Emotion regulation music therapy system based on biofeedback and method thereof

Abstract

The invention provides an emotion regulation music therapy system based on biological feedback and a method thereof, and aims to realize personalized emotion regulation and psychological rehabilitation by monitoring the emotion state of a user in real time and automatically regulating music rhythm and melody. The system collects physiological and behavioral data of a user by integrating a heart rate monitor, a skin electric response sensor, a facial expression recognition camera and other biosensors. By utilizing an advanced emotion analysis engine, the system can accurately identify the emotion change of the user and recommend appropriate music from a music database containing various styles and emotion tags. In addition, the system further has a personalized learning algorithm, music recommendation can be continuously optimized according to preference and feedback of the user, and customized music experience is provided. A user can interact with the system through an intuitive user interface, adjust settings and provide feedback. The system also pays attention to privacy protection of the user data, and ensures that all the collected data accord with privacy protection laws and regulations.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent emotion recognition and music therapy, and specifically to an intelligent emotion music therapy instrument and a dynamic music adjustment system thereof which integrate biosensor technology, facial expression recognition technology, machine learning algorithm and music therapy theory. Background Art

[0002] In the field of emotion recognition, with the continuous innovation and breakthroughs in biosensor technology and facial expression recognition technology, it has become possible to accurately identify the user's emotional state in real time. Biosensor technology provides a rich physiological signal basis for emotion recognition by non-invasively collecting a variety of users' physiological data, such as heart rate, skin conductance, respiratory rate, and body temperature. At the same time, facial expression recognition technology uses advanced image processing and machine learning algorithms to accurately capture and analyze changes in users' facial expressions, and then identify a variety of users' emotional states, such as happiness, sadness, anger, surprise, etc. The rapid development of these technologies has provided new and more accurate means for emotional management and mental health.

[0003] In the field of music therapy, music, as a universal language, has been widely recognized for its healing and relaxing effects. Since ancient times, music has been used in various healing practices to help people relieve stress, improve mood and promote psychological recovery. In recent years, with the in-depth research of music therapy theory and the continuous accumulation of clinical practice, the mechanism of music in emotion regulation and psychological rehabilitation has been more clearly revealed. However, traditional music therapy methods often lack personalization and real-time performance, and it is difficult to provide customized music therapy solutions based on the specific needs and real-time emotional states of different users.

[0004] Therefore, it is of great research significance and application value to develop an intelligent emotional music therapy instrument that can recognize the user's emotional state in real time and automatically adjust the music by combining emotion recognition technology and music therapy theory. This intelligent therapy instrument will integrate biosensor technology, facial expression recognition technology, machine learning algorithms, and music therapy theory to form a complete system that works together. It will be able to automatically select and adjust the rhythm, melody, volume, and timbre of the music according to the user's real-time emotional state to provide a personalized music therapy program. This intelligent emotional music therapy instrument is designed to help users achieve emotional regulation and psychological rehabilitation, improve the quality of life and mental health, and has broad application prospects and social value. Summary of the invention

[0005] The present invention proposes an innovative biofeedback-based emotion regulation music therapy system, which carefully designs a series of modules that work together to achieve real-time monitoring and regulation of the user's emotional state. The core of the system is the biofeedback monitoring module, which integrates a heart rate monitor, a galvanic skin response sensor, a facial expression recognition camera, a voice analysis module, a body temperature sensor, a motion sensor, and an eye tracker. These high-precision sensors work together to capture the user's physiological and behavioral data in real time. The collected data is then transmitted to the emotion analysis engine, which uses a deep learning algorithm to perform a comprehensive analysis of the data to identify the user's current emotional state. Once the emotional state is identified, the music recommendation system selects music that matches the user's emotions from the music database and automatically generates a personalized music playlist, which aims to regulate the user's emotions through the healing power of music. Users can provide feedback on the music played through the user interaction interface, evaluate its impact on emotions or express personal preferences. The personalized learning algorithm will analyze these feedbacks in real time, continuously learn and adapt to optimize future music recommendations. Throughout the entire process, privacy protection measures are implemented to ensure the security and privacy of all user data, from encrypted data storage to user identity authentication to explicit consent before data use, to fully protect user privacy. Through this series of sophisticated and coherent steps, the system not only provides immediate emotional support, but also provides increasingly accurate personalized music therapy over time through continuous learning and user interaction, thus building a scientific and humane emotional regulation system.

[0006] The biofeedback monitoring module not only contains a series of high-precision sensors, but also integrates an advanced synchronization and integration algorithm, which is specially designed to process data streams from different sensors such as heart rate monitors, galvanic skin response sensors, facial expression recognition cameras, voice analysis modules, body temperature sensors, motion sensors and eye trackers. This algorithm can pre-process, align and fuse these heterogeneous data sources in real time to ensure the synchronization and consistency of data, thereby providing high-quality and high-precision data input for the emotion analysis engine. In this way, the system can more accurately capture the subtle changes in the user's emotional state, because the integrated data can more comprehensively reflect the user's physiological and psychological state. Specifically, the algorithm first filters and de-noises the raw data collected by each sensor to eliminate environmental interference and measurement errors; then, the algorithm uses timestamp alignment technology to ensure that the data from different sensors are matched in time; then, the algorithm uses advanced data fusion techniques, such as Kalman filtering or neural networks, to fuse multi-source data into a comprehensive emotional state indicator; finally, this comprehensive indicator is passed to the emotion analysis engine for further emotion recognition and analysis. Through the application of this synchronization and integration algorithm, the system is able to significantly improve the accuracy of emotional state recognition and provide users with more precise emotion regulation music therapy.

[0007] Preferably, the emotion analysis engine analyzes the user's emotional state by using advanced machine learning algorithms. The engine first receives synchronized and integrated data from the biofeedback monitoring module, including heart rate (measurement range 1-200 beats / minute, accuracy ±1 beat / minute), skin galvanic response (measurement range 0-100μS, accuracy ±1%), facial expression (resolution 1920x1080, frame rate 30fps), voice characteristics (frequency response 20Hz-20kHz, signal-to-noise ratio>60dB), body temperature (measurement range 32-42℃, accuracy ±0.1℃), movement (measurement range 0-10m / s 2 , accuracy ±0.01m / s 2) and eye movement (tracking speed 200Hz, accuracy 0.5°). The emotion analysis engine then uses pre-trained machine learning models such as random forests, support vector machines (SVMs), deep neural networks (DNNs), or convolutional neural networks (CNNs) to perform feature extraction and pattern recognition on these multi-dimensional data. The emotion analysis engine may use a deep neural network with multiple hidden layers, each containing hundreds of neurons, to automatically learn to extract complex emotion-related features from physiological signals. Alternatively, it may use a support vector machine model that uses a radial basis function (RBF) kernel during the training phase to optimize classification performance by adjusting the values ​​of C (regularization parameter) and γ (kernel function parameter) to identify the user's emotional state, such as happiness, sadness, anger, stress, and relaxation, with an accuracy of more than 90%. During the analysis process, the emotion analysis engine also considers individual differences and environmental factors to improve the accuracy of emotion recognition through personalized model training and adjustment. For example, the engine may adjust the model's thresholds and weights based on the range of changes in the user's heart rate variability (HRV) and skin galvanic response to more accurately reflect the individual's emotional changes. The engine is able to recognize multiple emotional states and is able to self-optimize through continuous learning and user feedback. For example, if a user provides feedback about improved mood through the user interface after listening to a piece of music, the engine will record this data and consider this information in future analysis to further calibrate the emotion recognition model. The emotion analysis engine also has real-time processing capabilities, which can quickly respond to immediate changes in user emotions, ensuring that the music recommendation system can provide music that matches the user's current emotional state in a timely manner. This fast and accurate emotion analysis capability enables the system to provide users with timely and personalized emotion regulation support, thereby effectively promoting emotion management and mental health.

[0008] The music database is a key component of the system's personalized music therapy. It contains at least 1,000 carefully selected pieces of music, covering a variety of styles from classical to modern, from Eastern to Western, to ensure that the music tastes and mood regulation needs of different users can be met. Each piece of music has been carefully annotated by a professional team with emotional tags such as "relaxing", "motivating", "nostalgic", "cheerful" and "contemplative" to ensure that the music selection can match the user's emotional state. To further enhance the effectiveness of music therapy, the music database also contains a variety of rhythmic, melodic and harmonious musical elements that are scientifically combined to produce specific emotional responses. During the construction of the music database, strict screening and classification standards were adopted, and each piece of music was evaluated by music therapy experts to ensure its effectiveness in emotion regulation. In addition, the database is regularly updated to incorporate new music works and user feedback to keep music therapy modern and relevant. Through the output of the emotion analysis engine, the system is able to quickly retrieve and recommend music suitable for the user's current emotional state from this rich music library, thereby providing a personalized music experience. For example, when the emotion analysis engine recognizes that the user is under stress, the system will select music labeled "relaxing" and "meditative" from the database, such as slow-tempo classical music or natural sounds, to help the user relieve tension. Through this refined music selection and recommendation mechanism, the system can provide users with more accurate and effective emotional regulation support.

[0009] The music recommendation system is a key link in realizing personalized music therapy. It has advanced algorithmic capabilities and can dynamically adjust the music playlist according to the user's real-time emotional state and historical preference data. The system first receives the user's emotional state information through the emotion analysis engine, which includes but is not limited to emotion categories (such as happiness, sadness, anger, etc.), emotion intensity (such as mild, moderate, severe), and emotion change trends (such as emotion rise or fall). At the same time, the system also takes into account the user's historical preferences, which may include the type of music selected by the user in the past, the most frequent listening time, and the user's feedback on specific music (such as ratings, number of plays, skipped or repeated tracks). The music recommendation system inputs this data into a complex recommendation algorithm, which may be based on collaborative filtering, content recommendation, or hybrid recommendation strategies to ensure that the recommended music meets both the user's current emotional needs and his long-term music taste. For example, if the user tends to listen to relaxing music at night, the system will give priority to recommending related tracks that the user has listened to frequently and rated highly in the past at night when the user's emotional analysis shows that he needs to relax. In addition, the system will also perform adaptive learning based on real-time feedback from users. If a user gives positive feedback on a recommended song, the system will increase the frequency of recommending similar songs in the future. On the contrary, if a user skips a song, the system will reduce or stop recommending music of similar styles. The music recommendation system also has real-time processing capabilities, which can quickly respond to instant changes in user emotions and ensure that the adjustment of the music playlist is synchronized with the user's emotional state. This dynamic adjustment mechanism enables the system to provide users with a timely and personalized music experience, thereby effectively promoting emotional management and mental health. Through this refined music recommendation strategy, the system can not only meet the user's current emotional regulation needs, but also continuously optimize the recommendation quality over time and enhance the user's music experience.

[0010] The personalized learning algorithm uses deep learning technology to continuously optimize the music recommendation algorithm. The algorithm uses neural networks, especially recurrent neural networks (RNN) and long short-term memory networks (LSTM), to process and analyze time series data of user interactions with music. Through these networks, the system is able to learn the changing patterns of users' music preferences over time, as well as the complex relationship between users' emotional states and music choices. During the training process, the algorithm receives feedback from the user interface, such as users' ratings of recommended music, playback time, number of repeated playbacks, and skipping behaviors. These feedback data are converted into numerical features, such as ratings (1-5 points), playback time (seconds), number of repeated playbacks (times), and skip probability (0-1), and are input into the neural network as training signals. For example, if a user rates a track 4.5 points and plays 90% of the entire track, these data will be used to strengthen the system's understanding of the user's preference for similar music. The deep learning model also regularly updates itself, adjusting the network weights through the backpropagation algorithm to minimize prediction errors and improve recommendation accuracy. The system also implements regularization techniques, such as L1 and L2 regularization, to prevent model overfitting and ensure the generalization ability of the algorithm. In addition, the algorithm also includes a validation mechanism to regularly evaluate the performance of the model by dividing user data into training and test sets, ensuring that the recommendation system continues to improve as user behavior changes.

[0011] Preferably, the user interaction interface is a bridge for direct communication between the system and the user, which includes an intuitive graphical user interface (GUI) and a convenient voice control function. The graphical user interface is modern and responsive, providing clear music playback controls, such as play, pause, skip and repeat buttons, and dynamically displaying detailed information of the currently playing music, including the track name, artist, album cover and real-time emotional feedback slider. Users can slide the slider to give real-time feedback on the impact of music on their emotions. The scale of the slider is from 1 to 10, where 1 represents "very dislike" and 10 represents "very like". The interface also contains personalized music recommendation settings, allowing users to adjust the parameters of the recommendation algorithm according to their personal preferences, such as preferred music styles, activity types (such as work, exercise, relaxation) and emotional goals (such as refreshing, decompression, sleep). Users can also select different music tags such as "relax", "motivate", "nostalgia" through a drop-down menu, and set the time and volume preferences for music playback. The voice control function is implemented through an integrated natural language processing (NLP) engine, and users can control music playback through voice commands, such as saying "play jazz" or "volume up", and the system can recognize and perform the corresponding operations. The NLP engine supports multiple languages ​​and dialects, with a recognition accuracy of over 95% and a response time of less than 200 milliseconds, ensuring that users can smoothly interact with the system through voice. The user interaction interface also provides help and feedback options, allowing users to easily access help documents, learn how to use the system, and report problems or make suggestions for improvements to the developer through the built-in feedback system. The interface is designed with users of different ages and technical levels in mind, ensuring ease of use and accessibility, so that even first-time users can quickly get started and enjoy a personalized music therapy experience. Through this comprehensive user interaction design, the system is able to provide a richer and more intuitive user experience, while collecting valuable user feedback to further optimize the music recommendation algorithm and improve overall system performance.

[0012] Privacy protection measures are a key component to ensure user information security, which includes strict data encryption storage and user data use consent mechanism. In terms of data encryption storage, the system uses the Advanced Encryption Standard (AES) 256-bit encryption algorithm to protect user data stored locally and in the cloud, ensuring the security of data during transmission and static storage. All user data, including biofeedback monitoring data, emotion analysis results, music preferences, and user interaction records, are encrypted by encryption keys, and only authorized system components can decrypt and access these data. The user data use consent mechanism requires that the user's explicit consent must be obtained before collecting, processing, or sharing user data. When the user uses the system for the first time, the system displays the privacy policy and data use terms through the graphical user interface (GUI). The user must read and actively check the consent option before continuing to use the system. In addition, the system also provides a privacy setting interface that allows users to view, modify, or withdraw their consent at any time, as well as manage their data sharing and access rights. The system also implements the principle of data minimization, only collecting data necessary to achieve functionality and limiting access to these data. The system also conducts regular security audits and vulnerability assessments to ensure the effectiveness and timely updates of privacy protection measures. Through these comprehensive privacy protection measures, the system not only ensures the security of user data, but also builds user trust in the system, ensuring that users can enjoy a personalized music therapy experience with peace of mind.

[0013] The biofeedback-based emotion regulation music therapy approach carefully designs the following steps to achieve highly effective and personalized emotion regulation:

[0014] S1. Start the biofeedback monitoring module: The system activates a series of sensors, including a heart rate monitor (measuring range 1-200 times / minute, accuracy ±1 time / minute), a skin galvanic response sensor (measuring range 0-100μS, accuracy ±1%), a facial expression recognition camera (resolution 1920x1080, frame rate 30fps), a speech analysis module (frequency response 20Hz-20kHz, signal-to-noise ratio>60dB) and a body temperature sensor (measuring range 32-42℃, accuracy ±0.1℃) to collect the user's physiological and behavioral data in real time.

[0015] S2. Emotion Analysis: The collected data is transmitted to the emotion analysis engine, which uses deep learning algorithms and pattern recognition technology to analyze parameters such as the user's heart rate variability (HRV), skin conductivity, facial expressions, tone and speed of speech, body temperature changes, etc. to determine the user's emotional state, such as happiness, sadness, anger, stress and relaxation.

[0016] S3. Music recommendation: Based on the output of the sentiment analysis engine, the music recommendation system selects music that matches the user's current emotional state from a database of at least 1,000 music pieces of different styles and sentiment tags. For example, it recommends classical music with a slow rhythm and rich harmony when the user is under stress.

[0017] S4. Music Playing and Monitoring: The system plays the recommended music and continues to monitor the user's physiological and behavioral responses through the biofeedback monitoring module to evaluate the effect of music therapy, such as the smooth decrease in heart rate and the reduction in skin conductivity.

[0018] S5. User feedback collection: Through the user interaction interface, the system collects user feedback, including preferences for music (indicated by ratings of 1-5 points), self-assessment of emotional changes (indicated by an emotional feedback slider of 1-10), and evaluation of the effectiveness of music therapy (collected through a satisfaction questionnaire).

[0019] S6. Personalized learning: User feedback is input into a personalized learning algorithm, which analyzes user preferences and emotional responses, and continuously adjusts and optimizes the music recommendation algorithm to provide a more personalized music experience, such as increasing the recommendation weight of a certain type of music based on the frequency of the user's preference for that type of music.

[0020] S7. Continuous monitoring and adjustment: Repeat steps S1 to S6 to continuously monitor the user's emotional state and adjust the music therapy program in real time to ensure the effectiveness and adaptability of music therapy, such as adjusting the music recommendation strategy according to the user's emotional changes at different times of the day.

[0021] S8. Privacy protection implementation: When users use the system, privacy protection measures are implemented, including data encryption storage (using AES256-bit encryption algorithm) and user data usage consent mechanism, to ensure that all collected data complies with privacy protection regulations and is stored securely. Only with the user's explicit consent will their data be used for analysis and recommendation.

[0022] Through this series of steps, the method is able to provide users with a comprehensive, real-time and personalized emotion regulation music therapy experience while ensuring the security and privacy of user data.

[0023] A key innovation in the emotion regulation music therapy system is the personalized learning algorithm, which is specially designed to learn the user's music preferences at different times and in different environments, and adjust the music recommendations accordingly. The algorithm identifies and adapts to the user's personalized needs by analyzing the user's choice and feedback on music in specific situations, such as preferring energetic music to refresh in the morning, and soft music to help sleep at night. The algorithm uses machine learning techniques such as collaborative filtering and content recommendation systems, combined with user behavioral data and environmental variables such as time, location, and activity type, to predict the type of music that the user may prefer at a specific moment.

[0024] In addition, the algorithm will also take into account the long-term changes in users' preferences and update the recommendation model by continuously tracking and analyzing the user's music interaction history, including the number of plays, skipped songs, repeated tracks, and user ratings. For example, if the user prefers to listen to relaxing jazz during their leisure time on weekends, the algorithm will incorporate this preference into future recommendations. Through this continuous learning and adaptive adjustment, the personalized learning algorithm can ensure that music recommendations not only match the user's immediate emotional state, but also adapt to the user's living habits and personal preferences, thereby providing a more accurate and personalized music therapy experience. This dynamic recommendation mechanism enables the system to continuously optimize over time, providing users with continuously improved emotional support and music enjoyment. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 A schematic diagram of the music therapy system of the present invention

[0026] Figure 2 The music therapy system of the present invention uses a flow chart DETAILED DESCRIPTION

[0027] The following will be combined with the attached embodiment of the present invention Figure 1 and Figure 2 , the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0028] The present invention proposes an emotion regulation music therapy system based on biofeedback, which integrates a heart rate monitor (measuring range 1-200 times / minute, accuracy ±1 time / minute), a skin galvanic response sensor (measuring range 0-100μS, accuracy ±1%), a facial expression recognition camera (resolution 1920x1080, frame rate 30fps), a speech analysis module (frequency response 20Hz-20kHz, signal-to-noise ratio>60dB), a body temperature sensor (measuring range 32-42℃, accuracy ±0.1℃), a motion sensor (measuring range 0-10m / s 2 , accuracy ±0.01m / s 2 ) and eye trackers (tracking speed 200Hz, accuracy 0.5°) to capture the user's physiological and behavioral data in real time. These data are transmitted to the emotion analysis engine, which uses deep learning algorithms to accurately identify the user's emotional state by analyzing parameters such as heart rate variability, skin conductivity, facial expression characteristics, voice characteristics, body temperature changes, body movements and eye movement patterns. Based on the results of emotion analysis, the music recommendation system selects music that matches the user's emotions from a database of more than 1,000 music songs with different styles and emotion tags, and automatically generates a personalized music playlist. Users can provide feedback on the music played through user interaction interfaces such as touch screens or voice commands, evaluate its impact on emotions or express personal preferences. The personalized learning algorithm analyzes these feedbacks in real time, continuously learns and adapts to optimize future music recommendations. Throughout the process, privacy protection measures ensure the security and privacy of user data, including data encryption storage, user identity verification and explicit consent before data use, to protect user privacy in all aspects. Through this series of detailed and coherent steps, the system can not only provide immediate emotional support, but also provide increasingly accurate personalized music therapy over time through continuous learning and user interaction, thus building an emotion regulation system that is both scientific and humane.

[0029] The biofeedback monitoring module further integrates an advanced algorithm that is specifically designed to process and integrate data from a heart rate monitor (measuring range 1-200 beats / minute, accuracy ±1 beat / minute), a galvanic skin response sensor (measuring range 0-100μS, accuracy ±1%), a facial expression recognition camera (resolution 1920x1080, frame rate 30fps), a speech analysis module (frequency response 20Hz-20kHz, signal-to-noise ratio>60dB), a body temperature sensor (measuring range 32-42℃, accuracy ±0.1℃), a motion sensor (measuring range 0-10m / s 2 , accuracy ±0.01m / s 2) and eye trackers (tracking speed 200Hz, accuracy 0.5°). The algorithm first filters and de-noises the raw data of each sensor to eliminate environmental interference and measurement errors. For example, a bandpass filter is used to remove high-frequency noise and motion artifacts in the heart rate monitor data. Then, the algorithm ensures that the data from different sensors are synchronized in time through timestamp alignment technology, with alignment accuracy reaching millisecond level. Next, the algorithm uses advanced data fusion techniques, such as Kalman filter or neural network, to fuse the multi-source data into a comprehensive emotional state indicator. The fusion process takes into account the weight and reliability of each sensor data. For example, the weight of heart rate variability and skin conductivity in emotional state assessment is adjusted based on previous user feedback and validation studies. Finally, this comprehensive indicator is passed to the emotion analysis engine for further emotion recognition and analysis. The emotion analysis engine uses a support vector machine (SVM) classifier to identify the user's emotional state, such as happiness, sadness, anger, stress, and relaxation, with an accuracy of more than 90% based on the fused data. Through the application of this synchronization and integration algorithm, the system is able to significantly improve the accuracy of emotional state recognition and provide users with more precise emotion regulation music therapy.

[0030] The emotion analysis engine is the core component of the system. It analyzes the user's emotional state by using advanced machine learning algorithms. The engine first receives synchronized and integrated data from the biofeedback monitoring module, including heart rate (measurement range 1-200 beats / minute, accuracy ±1 beat / minute), skin galvanic response (measurement range 0-100μS, accuracy ±1%), facial expression (resolution 1920x1080, frame rate 30fps), voice characteristics (frequency response 20Hz-20kHz, signal-to-noise ratio>60dB), body temperature (measurement range 32-42℃, accuracy ±0.1℃), movement (measurement range 0-10m / s 2 , accuracy ±0.01m / s 2 ) and eye movements (tracking speed 200Hz, accuracy 0.5°) and other parameters. The sentiment analysis engine then uses pre-trained machine learning models such as random forests, support vector machines (SVMs), deep neural networks (DNNs), or convolutional neural networks (CNNs) to perform feature extraction and pattern recognition on these multi-dimensional data. The sentiment analysis engine may use a deep neural network with multiple hidden layers, each containing hundreds of neurons, to automatically learn to extract complex emotion-related features from physiological signals. Alternatively, it may utilize a support vector machine model that uses a radial basis function (RBF) kernel during the training phase to optimize classification performance by adjusting the values ​​of C (regularization parameter) and γ (kernel function parameter) to identify the user's emotional state, such as happiness, sadness, anger, stress, and relaxation, with an accuracy of over 90%.

[0031] During the analysis process, the emotion analysis engine also considers individual differences and environmental factors, and improves the accuracy of emotion recognition through personalized model training and adjustment. For example, the engine may adjust the threshold and weight of the model based on the range of changes in the user's heart rate variability (HRV) and skin galvanic response to more accurately reflect the individual's emotional changes. The engine is able to recognize multiple emotional states and is able to self-optimize through continuous learning and user feedback. For example, if a user provides feedback on mood improvement through the user interaction interface after listening to a piece of music, the engine will record this data and consider this information in future analysis to further calibrate the emotion recognition model. The emotion analysis engine also has real-time processing capabilities, which can quickly respond to immediate changes in user emotions, ensuring that the music recommendation system can provide music that matches the user's current emotional state in a timely manner. This fast and accurate emotion analysis capability enables the system to provide users with timely and personalized emotional regulation support, thereby effectively promoting emotional management and mental health.

[0032] The music database is the core of providing personalized music therapy. It contains at least 1,000 pieces of music with different styles and emotional labels. These pieces are carefully selected to cover a wide range of music types, including but not limited to classical music (such as Bach's Well-Tempered Clavier, each piece is about 2-5 minutes long), jazz (such as the pieces in Miles Davis' Kind of Blue album, with an average duration of 5-10 minutes), electronic music (such as electronic dance music of various rhythms, each piece is about 3-7 minutes long), pop music (such as popular singles from different decades, with an average duration of 3-5 minutes), etc. Each piece of music has been carefully annotated by a professional team and is accompanied by an emotional label, such as "relaxing" (which may be associated with slow-tempo, low-volume music), "motivating" (which may be associated with fast-tempo, high-energy music), "nostalgic" (which may be associated with classic old songs or music with the characteristics of a specific era), "cheerful" (which may be associated with music with bright melodies and positive lyrics) and "contemplative" (which may be associated with minimalist and introspective music). In order to ensure the effectiveness of music therapy, the works in the music database are also classified according to music parameters such as rhythm (BPM, beats per minute, such as 40-120BPM for slow rhythm, 120-160BPM for medium-fast rhythm, and 160BPM or above for fast rhythm), pitch (such as C major, A minor, etc., which may affect the brightness or melancholy of emotions), and harmony (such as major triads, minor triads, etc., which affect the emotional color of music). Through the output of the emotion analysis engine, the system can quickly retrieve and recommend music suitable for the user's current emotional state from this rich music library. For example, when the emotion analysis engine recognizes that the user is under stress, the system will select music with a BPM between 60-80, a pitch dominated by major keys, and a harmony dominated by major triads from the database to help users relieve tension. Through this refined music selection and recommendation mechanism, the system can provide users with more accurate and effective emotional regulation support.

[0033] The music recommendation system uses a sophisticated algorithm framework to adjust the music playlist in real time according to the user's emotional state and historical preferences. The system first receives data from the emotion analysis engine, including emotion categories (such as happiness, sadness, anger, etc., each category is subdivided into 5 levels, from 1 to 5 to indicate the increasing intensity of emotion), emotion intensity (expressed by numerical values, such as mild for 1-3, moderate for 4-6, severe for 7-10) and emotion change trends (expressed by percentage of change rate, such as 5% increase, 3% decrease). At the same time, the system also analyzes the user's historical preference data, including the user's past preferred music types (such as classical, jazz, electronic, pop, etc.), the most frequent listening time (such as 8 to 10 pm), and the user's feedback on specific music (such as an average rating of 4.5 / 5, more than 50 plays, and a skip rate of less than 5%).

[0034] The music recommendation system inputs this data into a recommendation algorithm based on machine learning. The algorithm uses collaborative filtering technology, combines the user's historical behavior and the data of similar users, and content recommendation strategies to analyze the matching degree of music features (such as rhythm, pitch, harmony) with the user's emotional state. For example, the system may recommend music with a slower tempo (BPM less than 70), a minor key, and a minor triad as the main harmony to a user with an emotional intensity of 5 to help the user relax. The system also conducts adaptive learning based on the user's real-time feedback. If the user gives positive feedback to a recommended track (such as a score of 4.8 / 5), the system will increase the frequency of recommending similar tracks in the future; on the contrary, if the user skips a track, the system will reduce or no longer recommend music of similar styles. The music recommendation system also has real-time processing capabilities, which can quickly respond to the immediate changes in the user's emotions and ensure that the adjustment of the music playlist is synchronized with the user's emotional state. This dynamic adjustment mechanism enables the system to provide users with timely and personalized music experience, thereby effectively promoting emotional management and mental health. Through this refined music recommendation strategy, the system can not only meet the user's current emotional regulation needs, but also continuously optimize the recommendation quality over time and enhance the user's music experience.

[0035] The personalized learning algorithm continuously optimizes the music recommendation algorithm through deep learning technology to provide a more accurate personalized music experience. The algorithm uses a long short-term memory network (LSTM), a special recurrent neural network (RNN), to process the time series data of user interaction with music. The network contains 3 LSTM layers, each with 100 neurons, which can capture the dynamic changes of user preferences over time. The algorithm receives feedback from the user interface, including the rating of the recommended music (1-5 points, where 1 point means very dislike and 5 points means very like), the playing time (in seconds, for example, the user played a song for 240 seconds), the number of repeated plays (for example, the user played a song three times in a week), and the skip behavior (represented by a value between 0-1, 0 means no skipping and 1 means skipping). During the training process, these feedback data are converted into numerical features and input into the LSTM network as training signals. For example, if a user rates a song 4.8 points and plays 90% of the entire song (i.e. 216 seconds), these data will be used to strengthen the system's understanding of the user's preference for similar music. The deep learning model is trained using the back-propagation algorithm and the mean square error (MSE) loss function, with a learning rate set to 0.001 to optimize network weights and reduce prediction errors. To prevent overfitting of the model, the system implements L2 regularization technology with a regularization coefficient of 0.01 to ensure the generalization ability of the algorithm. In addition, the algorithm also includes a validation mechanism that regularly evaluates the performance of the model by dividing user data into a training set (80%) and a test set (20%), such as a performance evaluation every 100 training cycles, to ensure that the recommendation system continues to improve as user behavior changes.

[0036] The user interface is the core of the system's interaction with users. It includes an intuitive graphical user interface (GUI) and advanced voice control functions. The graphical user interface is designed to be simple and intuitive, equipped with easy-to-identify playback control buttons, such as play, pause, skip and repeat buttons, with a size of at least 24x24 pixels to ensure easy operation on different devices. The interface also includes a real-time emotional feedback slider, ranging from 1 to 10 with an accuracy of 0.1. Users can slide this slider to provide real-time feedback on the impact of music on their emotions, where 1 represents "very dislike" and 10 represents "very like".

[0037] The interface also provides personalized music recommendation settings. Users can select different music styles such as "classical", "jazz", "electronic", and "pop" through a drop-down menu, and set preferences for music playback, such as activity type ("work", "exercise", "relax") and emotional goals ("refreshing", "stress relief", "sleep"). Each option has a clear icon and text label to improve recognition. Users can also set the volume of music playback through a slider, ranging from 0% to 100%, and adjust in 1% increments to ensure that users can adjust the volume according to the environment and personal preferences. The voice control function is implemented through an integrated natural language processing (NLP) engine, which is able to recognize and process voice commands entered by users through a microphone, such as "play jazz" or "volume up". The system is able to respond to voice commands within 200 milliseconds, with a recognition accuracy of more than 95%. The NLP engine supports multiple languages ​​and dialects, ensuring that users from different language backgrounds can interact with the system without obstacles. The user interaction interface also provides help and feedback options. Users can access the help document by clicking the "Help" icon in the upper right corner of the interface to learn how to use the system, and report problems or make suggestions for improvement to the developer through the built-in feedback system. The interface is designed with users of different ages and technical levels in mind, ensuring ease of use and accessibility, so that even first-time users can quickly get started and enjoy a personalized music therapy experience. Through this comprehensive user interaction design, the system is able to provide a richer and more intuitive user experience while collecting valuable user feedback to further optimize the music recommendation algorithm and improve overall system performance.

[0038] Privacy protection measures are the core link to ensure the security of user information. They include the use of industry-standard Advanced Encryption Standard (AES) 256-bit encryption algorithm to protect the security of user data. During the data transmission process, the system uses the Transport Layer Security Protocol (TLS) 1.2 or higher to ensure the integrity and confidentiality of the data during transmission. All user data, including biofeedback monitoring data, emotion analysis results, music preferences, and user interaction records, are encrypted by a randomly generated 256-bit encryption key. The generation of the key follows the random number generation rules of the National Security Standard (ANSI) to ensure the unpredictability of the key. The user data use consent mechanism requires the user's explicit consent before collecting, processing or sharing user data. When the user uses the system for the first time, the system displays the privacy policy and data use terms through the graphical user interface (GUI). The user must read and actively check the consent option before continuing to use the system. The system also provides a privacy settings interface that allows users to view, modify or withdraw their consent at any time, as well as manage their data sharing and access rights. Users can access these options at any time through the "Privacy Settings" button on the interface and make personalized settings. To further enhance privacy protection, the system also implements the principle of data minimization, only collects data necessary to achieve functionality, and limits access to this data. The system also conducts regular security audits and vulnerability assessments once a quarter to ensure the effectiveness and timely updates of privacy protection measures. Through these comprehensive privacy protection measures, the system not only ensures the security of user data, but also builds user trust in the system, ensuring that users can enjoy personalized music therapy experiences with peace of mind.

[0039] The biofeedback-based emotion regulation music therapy approach carefully designs the following steps to achieve highly effective and personalized emotion regulation:

[0040] S1. Start the biofeedback monitoring module: The system activates a series of sensors, including a heart rate monitor (measuring range 1-200 times / minute, accuracy ±1 time / minute), a skin galvanic response sensor (measuring range 0-100μS, accuracy ±1%), a facial expression recognition camera (resolution 1920x1080, frame rate 30fps), a speech analysis module (frequency response 20Hz-20kHz, signal-to-noise ratio>60dB) and a body temperature sensor (measuring range 32-42℃, accuracy ±0.1℃) to collect the user's physiological and behavioral data in real time.

[0041] S2. Emotion Analysis: The collected data is transmitted to the emotion analysis engine, which uses deep learning algorithms and pattern recognition technology to analyze parameters such as the user's heart rate variability (HRV), skin conductivity, facial expressions, tone and speed of speech, body temperature changes, etc. to determine the user's emotional state, such as happiness, sadness, anger, stress and relaxation.

[0042] S3. Music recommendation: Based on the output of the sentiment analysis engine, the music recommendation system selects music that matches the user's current emotional state from a database of at least 1,000 music pieces of different styles and sentiment tags. For example, it recommends classical music with a slow rhythm and rich harmony when the user is under stress.

[0043] S4. Music Playing and Monitoring: The system plays the recommended music and continues to monitor the user's physiological and behavioral responses through the biofeedback monitoring module to evaluate the effect of music therapy, such as the smooth decrease in heart rate and the reduction in skin conductivity.

[0044] S5. User feedback collection: Through the user interaction interface, the system collects user feedback, including preferences for music (indicated by ratings of 1-5 points), self-assessment of emotional changes (indicated by an emotional feedback slider of 1-10), and evaluation of the effectiveness of music therapy (collected through a satisfaction questionnaire).

[0045] S6. Personalized learning: User feedback is input into a personalized learning algorithm, which analyzes user preferences and emotional responses, and continuously adjusts and optimizes the music recommendation algorithm to provide a more personalized music experience, such as increasing the recommendation weight of a certain type of music based on the frequency of the user's preference for that type of music.

[0046] S7. Continuous monitoring and adjustment: Repeat steps S1 to S6 to continuously monitor the user's emotional state and adjust the music therapy program in real time to ensure the effectiveness and adaptability of music therapy, such as adjusting the music recommendation strategy according to the user's emotional changes at different times of the day.

[0047] S8. Privacy protection implementation: When users use the system, privacy protection measures are implemented, including data encryption storage (using AES256-bit encryption algorithm) and user data usage consent mechanism, to ensure that all collected data complies with privacy protection regulations and is stored securely. Only with the user's explicit consent will their data be used for analysis and recommendation.

[0048] Through this series of steps, the method is able to provide users with a comprehensive, real-time and personalized emotion regulation music therapy experience while ensuring the security and privacy of user data.

[0049] The personalized learning algorithm is implemented through a deep learning framework, which can learn users' music preferences at different times and in different environments, and adjust music recommendations accordingly. The algorithm analyzes users' choices and feedback on music in specific time periods (such as 6:00 to 8:00 in the morning and 20:00 to 22:00 in the evening), taking into account that users may prefer music with a fast rhythm and a volume of 70-85 decibels to refresh themselves in the morning, while they may prefer music with a slow rhythm and a volume of 40-60 decibels to help them sleep in the evening. The algorithm uses collaborative filtering technology, combined with user behavioral data and environmental variables such as time (marked by 24-hour timestamps), location (identified by GPS coordinates or indoor positioning systems), and activity type (inferred by user input or smart device sensors, such as exercise, reading, and rest) to predict the type of music that users may prefer at a specific moment.

[0050] The algorithm also takes into account the long-term changes in user preferences by continuously tracking and analyzing the user's music interaction history, including the number of plays (such as a song being played more than 15 times in a week), skipped songs (such as songs that the user skipped within 30 seconds), repeated tracks (such as users repeatedly playing the same track more than 3 times in an hour) and user ratings (such as users rating a song above 4.5 out of 5), to update the recommendation model. For example, if a user prefers to listen to relaxing jazz music during their leisure time on weekends, the algorithm will incorporate this preference into future recommendations, such as increasing the proportion of jazz tracks in the recommendation list on weekend afternoons.

[0051] Through this continuous learning and adaptive adjustment, the personalized learning algorithm can ensure that music recommendations not only match the user's immediate emotional state, but also adapt to the user's living habits and personal preferences, thereby providing a more accurate and personalized music therapy experience. This dynamic recommendation mechanism enables the system to continuously optimize over time, providing users with continuously improved emotional support and music enjoyment.

[0052] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A biofeedback-based emotion regulation music therapy system, characterized in that: include: Biofeedback monitoring module (heart rate monitor, galvanic skin response sensor, facial expression recognition camera voice analysis module, body temperature sensor, motion sensor, eye tracker), emotion analysis engine, music database, music recommendation system, personalized learning algorithm, user interaction interface, privacy protection measures. The biofeedback monitoring module serves as the data collection center of the system, and collects the user's physiological and behavioral data in real time through sensors such as heart rate monitor, galvanic skin response sensor, facial expression recognition camera, voice analysis module, body temperature sensor, motion sensor and eye tracker. These data are transmitted to the emotion analysis engine, which uses the deep learning algorithm to comprehensively analyze various parameters and accurately identify the user's emotional state. The music recommendation system selects music that matches the user's emotions from the music database based on the results of the emotion analysis, and automatically generates a playlist to regulate the user's emotions in the form of music. The user provides feedback on the music effect through the interaction interface, and the personalized learning algorithm processes these feedbacks in real time, continuously optimizes the music recommendation algorithm, and provides a more accurate personalized music experience. In the whole process, the privacy protection measures ensure the security and privacy of user data, thereby building a scientific and humane emotion regulation system.

2. The system according to claim 1, characterized in that The biofeedback monitoring module also includes an algorithm for synchronizing and integrating data from different sensors to improve the accuracy of emotional state recognition.

3. The system according to claim 1, characterized in that The emotion analysis engine is capable of analyzing the user's emotional state through a machine learning algorithm.

4. The system according to claim 1, characterized in that The music database contains at least 1,000 pieces of music with different styles and mood tags.

5. The system according to claim 1, characterized in that The music recommendation system can adjust the music playlist in real time according to the user's emotional state and historical preferences.

6. The system according to claim 1, characterized in that The personalized learning algorithm can continuously optimize the music recommendation algorithm through deep learning technology.

7. The system according to claim 1, characterized in that The user interaction interface includes a graphical user interface and a voice control function.

8. The system according to claim 1, characterized in that The privacy protection measures include data encryption storage and user data usage consent mechanism.

9. A biofeedback-based emotion regulation music therapy method, characterized in that: The following steps are involved: S1. Start the biofeedback monitoring module to collect the user's physiological and behavioral data in real time through the heart rate monitor, skin galvanic response sensor, facial expression recognition camera, voice analysis module and body temperature sensor; S2. The collected data is transmitted to the emotion analysis engine, which uses deep learning algorithms and pattern recognition technology to analyze the user's heart rate variability, skin conductivity, facial expressions, tone and speed of speech, body temperature changes, etc. to determine the user's emotional state; S3. Based on the output of the emotion analysis engine, the music recommendation system selects music that matches the user's current emotional state from the music database, such as recommending relaxing classical music when the user is under stress; S4. Play the recommended music and continue to monitor the user's emotional response through the biofeedback monitoring module to evaluate the effect of music therapy; S5. Collect user feedback through the user interaction interface, including music preferences, self-assessment of emotional changes, and evaluation of the effectiveness of music therapy; S6. Input user feedback into the personalized learning algorithm, which analyzes user preferences and emotional responses, and continuously adjusts and optimizes the music recommendation algorithm to provide a more personalized music experience; S7. Repeat steps 1 to 6 to continuously monitor the user's emotional state and adjust the music therapy program in real time to ensure the effectiveness and adaptability of the music therapy; S8. Implement privacy protection measures during the user's use of the system to ensure that all collected data complies with privacy protection regulations and is stored securely. The user's data is only used for analysis and recommendation with explicit consent from the user.

10. The method according to claim 9, characterized in that The personalized learning algorithm can learn the user's music preferences at different times and in different environments, and adjust the music recommendations accordingly.

Images