Multi-scene switching method, system and equipment based on song ordering table and medium

Abstract

The invention belongs to the field of man-machine interaction, and particularly discloses a multi-scene switching method, system, equipment and medium based on a song ordering table, and the method comprises the steps: collecting the behavior data of residence time, song ordering frequency and interaction intensity of a user in a virtual environment through a sensor, and processing the behavior data through a long and short term memory network, obtaining a user dynamic preference vector; matching is carried out according to the user dynamic preference vector and a preset scene type library, if the matching degree is higher than a threshold value, it is judged that the current scene preference is stable, otherwise, it is judged that potential switching intentions exist, and potential switching intention probability distribution is obtained through the judgment; after the probability distribution of the potential switching intention is obtained, a support vector machine classifier is adopted to train association features among multi-scene data; the invention aims to solve the problems of resource loading delay and unsmooth interface switching caused by rapid change of user scene preference in a virtual environment in the prior art.

Description

Technical Field

[0001] This invention relates to the field of human-computer interaction technology, specifically to a multi-scene switching method, system, device, and medium based on a karaoke system. Background Technology

[0002] As a popular form of online entertainment, song request platforms attract a large number of users to interact with music through virtual scenes. Their core function is to provide users with an immersive and personalized experience. Researching how to optimize scene transitions in song request platforms using user behavior prediction technology is crucial not only for the smoothness of the user experience but also for the platform's resource allocation efficiency and operational stability. With the growth of the user base and the increasing diversity of scenes, accurately predicting user behavior across different scenarios has become a key area for improving platform service quality and user satisfaction. The importance of this area lies in its ability to capture dynamic user needs through technological means, thereby enabling intelligent resource allocation and seamless interface transitions, resulting in a more natural interactive experience. However, existing methods often face the challenge of capturing dynamic user preferences when dealing with user behavior prediction. Many solutions rely on simple user profiles or static historical data analysis, ignoring the variability and real-time nature of user behavior in song request platform scenarios. For example, users may temporarily change their preferences due to current emotions, social interactions, or the atmosphere of the scene, and existing methods struggle to adapt quickly to such dynamic changes. Furthermore, while some solutions attempt to utilize real-time data, they are often limited to the analysis of a single scenario, lacking a holistic understanding of user behavior when switching between different scenarios. This makes it difficult for the system to accurately predict the next scenario a user might enter, thus affecting the efficiency of resource allocation and the consistency of the user experience. In the karaoke system scenario, the core technical challenge lies first and foremost in how to accurately quantify the user's scenario preferences. Summary of the Invention

[0003] This invention provides a multi-scene switching method, system, device, and medium based on a karaoke system, aiming to solve the problems in the prior art that cause resource loading delays and unsmooth interface switching due to rapid changes in user scene preferences in virtual environments.

[0004] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows: A multi-scene switching method based on a karaoke system includes: collecting user behavior data such as dwell time, song selection frequency, and interaction intensity in a virtual environment using sensors; processing the behavior data using a long short-term memory network to obtain a user dynamic preference vector; matching the user dynamic preference vector with a preset scene type library; if the matching degree is higher than a threshold, it is determined that the current scene preference is stable; otherwise, it is determined that a potential switching intention exists, and a probability distribution of the potential switching intention is obtained through the above judgment; after obtaining the probability distribution of the potential switching intention, a support vector machine classifier is used to train the correlation features between multi-scene data to obtain a list of possible target scenes for cross-scene switching; extracting the target scene with the highest probability from the list of possible target scenes, and combining it with real-time monitoring data... The system judges the trend of behavioral data changes. If the trend points to the chorus area, it determines that the user is about to switch to the chorus area; otherwise, it continues monitoring to obtain an accurate prediction of the switching demand. Based on the accurate prediction of the switching demand, the resource files and interface elements of the target scene are preloaded, and the server bandwidth allocation ratio is adjusted using a dynamic allocation algorithm to obtain an optimized resource configuration scheme. The interface layout parameters are updated using the optimized resource configuration scheme. If the prediction result matches the actual user operation, the layout parameters are directly applied; otherwise, the system rolls back to the original configuration to determine the final seamless interface state. Feedback data is extracted from the final seamless interface state and input into the iterative training process of the Long Short-Term Memory network to update the model parameters to adapt to the user behavior pattern.

[0005] In one aspect of this disclosure, the step of collecting user behavior data such as dwell time, song selection frequency, and interaction intensity in a virtual environment through sensors, and processing the behavior data using a long short-term memory network to obtain a user dynamic preference vector includes: User behavior data, including dwell time, song selection frequency, and interaction intensity, is collected from the virtual environment using sensors and stored as structured time-series data to obtain an initial behavior dataset. Data cleaning techniques were used to preprocess the initial behavioral dataset, removing noisy data and missing values. The dwell time, song request frequency, and interaction intensity were aligned using timestamps to obtain a standardized behavioral sequence. Based on the Long Short-Term Memory Network, time-series modeling of standardized behavioral sequences is performed to extract time-dependent features and generate user behavior pattern vectors. Based on user behavior pattern vectors, a clustering algorithm is used to group users, and user preference categories are obtained based on the similarity of behavior pattern vectors. If the user's preference category matches the preset music recommendation rules, the corresponding song list is retrieved from the music library based on the category characteristics to generate a personalized recommendation sequence; By comparing the personalized recommendation sequence with the frequency of user song requests, the matching degree between the recommendation sequence and the actual behavior is calculated, and the recommendation effect evaluation index is obtained. Based on the recommendation performance evaluation metrics, the gradient descent algorithm is used to optimize the parameters of the long short-term memory network and update the user behavior pattern vector.

[0006] In one aspect of this disclosure, the step of matching the user's dynamic preference vector with a preset scene type library, determining that the current scene preference is stable if the matching degree is higher than a threshold, and otherwise determining that a potential switching intention exists, and obtaining the probability distribution of the potential switching intention through the above determination, includes: Obtain user dynamic preference vectors and generate feature vectors by analyzing user historical interaction data; Based on the vector matching algorithm, the user's dynamic preference vector is compared with the standard vector in the preset scene type library to obtain the matching degree score; If the matching score is higher than a preset threshold, the current scene preference is determined to be stable, and a stable scene label is generated. Based on the stable scene tags, extract recommended content related to the current scene and generate a personalized recommendation list; If the matching score is lower than a preset threshold, the potential switching intent is analyzed using a Bayesian inference algorithm to obtain the intent probability distribution. Based on the intent probability distribution, a weighted sorting algorithm is used to generate a recommendation sequence of potential switching scenarios; Based on the recommended sequence, the user interface is dynamically adjusted, and optimized scene recommendation content is output.

[0007] In one aspect of this disclosure, the step of obtaining a list of possible target scenarios for cross-scenario switching by training a support vector machine classifier on the correlation features between multi-scenario data after obtaining the probability distribution of the potential switching intention includes: Acquire user interaction data and extract the probability distribution of potential switching intentions from it; By analyzing probability distributions, a support vector machine classifier is used to train the associated features of multi-scene data to obtain a classification model. If the prediction accuracy of the classification model reaches a preset threshold, the target scene list for cross-scene switching is determined based on the model output. Based on the list of target scenarios, extract the associated feature set for each scenario and generate a feature vector; Feature vectors are used to group target scenes using a clustering algorithm to obtain scene grouping results; By analyzing the scene grouping results, we can obtain the priority ranking of each group and get the final scene switching sequence. If the priority of the final scene switching sequence meets the preset conditions, the scene sequence is output; otherwise, the feature vectors are extracted again for clustering.

[0008] In one aspect of this disclosure, the step of extracting the highest probability target scenario from the list of possible target scenarios, combining it with the trend of changes in real-time monitored behavioral data, and determining that a switch to the chorus area is imminent if the trend points to the chorus area, otherwise continuing monitoring to obtain an accurate prediction result of the switching demand includes: The target scene is obtained from the scene list, and the highest probability target scene is extracted using a probability sorting algorithm to obtain the highest probability target scene identifier; Behavioral data is collected through a real-time monitoring system, and the changing trend of the behavioral data is calculated using time series analysis methods to obtain a behavioral data changing trend vector. If the trend vector of behavioral data changes points to the preset features of the chorus area, then the target scene is determined to be the chorus area, and a scene switching signal is generated. Based on the scene switching signal, obtain the relevant configuration parameters of the chorus area to determine the switching requirements; If the behavioral data change trend vector does not point to the preset features of the chorus area, the behavioral data will be updated through the continuous monitoring module to obtain a new behavioral data change trend vector. The random forest algorithm is used to classify the trend vector of new behavioral data to determine whether the chorus area switching condition is met, and the classification result is obtained. Based on the classification results, the final switching demand prediction results are determined, and the target scenario switching instructions are generated.

[0009] In one aspect of this disclosure, the step of preloading resource files and interface elements of the target scenario based on the precise switching demand prediction result, and adjusting the server bandwidth allocation ratio using a dynamic allocation algorithm to obtain an optimized resource configuration scheme includes: By analyzing user behavior data through predictive models, we can obtain prediction results of switching needs. Based on the prediction results, the priority and resource requirements of the target scenario are determined, and a resource loading sequence is generated. Based on the resource loading sequence, preload the resource files and UI elements of the target scene and store them in the local cache; A consistent hashing algorithm is used to dynamically allocate server bandwidth and adjust the bandwidth ratio for different scenarios. If the loading speed of a certain scenario is lower than the preset threshold after the bandwidth ratio is adjusted, the bandwidth will be redistributed through the load balancing mechanism to obtain an optimized bandwidth allocation scheme. For the optimized bandwidth allocation scheme, monitor the resource loading status in real time and obtain loading completion data; Based on the loading completion data, the timing of the preloading mechanism is dynamically adjusted to generate the final resource configuration scheme.

[0010] In one aspect of this disclosure, the step of updating the interface layout parameters through the optimized resource configuration scheme, and directly applying the layout parameters if the prediction result matches the actual user operation, otherwise rolling back to the original configuration, to determine the final seamless interface state, includes: By analyzing the resource allocation scheme, the currently available computing and storage resources are obtained from the preset resource pool to determine the real-time allocation status of the resource allocation scheme. Based on the real-time allocation status of the resource allocation scheme, the corresponding interface layout parameters are generated using a preset interface rendering algorithm to obtain the initially adjusted layout parameters. The layout parameters that have been initially adjusted are predicted by a machine learning classification model. The prediction results are then used to determine whether the prediction results match the expected user operation pattern. If the prediction results are consistent with the actual user operation, the initially adjusted layout parameters are deployed to the interface rendering module through parameter application logic to determine the update status of the interface layout. If the prediction result is inconsistent with the actual user operation, the original layout parameters are obtained from the original layout parameter repository, and the rollback operation mechanism is used to restore the interface to the initial state to obtain the rolled-back interface state. Based on the rolled-back interface state or updated state, a state synchronization algorithm is used to detect the latency and consistency during the interface switching process to determine the seamless switching state. By recording seamless state transitions and user operation data, the training dataset for the dynamic adjustment mechanism is updated to obtain an optimized resource configuration scheme and input parameters for the next adjustment.

[0011] In another aspect, this disclosure also relates to a multi-scene switching system based on a karaoke system, characterized in that it includes: The behavior data processing module is configured to collect user behavior data such as dwell time, song selection frequency, and interaction intensity in the virtual environment through sensors, process the behavior data using a long short-term memory network, and obtain a user dynamic preference vector. The scene preference judgment module is configured to match the user's dynamic preference vector with a preset scene type library. If the matching degree is higher than a threshold, it is judged that the current scene preference is stable; otherwise, it is judged that a potential switching intention exists. The module obtains the probability distribution of potential switching intentions through the above judgment. The associated feature training module is configured to, after obtaining the probability distribution of the potential switching intention, use a support vector machine classifier to train the associated features between multi-scene data to obtain a list of possible target scenes for cross-scene switching. The switching demand prediction module is configured to extract the target scenario with the highest probability from the list of possible target scenarios, and judge the trend of changes in real-time monitored behavioral data. If the trend of changes points to the chorus area, it is determined that the switch is about to be made to the chorus area; otherwise, monitoring continues to obtain an accurate switching demand prediction result. The resource configuration optimization module is configured to preload the resource files and interface elements of the target scenario based on the exact switching demand prediction results, and adjust the server bandwidth allocation ratio using a dynamic allocation algorithm to obtain an optimized resource configuration scheme. The interface connection determination module is configured to update the interface layout parameters through the optimized resource configuration scheme, and if the prediction result matches the actual user operation, directly apply the layout parameters; otherwise, roll back to the original configuration to determine the final seamless interface connection state. The model iteration update module is configured to extract feedback data in the final seamless interface state, input it into the iterative training process of the long short-term memory network, and update the model parameters to adapt to user behavior patterns.

[0012] In another aspect of this disclosure, there is also a karaoke device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the above-described multi-scene switching method based on a karaoke station.

[0013] In another aspect, this disclosure also relates to a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described multi-scene switching method based on a karaoke system.

[0014] Compared with the prior art, the present invention has the following beneficial effects: This invention achieves seamless scene switching by integrating behavioral data analysis, dynamic preference modeling, and resource optimization. First, it utilizes a Long Short-Term Memory (LSTM) network to process user dwell time, song selection frequency, and interaction intensity data, generating a dynamic preference vector. This vector is then matched against a scene type library to determine preference stability and potential switching intentions. Next, a Support Vector Machine (SVM) classifier is used to train cross-scene association features, identifying a list of possible target scenes. Specific switching needs are predicted based on real-time behavioral data trends. For the prediction results, this invention preloads target scene resources, employs a dynamic allocation algorithm to optimize server bandwidth, and finally updates interface layout parameters, verifying or rolling back the configuration based on actual operation. By iteratively updating model parameters through feedback data, this invention significantly improves the smoothness of scene switching and resource utilization efficiency, achieving a personalized, real-time responsive virtual environment experience. Attached Figure Description

[0015] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0016] Figure 1 This is a flowchart of the multi-scene switching method for the song selection system in this invention.

[0017] Figure 2 This is a schematic diagram of the multi-scene switching method for the song selection system in this invention. Detailed Implementation

[0018] The present invention will be further described below with reference to embodiments. These embodiments are merely some, not all, of the embodiments described. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the protection scope of the present invention.

[0019] Example 1 Please see Figure 1 and Figure 2As shown, this embodiment discloses a multi-scene switching method based on a karaoke system, including: collecting user behavior data such as dwell time, song selection frequency, and interaction intensity in a virtual environment through sensors; processing the behavior data using a long short-term memory network to obtain a user dynamic preference vector; matching the user dynamic preference vector with a preset scene type library; if the matching degree is higher than a threshold, it is determined that the current scene preference is stable; otherwise, it is determined that a potential switching intention exists, and a probability distribution of the potential switching intention is obtained through the above judgment; after obtaining the probability distribution of the potential switching intention, a support vector machine classifier is used to train the correlation features between multi-scene data to obtain a list of possible target scenes for cross-scene switching; extracting the target scene with the highest probability from the list of possible target scenes, and combining it with the actual... The system monitors behavioral data changes and trends. If the trend points to the chorus area, it determines that the user is about to switch to the chorus area; otherwise, monitoring continues to obtain a precise prediction of the switching demand. Based on this prediction, the system preloads the resource files and interface elements of the target scene and uses a dynamic allocation algorithm to adjust the server bandwidth allocation ratio to obtain an optimized resource configuration scheme. The system updates the interface layout parameters using the optimized resource configuration scheme. If the prediction results match the actual user operation, the layout parameters are applied directly; otherwise, the system rolls back to the original configuration to determine the final seamless interface state. Feedback data is extracted from the final seamless interface state and input into the iterative training process of the Long Short-Term Memory network to update the model parameters to adapt to user behavior patterns.

[0020] This invention achieves seamless scene switching by integrating behavioral data analysis, dynamic preference modeling, and resource optimization. First, it utilizes a Long Short-Term Memory (LSTM) network to process user dwell time, song selection frequency, and interaction intensity data, generating a dynamic preference vector. This vector is then matched against a scene type library to determine preference stability and potential switching intentions. Next, a Support Vector Machine (SVM) classifier is used to train cross-scene association features, identifying a list of possible target scenes. Specific switching needs are predicted based on real-time behavioral data trends. For the prediction results, this invention preloads target scene resources, employs a dynamic allocation algorithm to optimize server bandwidth, and finally updates interface layout parameters, verifying or rolling back the configuration based on actual operation. By iteratively updating model parameters through feedback data, this invention significantly improves the smoothness of scene switching and resource utilization efficiency, achieving a personalized, real-time responsive virtual environment experience.

[0021] Example 2 Please see Figure 1 and Figure 2 As shown, this embodiment is a further optimization based on embodiment one. In this embodiment, a multi-scene switching method based on a karaoke system is disclosed, which may specifically include: This invention provides a personalized music recommendation and scene switching method based on user behavior data. It aims to generate dynamic preferences and achieve accurate prediction and optimized resource allocation for scene switching by analyzing user behavior data in a virtual environment. The method is described in detail below with reference to specific implementation methods.

[0022] Step S1: Collect user behavior data such as dwell time, song selection frequency, and interaction intensity in the virtual environment through sensors to generate an initial behavior dataset.

[0023] Specifically, a virtual environment can be an interactive scenario based on virtual reality (VR) or augmented reality (AR) devices. For example, a user enters a virtual karaoke room through a head-mounted display (HUD) to select songs and interact with other users. Sensors, including but not limited to accelerometers, gyroscopes, eye trackers, or microphone arrays in the HUD, are used to capture real-time behavioral data of the user in the virtual environment. Dwell time refers to the duration a user remains in a specific virtual scene, such as a song selection interface or a virtual stage, typically recorded in seconds. Song selection frequency refers to the number of times a user selects a song per unit of time, such as the number of song selections per minute. Interaction intensity is measured by the frequency of the user's gestures, voice commands, or the use of virtual props, such as the number of interactions between the user and virtual characters in a virtual karaoke room or changes in the volume of voice commands.

[0024] In one possible implementation, the sensor collects data at a fixed frequency, such as 10 times per second, and stores the collected data as a structured time series with timestamps. For example, if a user stays in a virtual karaoke room for 5 minutes, selects songs 3 times, and interacts with the virtual scene 10 times via gestures, the sensor records these data in chronological order as an initial behavior dataset. It should be noted that the initial behavior dataset may contain noise, such as abnormal gesture data caused by device jitter or timestamp discrepancies due to network latency. To ensure data quality, filtering techniques can be used during the acquisition process to preliminarily process the sensor data, removing obvious outliers and generating a more reliable initial behavior dataset.

[0025] Step S11: Preprocess the initial behavior dataset to remove noisy data and missing values ​​to obtain a standardized behavior sequence.

[0026] Specifically, the initial behavioral dataset may contain noisy data due to equipment malfunctions or user errors, such as a song request record not being completed due to network interruption, or a period of dwell time being empty. The preprocessing process first identifies empty or incomplete records in the data using a missing value detection algorithm, such as checking whether timestamps are continuous. If a period of timestamps is found to be missing, the data is filled in using interpolation methods, such as filling in the missing dwell time based on the average of the preceding and following timestamps.

[0027] Noise removal can be achieved through statistical methods. For example, for song request frequency, the mean and standard deviation within a time window can be calculated. If the frequency of a song request exceeds the mean plus or minus three times the standard deviation, it is considered noise and removed. Furthermore, to ensure the comparability of different types of data, dwell time, song request frequency, and interaction intensity need to be normalized, mapping them to the same numerical range, such as 0 to 1.

[0028] For example, dwell time can be converted into a relative duration ratio through linear normalization, song request frequency can be normalized based on the maximum frequency, and interaction intensity is standardized based on the intensity value of gestures or voice. After preprocessing, a timestamp-aligned and numerically normalized behavior sequence is obtained, containing dwell time, song request frequency, and interaction intensity data arranged in chronological order.

[0029] Step S12: Perform time-series modeling on the standardized behavior sequence using a long short-term memory network, extract time-dependent features, and generate user behavior pattern vectors.

[0030] Long Short-Term Memory (LSTM) networks are a type of neural network suitable for processing time-series data, capable of capturing long-term dependencies in behavioral sequences. For example, in a virtual karaoke room, a user's song selection frequency might initially be low and then increase due to emotional changes; this temporal pattern can be modeled using an LTM network.

[0031] In one embodiment, a standardized sequence of behaviors is input into a Long Short-Term Memory (LSTM) network at time steps, with each time step containing a three-dimensional vector of dwell time, song selection frequency, and interaction intensity. The network uses forget gates, input gates, and output gates to determine which historical information to retain and which to update the current state. For example, the forget gate filters out irrelevant data from earlier time steps, such as random actions when the user initially enters the virtual environment; the input gate focuses on retaining recent high-frequency song selection behaviors; and the output gate generates the behavioral feature vector for the current time step.

[0032] It should be noted that the number of hidden layer nodes in the Long Short-Term Memory (LSTM) network can be adjusted according to the complexity of the virtual environment. For example, in a simple virtual karaoke scene, fewer hidden nodes, such as 64, can be set to improve computational efficiency; in a complex multi-scene virtual environment, the number of nodes can be increased, such as 256, to capture richer temporal features. After processing through multiple layers of the network, a fixed-dimensional user behavior pattern vector is generated, such as a 128-dimensional vector, representing the user's behavior patterns in the virtual environment.

[0033] For example, if a user repeatedly selects fast-paced songs in a virtual karaoke room and interacts with high intensity, the network can capture this pattern of preference for fast-paced music and generate a corresponding behavioral pattern vector. This vector not only reflects the user's current behavior but can also predict their potential preference trends through time-series modeling.

[0034] Step S2: Based on the user behavior pattern vector, a clustering algorithm is used to group users and determine user preference categories.

[0035] Specifically, clustering algorithms are used to group users with similar behavioral patterns into the same category, enabling personalized recommendations for users with different preferences. Clustering algorithms can be distance-based K-means or density-based DBSCAN.

[0036] In one possible implementation, the K-means algorithm is used to group user behavior pattern vectors. First, the number of clusters K is determined. For example, based on the user behavior characteristics of a virtual karaoke room, K is preset to 4, representing four types of users: those with a fast-paced preference, those with a lyrical preference, those who prefer random song selection, and those with low interaction.

[0037] The algorithm assigns users to the nearest cluster center by calculating the Euclidean distance between behavioral pattern vectors.

[0038] For example, if a user's behavior pattern vector shows that they frequently request songs and prefer fast-paced music, the algorithm might categorize them as a "fast-paced preference" user. Conversely, if another user spends a long time on the platform but requests songs infrequently and has weak interaction, they might be categorized as a "low-interaction" user. It's worth noting that the initialization of cluster centers can be optimized by random selection or based on prior knowledge of historical data to improve grouping accuracy.

[0039] In another embodiment, if user behavior in a virtual KTV scene is complex, the DBSCAN algorithm can be used to automatically identify the density distribution of behavior pattern vectors and generate a dynamic number of cluster categories. This method is suitable for processing irregularly distributed user behavior data, such as when some users concentrate on requesting songs during a specific time period.

[0040] Ultimately, the clustering results generate user preference category labels, such as fast-paced preferences or lyrical preferences, providing a basis for subsequent recommendations.

[0041] Step S21: If the user's preference category matches the preset music recommendation rules, then retrieve the corresponding song list from the music library and generate a personalized recommendation sequence.

[0042] Specifically, music recommendation rules are predefined mappings between categories and song features. For example, a fast-paced preference category corresponds to songs with a tempo exceeding 120 beats per minute, while a lyrical preference category corresponds to songs with a slower tempo and emotionally rich lyrics. Each song in the music library is pre-labeled with feature tags, such as tempo, style, and emotional attributes.

[0043] In one embodiment, if a user is classified as having a fast-paced preference, the system selects songs with a tempo of 120 beats per minute or higher from the music library according to music recommendation rules, and generates a recommendation sequence containing 10 songs. The recommendation sequence is sorted by the popularity of the songs or the user's historical song request records to improve the acceptance of the recommendations.

[0044] For example, the system can prioritize recommending singers or songs of similar style that the user has recently requested.

[0045] It should be noted that the music recommendation rules can be dynamically adjusted according to the virtual karaoke scenario. For example, in a group singing scenario, the rules can prioritize recommending popular songs suitable for a duet, while in a solo song selection scenario, more emphasis is placed on the user's personal preferences. The generation of personalized recommendation sequences enhances the user's immersion and satisfaction in the virtual environment.

[0046] Step S22: By comparing the personalized recommendation sequence with the frequency of user song requests, the matching degree between the recommendation sequence and the actual behavior is calculated to obtain the recommendation effect evaluation index.

[0047] Specifically, the matching degree is calculated by comparing the similarity between the songs in the recommended sequence and the user's actual song request records. For example, the system counts the overlap ratio between the features of the songs in the recommended sequence, such as rhythm and style, and the features of the user's recent song requests.

[0048] In one possible implementation, the matching degree is calculated using cosine similarity. The feature vectors of the songs in the recommended sequence are compared with the feature vectors of the user's recent song requests to generate a similarity score. For example, if the recommended sequence contains 5 fast-paced songs and the user's recent song requests contain 4 fast-paced songs, the matching degree is high. The matching degree score is further converted into a recommendation performance evaluation metric, such as a standardized score from 0 to 1, where a higher score indicates that the recommendation is more in line with the user's preferences.

[0049] In another embodiment, if the virtual KTV scene supports multi-person interaction, a weighted matching degree calculation can be introduced to comprehensively consider the user's personal song selection records and interaction behavior with virtual characters.

[0050] For example, if a user prefers a certain style when singing with a virtual character, the system increases the weight of that style, generating more accurate evaluation metrics. This method effectively improves the adaptability of recommendation systems in complex scenarios.

[0051] Step S23: Based on the recommendation effect evaluation index, the gradient descent algorithm is used to optimize the parameters of the long short-term memory network and update the user behavior pattern vector.

[0052] Specifically, the recommendation performance evaluation metric reflects the accuracy of the recommendation sequence. A low metric indicates that the behavioral pattern vectors generated by the Long Short-Term Memory network have failed to accurately capture user preferences. The gradient descent algorithm optimizes network parameters by minimizing a loss function, such as mean squared error.

[0053] In one embodiment, the system uses a recommendation performance evaluation metric as input to the loss function to calculate the deviation between the user behavior pattern vector predicted by the network and the actual behavior. The gradient descent algorithm adjusts the network's weights and biases, such as updating the parameters of the forget gate and input gate, making the network more inclined to retain behavioral features related to highly relevant recommendations.

[0054] For example, a user may recently prefer ballads, but the network-generated recommendation sequence favors fast-paced songs, resulting in a low match rate. The gradient descent algorithm, through multiple iterations, adjusts the network parameters to more accurately capture the user's preference for ballads and generates updated behavioral pattern vectors.

[0055] In another possible implementation, an adaptive learning rate adjustment mechanism can be introduced. For example, when the loss function decreases slowly, the learning rate can be dynamically reduced to improve optimization accuracy. This approach is particularly effective in the dynamic environment of a virtual karaoke room, as it can quickly adapt to changes in user behavior.

[0056] Step S3: Match the user's dynamic preference vector with the preset scene type library to determine whether the current scene preference is stable or whether there is a potential switching intention, and obtain the probability distribution of potential switching intention.

[0057] Specifically, the user dynamic preference vector is generated by a long short-term memory network, representing the user's behavioral tendencies in the current virtual environment. The scene type library contains standard vectors for various virtual KTV scenes, such as single-person song selection, multi-person chorus, and virtual stage performance. Each scene predefines typical behavioral characteristics, such as high song selection frequency and low interaction intensity.

[0058] In one embodiment, the system calculates the similarity between the user's dynamic preference vector and each standard vector in the scene type library using a vector matching algorithm. For example, using cosine similarity calculation, if the similarity between the user's dynamic preference vector and the standard vector of the multi-person chorus scene is higher than a preset threshold of 0.8, the current scene preference is determined to be stable, and a multi-person chorus scene label is generated. If the similarity is lower than the threshold, it indicates that the user may intend to switch to other scenes.

[0059] In another possible implementation, the system analyzes potential switching intentions using a Bayesian inference algorithm. For example, based on the user's historical interaction data, it calculates the conditional probability of the user switching from the current scene to another scene, generating an intention probability distribution. For instance, the probability distribution might show that the user has a 60% probability of switching to the virtual stage performance scene and a 30% probability of remaining in the multi-person chorus scene.

[0060] It's worth noting that the scene type library can be dynamically expanded based on the business needs of the virtual KTV. For example, if a new virtual DJ mixing scene is added, the system will update the standard vectors and recalculate the matching degree. This flexibility ensures that the recommendation system can adapt to the needs of different user groups.

[0061] Thus far, the first part of the method has achieved personalized recommendations and preliminary scene preference judgment by collecting user behavior data, generating dynamic preference vectors, and performing scene matching, laying the foundation for subsequent scene switching and resource optimization.

[0062] Step S4: After obtaining the probability distribution of potential switching intentions, a support vector machine classifier is used to train the correlation features between multi-scene data to determine the list of possible target scenes for cross-scene switching.

[0063] Specifically, the probability distribution of potential switching intentions is generated in step S3, representing the likelihood of a user switching from the current virtual KTV scene to other scenes. A support vector machine classifier constructs a classification model by analyzing the correlation features between data from multiple scenes, used to predict possible target switching scenes.

[0064] For example, in a virtual karaoke room, when a user switches from a solo song selection scenario to a group singing scenario, the behavioral characteristics that may be involved include an increase in the frequency of song selection or an increase in the intensity of interaction.

[0065] In one possible implementation, the system extracts probability distributions of potential switching intentions from user interaction data, such as probability distribution vectors for three scenarios: solo song selection, group singing, and virtual stage performance. A support vector machine (SVM) classifier uses these probability distributions as input features, combined with behavioral features from historical interaction data, such as dwell time, song selection frequency, and interaction intensity, to train a classification model. During training, the classifier generates decision boundaries that can distinguish between different scenarios by maximizing the margins between features. For example, if a user spends a short time in the solo song selection scenario but increases their interaction intensity, the classifier might predict that they are inclined to switch to the group singing scenario.

[0066] It should be noted that the training data includes user behavior samples from multiple scenarios, such as thousands of user interaction records extracted from the logs of a virtual karaoke platform. The sample data is labeled by scenario to ensure the classifier can learn the feature distribution of different scenarios. After training, the classifier outputs a list of possible target scenarios, such as a list containing two high-probability scenarios: multi-person chorus and virtual stage performance. This method improves the accuracy of cross-scenario switching prediction and provides a reliable basis for subsequent scenario grouping and resource preloading.

[0067] Step S41: Through probability distribution analysis, a support vector machine classifier is used to train the associated features of multi-scene data to obtain a classification model.

[0068] Specifically, the probability distribution analysis is based on the intent probability distribution generated in step S3, combined with the correlation features of multi-scenario data, such as the changing patterns of users' song selection preferences, interaction frequency, or dwell time in different scenarios. The support vector machine classifier uses kernel functions, such as radial basis functions, to map nonlinear features to a high-dimensional space to generate a classification model.

[0069] In one embodiment, the system extracts feature data from the interaction logs of the virtual karaoke platform. For example, a user might request songs once per minute in a solo karaoke scenario, indicating low interaction intensity, while in a group singing scenario, the frequency increases to three times per minute, significantly increasing interaction intensity. The classifier analyzes these features, learns the differences between scenarios, and generates a decision boundary that can distinguish between solo song requests and group singing. During training, the system can adjust the classifier's hyperparameters, such as the regularization parameter C, through cross-validation to improve the model's generalization ability.

[0070] In another possible implementation, if there are many types of virtual KTV scenes, a multi-class support vector machine can be introduced to train each scene combination separately. For example, when switching from "single-person song selection" to "virtual stage performance," the classifier focuses on analyzing the frequency of virtual prop usage that users may increase in the performance scene. Ultimately, the classification model outputs the predicted probability for each scene, forming the basis for the target scene list.

[0071] Step S42: If the prediction accuracy of the classification model reaches a preset threshold, then determine the target scene list for cross-scene switching based on the model output.

[0072] Specifically, the prediction accuracy of the classification model is evaluated using a validation set, such as calculating the model's classification accuracy on test data. If the accuracy reaches a preset threshold, such as 0.85, the system generates a list of target scenes for cross-scene switching based on the probability distribution output by the model. For example, the list might include three scenes: multi-person chorus, virtual stage performance, and virtual DJ mixing, sorted from high to low predicted probability.

[0073] In one embodiment, the system evaluates the classification model in a test environment of a virtual KTV platform, using 1000 user interaction records as a validation set. If the model correctly predicts more than 85% of the scene switching instances, the model is considered reliable, and a list of target scenes is output.

[0074] For example, if a user is currently in a solo song request scenario, the model predicts that there is a 70% probability that they will switch to a multi-person duet and a 20% probability that they will switch to a virtual stage performance. The system generates a list containing both scenarios.

[0075] It should be noted that if the accuracy does not reach the threshold, the system can be retrained by increasing the training samples or adjusting the classifier parameters. For example, supplementing the data with more interactive data from virtual DJ mixing scenarios can improve the model's ability to predict new scenarios. This method ensures the accuracy of the target scenario list and provides reliable input for subsequent scenario grouping.

[0076] Step S43: Based on the target scene list, extract the associated feature set of each scene and generate a feature vector.

[0077] Specifically, the target scene list contains multiple possible switching scenes, each associated with a set of features, such as song request frequency, interaction intensity, or scene-specific virtual prop usage. The system extracts these features from the interaction data to generate a feature vector for each scene.

[0078] In one possible implementation, for a multi-person chorus scenario, the system extracts features including the number of times users and virtual characters sing together, the frequency of song requests, and the average dwell time, generating a multi-dimensional feature vector, for example, a vector containing 10 features. Similarly, the feature vector for a virtual stage performance scenario might include the frequency of virtual prop usage and the intensity of user gesture interactions during the performance; the feature vector is normalized to ensure that the feature values ​​for different scenarios are within the same numerical range, facilitating subsequent cluster analysis.

[0079] In another embodiment, if the virtual KTV platform supports dynamic scene expansion, the system can dynamically generate feature vectors for new scenes, such as virtual DJ mixing. For example, it can extract the frequency and duration of sound effects adjusted by the user in the mixing scene to generate corresponding feature vectors. This method improves the system's adaptability to new scenes.

[0080] Step S44: Use a clustering algorithm to group the target scenes and obtain the scene grouping results.

[0081] Specifically, clustering algorithms are used to group similar target scenes together to provide users with more accurate switching recommendations. Clustering algorithms can be K-means or hierarchical clustering, grouping scenes based on the similarity of their feature vectors.

[0082] In one embodiment, the system employs the K-means algorithm to divide the scenes in the target scene list into three groups, such as high-interaction scenes, low-interaction scenes, and performance-oriented scenes. During the grouping process, the algorithm calculates the Euclidean distance between the feature vectors of each scene and assigns similar scenes to the same group. For example, multi-person chorus and virtual DJ mixing scenes are grouped together due to their high interaction intensity, while single-person song selection scenes are grouped separately due to their low interaction intensity.

[0083] In another possible implementation, if the scene features are complex, a hierarchical clustering algorithm can be used to generate a tree-like grouping structure. For example, the system first divides all scenes into two main categories: interactive-oriented and non-interactive-oriented, and then further subdivides the interactive-oriented category into chorus-oriented and mixing-oriented. The grouping results provide a basis for subsequent scene priority ranking.

[0084] Step S45: Based on the scene grouping results, obtain the priority ranking of each group and determine the final scene switching sequence.

[0085] Specifically, the scenario grouping results reflect the similarity of different scenarios, and the system assigns priority to each group based on the user's current behavior and historical preferences.

[0086] For example, priorities can be determined based on users' recent song request frequency or interaction intensity preferences.

[0087] In one embodiment, if a user recently prefers highly interactive scenarios, the system prioritizes the groups containing multi-person chorus and virtual DJ mixing, generating a switching sequence that includes these two scenarios. For example, the sequence could be multi-person chorus to virtual DJ mixing and then to virtual stage performance. During the sorting process, the system can use a weighted algorithm to comprehensively consider the predicted probability of the scenario and user preferences, for example, assigning higher weights to highly interactive scenarios.

[0088] In another possible implementation, the system dynamically adjusts priorities based on the user's historical switching records. For example, if a user frequently switches from solo song selection to group singing, the priority of the group singing is increased, thus enhancing the personalization of the switching sequence.

[0089] Step S46: If the priority of the final scene sequence meets the preset conditions, the scene sequence is output; otherwise, the feature vector is re-extracted for clustering.

[0090] Specifically, the preset condition can be that the predicted probability of the highest priority scene in the switching sequence exceeds a threshold such as 0.7. If the condition is met, the system outputs the final switching scene sequence, such as from a multi-person chorus to a virtual DJ mix. If the condition is not met, it means that the current grouping result may be inaccurate, and the system returns to step S43 to re-extract the feature vector and perform clustering.

[0091] In one embodiment, the system checks that the predicted probability of the highest priority scenario of multi-person chorus in the switching sequence is 0.75, which meets the threshold requirement, and directly outputs the sequence; if the probability is only 0.6, the system re-extracts the feature vector, for example by adding recent interaction data, and re-performs clustering and sorting. This iterative mechanism ensures the reliability of the switching sequence.

[0092] Step S5: Extract the target scene with the highest probability from the list of possible target scenes, and combine it with the real-time monitored behavioral data change trend to determine whether to switch to the chorus area, so as to obtain an accurate prediction result of the switching demand.

[0093] Specifically, the highest probability target scenario is determined by the target scenario list generated in step S4. The system further confirms the switching demand by monitoring the changing trends of user behavior data in real time. For example, the chorus area is a highly interactive scenario, with typical characteristics including increased frequency of song requests and multi-person voice interaction.

[0094] In one possible implementation, the system employs a probabilistic ranking algorithm to extract the highest-probability scene from the target scene list. For example, the probability of a multi-person chorus scene is 0.8. The real-time monitoring system continuously collects user behavior data through sensors, such as recording the frequency of song requests and the intensity of interaction once per second, generating a trend vector of behavioral data changes. If the trend vector shows a significant increase in the frequency of song requests and the user begins to interact with the virtual character via voice, the system confirms that the target scene is the chorus area and generates a switching signal.

[0095] In another embodiment, if the trend of user behavior data does not clearly point to the chorus area, for example, if the frequency of song requests does not increase, the system updates the behavior data through a continuous monitoring module and recalculates the trend vector. This dynamic monitoring mechanism improves the accuracy of switching prediction.

[0096] Step S51: Use a probability sorting algorithm to extract the target scene with the highest probability and obtain the identifier of the target scene with the highest probability.

[0097] Specifically, the probability ranking algorithm sorts the scenes in the target scene list from highest to lowest predicted probability and selects the one with the highest probability as the target scene. For example, if the probability of a multi-person chorus scene is 0.8 and that of a virtual stage performance is 0.15, then multi-person chorus is the highest probability target scene, and its identifier is generated.

[0098] In one embodiment, the system sorts scene probabilities using a quicksort algorithm to ensure computational efficiency. After the identifier is generated, the system stores the identifier of the multi-person chorus scene as input for subsequent switching signals. This method simplifies the scene selection process and improves real-time performance.

[0099] Step S52: Collect behavioral data through a real-time monitoring system, calculate the trend of behavioral data changes using time series analysis, and obtain a behavioral data trend vector.

[0100] Specifically, the real-time monitoring system collects user behavior data at a high frequency, such as 10 times per second, including dwell time, song request frequency, and interaction intensity. Time series analysis methods calculate data trends using a sliding window, for example, calculating the increment of song request frequency over the past 5 seconds.

[0101] In one possible implementation, the system uses the difference method to calculate the rate of change of behavioral data. For example, if the frequency of song requests increases from once per minute to three times per minute, a positive trend vector is generated. The trend vector comprehensively reflects the direction and magnitude of changes in dwell time, song request frequency, and interaction intensity, providing a basis for switching judgments.

[0102] In another embodiment, if the virtual karaoke scene supports real-time multi-user interaction, the system can incorporate weighted time series analysis, focusing on changes in interaction intensity. For example, as the number of times users sing along with virtual characters increases, the weight of interaction intensity in the trend vector increases. This approach improves the targeting of trend analysis.

[0103] Step S53: If the behavioral data change trend vector points to the preset feature of the chorus area, then the target scene is determined to be the chorus area, and a scene switching signal is generated.

[0104] Specifically, the preset characteristics of the chorus area include high song request frequency, multi-person voice interaction, and high interaction intensity. The system determines whether the switching conditions are met by comparing the similarity between the trend vector of behavioral data changes and the characteristics of the chorus area.

[0105] In one embodiment, if the trend vector shows that the frequency of song requests has increased to 3 times per minute, and the user has input multiple chorus commands through the microphone, the system confirms that the target scene is the chorus area and generates a switching signal. The switching signal includes the identifier and priority of the chorus area, which is used to trigger subsequent resource loading.

[0106] In another possible implementation, if the trend vector partially satisfies the characteristics of the chorus area, such as increased song request frequency but insufficient interaction intensity, the system can determine whether to generate a switching signal based on a threshold. This method avoids erroneous switching and improves the stability of scene switching.

[0107] Step S54: Based on the scene switching signal, obtain the relevant configuration parameters of the chorus area and determine the switching requirements.

[0108] Specifically, the configuration parameters for the chorus area include the audio processing modules required for the scene, the virtual character interaction logic, and the interface rendering requirements. The system extracts these parameters from a preset configuration library based on the switching signal.

[0109] In one embodiment, the chorus area requires high-fidelity audio processing and multi-person voice synchronization. The system extracts the corresponding audio bandwidth allocation parameters and virtual character animation configurations to generate a switching requirement description. The switching requirement clarifies the resource requirements and priorities of the chorus area, providing guidance for resource preloading.

[0110] In another possible implementation, if the virtual karaoke system supports customized chorus scenarios, the system can adjust configuration parameters based on users' historical preferences. For example, if a user prefers a specific style of virtual character, the system will prioritize loading related animation resources. This personalized configuration improves the user experience.

[0111] Step S55: If the behavioral data change trend vector does not point to the preset feature of the chorus area, the behavioral data is updated through the continuous monitoring module to obtain a new behavioral data change trend vector.

[0112] Specifically, if the trend vector does not meet the characteristics of the chorus area, such as the frequency of song requests not increasing significantly, the system will continue to collect behavioral data through the continuous monitoring module and update the trend vector.

[0113] In one embodiment, the continuous monitoring module updates the behavioral data every second, recalculates the rate of change of song request frequency and interaction intensity, and generates a new trend vector. For example, if a user starts requesting songs frequently after 5 seconds, the new trend vector may meet the characteristics of a chorus area, triggering a switching signal.

[0114] In another possible implementation, the system can extend the monitoring window, such as from 5 seconds to 10 seconds, to capture longer-term behavioral changes. This approach is suitable for situations where user behavior patterns are more complex.

[0115] Step S56: Use the random forest algorithm to classify the trend vector of the new behavioral data, determine whether the chorus area switching conditions are met, and obtain the classification result.

[0116] Specifically, the Random Forest algorithm classifies trend vectors using multiple decision trees to determine whether they meet the chorus region characteristics. The algorithm's input is the trend vector, and the output is the classification result, indicating whether the characteristics are met or not.

[0117] In one embodiment, the random forest comprises 100 decision trees, each trained based on a subset of the trend vector. For example, one tree analyzes changes in song request frequency, while another analyzes interaction intensity, and the combined votes generate a classification result. If 80% of the decision trees determine that the chorus area characteristics are met, the system confirms that the switching condition is met.

[0118] In another possible implementation, the system can optimize the random forest model through feature importance analysis. For example, if the frequency of song requests contributes the most to the classification result, the system prioritizes monitoring this feature, improving classification efficiency. This method improves the accuracy of switching decisions.

[0119] Step S57: Based on the classification results, determine the final switching demand prediction results and generate the target scenario switching instruction.

[0120] Specifically, the classification result determines whether to generate a switching instruction. If the chorus area characteristics are met, the system generates a switching instruction containing the chorus area identifier; otherwise, it continues to monitor or adjust the classification model.

[0121] In one embodiment, if the random forest algorithm confirms that the trend vector satisfies the characteristics of the chorus area, the system generates a switching instruction, which includes the resource loading priority of the chorus area and interface adjustment parameters. This instruction ensures the smoothness of the switching process and enhances the user's immersion in the virtual karaoke room.

[0122] Thus, the method, through scene switching prediction, classification, and priority ranking, has achieved a complete process from user behavior analysis to target scene determination, providing support for subsequent resource preloading and interface optimization.

[0123] It should be noted that, in this embodiment, multi-scenario data refers to a comprehensive data set that covers multiple different scenarios in the virtual environment of the karaoke machine and is related to user behavior and scenario attributes. Multi-scenario data is used to mine cross-scenario correlation features and support accurate prediction of scenario switching. Feedback data refers to the closed-loop data extracted from the seamless interface state after the scene switching process is completed. It is used to verify the accuracy of prediction and optimize the model. The purpose of feedback data is to realize the iterative upgrade of the long short-term memory network. Triggering timing refers to the timing of the preloading mechanism, that is, the specific time node when the system starts the preloading operation of resource files for the target scene, such as audio, images, interactive logic components and interface elements. Its essence is to solve the contradiction of bandwidth and cache redundancy caused by preloading too early and scene switching delay caused by preloading too late by dynamically adjusting the start time of preloading. Ultimately, it supports the core goal of seamless interface connection and ensures that when the user actually performs the switching operation, the resources of the target scene have been loaded, so as to achieve a smooth scene transition without lag.

[0124] In the description of this invention, it should be understood that the terms "coaxial," "bottom," "one end," "top," "middle," "other end," "upper," "side," "top," "inner," "front," "center," "both ends," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0125] Furthermore, the terms “first,” “second,” “third,” and “fourth” are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as “first,” “second,” “third,” or “fourth” may explicitly or implicitly include at least one of those features.

[0126] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "setting," "connection," "fixing," "screw connection," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal connection of two components or the interaction between two components. Unless otherwise explicitly limited, those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0127] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A multi-scene switching method based on a karaoke system, characterized in that, include: The system collects behavioral data on user dwell time, song selection frequency, and interaction intensity in the virtual environment using sensors, and processes the behavioral data using a long short-term memory network to obtain a dynamic user preference vector. The user's dynamic preference vector is matched with a preset scene type library. If the matching degree is higher than the threshold, it is determined that the current scene preference is stable; otherwise, it is determined that there is a potential switching intention. The probability distribution of potential switching intention is obtained through the above judgment. After obtaining the probability distribution of the potential switching intentions, a support vector machine classifier is used to train the correlation features between multi-scene data to obtain a list of possible target scenes for cross-scene switching. Extract the target scenario with the highest probability from the list of possible target scenarios, and judge it in combination with the trend of changes in real-time monitored behavioral data. If the trend of change points to the chorus area, it is determined that the switch to the chorus area is about to begin. Otherwise, continue monitoring to obtain an accurate prediction result of the switching demand. Based on the exact switching demand prediction results, the resource files and interface elements of the target scenario are preloaded, and the server bandwidth allocation ratio is adjusted using a dynamic allocation algorithm to obtain an optimized resource configuration scheme. The interface layout parameters are updated by the optimized resource configuration scheme. If the prediction result matches the actual user operation, the layout parameters are applied directly. If the prediction result does not match the actual user operation, the layout parameters are rolled back to the original configuration to determine the final seamless interface state. Feedback data is extracted from the final seamless interface state and input into the iterative training process of the Long Short-Term Memory network to update the model parameters to adapt to user behavior patterns.

2. The multi-scene switching method based on a karaoke system according to claim 1, characterized in that: The step of collecting user behavior data such as dwell time, song selection frequency, and interaction intensity in the virtual environment through sensors, and processing the behavior data using a long short-term memory network to obtain a user dynamic preference vector includes: User behavior data, including dwell time, song selection frequency, and interaction intensity, is collected from the virtual environment using sensors and stored as structured time-series data to obtain an initial behavior dataset. Data cleaning techniques were used to preprocess the initial behavioral dataset, removing noisy data and missing values. The dwell time, song request frequency, and interaction intensity were aligned using timestamps to obtain a standardized behavioral sequence. Based on the Long Short-Term Memory Network, time-series modeling of standardized behavioral sequences is performed to extract time-dependent features and generate user behavior pattern vectors. Based on user behavior pattern vectors, a clustering algorithm is used to group users, and user preference categories are obtained based on the similarity of behavior pattern vectors. If the user's preference category matches the preset music recommendation rules, the corresponding song list is retrieved from the music library based on the category characteristics to generate a personalized recommendation sequence; By comparing the personalized recommendation sequence with the frequency of user song requests, the matching degree between the recommendation sequence and the actual behavior is calculated, and the recommendation effect evaluation index is obtained. Based on the recommendation performance evaluation metrics, the gradient descent algorithm is used to optimize the parameters of the long short-term memory network and update the user behavior pattern vector.

3. The multi-scene switching method based on a karaoke system according to claim 1, characterized in that: The step of matching the user's dynamic preference vector with a preset scene type library, determining that the current scene preference is stable if the matching degree is higher than a threshold, and otherwise determining that a potential switching intention exists, and obtaining the probability distribution of the potential switching intention through the above determination, includes: Obtain user dynamic preference vectors and generate feature vectors by analyzing user historical interaction data; Based on the vector matching algorithm, the user's dynamic preference vector is compared with the standard vector in the preset scene type library to obtain the matching degree score; If the matching score is higher than a preset threshold, the current scene preference is determined to be stable, and a stable scene label is generated. Based on the stable scene tags, extract recommended content related to the current scene and generate a personalized recommendation list; If the matching score is lower than a preset threshold, the potential switching intent is analyzed using a Bayesian inference algorithm to obtain the intent probability distribution. Based on the intent probability distribution, a weighted sorting algorithm is used to generate a recommendation sequence of potential switching scenarios; Based on the recommended sequence, the user interface is dynamically adjusted, and optimized scene recommendation content is output.

4. The multi-scene switching method based on a karaoke system according to claim 1, characterized in that: The step of obtaining the potential switching intent probability distribution and then training a support vector machine classifier on the correlation features between multi-scene data to obtain a list of possible target scenes for cross-scene switching includes: Acquire user interaction data and extract the probability distribution of potential switching intentions from it; By analyzing probability distributions, a support vector machine classifier is used to train the associated features of multi-scene data to obtain a classification model. If the prediction accuracy of the classification model reaches a preset threshold, the target scene list for cross-scene switching is determined based on the model output. Based on the list of target scenarios, extract the associated feature set for each scenario and generate a feature vector; Feature vectors are used to group target scenes using a clustering algorithm to obtain scene grouping results; By analyzing the scene grouping results, we can obtain the priority ranking of each group and get the final scene switching sequence. If the priority of the final scene switching sequence meets the preset conditions, the scene sequence is output; otherwise, the feature vectors are extracted again for clustering.

5. The multi-scene switching method based on a karaoke system according to claim 1, characterized in that: The step of extracting the highest probability target scenario from the list of possible target scenarios, and judging the trend of behavioral data changes in real time, determining that the switch to the chorus area is imminent if the trend points to the chorus area, otherwise continuing monitoring, to obtain an accurate prediction result of the switching demand includes: The target scene is obtained from the scene list, and the highest probability target scene is extracted using a probability sorting algorithm to obtain the highest probability target scene identifier; Behavioral data is collected through a real-time monitoring system, and the changing trend of the behavioral data is calculated using time series analysis methods to obtain a behavioral data changing trend vector. If the trend vector of behavioral data changes points to the preset features of the chorus area, then the target scene is determined to be the chorus area, and a scene switching signal is generated. Based on the scene switching signal, obtain the relevant configuration parameters of the chorus area to determine the switching requirements; If the behavioral data change trend vector does not point to the preset features of the chorus area, the behavioral data will be updated through the continuous monitoring module to obtain a new behavioral data change trend vector. The random forest algorithm is used to classify the trend vector of new behavioral data to determine whether the chorus area switching condition is met, and the classification result is obtained. Based on the classification results, the final switching demand prediction results are determined, and the target scenario switching instructions are generated.

6. The multi-scene switching method based on a karaoke system according to claim 1, characterized in that: The steps of preloading resource files and interface elements of the target scenario based on the accurate switching demand prediction results, and adjusting the server bandwidth allocation ratio using a dynamic allocation algorithm to obtain an optimized resource configuration scheme include: By analyzing user behavior data through predictive models, we can obtain prediction results of switching needs. Based on the prediction results, the priority and resource requirements of the target scenario are determined, and a resource loading sequence is generated. Based on the resource loading sequence, preload the resource files and UI elements of the target scene and store them in the local cache; A consistent hashing algorithm is used to dynamically allocate server bandwidth and adjust the bandwidth ratio for different scenarios. If the loading speed of a certain scenario is lower than the preset threshold after the bandwidth ratio is adjusted, the bandwidth will be redistributed through the load balancing mechanism to obtain an optimized bandwidth allocation scheme. For the optimized bandwidth allocation scheme, monitor the resource loading status in real time and obtain loading completion data; Based on the loading completion data, the timing of the preloading mechanism is dynamically adjusted to generate the final resource configuration scheme.

7. The multi-scene switching method based on a karaoke system according to claim 1, characterized in that: The step of updating the interface layout parameters through the optimized resource configuration scheme, and directly applying the layout parameters if the prediction result matches the actual user operation, otherwise rolling back to the original configuration, to determine the final seamless interface state, includes: By analyzing the resource allocation scheme, the currently available computing and storage resources are obtained from the preset resource pool to determine the real-time allocation status of the resource allocation scheme. Based on the real-time allocation status of the resource allocation scheme, the corresponding interface layout parameters are generated using a preset interface rendering algorithm to obtain the initially adjusted layout parameters. The layout parameters that have been initially adjusted are predicted by a machine learning classification model. The prediction results are then used to determine whether the prediction results match the expected user operation pattern. If the prediction results are consistent with the actual user operation, the initially adjusted layout parameters are deployed to the interface rendering module through parameter application logic to determine the update status of the interface layout. If the prediction result is inconsistent with the actual user operation, the original layout parameters are obtained from the original layout parameter repository, and the rollback operation mechanism is used to restore the interface to the initial state to obtain the rolled-back interface state. Based on the rolled-back interface state or updated state, a state synchronization algorithm is used to detect the latency and consistency during the interface switching process to determine the seamless switching state. By recording seamless state transitions and user operation data, the training dataset for the dynamic adjustment mechanism is updated to obtain an optimized resource configuration scheme and input parameters for the next adjustment.

8. A multi-scene switching system based on a karaoke system, characterized in that, include: The behavior data processing module is configured to collect user behavior data such as dwell time, song selection frequency, and interaction intensity in the virtual environment through sensors, process the behavior data using a long short-term memory network, and obtain a user dynamic preference vector. The scene preference judgment module is configured to match the user's dynamic preference vector with a preset scene type library. If the matching degree is higher than a threshold, it is judged that the current scene preference is stable; otherwise, it is judged that a potential switching intention exists. The module obtains the probability distribution of potential switching intentions through the above judgment. The associated feature training module is configured to, after obtaining the probability distribution of the potential switching intention, use a support vector machine classifier to train the associated features between multi-scene data to obtain a list of possible target scenes for cross-scene switching. The switching demand prediction module is configured to extract the target scenario with the highest probability from the list of possible target scenarios, and judge the trend of changes in real-time monitored behavioral data. If the trend of changes points to the chorus area, it is determined that the switch is about to be made to the chorus area; otherwise, monitoring continues to obtain an accurate switching demand prediction result. The resource configuration optimization module is configured to preload the resource files and interface elements of the target scenario based on the exact switching demand prediction results, and adjust the server bandwidth allocation ratio using a dynamic allocation algorithm to obtain an optimized resource configuration scheme. The interface connection determination module is configured to update the interface layout parameters through the optimized resource configuration scheme. If the prediction result matches the actual user operation, the layout parameters are directly applied. If the prediction result does not match the actual user operation, the layout parameters are rolled back to the original configuration to determine the final seamless interface connection state. The model iteration update module is configured to extract feedback data in the final seamless interface state, input it into the iterative training process of the long short-term memory network, and update the model parameters to adapt to user behavior patterns.

9. A karaoke device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the multi-scene switching method based on the karaoke system as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the multi-scene switching method based on the karaoke system as described in any one of claims 1 to 7.

Images