousing the consultant's resonance and helping them relieve stress and dispel their inner distress.

[0082] As described above, the health information intelligent consulting system 100 based on interactive voice response according to the embodiment of the present application can be implemented in various wireless terminals, such as a server for health information intelligent consulting based on interactive voice response, etc. In one example, the health information intelligent consulting system 100 based on interactive voice response according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or hardware module. For example, the health information intelligent consulting system 100 based on interactive voice response can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the health information intelligent consulting system 100 based on interactive voice response can also be one of the many hardware modules of the wireless terminal.

[0083] Alternatively, in another example, the interactive voice response-based health information intelligent consulting system 100 and the wireless terminal may also be separate devices, and the interactive voice response-based health information intelligent consulting system 100 may be connected to the wireless terminal via a wired and / or wireless network and transmit interactive information in accordance with an agreed data format.

[0084] Exemplary Methods

[0085] Figure 6 The figure shows a flow chart of the health information intelligent consultation method based on interactive voice response according to an embodiment of the present application. Figure 6 As shown, the intelligent health information consultation method based on interactive voice response according to the embodiment of the present application includes: S1, obtaining the consultant's psychological consultation voice signal and obtaining multiple psychological health information data with different emotional labels from the database; S2, performing feature extraction and analysis on the consultant's psychological consultation voice signal and the psychological health information data with multiple different emotional labels obtained from the database to obtain the consultant's voice feature vector and the psychological health information feature matrix; S3, performing feature fusion on the consultant's voice feature vector and the psychological health information feature matrix to obtain a psychological health consultation information matching feature vector; S4, generating a personalized psychological health voice response based on the psychological health consultation information matching feature vector.

[0086] Here, those skilled in the art will appreciate that the specific functions and operations of each step in the above-mentioned health information intelligent consultation method based on interactive voice response have been referred to above. Figure 1 The description of the health information intelligent consultation system based on interactive voice response has been introduced in detail, and therefore, its repeated description will be omitted.

[0087] In the several embodiments provided herein, it should be understood that the disclosed systems, devices, apparatuses, and methods may be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the module division is merely a logical functional division, and actual implementation may employ other division methods.

[0088] In addition, the functional modules in the various embodiments of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0089] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0090] Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the claims are intended to be embraced therein. Any reference to a figure in a claim should not be construed as limiting the claim to which it relates.

[0091] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or systems recited in a system claim may also be implemented by a single unit or system through software or hardware. Second-order terms are used to indicate names and do not imply any particular order.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit of the technical solutions of the present invention.

Claims

1. An intelligent health information consultation system based on interactive voice response, characterized in that: include: The health information consultation data acquisition module is used to obtain the psychological consultation voice signal of the consultant and obtain psychological health information data with multiple different emotional labels from the database; A health information consultation feature acquisition module is used to extract and analyze the psychological consultation voice signal of the consultant and the psychological health information data with multiple different emotion labels obtained from the database to obtain the consultant's voice feature vector and the psychological health information feature matrix; A health information consultation feature fusion module is used to perform feature fusion on the consultant's voice feature vector and the mental health information feature matrix to obtain a mental health consultation information matching feature vector; The health information consultation feature classification result generation module is used to match feature vectors based on the mental health consultation information to generate personalized mental health voice responses.

2. The health information intelligent consulting system based on interactive voice response according to claim 1 is characterized in that: The health information consultation feature acquisition module includes: A consultant consultation information acquisition unit, configured to perform feature coding on the consultant's psychological consultation voice signal to obtain a consultant's voice feature vector; The mental health information acquisition unit is used to perform feature coding on the mental health information data with multiple different emotion tags obtained from the database to obtain the mental health information feature matrix.

3. The health information intelligent consulting system based on interactive voice response according to claim 2 is characterized in that: The consultant information acquisition unit includes: A speech denoising subunit, configured to pass the counsellor's psychological consultation speech signal through a speech denoising module based on an autoencoder to obtain a denoised counsellor's speech signal; A consultant voice feature acquisition subunit, configured to pass the waveform of the consultant voice signal after noise reduction through a consultant voice feature acquisition module based on a three-dimensional convolutional neural network model to obtain a consultant voice feature graph; The speech feature graph pooling subunit is used to perform a pooling operation on the consultant's speech feature graph to obtain the consultant's speech feature vector.

4. The health information intelligent consulting system based on interactive voice response according to claim 3 is characterized in that: The speech noise reduction subunit is used to: Extracting speech features from the counselor's psychological counseling speech signal using a feature encoder of the speech noise reduction module; as well as The feature decoder of the speech noise reduction module is used to decode the speech feature to obtain the consultant's speech signal after noise reduction.

5. The health information intelligent consulting system based on interactive voice response according to claim 4 is characterized in that: The mental health information acquisition unit includes: An embedding coding subunit, configured to embed and code the mental health information data with multiple different emotion tags obtained from the database to obtain multiple mental health information embedding vectors; The mental health information extraction subunit is used to arrange the multiple mental health information embedding vectors into a mental health information embedding matrix according to the emotional label dimension and then obtain the mental health information feature matrix through a mental health information feature acquisition module based on a feature encoder.

6. The health information intelligent consulting system based on interactive voice response according to claim 5 is characterized in that: The feature encoder includes a two-dimensional convolution kernel.

7. The health information intelligent consulting system based on interactive voice response according to claim 6 is characterized in that: The health information consultation feature classification result generation module includes: a health consultation information acquisition unit, configured to pass the mental health consultation information matching feature vector through a mental health consultation matching feature extraction module based on a convolutional neural network model to obtain a mental health consultation classification feature vector; a spatial mapping compensation unit, configured to perform a core spatial mapping adjustment based on feature basis regression on the mental health consultation classification feature vector to obtain a compensated mental health consultation classification feature vector; a health consultation information classification unit, configured to classify the compensated mental health consultation information into a feature vector to obtain a classification result, wherein the classification result is used to represent the psychological emotion label of the consultant; The health consultation information response unit is used to generate a personalized mental health voice response based on the psychological emotion label.

8. The intelligent health information consultation system based on interactive voice response according to claim 7 is characterized in that: The spatial mapping compensation unit is configured to: Constructing a pixel-level fine-grained correlation matrix of the mental health consultation classification feature vector; Performing core feature extraction on the pixel-level fine-grained correlation matrix based on a convolutional layer to obtain a core spatial nonlinear excitation matrix for mental health consultation classification; Performing spectral feature decomposition on the pixel-level fine-grained association matrix to obtain a set of mental health consultation classification feature primitive encoding vectors; Inputting each mental health consultation classification feature primitive encoding vector in the set of mental health consultation classification feature primitive encoding vectors into a feature prominence adjustment module based on a self-attention mechanism to obtain a set of enhanced mental health consultation classification feature primitive encoding vectors; Mapping each enhanced mental health consultation classification feature primitive coding vector in the set of enhanced mental health consultation classification feature primitive coding vectors to the mental health consultation classification core space nonlinear excitation matrix to obtain a set of mental health consultation classification feature primitive core mask coding vectors; The set of the mental health consultation classification feature primitive core mask coding vectors is fused to obtain the compensated mental health consultation classification feature vector.

9. A health information intelligent consultation method based on interactive voice response, characterized in that ,include: Obtain the counselor's psychological consultation voice signal and obtain psychological health information data with multiple different emotional labels from the database; Performing feature extraction and analysis on the counselor's psychological counseling voice signal and the mental health information data with multiple different emotion labels obtained from the database to obtain a counselor's voice feature vector and a mental health information feature matrix; Performing feature fusion on the consultant's voice feature vector and the mental health information feature matrix to obtain a mental health consultation information matching feature vector; Feature vectors are matched based on the mental health consultation information to generate a personalized mental health voice response.

Images