Music adaptive adjustment method and equipment

Abstract

The embodiment of the invention discloses a music adaptive adjustment method and equipment, belongs to the technical field of artificial intelligence, and solves the problems that background music causes interference to a learning process and the learning efficiency is reduced when the background music is not matched with a learning activity. Comprising the steps of performing task matching and task analysis according to learning scene information acquired by intelligent glasses to obtain learning task information; wherein the learning task information at least comprises one of a learning task type, a learning task duration and a learning task difficulty level; analyzing and detecting the real-time learning behavior information acquired by the intelligent glasses to obtain a real-time learning state; and matching corresponding background music in a music database based on the learning task information and / or the real-time learning state, and performing background music adaptive adjustment.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a method and device for adaptively adjusting music. Background Art

[0002] In modern learning environments, background music is widely used to optimize the learning experience. Appropriate background music can effectively improve learners' concentration by regulating the activity of brain nerves, while speeding up information processing efficiency and making the learning process more efficient.

[0003] However, existing music playback systems mostly use fixed playlists or make recommendations based on user preferences. This can lead to a mismatch between music and learning activities. For example, playing overly active music during deep reading that requires high concentration, or playing overly calm music when the user is tired, can disrupt the learning process and reduce learning efficiency. Summary of the Invention

[0004] The embodiments of the present application provide a music adaptive adjustment method and device for solving the following technical problem: when background music does not match learning activities, the background music may interfere with the learning process and reduce learning efficiency.

[0005] The embodiments of this application adopt the following technical solutions: The present application provides a method for adaptively adjusting music, including performing task matching and task analysis based on learning scenario information acquired by smart glasses to obtain learning task information, wherein the learning task information includes at least one of the following: learning task type, learning task duration, and learning task difficulty level; analyzing and detecting real-time learning behavior information acquired by the smart glasses to obtain real-time learning status; and matching corresponding background music in a music database based on the learning task information and / or the real-time learning status to perform adaptive adjustment of the background music.

[0006] In one implementation of the present application, task matching and task analysis are performed based on the learning scene information acquired by the smart glasses to obtain learning task information, specifically including: when determining the type of learning task: performing image recognition processing on the image data in the learning scene information to extract visual feature information; wherein the visual feature information includes at least one of the following: learning tools, learning materials, user learning posture, and text data generated during user learning; performing text recognition on the text data in the learning scene information through natural language processing technology to extract text feature information; fusing the visual feature information with the text feature information to obtain key learning information; matching the key learning information in a preset learning task database to determine the type of learning task; wherein the type of learning task includes at least one of the following: in-depth reading, language learning, math problem solving, and writing.

[0007] In one implementation of the present application, task matching and task analysis are performed based on the learning scene information obtained by the smart glasses to obtain learning task information, specifically including: when determining the learning task duration: determining the first reference learning task duration based on the amount of text content in the learning materials; obtaining the second reference learning task duration preset by the user; determining in the historical database the historical learning data whose similarity value with the key learning information meets the similarity conditions; inputting the key learning information, the first reference learning task duration, the second reference learning task duration and the historical learning data into a preset duration prediction model to output the learning task duration corresponding to the current learning task.

[0008] In one implementation of the present application, task matching and task analysis are performed based on the learning scene information obtained by the smart glasses to obtain learning task information, specifically including: when determining the difficulty level of the learning task: determining the complexity level of the learning task based on the type of text content in the learning materials; determining the historical learning data corresponding to the complexity level of the learning task in the historical database; constructing the learning portrait and learning curve corresponding to the user at the complexity level of the learning task based on the completion result data corresponding to the historical learning data, so as to determine the difficulty level of the learning task based on the user portrait and the learning curve; wherein the completion result data includes at least the task completion rate, the task error rate and the thinking time.

[0009] In one implementation of the present application, the real-time learning behavior information obtained by the smart glasses is analyzed and detected to obtain the real-time learning status, specifically including: obtaining the real-time learning behavior information sent by different information collection devices on the smart glasses; wherein the real-time learning behavior information includes at least one of the following: eye tracking information, head posture information, facial expression information and physiological characteristic parameters; comparing different types of real-time learning behavior information with corresponding state thresholds, and determining the state coefficient values ​​corresponding to each type of real-time learning behavior information based on the comparison difference; obtaining the real-time learning status based on the preset weight distribution parameters, the real-time learning behavior information and the state coefficient value; wherein the real-time learning status includes a distraction state and a fatigue state.

[0010] In one implementation of the present application, a real-time learning state is obtained based on preset weight distribution parameters, real-time learning behavior information, and state coefficient values, specifically including: screening out the real-time learning behavior information of the type required for distraction detection in the real-time learning behavior information, and constructing a distraction detection information set; performing weighted processing on each piece of real-time learning behavior information in the distraction detection information set based on the weight distribution parameters and state coefficient values ​​corresponding to the distraction detection information set, and obtaining the user's distraction state; screening out the real-time learning behavior information of the type required for fatigue detection in the real-time learning behavior information, and constructing a fatigue detection information set; performing weighted processing on each piece of real-time learning behavior information in the fatigue detection information set based on the weight distribution parameters and state coefficient values ​​corresponding to the fatigue detection information set, and obtaining the user's fatigue state.

[0011] In one implementation of the present application, based on learning task information and / or real-time learning status, corresponding background music is matched in a music database, and adaptive adjustment of the background music is performed, specifically including: based on the learning task type, determining a first music set in the music database; based on the learning task duration and the learning task difficulty level, constructing a music characteristic trend curve, and determining a second music set in the music database based on the music characteristic trend curve; wherein the music characteristic trend curve is divided into multiple playback stages based on changes in music characteristics; in multiple playback stages, based on the intersection result of the first music set and the second music set, a learning background music list is determined and played; based on the real-time learning status, a third music set is determined in the music data set; and the learning background music list is adjusted based on the third music set.

[0012] In one implementation of the present application, before determining a learning background music list and playing it, the method also includes: when there is no intersection between the first music set and the second music set, inputting the learning task information into a preset music screening model and outputting a reference music list; based on the user's historical music playback frequency, time decay factor and scene correlation, determining the historical preference score corresponding to each music in the reference music list; sorting the reference music list based on the historical preference score to construct a learning background music list.

[0013] In one implementation of the present application, after adjusting the learning background music list based on the third music set, the method also includes: during the playback of the learning background music, comparing the acquired real-time learning status with the preset status transition conditions, and regenerating the third music set when a change in the user status is detected; determining the playback volume based on the converted user status, and automatically adjusting the volume of the player; and playing the learning background music based on the regenerated third music set and the adjusted volume.

[0014] An embodiment of the present application provides a music adaptive adjustment device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can: perform task matching and task analysis based on learning scene information acquired by the smart glasses to obtain learning task information; wherein the learning task information includes at least one of the following: learning task type, learning task duration, and learning task difficulty level; analyze and detect the real-time learning behavior information acquired by the smart glasses to obtain a real-time learning status; based on the learning task information and / or the real-time learning status, match the corresponding background music in the music database to perform background music adaptive adjustment.

[0015] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects: the embodiments of the present application obtain learning scene information and user behavior information through smart glasses, realize multi-dimensional data fusion, and make music adjustment more intelligent and humane. Secondly, the embodiments of the present application perceive the user status in real time and adjust the music in time, effectively relieve distraction and fatigue, help users maintain a positive learning attitude, and enhance the overall learning experience. In addition, the embodiments of the present application dynamically adjust the music according to the type of learning task, task duration and difficulty, and the real-time status of the user, provide personalized and accurate music adjustment, reduce the interference of traditional fixed or preset music playback modes on the learning process, and improve learning efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments described in the present application. For those skilled in the art, other drawings can be obtained based on these drawings without inventive work. In the drawings: Figure 1 Schematic diagram of a system architecture in which the embodiments of the present application can be applied; Figure 2 A flowchart of a music adaptive adjustment method provided in an embodiment of the present application; Figure 3 A schematic diagram of a method for determining a learning task type provided in an embodiment of the present application; Figure 4 A schematic diagram of a method for determining the duration of a learning task provided in an embodiment of the present application; Figure 5 A schematic diagram of a method for determining the difficulty level of a learning task provided in an embodiment of the present application; Figure 6 A schematic diagram of a method for determining real-time learning status provided in an embodiment of the present application; Figure 7 A schematic diagram of a method for adaptively adjusting background music provided in an embodiment of the present application; Figure 8 A flowchart of a music adaptive adjustment process provided in an embodiment of the present application; Figure 9 A schematic structural diagram of a music adaptive adjustment device provided in an embodiment of the present application.

[0017] Reference numerals: 101: Smart glasses, 102: Music database, 103: Music playback application, 104: Communication network, 105: Cloud. DETAILED DESCRIPTION

[0018] The embodiments of the present application provide a method and device for adaptively adjusting music.

[0019] In order to enable those skilled in the art to better understand the technical solutions in this application, the following will clearly and completely describe the technical solutions in the embodiments of this application in conjunction with the drawings in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0020] The following describes in detail the technical solutions provided by various embodiments of the present application in conjunction with the accompanying drawings.

[0021] Figure 1 This is a schematic diagram of a system architecture in which the embodiments of the present application can be applied, such as Figure 1 As shown, the music adaptive adjustment method can be applied to Figure 1 In the environment shown, the application environment may include smart glasses 101, a music database 102, a music playback application 103, a communication network 104, and a cloud 105. The smart glasses 101 worn by the user are equipped with a variety of learning environment information collection devices, such as cameras, to collect learning environment information in the user's field of view. In addition, the smart glasses 101 may also be equipped with a variety of user behavior collection devices, such as head posture sensors and biosensors, to collect real-time learning behavior information of the user. The smart glasses 101 upload the collected various learning environment information to the cloud 105 via the communication network 104. The cloud server analyzes the learning task type, learning task duration, and learning task difficulty level in the learning environment information. Based on the analysis results, the corresponding learning background music is matched in the music database 102 and sent to the music playback application 103 via the communication network 104 to control the music playback application 103 to play the learning background music. Afterwards, the smart glasses 101 transmit the user learning behavior information acquired by the user behavior collection device to the cloud 105 via the communication network 104. The cloud server analyzes the user learning behavior information to determine the user's learning status. Based on the user's learning status, the cloud server matches the corresponding learning background music in the music database 102 and transmits the information back to the music playback application 103 via the communication network 104 to control the music playback application 103 to update the playback track. In addition, the communication network 104 can also transmit the playback sound control information sent by the cloud 105 to the smart glasses supporting application, which can then adjust the volume of the playback device on the smart glasses 101.

[0022] Figure 2 A flowchart of a music adaptive adjustment method provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the music adaptive adjustment method includes the following steps S201-S203: S201. Perform task matching and task analysis based on the learning scene information acquired by the smart glasses to obtain learning task information.

[0023] In one implementation of the present application, the learning task information includes at least one of the following: the type of learning task, the duration of the learning task, and the difficulty level of the learning task. When determining the type of learning task, the smart glasses have a built-in high-resolution camera, which uses a computer vision algorithm to capture the learning environment information in the user's field of view in real time, and then uses image recognition and natural language processing technology to identify the type of learning activity task the user is currently performing. Among them, the acquired learning environment information can be large text contents in books, e-readers or screens, textbook contents, math problems, formulas, calculation processes on draft paper, etc., as well as the user's head posture, lip shape when the user is practicing oral English, pronunciation movements, etc. Figure 3 A schematic diagram of a method for determining the type of learning task provided in an embodiment of the present application is shown in FIG. Figure 3 As shown, the method for determining the type of learning task based on the acquired learning scenario information includes steps S301-S304: S301: Perform image recognition processing on image data in the learning scene information to extract visual feature information.

[0024] The acquired image data from the learning scene is decomposed into multiple key regions, such as a book, pen, user's hand, head, and body parts, and these regions are used as nodes in a graph structure. The association weights between each node are calculated, for example, using a graph attention model. This focuses on core objects related to the learning scene, such as learning tools, teaching materials, and the user's learning posture. Furthermore, the text regions contained in the image are located and labeled, ultimately extracting various visual feature information from the image.

[0025] S302: Perform text recognition on the text data in the learning scenario information through natural language processing technology to extract text feature information.

[0026] Analyze the acquired text data from the learning scenario. If the text data comes from a text area within an image, optical character recognition technology is first used to convert the text in the image into editable text. If the text is an existing electronic document, it can be processed directly. The text is then preprocessed by performing word segmentation and stop word removal. The text is then converted into a vector representation using a word embedding model. Semantic features of the text, such as keywords, sentence structure, and topic orientation, are further extracted to form text feature information.

[0027] S303: Fusing the visual feature information with the text feature information to obtain key learning information.

[0028] The extracted visual feature information is fused with the textual feature information to obtain more comprehensive key learning information. This fusion process requires establishing a correlation between the two features. For example, the math workbook in the visual feature can be associated with keywords such as functions and geometry that appear in the text, or the user's head-down writing posture can be associated with the handwritten notes in the text. This feature fusion is performed through the attention fusion mechanism to obtain key learning information.

[0029] In one implementation of the present application, when feature fusion is performed through the attention fusion mechanism, first, for each group of associated visual-text features, the attention weight distribution is determined according to its relevance to the learning task, and based on the attention weight distribution, the visual features and text features are weightedly fused, and the weighted visual features and text features are added or concatenated to form a comprehensive vector. By setting a threshold, the feature dimensions in the comprehensive vector whose weights are higher than the threshold are screened out, thereby obtaining key learning information.

[0030] S304: Match the key learning information in a preset learning task database to determine the learning task type.

[0031] The embodiment of the present application pre-constructs a preset learning task database, which stores feature templates corresponding to different types of learning tasks. Among them, the learning task types include at least one of the following: in-depth reading, language learning, math problem solving, and writing. For example, the feature template of in-depth reading may include: long text teaching materials, user focused reading posture (such as lowering the head for a long time, eyeballs fixed on the text area), etc.; the feature template of language learning may include: language learning application interface, word list, dialogue text, user lip shape and pronunciation movements when practicing oral language, etc.; the feature template of math problem solving may include: math teaching materials, calculators, formula symbols in the text, the posture of the user writing the calculation process, etc. The writing feature template may include: the user's behavior of inputting text on a computer or paper, the action of conceiving, and the writing interface, etc.

[0032] The fused key learning information is compared with the feature templates of each task type in the database, the similarity is calculated, and the task type corresponding to the feature template with the highest similarity is selected as the final determined user's current learning task type.

[0033] Furthermore, if the learning task involves in-depth reading, the system will prioritize music like instrumental music and light music. These music typically have low rhythmic and melodic complexity, fostering a calm and focused atmosphere and preventing distractions. If the learning task involves language learning, the system will select music with a moderate rhythm and clear, but not loud, melodies. These include upbeat music, background music with a hint of human humming, or white noise of a specific frequency. This can help users stay alert and aid in memorization and pronunciation practice. If the learning task involves math problem solving, the system will tend to recommend classical music and natural sounds like birdsong in the forest or the gurgling of water. This type of music fosters a rational and logical mindset, promoting deep thinking and logical coherence when solving complex problems. If the learning task involves writing, the system will recommend different types of music based on the nature of the writing, such as creative writing or technical reporting, and the user's current writing stage, such as conception, first draft, or revision. For example, more inspiring music may be needed during the conception phase, while more calming music may be needed when organizing thoughts.

[0034] Secondly, the embodiment of the present application determines the duration of the learning task based on the learning scene information obtained by the smart glasses. Figure 4 A schematic diagram of a method for determining the duration of a learning task provided in an embodiment of the present application is shown as follows: Figure 4 As shown, the process of determining the duration of a learning task includes the following steps S401-S404: S401. Determine a first reference learning task duration based on the amount of text content in the learning material.

[0035] In one implementation of the present application, the camera of the smart glasses can capture the content of the learning materials, analyze the number of questions, article length, number of chapters, etc. through image recognition and optical character recognition technology, and thus preliminarily estimate the time required to complete the task.

[0036] Specifically, the total amount of text in the filmed learning materials is counted, which can be quantified by indicators such as the number of words, paragraphs, or pages. Then, combined with pre-set text reading and comprehension rate indicators such as the average number of words read per minute and the average processing time per paragraph, the first reference learning task duration required to complete the learning of the text content is calculated.

[0037] S402: Obtain a second reference learning task duration preset by the user.

[0038] In one implementation of the present application, the second reference learning task duration set by the user can be collected through the supporting application of the smart glasses. The duration is manually input by the user based on his or her own plan, learning goals or time schedule.

[0039] S403: Determine in the historical database the historical learning data whose similarity value with the key learning information meets the similarity condition.

[0040] In one implementation of this application, a similarity algorithm is used to calculate the similarity between the extracted core features of the current learning key information, such as the learning task type, the type of teaching materials involved, and the user's posture characteristics, and the historical data in the database. A similarity threshold is set to filter out historical learning data with similarity values ​​that meet the conditions. The filtered historical learning data must include information such as the corresponding learning task duration and user completion status.

[0041] S404: Input the key learning information, the first reference learning task duration, the second reference learning task duration, and the historical learning data into a preset duration prediction model to output the learning task duration corresponding to the current learning task.

[0042] The key learning information, the duration of the first and second reference learning tasks, and the matched historical learning data are integrated to form an input feature set. This input feature set is fed into a preset duration prediction model, which uses the correlation between learning features to predict the duration of the current learning task.

[0043] Among them, when training the preset duration prediction model, historical learning key information samples, historical first reference duration samples, historical second reference duration samples, and historical data statistical feature samples are used as input samples, and the actual completion duration samples of the corresponding historical learning tasks are used as outputs to train the neural network model to obtain the preset duration prediction model.

[0044] Secondly, the embodiment of the present application determines the difficulty level of the learning task based on the learning scene information obtained by the smart glasses. Figure 5 A schematic diagram of a method for determining the difficulty level of a learning task provided in an embodiment of the present application is shown in FIG. Figure 5 As shown, the process of determining the difficulty level of a learning task includes the following steps S501-S503: S501. Determine the complexity level of the learning task based on the type of text content in the learning material.

[0045] The complexity level of the learning task is determined by identifying the density of professional terms, complexity of formulas, amount of information in charts, etc. in the learning materials.

[0046] S502: Determine historical learning data corresponding to the complexity level of the learning task in a historical database.

[0047] In the historical database, the determined learning task complexity level is used as the search condition to filter out historical learning data of users who have completed tasks of the same complexity level. The historical learning data must include the text type corresponding to the task, the user's operation process, and completion result data, such as task completion rate, error rate, and thinking time.

[0048] S503. Based on the completion result data corresponding to the historical learning data, construct the learning profile and learning curve corresponding to the user under the complexity level of the learning task, so as to determine the difficulty level of the learning task according to the user profile and learning curve.

[0049] In one implementation of this application, a learning profile is constructed based on the filtered historical learning data. The learning profile covers: user strengths in learning, such as a high completion rate for a certain type of text; user weaknesses in learning, such as a high error rate for a certain type of question; and average thinking speed.

[0050] At the same time, with time as the horizontal axis and completion result data, such as accuracy and efficiency, as the vertical axis, a learning curve is drawn to present the user's learning task trends or fluctuations at this complexity level.

[0051] Furthermore, a comprehensive analysis of the user profile and learning curve is conducted to determine the actual difficulty of the current task for the user. Specifically, if the user profile shows a high completion rate and low error rate for tasks of similar complexity, and the curve shows a continuously rising trend, the current task is relatively easy. If the profile shows a high error rate and long thinking time for similar tasks, and the curve fluctuates greatly or rises slowly, the current task is relatively difficult. Combined with the preset difficulty grading standards (e.g., easy, medium, and difficult), the difficulty level of the learning task is ultimately determined.

[0052] The embodiment of the present application assigns different music adjustment strategies to different learning task durations and learning difficulty levels. When the system estimates that the user's learning task is long and difficult, it will generate a series of soothing and layered music. This music adjustment is gradual, starting from the relatively calm, background music at the beginning of the task, and gradually transitioning to music with moderate rhythm changes but without interrupting thinking. For example, you can start with pure ambient sound, gradually add soft melodies, and then instrumental music with faint beats, to ensure that music is always used as an aid rather than a distraction. This gradual adjustment helps users maintain concentration during long learning processes and relieves cognitive fatigue. For short, low-difficulty tasks, the system may choose lighter and more refreshing music to quickly stimulate the user's learning interest and efficiency.

[0053] S202: Analyze and detect the real-time learning behavior information acquired by the smart glasses to obtain the real-time learning status.

[0054] In one implementation of the present application, the user's learning behavior information obtained by smart glasses is used to detect in real time whether the user is distracted or tired, and the generated background music is automatically adjusted according to the user's current state. For example, when the user is distracted: classical music and light music are generated to help the user concentrate and relieve distraction. When the user is tired: pop music with a light but not overly complex rhythm is generated to help the listener divert attention from the negative emotions of fatigue and face the fatigue state with a more positive attitude. At the same time, the light rhythm can also invigorate the spirit to a certain extent and relieve physical fatigue. Figure 6 A schematic diagram of a method for determining real-time learning status provided in an embodiment of the present application is shown as follows: Figure 6 As shown, the method for determining the real-time learning status includes steps S601-S603: S601. Acquire real-time learning behavior information sent by different information collection devices on the smart glasses.

[0055] The real-time learning behavior information includes at least one of the following: eye tracking information, head posture information, facial expression information, and physiological characteristic parameters.

[0056] In one implementation of the present application, the real-time learning behavior information of the user during the learning process is obtained through the information collection device equipped with smart glasses. For example, eye tracking information is captured by an eye movement sensor, including the position of the gaze, the duration of gaze, the frequency of blinking, etc.; head posture information is recorded by a gyroscope and an accelerometer, such as the head rotation angle, tilt amplitude, stability duration, etc., and fatigue postures such as head drooping and frequent nodding are monitored; facial expression information is captured and extracted by a camera, including features such as the curvature of the mouth corners, the frequency of frowning, and changes in eye expression, and signs of fatigue such as drooping eyelids, frequent eye rubbing, and eye congestion are identified by the camera; physiological characteristic parameters are collected by biosensors, such as heart rate changes and skin electrical response. After the collection is completed, it is classified and stored to ensure that each type of data is independent and complete.

[0057] S602: Compare different types of real-time learning behavior information with corresponding state thresholds, and determine state coefficient values ​​corresponding to each type of real-time learning behavior information based on comparison differences.

[0058] In one implementation of the present application, for each type of classified real-time learning behavior information, it is compared with the preset state threshold to calculate the corresponding state coefficient value. For example, for eye tracking information, if the length of time the line of sight stays outside the learning material exceeds the distraction state threshold, or the gaze duration is lower than the focus state threshold, a distraction coefficient between 0 and 1 is assigned based on the degree of excess or deficiency; for physiological characteristic parameters, if the heart rate fluctuation amplitude or blinking frequency exceeds the fatigue state threshold, a fatigue coefficient between 0 and 1 is assigned based on the degree of deviation. Each type of information uses similar comparison logic to obtain a state coefficient value that reflects the degree of distraction or fatigue under that dimension. The closer the coefficient is to 1, the more obvious the corresponding state is.

[0059] S603: Obtain a real-time learning state based on preset weight distribution parameters, real-time learning behavior information, and state coefficient values.

[0060] In one implementation of the present application, real-time learning behavior information is filtered to identify the type of real-time learning behavior information required for distraction detection, thereby constructing a distraction detection information set. Based on the weight distribution parameters and state coefficient values ​​corresponding to the distraction detection information set, each piece of real-time learning behavior information in the distraction detection information set is weighted to determine the user's distraction state.

[0061] Specifically, from the real-time learning behavior information, the information types that are highly correlated with the distraction state are screened out to construct a distraction detection information set. The distraction detection information set includes at least one of the following: eye tracking information, head posture information, and facial expression information. For each type of information in the distraction detection information set, the corresponding weight allocation parameters are called, and weighted processing is performed in combination with the calculated state coefficient value. For example, the distraction coefficient of eye tracking is multiplied by its weight, and the distraction coefficient of head posture is multiplied by the corresponding weight, and then all weighted results are added together to obtain a comprehensive score of the distraction state. If the score exceeds the preset distraction judgment threshold, it is determined that the user is in a distracted state. The higher the score, the more serious the distraction.

[0062] When the system detects a user's distraction, it immediately adjusts the music style, such as switching to classical music, instrumental music, or light music. These types of music typically feature rigorous structure, smooth melodies, and steady rhythms, helping users refocus their attention and guide their thinking back to the learning task.

[0063] In one implementation of the present application, real-time learning behavior information is filtered out to identify the type of real-time learning behavior information required for fatigue detection, thereby constructing a fatigue detection information set. Based on the corresponding weight allocation parameters and state coefficient values ​​of the fatigue detection information set, each piece of real-time learning behavior information in the fatigue detection information set is weighted to determine the user's fatigue state.

[0064] Specifically, from real-time learning behavior information, information types highly correlated with fatigue status are screened to construct a fatigue detection information set. This fatigue detection information set includes at least one of the following: eye features, head posture, and physiological characteristic parameters. Eye features can include signs of fatigue such as drooping eyelids, frequent eye rubbing, and bloodshot eyes; head posture can include fatigue gestures such as drooping head and frequent nodding.

[0065] Furthermore, for each type of information in the fatigue detection information set, the corresponding weighting parameters are called and combined with the existing state coefficient values ​​for weighted processing. For example, the fatigue coefficient of the physiological characteristics is multiplied by the weight, and the fatigue coefficient of the blink frequency is multiplied by the corresponding weight. All weighted results are summed up to obtain a comprehensive fatigue score. If this score exceeds the preset fatigue determination threshold, the user is determined to be in a fatigued state. The higher the score, the more obvious the fatigue level.

[0066] When the system detects user fatigue, it generates upbeat, yet not overly complex, pop, light rock, or electronic music. This type of music typically features a lively beat and positive melodies, helping to distract the listener from the negative emotions of fatigue and fostering a positive mindset. At the same time, a light beat can invigorate the spirit and alleviate physical fatigue, while avoiding overly intense music that could cause further fatigue or distraction.

[0067] S203: Based on the learning task information and / or the real-time learning status, the corresponding background music is matched in the music database, and the background music is adaptively adjusted.

[0068] The embodiment of the present application dynamically adjusts music according to the learning scenario, task duration and difficulty, and the user's real-time status, providing personalized and accurate music adjustment services. Figure 7 A schematic diagram of a method for adaptively adjusting background music provided in an embodiment of the present application is shown in FIG. Figure 7 As shown, the method for adaptively adjusting background music includes the following steps S701-S705: S701. Determine a first music set in a music database based on a learning task type.

[0069] Based on the determined learning task type, the first music collection suitable for the task type is screened from the music database. For example, if the task is identified as "deep reading" and the user is currently in a good state, pure music or light music is matched.

[0070] S702. Construct a music characteristic trend curve based on the learning task duration and the learning task difficulty level, and determine a second music set in the music database based on the music characteristic trend curve; wherein the music characteristic trend curve is divided into multiple playback stages based on changes in music characteristics.

[0071] In one implementation of the present application, a music characteristic trend curve is constructed based on the duration of the learning task and the difficulty level of the learning task to preset and adjust the learning background music. For example, the learning process can be divided into multiple playback stages according to the duration, such as the first 20% of the duration is the introductory stage, the middle 60% is the concentration stage, and the last 20% is the finishing stage. Then, for each stage, the music characteristic change rules are set in combination with the difficulty level. For example, the music rhythm in the concentration stage of low-difficulty tasks can be slightly lighter, while that of high-difficulty tasks needs to be smoother; when the duration is longer, the transition of the music characteristics of each stage needs to be more natural. Based on this curve, music that meets the characteristic requirements of each stage is screened out from the music database, and the second music set is compiled to ensure that the music can adapt to the task requirements as the learning progresses.

[0072] For example, for long, challenging tasks, the music starts out calm and low-energy, gradually introducing richer melodies and moderate rhythms, but always maintaining the principle of not interfering with thinking.

[0073] S703. In multiple playing stages, based on the intersection result of the first music set and the second music set, a learning background music list is determined and played.

[0074] In one implementation of this application, within the multiple playback stages defined by the music characteristic trend curve, the intersection of the first and second music sets is determined to obtain candidate music for each stage. These candidate music pieces are then organized into a list of learning background music, ensuring that the music in the list matches both the learning task type and the music characteristic requirements of each stage. The background music is then automatically played in the order listed, creating a suitable learning atmosphere for the user.

[0075] In one implementation of the present application, if the first and second music collections do not intersect, the learning task information is input into a preset music screening model, which outputs a reference music list. Based on the user's historical music playback frequency, time decay factor, and scene relevance, a historical preference score corresponding to each piece of music in the reference music list is determined. The reference music list is sorted based on the historical preference scores to construct a learning background music list.

[0076] When the first and second music collections do not overlap, the learning task information is input into a pre-set music screening model. The input samples for training this pre-set music screening model can include: learning task type samples, duration samples, and difficulty level samples; the output samples can be samples of a music list that meets the requirements. This pre-set music screening model analyzes the correlation between task information and musical characteristics, such as style, rhythm, and melodic complexity, to select music from the music database that is suitable for the current learning scenario and generate a reference music list. Based on the user's historical music play data, a historical preference score is calculated for each piece in the reference music list. Specifically, the historical play frequency of each piece of music is calculated; higher play frequency results in a higher base score. A time decay factor is introduced to assign higher weight to recently played music and lower weight to more recently played music to prevent outdated preferences from influencing the results. The model also considers scenario relevance, i.e., the number of times the piece of music has been played in similar learning scenarios in the past; higher relevance results in a greater score bonus. The values ​​of these three dimensions are integrated according to a preset ratio to obtain a historical preference score for each piece of music, reflecting the user's potential acceptance of the piece of music. The music in the reference music list is sorted in descending order based on the calculated historical preference scores, with the highest-scoring music at the top and the rest of the list in descending order. Once sorted, the sorted music is assigned to each stage based on the duration of the learning task and the designated playback phases, ensuring that the number of songs in each stage matches the duration and that the overall style remains consistent.

[0077] S704: Based on the real-time learning state, determine a third music set in the music data set.

[0078] In one implementation of the present application, based on the real-time learning status, a third music set that can improve the status is screened out from the music database.

[0079] For example, if a user becomes distracted, the system will immediately adjust the music style, switching to classical music, instrumental music, or light music with a rigorous structure, smooth melody, and steady rhythm. The volume can be lowered appropriately to reduce external distractions, helping the user refocus and guide their thoughts back to the learning task.

[0080] When the user is fatigued, the system generates upbeat, yet not overly complex, pop, light rock, electronic music, or instrumental music with positive elements. This type of music typically features a lively beat and upbeat melody, helping the listener distract from the negative emotions of fatigue and fostering a positive mindset. At the same time, a light rhythm can invigorate the mind and alleviate physical fatigue, while avoiding the cognitive overload that might otherwise be caused by overly intense or complex music.

[0081] If the user is anxious or irritable: the system can switch to soothing natural sounds, meditation music or low-frequency white noise to help the user calm down.

[0082] The user maintains high concentration for a long time: the system can maintain the current music style or make fine adjustments to avoid listening fatigue.

[0083] S705: Adjust the learning background music list based on the third music set.

[0084] If it is detected that the user is distracted or tired, appropriate music is selected from the third music collection to replace the music of the corresponding stage in the list, or the playback parameters of the music, such as volume and tempo, are adjusted.

[0085] In one implementation of the present application, during the playback of learning background music, the real-time learning state is compared with preset state transition conditions. If a user state transition is detected, a new third music set is generated. Based on the transitioned user state, the playback volume is determined, and the player volume is automatically adjusted. The learning background music is then played based on the newly generated third music set and the adjusted volume.

[0086] Specifically, while background music is playing, the user's real-time learning status is continuously captured and compared against pre-set state transition conditions. These transition conditions include trigger thresholds for transitioning from focused to distracted, and criteria for transitioning from alert to fatigue. This comparison determines whether the user's current state has transitioned. If so, the subsequent music adjustment process is initiated; if not, the current playback state is maintained.

[0087] When a user's state transition is detected, a third music collection is generated based on the new state. For example, when the user transitions from distraction to focus, soothing music that can reinforce focus is selected; when the user transitions from fatigue to alertness, music with a light but not overly stimulating rhythm is retained.

[0088] Furthermore, the playback volume is determined based on the converted user state and pre-set volume adjustment rules. For example, if the user is distracted and needs to improve their attention, the volume can be set slightly higher than normal to moderately stimulate their hearing. If the user is tired and needs to relieve fatigue, the volume can be adjusted slightly lower than normal to avoid excessive volume that aggravates fatigue.

[0089] Based on the determined playback volume, the background music player is automatically adjusted to ensure a smooth transition to the target volume, avoiding sudden volume changes that disrupt the user. Simultaneously, music adapted to the current session is selected from the newly generated third music collection and replaced with the corresponding tracks in the original playlist, ensuring that music switching and volume adjustment occur synchronously.

[0090] The tuned music in the embodiment of the present application is played to the user through the built-in micro-speaker of the smart glasses or headphones connected via Bluetooth. The system continuously monitors the user's status, forming a closed-loop feedback loop to continuously optimize the music tuning effect. All tuning processes strive to be smooth and seamless, avoiding abrupt music switching that disrupts the user. The system can also record the effect of each music adjustment and incorporate it into the training data of the machine learning model, further improving the accuracy and intelligence of the adaptive tuning.

[0091] Figure 8 A music adaptive adjustment flow chart provided in the embodiment of the present application is as follows: Figure 8 As shown, the embodiment of the present application collects learning scene information through the visual recognition technology of smart glasses, and matches the corresponding learning task type by performing feature recognition on the learning scene. Afterwards, the duration of the current learning task and the difficulty level of the learning task are determined by collecting and analyzing the learning content, the user's learning posture, and the physiological characteristic parameters. Secondly, by obtaining the user's behavioral information during learning, the user's real-time status is detected, and the learning background music is adjusted based on the real-time status. For example, if the user is in a distracted state, the system automatically switches to classical or light music; if the user is in a fatigued state, the system automatically switches to light pop music; if the user is in a normal learning state, the background music matched by the learning task type, duration, and difficulty level continues to be used. Finally, through music playback and feedback, the music adjustment effect is continuously optimized.

[0092] Figure 9 This is a structural diagram of a music adaptive adjustment device provided in an embodiment of the present application. Figure 9 As shown, the music adaptive adjustment device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute any of the above-mentioned music adaptive adjustment methods.

[0093] The various embodiments in this application are described in a progressive manner. Similar portions between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from the other embodiments. In particular, the device, apparatus, and non-volatile computer storage medium embodiments are generally similar to the method embodiments, so their descriptions are relatively simple. For relevant portions, refer to the descriptions of the method embodiments.

[0094] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. It will be apparent to those skilled in the art that various modifications and variations may be made to the embodiments of the present application. However, such modifications or substitutions do not deviate from the spirit and scope of the technical solutions of the various embodiments of the present application.

Claims

1. A music adaptive adjustment method, characterized in that: The method comprises: Performing task matching and task analysis based on the learning scenario information acquired by the smart glasses to obtain learning task information; wherein the learning task information includes at least one of the following: learning task type, learning task duration, and learning task difficulty level; Analyzing and detecting the real-time learning behavior information acquired by the smart glasses to obtain a real-time learning status; Based on the learning task information and / or the real-time learning status, corresponding background music is matched in a music database and adaptive adjustment of the background music is performed.

2. The method according to claim 1, characterized in that The step of performing task matching and task analysis based on the learning scenario information acquired by the smart glasses to obtain learning task information specifically includes, when determining the type of the learning task: Performing image recognition processing on the image data in the learning scene information to extract visual feature information; wherein the visual feature information includes at least one of the following: learning tools, learning materials, user learning posture, and text data generated by the user during learning; Performing text recognition on the text data in the learning scenario information by using natural language processing technology to extract text feature information; fusing the visual feature information with the text feature information to obtain key learning information; Matching the key learning information in a preset learning task database to determine the type of the learning task; The learning task types include at least one of the following: in-depth reading, language learning, math problem solving, and writing.

3. The method according to claim 2, characterized in that The step of performing task matching and task analysis based on the learning scenario information acquired by the smart glasses to obtain learning task information specifically includes, when determining the duration of the learning task: Determine the duration of the first reference learning task based on the amount of text content in the learning materials; Obtain the second reference learning task duration preset by the user; In the historical database, determining historical learning data whose similarity value with the key learning information meets the similarity condition; The key learning information, the first reference learning task duration, the second reference learning task duration and the historical learning data are input into a preset duration prediction model to output the learning task duration corresponding to the current learning task.

4. The method according to claim 1, characterized in that The performing of task matching and task analysis based on the learning scenario information acquired by the smart glasses to obtain learning task information specifically includes, when determining the difficulty level of the learning task: Determine the level of complexity of the learning task based on the type of text content in the learning materials; Determining historical learning data corresponding to the complexity level of the learning task in a historical database; Based on the completion result data corresponding to the historical learning data, construct a learning profile and a learning curve corresponding to the user at the complexity level of the learning task, so as to determine the difficulty level of the learning task according to the user profile and the learning curve; The completion result data includes at least the task completion rate, the task error rate and the thinking time.

5. The method according to claim 1, characterized in that: The analyzing and detecting the real-time learning behavior information acquired by the smart glasses to obtain the real-time learning status specifically includes: Acquiring real-time learning behavior information sent by different information collection devices on the smart glasses; wherein the real-time learning behavior information includes at least one of the following: eye tracking information, head posture information, facial expression information, and physiological characteristic parameters; Comparing different types of real-time learning behavior information with corresponding state thresholds, and determining state coefficient values ​​corresponding to each type of real-time learning behavior information based on comparison differences; Obtaining the real-time learning state based on preset weight distribution parameters, the real-time learning behavior information, and the state coefficient value; The real-time learning state includes a distraction state and a fatigue state.

6. The method according to claim 5, characterized in that The obtaining of the real-time learning state based on the preset weight distribution parameter, the real-time learning behavior information, and the state coefficient value specifically includes: Filtering the real-time learning behavior information of the type required for distraction detection from the real-time learning behavior information to construct a distraction detection information set; Based on the weight distribution parameter corresponding to the distraction detection information set and the state coefficient value, weighting each piece of real-time learning behavior information in the distraction detection information set is performed to obtain the user distraction state; Filtering out the real-time learning behavior information of the type required for fatigue detection from the real-time learning behavior information, and constructing a fatigue detection information set; Based on the weight distribution parameter corresponding to the fatigue detection information set and the state coefficient value, each piece of real-time learning behavior information in the fatigue detection information set is weighted to obtain the user fatigue state.

7. The method according to claim 1, characterized in that: The matching of corresponding background music in a music database based on the learning task information and / or the real-time learning status and performing adaptive adjustment of the background music specifically includes: determining a first music collection in the music database based on the learning task type; constructing a music characteristic trend curve based on the learning task duration and the learning task difficulty level, and determining a second music set in the music database based on the music characteristic trend curve; wherein the music characteristic trend curve is divided into a plurality of playback stages based on changes in music characteristics; In the plurality of playing stages, a learning background music list is determined based on the intersection result of the first music set and the second music set, and the list is played; determining a third music set in the music data set based on the real-time learning state; The learning background music list is adjusted based on the third music set.

8. The method according to claim 7, characterized in that: Before determining the learning background music list and playing it, the method further includes: In the case where the first music set and the second music set do not have an intersection, inputting the learning task information into a preset music screening model and outputting a reference music list; Determining the historical preference score corresponding to each piece of music in the reference music list based on the user's historical music playing frequency, time decay factor, and scene relevance; The reference music list is sorted based on the historical preference scores to construct the learning background music list.

9. The method according to claim 7, characterized in that: After adjusting the learning background music list based on the third music set, the method further includes: During the learning background music playback, the real-time learning state obtained is compared with a preset state transition condition, and when a user state transition is detected, the third music set is regenerated; Determining a playback volume based on the converted user state and automatically adjusting the volume of the player; Based on the regenerated third music set and the adjusted volume, background music playback is learned.

10. A music adaptive adjustment device, characterized in that: The device comprises: at least one processor; and, a memory communicatively coupled to the at least one processor; Wherein, the memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can: execute the music adaptive adjustment method described in any one of claims 1-9 above.

Images