A music-based counselor matching method and system

Abstract

The application discloses a music-based psychological consultant matching method and system, which constructs a music-psychology mapping database, extracts music multi-level audio features to train a supervised learning model, and establishes a mapping relationship between the features and psychological dimensions. The music selection data of the visitor is obtained through an adaptive pairing comparison method, the presentation strategy is dynamically adjusted, the response delay is recorded, the psychological dimension vector of the selected music is weighted and aggregated, and a structured psychological portrait containing continuous value coordinates and clinical labels is decoded and generated. The psychological portrait and the consultant style vector are mixed and matched in a unified feature space, the cosine similarity and the Euclidean distance are combined to analyze the emotional resonance and the feature complementarity, the recommendation list is output after being filtered by the clinical knowledge weighting and the hard rule. The application can non-verbally capture the real-time and deep psychological state of the visitor through music preference, construct a multi-dimensional psychological vector, realize high-fitting-degree matching with the consultant, and improve the matching accuracy and the first consultation success rate.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a music-based method and system for matching psychological counselors. Background Technology

[0002] The psychological counselor matching system, as a key carrier for the integration of mental health services and artificial intelligence technology, is widely used in online psychological service platforms. Among related technologies, a matching system centered on problem tags, personality traits, and text features has been constructed through the collaborative operation of structured information matching, psychological scale analysis, and natural language processing (NLP). Specifically, this technical system covers the entire process from user data collection to counselor recommendation, including typical solutions such as CN202110242841.3 based on a decision tree model and CN202110170440.1 employing a secondary matching architecture.

[0003] However, existing methods for matching therapists with counselors often rely on structured information matching, scale tests, or text analysis without employing nonverbal techniques to dynamically capture the user's immediate psychological state. This can lead to issues such as information distortion, missing dimensions, or excessive interaction burden. Specifically, decision tree models based on historical counseling data (e.g., CN202110242841.3) simplify complex individuals into static information points, only addressing the surface logic of "what problems the client has," failing to address deeper needs such as "what kind of person the client is" and "what they are feeling at this moment." Secondary matching architectures (e.g., CN202110170440.1) mechanically sever the connection between counseling methods and therapist traits, ignoring the therapeutic effect of the therapist's personal traits on the delivery of the method. Existing technologies typically use tools like the MBTI scale for personality testing, but their results are easily influenced by the social approval effect and fail to reflect dynamic emotional changes. While NLP text analysis can extract emotional information, it carries a significant risk of bias in decoding nonverbal psychological states. These technological deficiencies result in a lack of accurate capture of the client's immediate and deep psychological characteristics during the matching process, which in turn affects the efficiency of establishing a therapeutic relationship and the effectiveness of counseling. Summary of the Invention

[0004] Addressing the core shortcomings of existing therapist matching technologies, such as limited matching dimensions, neglect of the client's immediate internal state, poor user experience, and incomplete and distorted information acquisition, this invention aims to provide a music-based therapist matching method. This method transcends the limitations of language and text, using non-verbal media (music) to deeply and dynamically understand the client's internal psychological state (emotions, cognition, energy), thereby achieving a multi-dimensional and highly compatible match between the client and the therapist.

[0005] Another objective of this invention is to propose a music-based psychological counselor matching system.

[0006] To achieve the above objectives, a first aspect of the present invention proposes a music-based method for matching psychological counselors, comprising: S1. Construct a database of music features and psychological mappings. By extracting multi-level audio feature vectors from music segments and labeling them with psychological continuous value labels, train a supervised learning model to establish the mapping relationship between audio feature vectors and psychological dimension vectors. S2. Adaptive pairwise comparison method is used to obtain music selection data of visitors. Based on their selection behavior in the preset psychological mapping space, the presentation strategy of music segments is dynamically adjusted, and the selection results and response delays are recorded as the basis for psychological state analysis. S3. Based on the music selection data decoding, generate an instant psychological profile vector of the visitor. By weighted aggregating the psychological dimension vectors of music segments and mapping them to predefined clinical meaning labels, a structured psychological profile containing continuous value coordinates and state labels is formed. S4. The psychological profile vector and the counselor's professional style vector are mixed and matched in a unified psychological feature space. The cosine similarity is used to evaluate the emotional resonance and the Euclidean distance is used to analyze the feature complementarity. The matching results are generated by a weighted function driven by clinical knowledge and a recommendation list is output by applying hard rule filtering conditions.

[0007] In one embodiment of the present invention, the construction of the music feature and psychological mapping database further includes: S11 uses the Librosa audio processing library to extract low-level features of music clips, including Mel frequency cepstral coefficients (MFCCs), chromaticity features (Chroma), spectral centroids, and zero crossover rate. S12 calculates the musical structure complexity of high-level features using a supervised learning algorithm, employing a fractal dimension formula. Quantifying the complexity of musical fragments, among which The number of repeating structural units. is the scale factor.

[0008] In one embodiment of the present invention, the step of obtaining visitor music selection data using the adaptive paired comparison method further includes: S21, dynamically adjust the difference range of the next set of music segments based on the user's selection of continuity. If the user continues to select high-arousal music, then... Options for generating more subtle differences within the interval, among which The preset adaptive step size; S22 will delay the response. The exponential decay function is used as the basis for calculating the selection weights. The selection results are weighted, where This is the preset attenuation coefficient.

[0009] In one embodiment of the present invention, the step of generating an instant psychological profile vector of a visitor based on music selection data decoding further includes: S31, calculate the user psychological profile vector through weighted vector aggregation. ,in For the first The next selection corresponds to the musical psychological dimension vector. The weights are calculated based on the response latency; S32, continuous value vector When mapping to clinically meaningful labels, a piecewise function is used. ,in Valence value, This is the Arousal value.

[0010] In one embodiment of the present invention, the step of performing a mixed matching calculation of the psychological profile vector and the counselor's professional style vector further includes: S41, Calculate cosine similarity Assess emotional resonance and calculate Euclidean distance. Analyze feature complementarity, among which The number of dimensions of the psychological dimension vector; S42, When applying hard rule filtering, prioritize retaining those that meet the criteria. The consultants, among them This is the preset threshold.

[0011] In one embodiment of the present invention, it further includes: S5, collect feedback data from visitors regarding the matching results and update the weight parameters of the matching model, specifically including: S51, calculate the reward value of the feedback data using a reinforcement learning algorithm. ,in and This is the preset reward coefficient; S52, based on reward value Adjusting the clinical knowledge weights in the weighted matching function Optimization using gradient descent method To minimize matching error ,in To match the number of samples.

[0012] To achieve the above objectives, a second aspect of the present invention provides a music-based psychological counselor matching system, comprising: The module for constructing a music feature and psychological mapping database is used to extract multi-level audio feature vectors from music segments and label them with psychological continuous value labels, and to train a supervised learning model to establish a mapping relationship between audio feature vectors and psychological dimension vectors. The adaptive pairing comparison data acquisition module is used to acquire music selection data of visitors using the adaptive pairing comparison method. It dynamically adjusts the presentation strategy of music segments based on the selection behavior in the preset psychological mapping space and records the selection results and response delays as the basis for psychological state analysis. The psychological profile decoding and generation module is used to decode and generate an instant psychological profile vector of the visitor based on the music selection data. It aggregates the psychological dimension vectors of music fragments by weighting and mapping them to predefined clinical meaning labels to form a structured psychological profile containing continuous value coordinates and state labels. The hybrid matching calculation module is used to perform hybrid matching calculations on the psychological profile vector and the counselor's professional style vector in a unified psychological feature space. It combines cosine similarity to assess emotional resonance and Euclidean distance to analyze feature complementarity. The module generates matching results through a weighted function driven by clinical knowledge and applies hard rule filtering conditions to output a recommendation list.

[0013] The music-based psychological counselor matching method and system of this invention captures the user's immediate and deep psychological state through music preferences, constructs a multi-dimensional psychological vector to achieve a high degree of compatibility with the counselor, and improves the accuracy of matching and the success rate of the first consultation.

[0014] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.

[0015] The beneficial effects of this invention are as follows: Breaking through the matching dimension, it achieves a deep fit from "treating the symptoms" to "treating the person": transcending the superficial logic of existing technology of "problem to profession", it realizes a deep "person to person" matching based on intrinsic characteristics, fundamentally improving the possibility of establishing an effective treatment relationship.

[0016] Excellent user experience, significantly lowering the threshold for seeking help: transforming the information collection process into an easy and intuitive musical exploration, especially suitable for visitors who are not good at speaking or are emotionally exhausted, solving the problem of secondary consumption caused by scales and text descriptions.

[0017] Capturing real-time dynamic psychological states for precise targeted matching: Unlike the static personality measured by traditional scales, this invention can capture the dynamic psychological state of users seeking help "at this moment," making matching more timely and targeted, and effectively improving the success rate of the first consultation.

[0018] The information is more authentic and objective: Music preferences, as a more fundamental projective response, are harder to fake than self-reports, resulting in more authentic and objective data that ensures the accuracy of the matching results. Attached Figure Description

[0019] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 A flowchart illustrating a music-based psychological counselor matching method provided in an embodiment of the present invention; Figure 2 This is a structural diagram of a music-based psychological counselor matching system provided in an embodiment of the present invention. Detailed Implementation

[0020] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0021] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0022] The following describes a music-based psychological counselor matching method and system according to an embodiment of the present invention, with reference to the accompanying drawings.

[0023] This embodiment provides a music-based method for matching therapists with therapists. For example... Figure 1 As shown, it includes: S1. Construct a database of music features and psychological mappings. By extracting multi-level audio feature vectors from music segments and labeling them with psychological continuous value labels, a supervised learning model is trained to establish the mapping relationship between audio feature vectors and psychological dimension vectors.

[0024] Specifically, constructing a database of music features and psychological mappings is a fundamental step in achieving deep matching between clients and therapists. The core of this step lies in establishing a quantitative mapping relationship between music fragments and psychological dimensions through audio feature engineering and psychological labeling, thereby providing high-precision data support for subsequent psychological profiling and matching algorithms.

[0025] First, multi-level audio feature extraction is performed on the music clip (suggested duration 15-30 seconds). Low-level features include Mel-frequency cepstral coefficients (MFCCs), chroma features, spectral centroid, and zero-crossing rate, which effectively characterize the timbre, harmony, and rhythmic properties of the music. Mid-level features cover beat (BPM), mode (Major / Minor), etc., used to describe the emotional tone of the music. High-level features, such as structural complexity and genre label, are obtained through pre-trained models or expert annotations to capture the abstract stylistic features of the music.

[0026] Subsequently, the psychological labels were labeled using a two-dimensional Valence-Arousal emotion model, where Valence represents the positive or negative nature of the emotion (ranging from -1 to +1), and Arousal represents the level of arousal (also ranging from -1 to +1). The labeling process can be completed by psychology experts or participants with standardized training to ensure the reliability and validity of the ratings. In some implementations, this can be extended to dimensions such as cognitive load and tension to enrich the psychological representation.

[0027] Audio feature vectors typically have 60-120 dimensions, depending on the combination of selected features. Psychological labeling should follow a standardized process, such as mapping a Likert 7-point scale to a continuous value range, to improve the stability of model training. Supervised learning models can include XGBoost, Support Vector Regression (SVR), or Deep Neural Networks (DNNs), with the audio feature vector as input. The output is a psychological rating vector. The training objective is to minimize the prediction error, such as the mean squared error (MSE).

[0028] In practical applications, this step can be deployed in the backend data processing system of a psychological counseling platform to batch process music libraries and generate a structured database. Each record contains a music ID, an audio feature vector, and a corresponding psychological mapping vector, providing a data foundation for subsequent user interaction and profile generation.

[0029] This step, by constructing a high-precision music-psychology mapping model, achieves an interpretable transformation from non-verbal music preferences to psychological states, providing the system with an objective, dynamic, and non-invasive means of capturing psychological states, and significantly improving the depth and accuracy of the matching system.

[0030] Furthermore, S1 includes: S11 uses the Librosa audio processing library to extract low-level features of music clips, including Mel frequency cepstral coefficients (MFCCs), chroma features, spectral centroids, and zero crossover rate.

[0031] Specifically, in some implementations, the Librosa audio processing library is used to extract the low-level features of music clips. This step uses time-frequency analysis of the audio signal to extract feature vectors that reflect the acoustic characteristics of the music, such as timbre, harmony, and rhythm, providing high-dimensional, quantifiable input data for subsequent psychological state mapping.

[0032] Specifically, the Librosa library processes audio signals based on Short-Time Fourier Transform (STFT) and Mel Filterbank. Mel frequency cepstral coefficients (MFCCs) capture the nonlinear perceptual characteristics of timbre by calculating the cepstral features of the audio signal at the Mel scale. Typically, 12-13 MFCC coefficients are extracted, covering the main dimensions of timbre variation, and their calculation formula is as follows: ,in For spectral amplitude, For the first The response of a Mel filter.

[0033] Chroma is used to characterize harmonic information in music by mapping the spectrum onto a 12-chromatic scale to extract the harmonic contours. The spectral centroid reflects the center of the frequency distribution of audio energy; a higher value indicates a brighter timbre. The calculation formula is: ,in For spectral amplitude, For frequency indexing.

[0034] Zero-Crossing Rate (ZCR) measures the frequency at which an audio signal crosses zeros in the time domain, reflecting the "noise" or "dynamics" of the audio. Its calculation formula is as follows: ,in The first audio signal One sampling point.

[0035] In practical applications, this step typically processes 15-30 second music clips, with a sampling rate set to 44.1kHz or 22.05kHz, a window length of 2048 points, and a hop length of 512 points to ensure temporal continuity of features and computational efficiency. The extracted feature vectors will serve as input to the music-psychology mapping model to train the regression model, thereby achieving a quantitative association between audio features and psychological states.

[0036] The technical value of this step lies in providing the system with a non-verbal, non-invasive means of perceiving psychological states through a standardized and reproducible audio feature extraction method. Its high-dimensional feature space can capture the complexity of music, providing a solid data foundation for subsequent user psychological profiling and counselor matching.

[0037] S12 calculates the musical structure complexity of high-level features using a supervised learning algorithm, employing a fractal dimension formula. Quantifying the complexity of musical fragments, among which The number of repeating structural units. is the scale factor.

[0038] Specifically, in some implementations, this invention calculates the musical structural complexity of high-level features using a supervised learning algorithm and employs a fractal dimension formula. Quantifying the complexity of musical fragments, among which Indicates the number of repeating structural units. This represents the scale factor. This step is a crucial part of the music feature and psychological mapping database construction module, used to extract high-level features from music fragments, thereby providing structured and interpretable feature inputs for subsequent psychological state modeling.

[0039] Musical structural complexity is assessed by analyzing the repetition and variability of musical segments over time to evaluate the non-linear characteristics of their organizational structure. Specifically, the system first segments the musical segments, extracts repeating units (such as phrases, motives, rhythmic patterns, etc.) at different time scales, and counts the total number of repeating units at each scale. Scale factor This indicates the scaling ratio of the time window, for example, dividing a music segment into... "Time" indicates that structural analysis is performed at twice the time granularity. This is achieved by calculating different scales. By substituting the value into the fractal dimension formula, the system can obtain a continuous complexity index. It is used to describe the degree of non-linear organization in the structure of a musical segment.

[0040] The statistics need to be based on automatic music structure recognition algorithms, such as music structure segmentation models based on Dynamic Time Warping (DTW) or deep learning (such as Music Transformer). Scale factor The value is usually taken as , , etc., to cover structural changes from short-term to long-term. Fractal dimensions The value range is usually in arrive Between these values, the higher the value, the more complex the musical structure, and the more likely it is to trigger cognitive load or emotional fluctuations.

[0041] This step primarily aims to establish a quantitative correlation between music and psychological states. By using the structural complexity of music as one of the input features, the system can more comprehensively characterize the potential impact of musical fragments on emotional arousal, cognitive tension, and other aspects, thereby improving the accuracy and richness of psychological profiles. For example, in highly arousing music, if its structural complexity... A higher level of intensity may be more suitable as a medium for emotional regulation, rather than simply relying on rhythm or mode.

[0042] Furthermore, the technical advantage of this step lies in its ability to capture the deep temporal organization of music by introducing the non-linear characteristic of musical structural complexity, thereby enhancing the system's ability to recognize users' emotional states. Compared to traditional feature extraction methods that rely solely on beat, pitch, or timbre, this step significantly improves the expressiveness and generalization ability of the mental mapping model, providing more accurate feature support for achieving deep "person-to-person" matching.

[0043] S2 uses an adaptive paired comparison method to obtain music selection data from visitors. Based on their selection behavior in a preset psychological mapping space, the presentation strategy of music segments is dynamically adjusted, and the selection results and response delays are recorded as the basis for psychological state analysis.

[0044] Specifically, in some implementations, the "adaptive pairing comparison method" of the present invention is the core mechanism for acquiring music selection data of visitors. Its technical principle is based on the interactive logic of Computerized Adaptive Testing (CAT) and combines the multidimensional features of music fragments in the psychological mapping space to achieve efficient and dynamic capture of the user's real-time psychological state.

[0045] The specific operation is as follows: Within a preset psychological mapping space, the system selects two music segments that differ significantly in a certain dimension (e.g., one with high arousal and the other with low arousal) based on the Psychological_Mapping_Vector of the music segments (such as dimensions like Valence and Arousal). These segments are then presented to the user as 15-30 second audio clips. Each time the user is presented, they must choose "which music is closer to your current mood." The system dynamically adjusts the presentation strategy of subsequent music segments based on the user's selection, gradually approximating the user's true position within the psychological space.

[0046] The initial selection of music clips is typically based on uniform distribution or Latin hypercube sampling to ensure coverage of multiple regions of the mental space. In each pairing comparison, the difference between music clips in the target dimension is usually controlled between 0.2 and 0.4 (e.g., Valence from -0.8 to -0.4) to ensure that users can perceive the difference without making it overly complex. Response latency (Response_Latency_ms) is recorded as an auxiliary indicator of user selection behavior, used to assess the decisiveness of their choice; the shorter the latency, the more closely the music clip matches the user's current mental state.

[0047] In practical applications, this step can be deployed on mobile or web platforms. When a user enters the psychological counseling matching process, they first complete a music selection test. Based on their selection sequence and response time, the system generates a psychological profile vector P_user in real time. This method is particularly suitable for clients with limited verbal expression or suppressed emotions because it collects psychological states through non-verbal media, resulting in higher authenticity and non-invasiveness.

[0048] This step, through quantitative analysis of music selection behavior, achieves multidimensional dynamic modeling of users' emotions, cognition, and energy states, providing a high-precision and highly interpretable data foundation for subsequent matching with counselor profiles. Compared to traditional scales or text analysis methods, it is more efficient in information acquisition and less susceptible to the influence of social desirability effects, significantly improving the accuracy of the matching system and the user experience.

[0049] Furthermore, S2 includes: S21, dynamically adjust the difference range of the next set of music segments based on the user's selection of continuity. If the user continues to select high-arousal music, then... Options for generating more subtle differences within the interval, among which This is the preset adaptive step size.

[0050] Specifically, in some implementations, the difference range of the next set of music segments is dynamically adjusted based on the continuity of the user's selections. The core technical principle of this step is based on the logic of Computerized Adaptive Testing (CAT). By analyzing the user's selection behavior in real time, the difference range of subsequent music segments in the mental mapping space is dynamically adjusted, thereby more efficiently and accurately locating the user's current mental state coordinates.

[0051] The specific implementation is as follows: Each time a music segment pair is presented, the system adjusts the psychological dimension difference range of the next music segment based on the previous selection result. For example, if a user selects a highly arousing music segment in several consecutive interactions, the system will assume the user is currently in a highly aroused state and, accordingly, filter music segments from the music feature and psychological mapping database in subsequent selections. Musical fragments within the interval, among which This represents the center value of the current high wakefulness level. The preset adaptive step size is usually set to... This is to ensure that the range of estimates of psychological states is gradually narrowed while keeping the differences perceptible.

[0052] Furthermore, this adaptive step size The response latency can be dynamically adjusted based on the user's selection. A shorter response latency indicates greater user engagement with the current option, allowing the system to reduce latency accordingly. This improves the resolution and matching accuracy of subsequent selections. Conversely, if the user is hesitant, the resolution can be increased appropriately. This expands the scope of exploration and avoids premature convergence to local estimates.

[0053] In practical applications, this step is typically deployed in the backend logic of the user interface, combined with the frontend music playback and selection feedback mechanisms, to achieve non-verbal, non-intrusive, and real-time capture of the user's psychological state. This mechanism is particularly suitable for visitors who are emotionally suppressed, have difficulty expressing themselves verbally, or experience significant fluctuations in their psychological state, effectively lowering the information collection threshold and improving the adaptability and accuracy of the matching system.

[0054] By dynamically adjusting the presentation strategy of music segments, the efficiency and accuracy of psychological state estimation are significantly improved. Compared with the traditional fixed-difference matching method, the adaptive mechanism of this invention can achieve high-resolution modeling of user psychological profiles with fewer interactions, thereby providing a more reliable data foundation for subsequent counselor matching.

[0055] S22 will delay the response. The exponential decay function is used as the basis for calculating the selection weights. The selection results are weighted, where This is the preset attenuation coefficient.

[0056] Specifically, in the "Real-time Psychological Profile Generation Module for Visitors" of this invention, the step "delaying the response" The exponential decay function is used as the basis for calculating the selection weights. Weighting the selection results is a key step in achieving dynamic modeling of users' psychological states. This step quantifies the user's reaction speed during the music pairing comparison process, further extracting the stability of their psychological state and the decisiveness of their emotional expression, thereby improving the accuracy and representativeness of the psychological profile.

[0057] This step is based on an adaptive paired comparison mechanism, presenting the user with two musical fragments that differ in their mental mapping space each time. The user indirectly expresses their emotional state by choosing the fragment that better reflects their current mood. The system records the response delay for each selection. This refers to the time interval (in milliseconds) between the presentation of a music clip and the user's selection. Subsequently, an exponential decay function is employed. Each selection result is weighted, where The preset attenuation coefficient typically ranges from [value range missing]. This is used to adjust the degree to which latency affects weighting. A shorter response latency indicates a stronger user acceptance of the option, and therefore assigns it a higher weight. Conversely, the weight decreases.

[0058] The data collection must meet the following standards: the sampling frequency is not less than 100Hz, and the timestamp accuracy is in the millisecond range to ensure the reliability of the delayed data. The settings need to be optimized based on user behavior experimental data, such as through A / B testing or cross-validation, to determine the optimal value, balancing response speed and sensitivity to psychological state. Furthermore, the weighted psychological mapping vector... Generated through vector aggregation algorithms (such as weighted average), its dimensions include valence, arousal, etc., and it is usually a 2- to 5-dimensional continuous value vector.

[0059] This step is applicable to scenarios such as initial user registration on a psychological counseling platform, pre-consultation preparation, or rapid assessment of emotional state. Through non-verbal interaction via music selection, the system can efficiently acquire the user's emotional state, especially suitable for individuals with limited verbal expression abilities or suppressed emotions. This method avoids the verbosity and subjectivity of traditional scales while enhancing the naturalness of data collection and user engagement.

[0060] This step enhances the dynamism of the mental profile and its ability to identify individual differences by introducing response delay as a weighting factor. Compared to the "hard matching" method in traditional approaches that rely solely on selection results, this invention... The weighted mechanism enables "soft decoding" of users' psychological states, thereby improving the sensitivity and personalization of the matching system and providing more accurate input basis for subsequent counselor matching.

[0061] S3. Based on the music selection data decoding, generate an instant psychological profile vector of the visitor. By weighted aggregating the psychological dimension vectors of music fragments and mapping them to predefined clinical meaning labels, a structured psychological profile containing continuous value coordinates and state labels is formed.

[0062] Specifically, in some implementations, the real-time psychological profile generation module for visitors constructs a structured psychological profile containing continuous value coordinates and state labels by weighted aggregation of psychological dimension vectors from music clips and mapping them to predefined clinical meaning labels. This step is the core component of the system's "non-verbal psychological state capture," and its technical implementation is based on the mapping relationship between the psychological label space and the music feature vectors.

[0063] Specifically, the system first extracts the Psychological_Mapping_Vector corresponding to the selected music segment based on the user's choice between two music segments in the interactive interface. This vector typically contains two core psychological dimensions: Valence and Arousal, ranging from [-1, +1] to [0, 1], and can also be extended to other dimensions such as cognitive load and tension. Each selection action records the response latency (Response_Latency_ms) as a quantitative indicator of the user's decisiveness in the selection, which is used in subsequent weighted calculations.

[0064] In the weighted vector aggregation stage, the system uses response latency as a weighting factor to construct a weighted average model. Assume users collectively select... There are music fragments, and the psychological vector corresponding to each fragment is . The response latency is The final psychological profile vector It can be represented as:

[0065] This formula ensures that more decisive choices by users have a higher weight in the final profile, thereby improving the accuracy of psychological state capture.

[0066] Furthermore, the system will The system maps values ​​to a predefined clinically meaningful label space. For example, when Valence < -0.5 and Arousal > 0.7, the system can label it as "Agitated-Anxious" and retain the original vector for subsequent matching calculations. This labeling system can be extended and customized based on clinical psychology standards (such as the DSM-5 emotion classification).

[0067] In practical applications, this step can be deployed in the front-end interactive process of a psychological counseling platform. When users enter the counseling matching interface, they can complete the non-invasive collection of their psychological state through a quick music selection. Its technical value lies in achieving high-precision modeling of the user's immediate, deep, and dynamic psychological state through music, a non-verbal medium. This provides a scientific basis for subsequent multi-dimensional matching with counselors, significantly improving the compatibility of the match and the effectiveness of the counseling.

[0068] Furthermore, S3 includes: S31, calculate the user psychological profile vector through weighted vector aggregation. ,in For the first The next selection corresponds to the musical psychological dimension vector. The weights are calculated based on the response latency.

[0069] Specifically, in the "Visitor Real-Time Psychological Profile Generation Module" of this invention, weighted vector aggregation is the core algorithmic step that transforms the user's behavioral sequence during the music selection process into a multi-dimensional vector representation of their psychological state. This step is achieved through a formula... Implementation, in which Indicates the user's position in the first month. The psychological dimension vector corresponding to the music segment selected in the next selection usually includes emotional dimensions such as valence and arousal, as well as psychological characteristics such as cognitive load and tension that may be extended. The weighting coefficient is calculated based on the response latency of each user's selection.

[0070] At the technical implementation level, this step first relies on the user selection behavior data collected in Module 2, including the music ID selected each time, the selection result (A / B), and the response latency (in milliseconds). Based on the user's selection, the system extracts the corresponding psychological dimension vector of the music from the "music-psychological mapping database" constructed in Module 1. Subsequently, through response delay Calculate weights Its calculation method can be based on an exponential decay function or a linear normalization method. For example, it can be achieved using... ,in To adjust the coefficient that controls the impact of response latency on weights, it is typically set to... This is to ensure that choices with shorter delays have higher weight.

[0071] Dimensions and To maintain consistency, it is typically 2-5 dimensions, covering psychological dimensions such as Valence, Arousal, Cognitive Load, and Tension. Weighting coefficients. The value range is [0,1], and the sum of all weights is normalized to 1 to ensure the interpretability and stability of the aggregation results. Furthermore, the system can set a threshold. and , representing the minimum and maximum response latency, respectively. Responses exceeding these ranges will be filtered or subject to special processing to prevent outliers from interfering with profile generation.

[0072] This step is widely used in the user psychological state collection process of psychological counseling platforms. When users select music pairings in the interactive interface, the system records their selection behavior and response latency in real time, and after completing a round of testing, generates their current psychological profile vector through weighted vector aggregation. This vector is not only used for subsequent matching calculations with the counselor's profile, but also serves as an instant assessment tool for the user's emotional state, helping counselors understand the client's psychological coordinates before the first meeting.

[0073] This step, by introducing response delay as a weighting factor, effectively distinguishes the intensity of users' selection tendencies across different psychological dimensions, thereby improving the dynamism and accuracy of psychological profiling. Compared to traditional scales or text analysis methods, this technical solution can capture users' true psychological state at the moment of seeking help in a more natural and non-invasive way, providing a reliable data foundation for achieving deep "person-to-person" matching.

[0074] S32, continuous value vector When mapping to clinically meaningful labels, a piecewise function is used. ,in Valence value, This is the Arousal value.

[0075] Specifically, in the visitor instant psychological profile generation module, continuous value vectors are used. Mapping to clinically meaningful emotion labels is one of the key steps in achieving personalized matching. This step employs a piecewise function. ,in Indicates valence value. The values ​​represent arousal, and both are normalized to the range of [-1, +1], reflecting the positive or negative tendency of emotions and the degree of activation, respectively.

[0076] In some implementations, this piecewise function is based on the widely used two-dimensional valence-arousal model in psychology, which structurally categorizes the emotional states exhibited by users during music selection. Specifically, when a user exhibits a moderate to low negative value on the valence dimension ( ), and at the same time, it is at a high level in terms of arousal ( When the valence is close to neutral or slightly positive, the system labels the emotional state as "Agitated-Anxious," indicating that the user is currently in a state of anxiety, irritability, or emotional agitation. Conversely, when the valence is close to neutral or slightly positive... ), and low arousal level ( When a user is in a calm, introspective, or emotionally stable state, the system marks them as "Calm-Reflective," indicating that the user is in a calm, introspective, or emotionally stable state.

[0077] This mapping process relies on music selection data collected in the preceding module using an adaptive pairwise comparison method. Combined with a weighted vector aggregation algorithm, it transforms the user's selection behavior into continuous emotional coordinates. By setting a clear threshold range, the system can convert abstract continuous value vectors into clinically interpretable discrete labels, thus providing an actionable description of emotional state for subsequent therapist matching. In practical applications, this step can be deployed in the front-end or back-end services of a psychological counseling platform as a real-time decoding mechanism for user emotional states, particularly suitable for clients with limited emotional expression or communication difficulties. Its technical value lies in improving the interpretability and clinical applicability of the matching system, providing therapists with more context-aware recommendations.

[0078] S4. The psychological profile vector and the counselor's professional style vector are mixed and matched in a unified psychological feature space. The cosine similarity is used to evaluate the emotional resonance and the Euclidean distance is used to analyze the feature complementarity. The matching results are generated by a weighted function driven by clinical knowledge and a recommendation list is output by applying hard rule filtering conditions.

[0079] Specifically, in some implementations, hybrid matching calculation is a core step in this invention to achieve deep matching between clients and psychological counselors. This step involves using psychological profile vectors... Vector of Consultant Practice Style Mapped to a unified psychological feature space, the system combines cosine similarity and Euclidean distance to comprehensively evaluate the system from two dimensions: emotional resonance and feature complementarity. A final matching score is generated through a weighted function driven by clinical knowledge, and then a recommendation list is output by applying hard rule filtering conditions.

[0080] First, the system generates a multi-dimensional psychological profile vector based on the user's music selection. Vector of Consultant's Practice Style Alignment is performed to the same psychological feature space. This space typically includes dimensions such as valence, arousal, cognitive load, and emotional tension, with each dimension ranging from [-1, +1] to [0, 1], depending on the annotation criteria. Emotional resonance is assessed using cosine similarity. The closer the value is to 1, the higher the compatibility of the two parties in terms of emotional characteristics.

[0081] Furthermore, feature complementarity is expressed through Euclidean distance. To assess this, the smaller the distance, the smaller the differences in psychological characteristics between the two parties, and the more likely they are to complement each other in certain dimensions. For example, a client with high arousal may be more suitable for a therapist with high emotional regulation skills, even if the client has lower arousal.

[0082] The system uses a weighting function defined by clinical psychology experts. ,in and These are adjustable parameters, representing the weights of emotional resonance and feature complementarity, respectively. In actual deployment, It is usually set to 0.6. The value is set at 0.4 to balance the matching requirements of emotional connection and style differences.

[0083] This step is widely used in the intelligent recommendation system of psychological counseling platforms, especially suitable for users who are matching for their first consultation, experiencing significant emotional fluctuations, or having limited verbal expression. After receiving the user's music selection data, the system generates a psychological profile in real time and performs batch matching with the professional style vectors in the counselor database, outputting a sorted recommendation list.

[0084] This hybrid matching algorithm significantly improves the accuracy and personalization of matching. By capturing users' immediate psychological state through nonverbal music preferences and combining this with the multidimensional style characteristics of counselors, the system can shift from "problem-to-professional" to "person-to-person," enhancing the efficiency and stability of establishing counseling relationships and thus improving overall counseling outcomes.

[0085] Furthermore, S4 includes: S41, Calculate cosine similarity Assess emotional resonance and calculate Euclidean distance. Analyze feature complementarity, among which Let be the number of dimensions of the psychological dimension vector.

[0086] Specifically, in the intelligent matching engine module of this invention, cosine similarity is calculated. Euclidean distance This is the core algorithmic step for achieving deep matching between clients and therapists. This step quantifies the differences in direction and distance between psychological profile vectors, assessing emotional resonance and feature complementarity respectively, thereby providing a scientific and interpretable basis for the recommendation system's decision-making.

[0087] Cosine similarity is used to measure a user's psychological vector. Psychological Vectors of Counselors Directional consistency in the psychological dimension space. Its calculation is based on the ratio of the vector dot product to the magnitude, with a value ranging from [-1, 1]. A value closer to 1 indicates a more similar emotional state, while a value closer to -1 indicates a more opposing emotional state. This indicator is particularly suitable for assessing the emotional resonance ability between the client and the therapist, such as whether the therapist possesses a corresponding empathetic style in a high-arousal, low-valence emotional state.

[0088] Euclidean distance measures the absolute difference between two vectors in psychological space; a smaller value indicates closer similarities in characteristics, while a larger value indicates stronger complementarity. In a psychological counseling setting, this complementarity may manifest as differences in energy levels, cognitive styles, or coping mechanisms, thus providing more appropriate guidance and support to the client. For example, a client in a high-arousal state may need a counselor with extensive experience in low-arousal states to guide their emotional regulation.

[0089] The number of dimensions in a psychological dimension vector typically includes valence, arousal, cognitive load, and emotional tension, etc. Specific dimensions can be expanded based on clinical psychology models. Each component in the vector... and All values ​​are continuous, typically ranging from [-1, +1] to [0, 1], and are output by the music mapping model.

[0090] This step, embedded in the recommendation system's matching engine, is used to rank the counselors after initial screening. The system first calculates the cosine similarity and Euclidean distance between all candidate counselors and clients in the psychological vector space, and then combines this with a weighted function based on clinical knowledge to generate the final matching score. This process supports real-time calculations and can output recommendation results within seconds of the user making a music selection.

[0091] This step shifts the focus from "problem-to-professional" to "person-to-person," allowing the matching process to consider both emotional resonance and complementary characteristics. By introducing nonverbal music preferences as a projection of psychological state, the system can more realistically and dynamically reflect the user's current psychological characteristics, thereby improving the success rate of the first consultation and user satisfaction.

[0092] S42, When applying hard rule filtering, prioritize retaining those that meet the criteria. The consultants, among them This is the preset threshold.

[0093] Specifically, in the "Consultant Multidimensional Feature Database and Intelligent Matching Engine Module" of this invention, when applying hard rule filtering, priority is given to retaining those that meet the criteria. Finding a suitable counselor is a crucial step in achieving precise matching. This step is based on the widely used two-dimensional emotion model (Valence-Arousal model) in psychology, where Valence represents the positive or negative valence of emotion (a continuous value from -1 to +1), used to measure the degree of pleasure or enjoyment of a musical passage or an individual's emotional state. This is achieved by setting a threshold... The system quantifies the difference in emotional valence between counselors and clients during the matching process, thereby selecting candidate counselors who have a high degree of compatibility in emotional tone.

[0094] This step first extracts the effectiveness value from the visitor's psychological profile generated in Module 3. And extract the corresponding effectiveness value from the consultant's professional style vector in Module 4. The system then calculates the absolute difference between the two. and compare it with a preset threshold. Compare them. If the difference is less than... If the counselor is deemed to have a high potential to match the client in terms of emotional valence, they will be retained for further matching calculation.

[0095] threshold The value of needs to be adjusted according to the actual application scenario. In some implementations, Can be set to This indicates that a difference of no more than 20% in valence between the therapist and the client is permissible. This parameter needs to be set in conjunction with clinical psychology knowledge to ensure that a sufficient number of candidate therapists are retained without the match quality being negatively impacted by excessive differences. Furthermore, The threshold can be dynamically adjusted based on the stability of the visitor's current psychological state. For example, in cases of high anxiety or significant mood swings, the threshold can be appropriately relaxed to increase the inclusiveness of the match.

[0096] This step is primarily applied in the recommendation system of a psychological counseling platform. After a user completes a music selection test, the system uses their psychological profile vector to filter candidates from the counselor database who meet the hard rules, serving as the basis for subsequent calculations of cosine similarity and Euclidean distance. This process is executed during the pre-filtering stage of the recommendation engine, significantly reducing computational complexity and improving recommendation efficiency.

[0097] Also includes: S5, collect feedback data from visitors regarding the matching results and update the weight parameters of the matching model, specifically including: Specifically, the step of "collecting feedback data from visitors on the matching results and updating the weight parameters of the matching model" in this invention is a key step in achieving closed-loop optimization and adaptive evolution of the model. In some implementations, this step, by constructing a feedback data acquisition mechanism and a model parameter update algorithm, enables the matching model to dynamically adjust according to the actual matching effect, thereby continuously improving matching accuracy and user experience.

[0098] After each match is completed, the system sends a structured feedback questionnaire to the visitor. The questionnaire includes dimensions such as satisfaction rating of the recommended consultant (e.g., on a 1-5 scale), emotional resonance, and compatibility of communication styles. Simultaneously, the system records user behavioral data during the matching process, such as response latency (Response_Latency_ms) for music selection, choice consistency (Choice_Consistency), and click-through rate (Click_Through_Rate) for the matching results. These data collectively constitute the feedback vector F = [S, C, L, R], where... Indicates matching satisfaction. This indicates consistency in choice. This indicates a response delay. This indicates the recommended click-through rate.

[0099] The collection of feedback data must meet certain quality standards. For example, response latency. The threshold can be set to To ensure effective feedback and exclude invalid or erroneous data, match satisfaction levels. The weighting coefficient can be set to Emotional resonance The weighting coefficients are Response delay The normalization process adopts The format reflects the decisiveness of the user's choice. Model updates employ gradient descent or policy gradient algorithms in reinforcement learning, adjusting the weight parameters in the matching model. Perform iterative optimization, with the objective function being: ,in This is the feedback value predicted by the model.

[0100] This step can be deployed on mobile or web-based mental health counseling platforms. After a user completes the matching process, the system automatically pushes a feedback questionnaire and performs real-time data processing and model updates in the background. This mechanism is particularly suitable for high-frequency, multi-round matching scenarios, such as users with regular mental health counseling needs or those experiencing significant mood swings, and can significantly improve the system's personalized recommendation capabilities.

[0101] This step, by incorporating user feedback data, enables the matching model to continuously learn and adapt to users' true preferences and psychological states, thereby improving matching accuracy and user engagement. Furthermore, this mechanism enhances the system's self-optimization capabilities, allowing it to maintain high matching quality over long-term operation and resolving the performance degradation problem caused by outdated data in traditional matching systems.

[0102] S51, calculate the reward value of the feedback data using a reinforcement learning algorithm. ,in and This is the preset reward coefficient.

[0103] Specifically, in the closed-loop feedback and model self-optimization module of this invention, calculating reward values ​​for feedback data using reinforcement learning algorithms is a key technical step in achieving continuous system optimization and improved personalized matching capabilities. The core of this step lies in constructing a quantifiable reward function. ,in and These are preset reward coefficients, used to adjust the relative weights of matching success rate and user satisfaction in the reward function.

[0104] The reward function is calculated based on the actual feedback data from the client after completing the psychological counseling matching. This indicates the proportion of visitors who complete their first consultation with the recommending counselor and confirm the establishment of a continuing counseling relationship. It is usually represented by a continuous value between 0 and 1, such as whether the user schedules their next consultation within 72 hours. The satisfaction rating is obtained by users based on their satisfaction with the matching results. The rating mechanism can use a five-point scale (1-5 points) and normalize it to a range of 0 to 1 for easy and uniform calculation.

[0105] Reinforcement learning algorithms employ policy gradient methods or Q-learning frameworks in this process to assign reward values... As environmental feedback, it is used to update the parameters of the matching model. Specifically, the system uses the user's mental profile vector... Vector of Consultant's Practice Style Input matching model, output matching score And calculate the reward value based on the actual matching results. This allows for adjustments to the model's matching weight function, thereby improving the accuracy and relevance of subsequent recommendations.

[0106] In practical applications, this step can be deployed in the system's backend matching engine. By collecting user feedback data in real time and combining it with reinforcement learning algorithms, the model can be updated online. Its technical value lies in the fact that, through a dynamic feedback mechanism, the system can continuously adapt to changes in user behavior, improve the robustness and personalization capabilities of the matching model, and thus significantly increase the success rate of psychological counseling matching and user satisfaction.

[0107] S52, based on reward value Adjusting the clinical knowledge weights in the weighted matching function Optimization using gradient descent method To minimize matching error ,in To match the number of samples.

[0108] Specifically, in some implementations, the present invention employs a reward-value-based approach. Gradient descent method for clinical knowledge weights in weighted matching functions Optimize to minimize matching error ,in This indicates the number of matched samples. This step is a key optimization step in the intelligent matching engine module, used to improve the accuracy of the system's modeling of the matching relationship between clients and counselors in a multi-dimensional psychological space.

[0109] Matching function It is a weighted system based on clinical knowledge. The weighted function for regulation takes the visitor's psychological profile vector as input. Vector of Consultant Practice Style The output is the predicted matching score. This function typically combines cosine similarity and Euclidean distance, in the form of... ,in Representing vector similarity, Represents vector distance. This is achieved by introducing... The system can dynamically balance similarity and complementarity, thus better meeting the matching needs of "resonance" and "guidance" in clinical psychology.

[0110] The value of is typically limited to the interval [0, 1], representing the influence of clinical knowledge on the matching results. The initial value can be set to . This indicates that similarity and complementarity have equal weight. Gradient descent updates the algorithm iteratively. Its learning rate Normal settings arrive To ensure convergence speed and stability, the loss function is used. Using the mean squared error (MSE) method can effectively measure the deviation between the predicted matching score and the actual matching feedback.

[0111] This optimization step applies to the stage where the system provides personalized recommendations after constructing the client's psychological profile and the counselor's professional style vector. By collecting user selection behavior for music clips in real time, the system can quickly generate a psychological profile and match it with the professional style vector in the counselor database. During this process... The dynamic adjustment makes the matching results more in line with the guiding principles of clinical psychology, thereby improving the accuracy and interpretability of the recommendations.

[0112] Furthermore, the technical effect of this step is to significantly improve the adaptability and accuracy of the matching model. By minimizing matching errors, the system can more accurately identify the compatibility between clients and therapists, especially in key dimensions such as emotional resonance and energy complementarity. In addition, this method supports a closed-loop feedback mechanism, which can be combined with reinforcement learning strategies to enable the system to continuously optimize, thereby achieving personalized, dynamic, and highly compatible therapist matching.

[0113] The counselor matching method based on music-psychological mapping in this invention collects client feedback data on the matching results and dynamically updates the weight parameters of the matching model using a reinforcement learning algorithm, especially optimizing the weight of clinical knowledge based on reward values. This further enhances the adaptability and accuracy of the matching system, thereby achieving more personalized counselor matching that better meets individual psychological needs, and improving user experience and counseling effectiveness.

[0114] To implement the methods of the above embodiments, the present invention also provides a music-based psychological counselor matching system 10, such as... Figure 2 As shown, it includes: The music feature and psychological mapping database construction module 100 is used to extract multi-level audio feature vectors of music segments and label them with psychological continuous value labels, and train a supervised learning model to establish a mapping relationship between audio feature vectors and psychological dimension vectors. The adaptive pairing comparison data acquisition module 200 is used to acquire the music selection data of visitors using the adaptive pairing comparison method, dynamically adjust the presentation strategy of music segments according to their selection behavior in the preset psychological mapping space, and record the selection results and response delay as the basis for psychological state analysis. The psychological profile decoding and generation module 300 is used to decode and generate an instant psychological profile vector of the visitor based on the music selection data. It aggregates the psychological dimension vectors of music fragments by weighting and mapping them to predefined clinical meaning labels to form a structured psychological profile containing continuous value coordinates and state labels. The hybrid matching calculation module 400 is used to perform hybrid matching calculations on the psychological profile vector and the counselor's professional style vector in a unified psychological feature space. It combines cosine similarity to evaluate emotional resonance and Euclidean distance to analyze feature complementarity. The matching results are generated through a weighted function driven by clinical knowledge, and a recommendation list is output by applying hard rule filtering conditions.

[0115] Furthermore, the music feature and psychological mapping database construction module is also used for: The Librosa audio processing library was used to extract low-level features of the music clips, including Mel frequency cepstral coefficients (MFCCs), chroma features (Chroma), spectral centroids, and zero crossover rate. The music structure complexity of high-level features is calculated using a supervised learning algorithm, employing a fractal dimension formula. Quantifying the complexity of musical fragments, among which The number of repeating structural units. is the scale factor.

[0116] Furthermore, the adaptive pairing comparison data acquisition module is also used for: The difference range of the next set of music segments is dynamically adjusted based on the user's selection of continuity. If the user continues to select high-arousal music, then... Options for generating more subtle differences within the interval, among which The preset adaptive step size; Response delay The exponential decay function is used as the basis for calculating the selection weights. The selection results are weighted, where This is the preset attenuation coefficient.

[0117] Furthermore, the psychological profile decoding and generation module is also used for: Calculate user psychological profile vectors using weighted vector aggregation. ,in For the first The next selection corresponds to the musical psychological dimension vector. The weights are calculated based on the response latency; continuous value vector When mapping to clinically meaningful labels, a piecewise function is used. ,in Valence value, This is the Arousal value.

[0118] The music-based psychological counselor matching system of this invention collects feedback data from clients on the matching results and uses a reinforcement learning algorithm to dynamically update the weight parameters of the matching model. In particular, it optimizes the weight of clinical knowledge based on reward values, which further improves the adaptability and accuracy of the matching system, thereby achieving more personalized counselor matching that meets individual psychological needs and enhancing user experience and counseling effectiveness.

[0119] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0120] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0121] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

Claims

1. A music-based method for matching therapists, characterized in that, include: S1. Construct a database of music features and psychological mappings. By extracting multi-level audio feature vectors from music segments and labeling them with psychological continuous value labels, train a supervised learning model to establish the mapping relationship between audio feature vectors and psychological dimension vectors. S2. Adaptive pairwise comparison method is used to obtain music selection data of visitors. Based on their selection behavior in the preset psychological mapping space, the presentation strategy of music segments is dynamically adjusted, and the selection results and response delays are recorded as the basis for psychological state analysis. S3. Based on the music selection data decoding, generate an instant psychological profile vector of the visitor. By weighted aggregating the psychological dimension vectors of music segments and mapping them to predefined clinical meaning labels, a structured psychological profile containing continuous value coordinates and state labels is formed. S4. The psychological profile vector and the counselor's professional style vector are mixed and matched in a unified psychological feature space. The cosine similarity is used to evaluate the emotional resonance and the Euclidean distance is used to analyze the feature complementarity. The matching results are generated by a weighted function driven by clinical knowledge and a recommendation list is output by applying hard rule filtering conditions.

2. The method as described in claim 1, characterized in that, The construction of the music feature and psychological mapping database also includes: S11 uses the Librosa audio processing library to extract low-level features of music clips, including Mel frequency cepstral coefficients (MFCCs), chromaticity features (Chroma), spectral centroids, and zero crossover rate. S12 calculates the musical structure complexity of high-level features using a supervised learning algorithm, employing a fractal dimension formula. Quantifying the complexity of musical fragments, among which The number of repeating structural units. is the scale factor.

3. The method as described in claim 1, characterized in that, The method of obtaining visitor music selection data using adaptive pairwise comparison also includes: S21, dynamically adjust the difference range of the next set of music segments based on the user's selection of continuity. If the user continues to select high-arousal music, then... Options for generating more subtle differences within the interval, among which The preset adaptive step size; S22 will delay the response. The exponential decay function is used as the basis for calculating the selection weights. The selection results are weighted, where This is the preset attenuation coefficient.

4. The method as described in claim 1, characterized in that, The method for generating an instant psychological profile vector of a visitor based on music selection data decoding also includes: S31, calculate the user psychological profile vector through weighted vector aggregation. ,in For the first The next selection corresponds to the musical psychological dimension vector. The weights are calculated based on the response latency; S32, continuous value vector When mapping to clinically meaningful labels, a piecewise function is used. ,in Valence value, This is the Arousal value.

5. The method as described in claim 1, characterized in that, The method of performing a mixed matching calculation between the psychological profile vector and the counselor's professional style vector also includes: S41, Calculate cosine similarity Assess emotional resonance and calculate Euclidean distance. Analyze feature complementarity, among which The number of dimensions of the psychological dimension vector; S42, When applying hard rule filtering, prioritize retaining those that meet the criteria. The consultants, among them This is the preset threshold.

6. The method as described in claim 1, characterized in that, Also includes: S5, collect feedback data from visitors regarding the matching results and update the weight parameters of the matching model, specifically including: S51, calculate the reward value of the feedback data using a reinforcement learning algorithm. ,in and This is the preset reward coefficient; S52, based on reward value Adjusting the clinical knowledge weights in the weighted matching function Optimization using gradient descent method To minimize matching error ,in To match the number of samples.

7. A music-based psychotherapist matching system, characterized in that, include: The module for constructing a music feature and psychological mapping database is used to extract multi-level audio feature vectors from music segments and label them with psychological continuous value labels, and to train a supervised learning model to establish a mapping relationship between audio feature vectors and psychological dimension vectors. The adaptive pairing comparison data acquisition module is used to acquire music selection data of visitors using the adaptive pairing comparison method. It dynamically adjusts the presentation strategy of music segments based on the selection behavior in the preset psychological mapping space and records the selection results and response delays as the basis for psychological state analysis. The psychological profile decoding and generation module is used to decode and generate an instant psychological profile vector of the visitor based on the music selection data. It aggregates the psychological dimension vectors of music fragments by weighting and mapping them to predefined clinical meaning labels to form a structured psychological profile containing continuous value coordinates and state labels. The hybrid matching calculation module is used to perform hybrid matching calculations on the psychological profile vector and the counselor's professional style vector in a unified psychological feature space. It combines cosine similarity to assess emotional resonance and Euclidean distance to analyze feature complementarity. The module generates matching results through a weighted function driven by clinical knowledge and applies hard rule filtering conditions to output a recommendation list.

8. The system as described in claim 7, characterized in that, The music feature and psychological mapping database construction module is also used for: The Librosa audio processing library was used to extract low-level features of the music clips, including Mel frequency cepstral coefficients (MFCCs), chroma features (Chroma), spectral centroids, and zero crossover rate. The music structure complexity of high-level features is calculated using a supervised learning algorithm, employing a fractal dimension formula. Quantifying the complexity of musical fragments, among which The number of repeating structural units. is the scale factor.

9. The system as described in claim 7, characterized in that, The adaptive pairing comparison data acquisition module is also used for: The difference range of the next set of music segments is dynamically adjusted based on the user's selection of continuity. If the user continues to select high-arousal music, then... Options for generating more subtle differences within the interval, among which The preset adaptive step size; Response delay The exponential decay function is used as the basis for calculating the selection weights. The selection results are weighted, where This is the preset attenuation coefficient.

10. The system as described in claim 7, characterized in that, The psychological profile decoding and generation module is also used for: Calculate user psychological profile vectors using weighted vector aggregation. ,in For the first The next selection corresponds to the musical psychological dimension vector. The weights are calculated based on the response latency; continuous value vector When mapping to clinically meaningful labels, a piecewise function is used. ,in Valence value, This is the Arousal value.

Images