Soundscape control method and system for virtual reality scenes

Abstract

The present invention provides a soundscape control method and system for a virtual reality scene, which relates to virtual reality technology, including: dividing VR audio and video into slices with equal time intervals; encapsulating heart rate variability feature data and VR audio and video into attribute data of the slices according to the time sequence of the slices; randomly adjusting the audio attributes of the slices, and collecting heart rate data of different experiencers when experiencing the slices after the audio attributes are adjusted; extracting the heart rate variability feature data and inputting it into an emotion recognition classifier to identify the emotional state of the experiencer as training data; calculating the Youden index of different audio attributes according to the target emotional state and the statistical emotional response to determine the optimal threshold interval of the audio attribute; judging whether the audio attribute needs to be adjusted according to the correlation between the audio attribute in the optimal threshold interval and the emotional state of the experiencer, thereby realizing adaptive soundscape control, solving the problem of mismatch between sound field creation and visual experience, and improving user experience.

Description

Technical Field

[0001] The present invention relates to the field of virtual reality technology, and in particular to a soundscape control method and system for a virtual reality scene. Background Art

[0002] VR experience hall is an entertainment venue that combines virtual reality (VR) technology. It creates a three-dimensional virtual environment through head-mounted display (HMD) and other sensor devices. It provides customers with a new immersive experience. The concept of "soundscape" is different from the understanding of sound at the physical level. Sound and scenery are unified as landscape. Soundscape includes three elements: sound, environment, and people. The existing research on soundscape regulation is to explore the interactive relationship between sound field, people, and landscape, improve soundscape, and apply it to urban landscape design and road construction planning. The existing soundscape regulation focuses on the analysis of sound field characteristics, subjective evaluation of soundscape, systematic establishment of the connection between objective acoustic data and subjective perception, and the use of landscape elements and sound field management to regulate the objective environment soundscape. VR experience can provide a realistic virtual scene experience. Different from the real environment, the sound field is fully controllable, and with the development of new sensor technology, the perceptual information of user experience can be obtained in real time. Therefore, the user's subjective feelings can be portrayed by objective perceptual information, and the sound field in different visual environments can be evaluated and improved based on the application of user perceptual information.

[0003] Different from traditional cultural consumption, in VR experiential cultural consumption, users not only participate in the experience, but are also part of the work. The interactive process in VR experiential cultural consumption can provide users with multi-sensory experiences - such as vision and hearing, providing more dimensions for narrative and aesthetics using VR technology. However, limited by existing technical methods, many VR experiences have problems such as serious homogeneity and poor experience. In terms of content presentation, the content and experience forms provided are too similar, and there is a lack of consideration for user needs. The user's subjective initiative in the VR experience has not been fully utilized, making it difficult to provide personalized services. In terms of experience effect, the sound field creation of the VR experience is not highly matched with the visual experience, resulting in insufficient immersion of users during the experience, thus affecting the effect of soundscape creation.

[0004] Therefore, how to improve the matching degree between sound field creation and visual experience becomes a technical problem that needs to be solved. Summary of the invention

[0005] The present invention aims to solve at least one of the technical problems existing in the prior art or related technology, and discloses a soundscape control method and system for virtual reality scenes. According to the user experience effect evaluation, the sound field adjustment direction is clarified, the sound field adaptive control is realized in a specific visual scene, the problem of mismatch between sound field creation and visual experience is solved, and personalized experience services are provided in full combination with the characteristics of user groups.

[0006] The first aspect of the present invention discloses a soundscape control method for a virtual reality scene, comprising: S1, using a heart rate monitoring device to collect heart rate data of a human body in different emotional states, extracting heart rate variability feature data from the heart rate data and annotating it according to the emotional state annotation, and using the annotated feature data for machine learning to construct an emotion recognition classifier; S2, collecting heart rate data of different experiencers when experiencing a virtual reality scene and corresponding VR videos and VR audios; extracting heart rate variability feature data of the heart rate data; extracting audio attributes of the VR audio; dividing the VR video and VR audio into slices with equal time intervals; encapsulating the heart rate variability feature data and the VR audio and VR video into attribute data of the slices according to the time order of the slices; S3, randomly adjusting the audio attributes of the slices, collecting heart rate data of different experiencers when experiencing the slices after the audio attributes are adjusted; extracting heart rate variability feature data of the VR audio and VR video; The feature data is input into the emotion recognition classifier to identify the emotional state of the experiencer as training data; S4, set the target emotional state according to the content of the VR video of the slice; count the various emotional states reflected by the experiencer when experiencing the slice under different audio attributes; calculate the Youden index of different audio attributes according to the target emotional state and the counted emotional response, and the threshold interval of the audio attribute corresponding to the maximum value of the Youden index is the optimal threshold interval; according to the correlation between the audio attribute in the optimal threshold interval and the emotional state of the experiencer, determine whether the audio attribute needs to be adjusted; S5, if the audio attribute needs to be adjusted, adjust the audio attribute of the slice to the optimal threshold interval and generate VR audio, re-collect and extract the heart rate variability feature data of the experiencer's VR experience slice content and include it in the training data; repeat steps S3 and S4, continue to adjust the audio attributes of other slices, and realize adaptive control of soundscape.

[0007] In this technical solution, the human body's perception of sound is highly subjective. The application of portable wearable ECG acquisition equipment can realize real-time monitoring and analysis of emotions. The use of machine learning methods to perform emotion recognition through heart rate data can obtain more objective and effective results, realize the user's VR experience throughout the process of emotion monitoring, and provide data support for the evaluation of VR experience effects. VR sound scene (soundscape) has a high degree of controllability. The audio attributes of VR sound scene can be finely designed to evaluate the impact of VR audio attributes on user (experiencer) emotions and obtain the optimal threshold interval of audio attributes, which can make full use of the user's VR experience heart rate data collected in real time. VR audio attribute adjustment involves parameters of multiple dimensions. These attributes are interrelated. The change of one attribute may affect the performance of other attributes, thereby having an unpredictable impact on the overall audio effect. The present invention analyzes the patterns and relationships in the data by analyzing a large amount of data, directly establishes the association between audio attributes and user emotional state, and determines the adjustment scheme of audio attributes by updating training data to achieve adaptive control of soundscape. There is no need to conduct correlation research between various audio attributes, which effectively improves the feasibility of the scheme.

[0008] According to the soundscape control method for a virtual reality scene disclosed in the present invention, preferably, the calculation process of the Youden index specifically includes:

[0009] Determine target emotional state based on sliced ​​VR video content ;

[0010] Statistics for different threshold intervals Audio properties The total number of emotional states reflected by the experiencer ,in for emotional states;

[0011] Divide into multiple threshold intervals and calculate , , , :

[0012] Represents the true positive number, which is the threshold interval The total number of emotional states that match the target emotional state:

[0013]

[0014] Represents the number of false positive cases, which is the threshold interval Total number of emotional states that do not match the target emotional state:

[0015]

[0016] Represents the number of false negative examples, which is the threshold interval The total number of emotional states that match the target emotional state:

[0017]

[0018] Represents the number of true negative examples, which is the threshold interval The total number of emotional states that do not match the target emotional state:

[0019]

[0020] Calculate each threshold interval The corresponding true positive rate True negative rate :

[0021]

[0022]

[0023] Calculate the Youden index corresponding to each threshold interval :

[0024] .

[0025] According to the soundscape control method for virtual reality scenes disclosed in the present invention, preferably, the step of judging whether the audio attributes need to be adjusted according to the correlation between the audio attributes in the optimal threshold interval and the emotional state of the experiencer specifically includes:

[0026] Computing Audio Properties The accuracy corresponding to the optimal threshold interval :

[0027]

[0028] If the accuracy , then determine the audio attributes The slice experience is not related to the emotional state of the experiencer, and there is no need to adjust the audio attributes later. Make adjustments.

[0029] According to the soundscape control method for virtual reality scenes disclosed in the present invention, preferably, extracting heart rate variability characteristic data specifically includes:

[0030] Apply peak detection algorithm to the collected heart rate data Wave sequence is used to statistically analyze the characteristic data of heart rate variability and obtain the time domain and frequency domain indicators of heart rate variability.

[0031] According to the soundscape control method for virtual reality scenes disclosed in the present invention, preferably, the audio attributes include timbre and technical parameters of the timbre, the timbre includes ambient sound, human voice or sound generated by the interaction between the human body and the environment, and the technical parameters include loudness, pitch, duration, roughness, sharpness, pitch prominence rate and pitch-to-noise ratio.

[0032] According to the soundscape control method for a virtual reality scene disclosed in the present invention, preferably, the emotional state includes happiness, sadness, anger, fear, disgust, boredom and calmness.

[0033] According to the soundscape control method for virtual reality scenes disclosed in the present invention, preferably, the emotion recognition classifier is constructed based on a support vector machine algorithm, a random forest algorithm or a K nearest neighbor algorithm.

[0034] The second aspect of the present invention also discloses a soundscape control system for a virtual reality scene, comprising: a memory for storing program instructions; a processor for calling the program instructions stored in the memory to implement the soundscape control method for a virtual reality scene provided by any of the above technical solutions.

[0035] The beneficial effects of the present invention include at least: timely obtaining and evaluating the user experience effect by monitoring the experiencer's heart rate, clarifying the sound field adjustment direction based on the user experience effect evaluation, and realizing adaptive control of the sound field in a specific visual scene; solving the problem of mismatch between sound field creation and visual experience, and enhancing the experiencer's sense of immersion; and helping to fully combine the characteristics of the user group to provide personalized experience services. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 The figure shows an overall flow chart of a soundscape control method for a virtual reality scene according to an embodiment of the present invention.

[0037] Figure 2 A schematic diagram of a soundscape adaptive control process according to an embodiment of the present invention is shown.

[0038] Figure 3 A schematic block diagram of a soundscape control system for a virtual reality scene according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0039] In order to more clearly understand the above-mentioned objects, features and advantages of the present invention, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0040] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0041] like Figure 1 and Figure 2 As shown, according to one embodiment of the present invention, a soundscape control method for a virtual reality scene is disclosed, comprising:

[0042] S1, constructing an emotion recognition classifier: using a heart rate monitoring device to collect heart rate data of individuals in different emotional states, extracting heart rate variability feature data from the heart rate data and labeling the feature data according to the emotional state, and using the labeled feature data for machine learning to construct an emotion recognition classifier;

[0043] S2, collect user VR experience data: collect data from different users ( =1,2,3...) during the VR experience and the corresponding VR video and audio, extract the heart rate variability feature data and VR audio attribute data, and divide the collected data into slices with equal time intervals. The attribute data of the slices includes the user's heart rate data within the corresponding time interval. Heart rate variability feature data, VR video and audio;

[0044] S3, obtain training data: randomly adjust the audio properties of the slices and generate VR audio, re-collect different users The heart rate data during the VR experience slicing process is used to extract the heart rate variability feature data and input it into the emotion recognition classifier to identify the user's emotional state corresponding to the slice. As training data;

[0045] S4, soundscape matching evaluation: determine the target emotion based on the VR video content of the slice, calculate the Youden index of the audio attribute under different threshold ranges, determine the optimal threshold range of the slice audio attribute, and decide whether to adjust the audio attribute based on the correlation between the audio attribute and the user's emotional state during the slice VR experience;

[0046] In this step, heart rate variability (HRV) refers to the small changes in the intervals between adjacent heartbeats in a normal heart rate rhythm. It is an important bridge between emotional states and physiological responses, and it provides a strong physiological basis for emotion recognition. The application of machine learning algorithms in emotion recognition is an important branch of the field of affective computing. It uses algorithms to learn and identify emotional states from physiological signals. Support vector machines, artificial neural networks, K-nearest neighbors, Bayesian classifiers, etc. are commonly used algorithms. By designing the target emotional state and comparing the relationship between the True Positive Rate (TPR) and the True Negative Rate (TNR), the optimal design of VR audio can be determined and the soundscape matching evaluation can be achieved.

[0047] S5, adaptive control of sound field: adjust the VR audio properties of the slice to the optimal threshold range and generate VR audio, re-collect and extract the heart rate variability feature data of the user's VR experience slice content into the training data, repeat S3 and S4, continue to adjust the audio properties of other slices, and realize adaptive control of soundscape (sound field). The specific process of adaptive control is as follows: Figure 2 shown.

[0048] According to the above embodiment, further, step S1 specifically includes:

[0049] Step S11, using a heart rate monitoring device to collect heart rate data under different emotional states, and applying a peak detection algorithm to the collected heart rate data to extract an R wave sequence, applying a statistical method to extract heart rate variability feature data, including time domain and frequency domain indicators of heart rate variability, annotating the heart rate variability feature data according to the emotional state, and forming a heart rate variability feature data set for different emotional states;

[0050] Step S12, divide the heart rate feature data set of different emotional states into training data and test data, use different machine learning algorithms to train the training data to obtain an emotion recognition classifier, use the test data to test the performance of the emotion recognition classifier, and select the emotion recognition classifier with the highest accuracy.

[0051] According to the above embodiment, further, step S2 specifically includes:

[0052] Step S21, using a heart rate monitoring device to collect the user's heart rate data during the VR experience, and applying a peak detection algorithm to the collected heart rate data Wave sequence, using statistical methods to extract heart rate variability feature data, including time domain indicators of heart rate variability , frequency domain indicators ;

[0053] Step S22, extracting VR audio attribute data and video data of the user experience, VR audio attribute data Including technical parameters of the timbre contained in the audio, VR video data is video information stored and represented in digital form;

[0054] Step S23, dividing the collected heart rate variability characteristic data and VR audio and video data into slices at equal time intervals, and encapsulating the divided heart rate variability characteristic data and VR audio and video data into slice attribute data in a slice order.

[0055] According to the above embodiment, further, step S4 specifically includes:

[0056] Step S41: Count the specific audio attributes of the user's VR experience slices The total number of various emotional states ,in for emotional states;

[0057] Step S42: determining the target emotion based on the sliced ​​VR video content , set VR audio properties The optimal threshold range is , and Divide the threshold interval , respectively calculate , , , :

[0058]

[0059]

[0060]

[0061] in, Represents the true positive number, which is the threshold interval The total number of emotional states in the range that match the target emotional state; Represents the number of false positive cases, which is the threshold interval The total number of emotional states in the range that do not match the target emotional state; Represents the number of false negative examples, which is the threshold interval the total number of emotion states outside the range that match the target emotion state; Represents the number of true negative examples, which is the threshold interval The total number of out-of-range emotional states that do not match the target emotional state.

[0062] Step S43, calculate all threshold intervals The corresponding true positive rate True negative rate , the calculation formula is as follows:

[0063]

[0064] .

[0065] Step S44, adjusting the optimal threshold range Repeat S42-S43 to calculate the Youden index corresponding to each threshold interval , based on the Youden index The threshold interval corresponding to the maximum value is the audio attribute The optimal threshold interval, Youden index The calculation formula is as follows:

[0066] .

[0067] Step S45, calculating audio attributes The accuracy corresponding to the optimal threshold interval :

[0068]

[0069] If the accuracy , then the audio attribute is considered In the process of slicing VR experience, there is no correlation with the user's emotional state, and step S5 does not Make adjustments.

[0070] According to another embodiment of the present invention, a specific application of the soundscape control method for a virtual reality scene is also disclosed. The adjustable items include the loudness, pitch, duration, roughness, sharpness, pitch prominence, pitch-to-noise ratio and other technical parameters of the timbre of breathing, wind, crying, footsteps, etc. This embodiment only takes the loudness of footsteps as an example for explanation, and the specific process includes:

[0071] S1, build an emotion recognition classifier: use heart rate monitoring equipment to collect heart rate data of individuals in different emotional states, including happiness, sadness, anger, fear, disgust, boredom, and calmness, extract heart rate variability feature data from the heart rate data and annotate the feature data according to the emotional state, and use the annotated feature data for machine learning to build an emotion recognition classifier. Specifically:

[0072] S11, using a heart rate monitoring device to collect heart rate data under different emotional states, and applying a peak detection algorithm to the collected heart rate data to extract the R wave sequence, and applying a statistical method to extract heart rate variability feature data, including time domain and frequency domain indicators of heart rate variability. The specific indicator meanings are shown in Table 1:

[0073] Table 1 Significance of time domain and frequency domain indicators of heart rate variability

[0074]

[0075] The heart rate variability feature data is labeled according to the emotional state to form a heart rate variability feature data set of different emotional states. The heart rate variability feature data set is shown in Table 2:

[0076] Table 2 Part of the heart rate variability feature dataset

[0077]

[0078] Step S12, divide the heart rate feature data sets of different emotional states into training data and test data, select 20 examples of heart rate variability feature data of each emotional state, a total of 140 examples of data as training data, and 10 examples of heart rate variability feature data of each emotional state, a total of 70 examples of data as test data. In this embodiment, three classification models, namely support vector machine, random forest, and K nearest neighbor algorithm, are selected to train with the training data to obtain an emotion recognition classifier, and the performance of the emotion recognition classifier is tested with the test data, and the emotion recognition classifier with the highest accuracy is selected. In this example, the emotion recognition classifier constructed by the support vector machine algorithm is finally selected.

[0079] S2, collect user VR experience data: collect data from different users ( =1,2,3...) heart rate data during the VR experience and the corresponding VR video and audio, extract heart rate variability feature data and VR audio attribute data, and divide the collected data into slices with equal time intervals. The attribute data of the slices include the heart rate variability feature data, VR video and audio of user u in the corresponding time interval. Specifically:

[0080] S21, using heart rate monitoring equipment to collect user Heart rate data during VR experience, peak detection algorithm is applied to the collected heart rate data Wave sequence, using statistical methods to extract heart rate variability feature data, including time domain indicators of heart rate variability , frequency domain indicators ;

[0081] S22, extracting VR audio attribute data and video data of user experience, VR audio attribute data It includes technical parameters of the timbre contained in the audio, such as loudness, pitch, duration, roughness, sharpness, pitch prominence, and pitch-to-noise ratio. VR video data is video information stored and represented in digital form;

[0082] S23, dividing the collected heart rate variability characteristic data and VR audio and video data into slices at equal time intervals, and encapsulating the divided heart rate variability characteristic data and VR audio and video data into attribute data of the slices in a slice order.

[0083] S3, obtain training data: adjust the loudness of footsteps in the sliced ​​audio within the range of 5-15db, and re-collect different users The heart rate data during the VR experience slicing process is used to extract the heart rate variability feature data and input it into the emotion recognition classifier to identify the user's emotional state corresponding to the slice. , as training data.

[0084] S4, soundscape matching evaluation: determine the target emotional state based on the VR video content of the slice, calculate the Youden coefficient of different audio attributes under different threshold ranges, determine the optimal threshold range of the slice audio attribute, and decide whether to adjust the audio attribute based on the correlation between the audio attribute and the user's emotional state during the slice VR experience. Specifically:

[0085] S41, VR audio attribute is footstep loudness, count the total number of various emotional states of user VR experience slice content under different footstep loudness ,in is the emotional state type. The specific results are shown in Table 3:

[0086] Table 3 The total number of various emotional states at different loudness

[0087]

[0088] S42, determining the target emotional state according to the sliced ​​VR video content (For example, the target emotion of a horror video is fear, and the target emotion of a pastoral video is calmness). Determine the fear and set the VR audio properties The optimal threshold range is , and Divide the threshold interval , respectively calculate , , , :

[0089]

[0090]

[0091]

[0092]

[0093] in, Represents the true positive number, threshold interval The total number of emotional states in the range that match the target emotional state, Represents the number of false positive cases, which is the threshold interval The total number of emotional states in the range that do not match the target emotional state, Represents the number of false negative examples, which is the threshold interval The total number of emotional states outside the range that match the target emotional state, Represents the number of true negative examples, which is the threshold interval The total number of emotion states outside the range that do not match the target emotion state;

[0094] S43, calculate all threshold intervals The corresponding true positive rate True negative rate :

[0095]

[0096] ;

[0097] by For the optimal threshold range, multiple different threshold intervals are divided , each threshold interval Corresponding , , , , , The values ​​are shown in Table 4:

[0098] Table 4 Index values ​​corresponding to different threshold intervals

[0099]

[0100] S44, adjust the optimal threshold range ,Will Adjust to 2db and 3db, repeat S42-S43, and calculate the Youden index corresponding to each threshold interval , based on the Youden index The threshold interval corresponding to the maximum value is the audio attribute The optimal threshold interval, Youden index The calculation formula is as follows:

[0101] ;

[0102] The Youden index corresponding to each threshold interval in this embodiment As shown in Table 5:

[0103] Table 5 Youden index corresponding to each threshold interval

[0104]

[0105] It can be seen that the optimal threshold range for footsteps loudness is 9-12db.

[0106] S45, Calculate audio properties The accuracy corresponding to the optimal threshold interval :

[0107] ;

[0108] If the accuracy , then the audio attribute is considered In the process of slicing VR experience, there is no correlation with the user's emotional state, and the audio attribute is not processed in step S5. In this embodiment, the optimal threshold range is 9-12db corresponding to the accuracy =0.736>0.7, indicating that the loudness of footsteps in the sliced ​​audio is related to the user's emotional state during the sliced ​​VR experience, and the audio attribute needs to be adjusted in step S5.

[0109] S5, adaptive soundscape control: adjust the loudness of footsteps in the sliced ​​VR audio to the range of 9-12db and generate VR audio, re-collect and extract the heart rate variability feature data of the user's VR experience slice content and include it in the training data, repeat S3 and S4, continue to adjust other slice audio attributes, and generate VR audio according to the VR audio attributes. The optimal threshold range continues to adjust the VR audio to achieve adaptive control of the soundscape.

[0110] like Figure 3 As shown, according to another embodiment of the present invention, a soundscape control system 300 for a virtual reality scene is also disclosed, including: a memory 301, used to store program instructions; a processor 302, used to call the program instructions stored in the memory to implement the soundscape control method for a virtual reality scene as in the above embodiment.

[0111] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A soundscape control method for a virtual reality scene, characterized in that: include: S1, using a heart rate monitoring device to collect heart rate data of a human body under different emotional states, extracting heart rate variability feature data from the heart rate data and annotating it according to the emotional state, and using the annotated feature data for machine learning to construct an emotion recognition classifier; S2, collects heart rate data of different experiencers when experiencing virtual reality scenes and the corresponding VR videos and VR audios; extracts heart rate variability feature data of heart rate data; extracts audio attributes of VR audio; divides VR video and VR audio into slices with equal time intervals; encapsulates heart rate variability feature data, VR audio and VR video into slice attribute data according to the time order of slices; S3, randomly adjusting the audio attributes of the slice, and collecting heart rate data of different experiencers when experiencing the slices after the audio attributes are adjusted; Extracting heart rate variability feature data and inputting it into the emotion recognition classifier to identify the emotional state of the experiencer as training data; S4, setting a target emotional state according to the content of the VR video of the slice; counting the various emotional states reflected by the experiencer when experiencing the slice under different audio attributes; calculating the Youden index of different audio attributes according to the target emotional state and the statistical emotional response, and the threshold interval of the audio attribute corresponding to the maximum value of the Youden index is the optimal threshold interval; judging whether the audio attribute needs to be adjusted according to the correlation between the audio attribute in the optimal threshold interval and the emotional state of the experiencer; S5, if it is necessary to adjust the audio attribute, adjust the audio attribute of the slice to the optimal threshold range and generate VR audio, and re-collect and extract the heart rate variability feature data of the experiencer's VR experience slice content into the training data; Repeat steps S3 and S4 to continue adjusting the audio properties of other slices to achieve adaptive control of the soundscape; The calculation process of Youden Index includes: Determine the target emotional state based on the sliced ​​VR video content; Divide the audio attributes into multiple threshold intervals, and count the different threshold intervals separately. Audio properties of p C The total number of each emotional state reported by the experiencer; Calculate different threshold intervals respectively Corresponding Said Represents the true positive number, which is the threshold interval The total number of emotional states within the group that match the target emotional state; Said Represents the number of false positive cases, which is the threshold interval The total number of emotional states that do not match the target emotional state; Said Represents the number of false negative examples, which is the threshold interval The total number of emotional states that match the target emotional state; Said Represents the number of true negative examples, which is the threshold interval The total number of emotional states that do not match the target emotional state; Calculate each threshold interval The corresponding true positive rate True negative rate Calculate each threshold interval The corresponding Youden index is:

2. The soundscape control method for a virtual reality scene according to claim 1, characterized in that: The step of judging whether the audio attribute needs to be adjusted according to the correlation between the audio attribute in the optimal threshold interval and the emotional state of the experiencer specifically includes: Calculate audio attribute p C The accuracy corresponding to the optimal threshold interval If the accuracy Then determine the audio attribute p C The slice experience is not related to the emotional state of the experiencer, and there is no need to adjust the audio attribute p C Make adjustments.

3. The soundscape control method for a virtual reality scene according to claim 1, characterized in that: Extracting heart rate variability feature data specifically includes: The peak detection algorithm is applied to the collected heart rate data to detect the R wave sequence, so as to statistically calculate the characteristic data of heart rate variability and obtain the time domain index and frequency domain index of heart rate variability.

4. The soundscape control method for a virtual reality scene according to claim 1, characterized in that: The audio attributes include timbre and technical parameters of the timbre. The timbre includes ambient sound, human voice or sound generated by the interaction between the human body and the environment. The technical parameters include loudness, pitch, duration, roughness, sharpness, pitch prominence and pitch-to-noise ratio.

5. The soundscape control method for a virtual reality scene according to claim 1, characterized in that: The emotional states include happiness, sadness, anger, fear, disgust, boredom, and calmness.

6. The soundscape control method for a virtual reality scene according to claim 1, characterized in that: The emotion recognition classifier is constructed based on a support vector machine algorithm, a random forest algorithm or a K nearest neighbor algorithm.

7. A soundscape control system for a virtual reality scene, characterized in that: include: A memory for storing program instructions; A processor, configured to call the program instructions stored in the memory to implement the soundscape control method for a virtual reality scene according to any one of claims 1 to 6.

Images