Audio control method and system for stage performance

Abstract

The invention relates to the technical field of artificial intelligence and stages, and provides an audio control method and system for stage performance, and the method comprises the steps: collecting real-time environment parameters of an outdoor stage, the environment parameters comprising at least two of rainfall, wind speed, environment noise, temperature and humidity, and audience noise; inputting the real-time environment parameters into a preset dynamic interaction model, and outputting the disturbed quantity of each frequency band; the dynamic interaction model is a nonlinear model obtained through multi-factor combination sample training, and a multi-factor combination comprises the actual disturbed quantity of each frequency band under different environmental parameter combinations; calculating the compensation gain of each frequency band based on the disturbed quantity of each frequency band; and compensating each frequency band of the to-be-output stage performance audio signal based on the compensation gain to obtain a compensation audio signal. According to the invention, the system adapts to environmental changes through learning and feedback, can more accurately model a complex interaction relationship, and realizes that key audio contents are always clear and distinguishable no matter how environmental factors are superposed.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of artificial intelligence and stage technology, and particularly relates to an audio control method and system for stage performance. BACKGROUND

[0002] In an open-air stage performance, there are environmental factors that mask key audio content such as human voice and instrument details. In order to reduce the influence of environmental factors, the existing method calculates the total gain of each frequency band according to the preset gain and preset weight of each frequency band for each environmental factor, and uses the total gain of each frequency band to compensate the gain of each frequency band volume. However, in actual scenarios, the influence of each environmental factor on different frequency bands may interfere with each other, and the existing method cannot accurately model this complex interaction, resulting in poor compensation effect. In addition, the air humidity increases in rainy days, resulting in changes in sound wave propagation, and the absorption of water to high-frequency sound waves is enhanced, causing the frequency response characteristics of the audio system to shift, resulting in poor performance. SUMMARY

[0003] In view of the above technical problems, the purpose of the present application is to provide an audio control method and system for stage performance, which aims to solve at least one of the above technical problems.

[0004] In a first aspect, an audio control method for stage performance is provided, the method comprising:

[0005] Collecting real-time environmental parameters of an open-air stage, the environmental parameters including at least two of rainfall, wind speed, environmental noise, temperature and humidity, and audience noise;

[0006] Inputting the real-time environmental parameters into a preset dynamic interaction model to output the disturbed amount of each frequency band; the dynamic interaction model is a nonlinear model trained by multi-factor combination samples, and the multi-factor combination includes the actual disturbed amount of each frequency band under different environmental parameter combinations;

[0007] Calculating the compensation gain of each frequency band based on the disturbed amount of each frequency band;

[0008] Compensating each frequency band of the stage performance audio signal to be output based on the compensation gain to obtain a compensated audio signal.

[0009] Further, the disturbed amount includes amplitude attenuation amount and signal-to-noise ratio deterioration amount, and the training process of the dynamic interaction model includes:

[0010] Collecting N environmental parameter combination samples, each sample including specific values of M environmental parameters and actual disturbed amounts of each frequency band under the corresponding scene;

[0011] Dividing the environmental parameter combination samples into a training set and a validation set;

[0012] The neural network model is trained by taking the environmental parameters in the training set as input, the disturbed quantities of each frequency band as output, and a preset loss function.

[0013] The trained model is verified for accuracy by the verification set, and when the verification accuracy is greater than a preset threshold, the dynamic interaction model is determined.

[0014] Further, the loss function for training the dynamic interaction model is:

[0015] ;

[0016] wherein, ;

[0017] ;

[0018] ;

[0019] ;

[0020] wherein, is the total loss, is the L1 loss of the amplitude attenuation of each frequency band, is the MSE loss of the signal-to-noise ratio degradation of each frequency band, is a high-priority frequency band prediction deviation penalty term, 、 、 is a weight coefficient, and ; K is the number of frequency bands, and k is the frequency band index, is the weight coefficient of the kth frequency band, and the weight coefficient of the high-priority frequency band is greater than the weight coefficient of the low-priority frequency band, is the real signal-to-noise ratio degradation of the kth frequency band of the ith sample, is the model predicted signal-to-noise ratio degradation of the kth frequency band in the ith sample; is the real amplitude attenuation of the kth frequency band of the ith sample, is the amplitude attenuation prediction value of the kth frequency band of the ith sample by the model; is a deviation threshold, is the comprehensive prediction deviation of the ith sample in the mth high-priority frequency band, and m is the high-priority frequency band index, is the real amplitude attenuation of the mth high-priority frequency band of the ith sample, is the amplitude attenuation prediction value of the mth high-priority frequency band of the ith sample by the model; is the real signal-to-noise ratio degradation of the mth high-priority frequency band of the ith sample, a predicted signal-to-noise ratio degradation amount of the mth high-priority frequency band of the ith sample; a penalty coefficient.

[0021] Further, after the step of compensating each frequency band of the stage performance audio signal to be output based on the compensation gain to obtain a compensated audio signal, the method further comprises:

[0022] collecting air humidity and air temperature;

[0023] calculating a time compensation value of each sound box in the multi-sound box system based on the air humidity and the air temperature;

[0024] controlling the compensated audio signal to be played in each sound box based on the time compensation value of each sound box.

[0025] Further, the step of calculating the time compensation value of each sound box in the multi-sound box system based on the air humidity and the air temperature comprises:

[0026] calculating a current sound speed according to the air humidity and the air temperature;

[0027] calculating a time compensation value of each sound box in the multi-sound box system based on the sound speed when the air is dry and the current sound speed.

[0028] Further, the step of calculating the current sound speed according to the air humidity and the air temperature comprises:

[0029] calculating the current sound speed according to the formula ; wherein v is the current sound speed, T is the air temperature in Celsius, and h is the relative humidity percentage.

[0030] Further, the step of calculating the time compensation value of each sound box in the multi-sound box system based on the air humidity and the air temperature comprises:

[0031] calculating the time compensation value of each sound box in the multi-sound box system according to the formula ; wherein Δt is the time compensation value of a target sound box in the multi-sound box system, d is the distance between the target sound box and a reference point of the audience, v is the current sound speed, is the sound speed when the air is dry.

[0032] In a second aspect, the embodiments of the present application provide an audio control system for stage performance, which comprises:

[0033] a collecting module, configured to collect real-time environmental parameters of an open-air stage, the environmental parameters comprising at least two of rainfall, wind speed, environmental noise, temperature and humidity, and audience noise;

[0034] The input module is configured to input the real-time environmental parameters into a preset dynamic interaction model, and output disturbed amounts of each frequency band; the dynamic interaction model is a nonlinear model trained by a plurality of factor combination samples, and the plurality of factor combinations include actual disturbed amounts of each frequency band under different environmental parameter combinations;

[0035] The calculation module is configured to calculate compensation gains of each frequency band based on the disturbed amounts of each frequency band;

[0036] The compensation module is configured to compensate each frequency band of the stage performance audio signal to be output based on the compensation gains, to obtain a compensated audio signal.

[0037] Further, the disturbed amounts include amplitude attenuation amounts and signal-to-noise ratio (SNR) deterioration amounts, and the training process of the dynamic interaction model includes:

[0038] N environmental parameter combination samples are collected, each sample including specific values of M environmental parameters and actual disturbed amounts of each frequency band under a corresponding scene;

[0039] The environmental parameter combination samples are divided into a training set and a verification set;

[0040] A neural network model is trained by taking the environmental parameters in the training set as input, the disturbed amounts of each frequency band as output, and a preset loss function;

[0041] The trained model is verified for accuracy by the verification set, and when the verification accuracy is greater than a preset threshold, the model is determined as the dynamic interaction model.

[0042] Further, the loss function for training the dynamic interaction model is:

[0043] ;

[0044] wherein, ;

[0045] ;

[0046] ;

[0047] ;

[0048] wherein, is a total loss, is an amplitude attenuation amount L1 loss of each frequency band, is an MSE loss of an SNR deterioration amount of each frequency band, is a high-priority frequency band prediction deviation penalty term, 、 、 is a weight coefficient, and ; K is the number of frequency bands, k is the frequency band index, is the weight coefficient of the kth frequency band, the weight coefficient of the high priority frequency band is greater than the weight coefficient of the low priority frequency band, is the real signal-to-noise ratio degradation of the kth frequency band of the ith sample, is the model predicted signal-to-noise ratio degradation of the kth frequency band in the ith sample; is the real amplitude attenuation of the kth frequency band of the ith sample, is the model predicted amplitude attenuation of the kth frequency band of the ith sample; is the deviation threshold, is the comprehensive predicted deviation of the ith sample in the mth high priority frequency band, m is the high priority frequency band index, is the real amplitude attenuation of the mth high priority frequency band of the ith sample, is the model predicted amplitude attenuation of the mth high priority frequency band of the ith sample; is the real signal-to-noise ratio degradation of the mth high priority frequency band of the ith sample, is the predicted signal-to-noise ratio degradation of the mth high priority frequency band of the ith sample; is the penalty coefficient.

[0049] The present application has the following technical effects:

[0050] (1) The stage performance audio control method provided by the embodiment of the present application comprises: collecting real-time environmental parameters of an open-air stage, the environmental parameters comprising at least two of rainfall, wind speed, environmental noise, temperature and humidity, and audience noise; inputting the real-time environmental parameters into a preset dynamic interaction model to output disturbed quantities of each frequency band; the dynamic interaction model is a nonlinear model trained through multi-factor combined samples, and the multi-factor combination contains actual disturbed quantities of each frequency band under different environmental parameter combinations; calculating compensation gains of each frequency band based on the disturbed quantities of each frequency band; and compensating each frequency band of a stage performance audio signal to be output based on the compensation gains to obtain a compensated audio signal. Compared with the prior art, the present application replaces the fixed overlap of preset gains and preset weights with a data-driven dynamic interaction model, no longer uses artificial rule hard solution for complex interaction, but lets the system adapt to environmental changes through learning and feedback, which can more accurately model complex interaction relationships and realize clear and distinguishable key audio content regardless of how environmental factors are superimposed.

[0051] (2) Because the air humidity and temperature will change the sound propagation speed in the air during rainfall, the time compensation value of each sound box in the multi-sound box system is calculated based on the air humidity and air temperature, so that the audio can be better controlled to be synchronized with other stage elements (such as actions), and the stage performance effect is ensured. At the same time, the sound signals of the sound boxes at different positions are ensured to be synchronized to reach the audience position, the phase difference caused by the change of the sound speed is eliminated, and the phenomenon of "tail" or "overlap" of the sound is avoided. BRIEF DESCRIPTION OF DRAWINGS

[0052] In order to more clearly illustrate the technical solutions of the present application, the drawings needed in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0053] Figure 1 is a flowchart of the audio control method for stage performance provided by the embodiments of the present application. DETAILED DESCRIPTION

[0054] In order to make the purposes, technical solutions and advantages of the present application more clear, the present application will be further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application.

[0055] Those skilled in the art can understand that, unless otherwise defined, all the terms (including technical terms and scientific terms) used herein have the same meaning as that generally understood by those skilled in the art in the field to which the present application belongs. It should also be understood that, those terms such as those defined in a general dictionary should be understood as having the same meaning as in the context of the prior art, and should not be interpreted in an idealized or overly formal sense unless specifically defined as such.

[0056] It should be noted that, in this document, the term "comprise", "include" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, device, article or method comprising a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such a process, device, article or method. Without more limitations, the element defined by the statement "comprises a" does not exclude the presence of additional identical elements in the process, device, article or method comprising the element.

[0057] As shown in Figure 1 The embodiments of the present application provide an audio control method for stage performance, which comprises:

[0058] S1, collect real-time environment parameters of an open-air stage, the environment parameters including at least two of rainfall, wind speed, environmental noise, temperature and humidity, and audience noise;

[0059] S2, input the real-time environment parameters into a preset dynamic interaction model, and output disturbed quantities of each frequency band; the dynamic interaction model is a nonlinear model trained through multi-factor combined samples, and the multi-factor combination includes actual disturbed quantities of each frequency band under different environment parameter combinations;

[0060] S3, calculate compensation gains of each frequency band based on the disturbed quantities of each frequency band;

[0061] S4, compensate each frequency band of a stage performance audio signal to be output based on the compensation gains, and obtain a compensated audio signal.

[0062] Compared with the prior art, the application replaces the fixed overlap of preset gains and preset weights with a data-driven dynamic interaction model, no longer uses artificial rule hard solution for complex interaction, but makes the system adapt to environmental changes through learning and feedback, so that the complex interaction relationship can be accurately modeled, and the key audio content can be clearly identified regardless of the superposition of environmental factors.

[0063] In an embodiment, the disturbed quantity includes an amplitude attenuation quantity and a signal-to-noise ratio deterioration quantity, and the training process of the dynamic interaction model includes:

[0064] Collect N environment parameter combination samples, each sample including specific values of M environment parameters and actual disturbed quantities of each frequency band in the corresponding scene;

[0065] Divide the environment parameter combination samples into a training set and a validation set;

[0066] Train a neural network model using the environment parameters in the training set as input, the disturbed quantities of each frequency band as output, and a preset loss function to train the neural network model;

[0067] Verify the accuracy of the trained model through the validation set, and determine the dynamic interaction model when the verification accuracy is greater than a preset threshold.

[0068] In the embodiments of the application, the environment parameters can include rainfall, wind speed, environmental noise, temperature and humidity, and audience noise. The neural network model can use a multi-layer perceptron, a recurrent neural network, etc. The amplitude attenuation quantity refers to the difference between the ideal output amplitude and the actual output amplitude of a certain frequency band. The signal-to-noise ratio deterioration quantity refers to the difference between the ideal signal-to-noise ratio and the disturbed signal-to-noise ratio of a certain frequency band.

[0069] In order to make the trained model more accurately predict the disturbed quantity of the key audio, in an embodiment, the loss function for training the dynamic interaction model is:

[0070] ;

[0071] wherein, ;

[0072] ;

[0073] ;

[0074] ;

[0075] wherein, is the total loss, is the amplitude attenuation loss of each frequency band, is the MSE loss of the signal-to-noise ratio deterioration of each frequency band, is the prediction deviation penalty term of the high-priority frequency band, , , is a weight coefficient, and ; K is the number of frequency bands, and k is the frequency band index, is the weight coefficient of the kth frequency band, and the weight coefficient of the high-priority frequency band is greater than the weight coefficient of the low-priority frequency band, is the real signal-to-noise ratio deterioration of the kth frequency band of the ith sample, is the model-predicted signal-to-noise ratio deterioration of the kth frequency band in the ith sample; is the real amplitude attenuation of the kth frequency band of the ith sample, is the model-predicted amplitude attenuation of the kth frequency band of the ith sample; is a deviation threshold value, is the comprehensive prediction deviation of the ith sample in the mth high-priority frequency band, and m is the high-priority frequency band index, is the real amplitude attenuation of the mth high-priority frequency band of the ith sample, is the model-predicted amplitude attenuation of the mth high-priority frequency band of the ith sample; is the real signal-to-noise ratio deterioration of the mth high-priority frequency band of the ith sample, is the predicted signal-to-noise ratio deterioration of the mth high-priority frequency band of the ith sample; is a penalty coefficient.

[0076] In an embodiment, the disturbed quantity includes an amplitude attenuation and a signal-to-noise ratio deterioration, and the step of calculating a compensation gain of each frequency band based on the disturbed quantity of each frequency band includes:

[0077] The compensation gain of each frequency band is calculated according to the following formula:

[0078] ;

[0079] ;

[0080] ;

[0081] wherein, is the compensation gain of a certain frequency band, is the weight coefficient of the amplitude compensation gain, is the weight coefficient of the signal-to-noise ratio compensation gain, is the amplitude compensation gain, is the amplitude attenuation of a certain frequency band, is the amplitude compensation coefficient, the value range is generally between 1.0-1.3, which needs to be adjusted according to the acoustic characteristics of the open-air stage. For example, when there are many reflecting objects on the open-air stage, the acoustic environment is relatively complex, in order to avoid excessive compensation, A can be taken as 1.0-1.1; in the open square and other environments where sound waves are easy to diffuse, in order to make up for the additional signal attenuation, A can be taken as 1.2-1.3; is the signal-to-noise ratio compensation gain, is the signal-to-noise ratio deterioration of a certain frequency band, is the signal-to-noise ratio threshold, the value range is generally between 3-5dB. When , it means that the noise interference is relatively weak, and the influence on the signal clarity is small, and no additional compensation is needed to avoid unnecessary compensation fluctuations caused by small disturbances; B is the signal-to-noise ratio compensation coefficient; the value range is between 0.8-1.2, which needs to be adjusted according to the signal type. For the human voice frequency band (1-4kHz), since the human ear is sensitive to speech clarity, in order to prioritize the guarantee of speech clarity and intelligibility, B can be taken as 1.1-1.2; while for the low frequency band of musical instruments (20-200Hz), the sensitivity of low frequency signals to signal-to-noise ratio is relatively low, in order to avoid excessive compensation leading to audio turbidity, B can be taken as 0.8-0.9.

[0082] In an embodiment, after the step of compensating each frequency band of the stage performance audio signal to be output based on the compensation gain to obtain a compensated audio signal, the method further comprises:

[0083] S51, collecting air humidity and air temperature;

[0084] S52, calculating a time compensation value of each sound box in a multi-sound box system based on the air humidity and the air temperature;

[0085] S53, controlling the compensated audio signal to be played in each sound box based on the time compensation value of each sound box.

[0086] In step S52, the step of calculating the time compensation value of each sound box in the multi-sound box system based on the air humidity and air temperature comprises:

[0087] S521, calculating the current sound speed according to the air humidity and air temperature;

[0088] S522, calculating the time compensation value of each sound box in the multi-sound box system based on the sound speed when the air is dry and the current sound speed.

[0089] In step S521, the step of calculating the current sound speed according to the air humidity and air temperature comprises:

[0090] According to the formula the current sound speed is calculated; wherein v is the current sound speed, T is the air temperature in Celsius, and h is the relative humidity percentage.

[0091] In step S522, the step of calculating the time compensation value of each sound box in the multi-sound box system based on the air humidity and air temperature comprises:

[0092] According to the formula the time compensation value of each sound box in the multi-sound box system is calculated; wherein Δt is the time compensation value of the target sound box in the multi-sound box system, d is the distance between the target sound box and the reference point of the audience, v is the current sound speed, and v is the sound speed when the air is dry.

[0093] In the embodiments of the present application, since the air humidity and temperature change the sound propagation speed in the air during rainfall, calculating the time compensation value of each sound box in the multi-sound box system based on the air humidity and air temperature can better control the synchronization of audio with other stage elements (such as actions), ensuring the stage performance effect. At the same time, it ensures that the sound signals of different position sound boxes reach the audience position synchronously, eliminates the phase difference caused by the change of sound speed, and avoids the phenomenon of "tail" or "overlap" of sound.

[0094] The embodiments of the present application also provide an audio control system for stage performance, the system comprising:

[0095] The acquisition module 1 is used to acquire real-time environmental parameters of an open-air stage, and the environmental parameters include at least two of rainfall, wind speed, environmental noise, temperature and humidity, and audience noise;

[0096] The input module 2 is used to input the real-time environmental parameters into a preset dynamic interaction model to output the disturbed amount of each frequency band; the dynamic interaction model is a nonlinear model trained by a multi-factor combination sample, and the multi-factor combination includes the actual disturbed amount of each frequency band under different environmental parameter combinations;

[0097] a calculation module 3, configured to calculate a compensation gain of each frequency band based on the disturbance of each frequency band;

[0098] a compensation module 4, configured to compensate each frequency band of the stage performance audio signal to be output based on the compensation gain, to obtain a compensated audio signal.

[0099] In an embodiment, the disturbance includes an amplitude attenuation and a signal-to-noise ratio (SNR) degradation, and the training process of the dynamic interaction model includes:

[0100] collecting N combinations of environmental parameters, each combination including specific values of M environmental parameters and actual disturbances of each frequency band in a corresponding scene;

[0101] dividing the combination samples of the environmental parameters into a training set and a validation set;

[0102] training a neural network model using the environmental parameters in the training set as input, the disturbances of each frequency band as output, and a preset loss function to train the neural network model;

[0103] verifying the accuracy of the trained model through the validation set, and determining the dynamic interaction model when the verification accuracy is greater than a preset threshold.

[0104] In an embodiment, the loss function for training the dynamic interaction model is:

[0105]

[0106] wherein,

[0107]

[0108]

[0109]

[0110] wherein, is the total loss, is the amplitude attenuation loss L1 of each frequency band, is the MSE loss of the SNR degradation of each frequency band, is a high-priority frequency band prediction deviation penalty term, , , is a weight coefficient, and ; K is the number of frequency bands, and k is the frequency band index, is the weight coefficient of the kth frequency band, and the weight coefficient of the high-priority frequency band is greater than the weight coefficient of the low-priority frequency band, is the actual SNR degradation of the kth frequency band of the ith sample, ​​​​​a model predicted signal-to-noise ratio degradation amount for the kth frequency band in the ith sample; a real amplitude attenuation amount for the kth frequency band in the ith sample, a model predicted amplitude attenuation amount for the kth frequency band in the ith sample; a deviation threshold value, a comprehensive predicted deviation of the ith sample in the mth high-priority frequency band, m being a high-priority frequency band index, a real amplitude attenuation amount for the mth high-priority frequency band in the ith sample, a model predicted amplitude attenuation amount for the mth high-priority frequency band in the ith sample; a real signal-to-noise ratio degradation amount for the mth high-priority frequency band in the ith sample, a predicted signal-to-noise ratio degradation amount for the mth high-priority frequency band in the ith sample; a penalty coefficient.

[0111] The above merely describes preferred embodiments of the present application, and is not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation, or direct or indirect application in other related technical fields, which is made by using the content of the present application specification and drawings, is also included in the patent protection scope of the present application.

Claims

1. An audio control method for stage performances, characterized in that, The method includes: Real-time environmental parameters of the open-air stage are collected, including at least two of the following: rainfall, wind speed, ambient noise, temperature and humidity, and audience noise. The real-time environmental parameters are input into a preset dynamic interaction model, which outputs the disturbance amount of each frequency band. The dynamic interaction model is a nonlinear model trained by multi-factor combination samples, and the multi-factor combination includes the actual disturbance amount of each frequency band under different combinations of environmental parameters. The compensation gain for each frequency band is calculated based on the amount of disturbance in each frequency band. The compensated audio signal is obtained by compensating each frequency band of the stage performance audio signal to be output based on the compensation gain.

2. The audio control method for stage performance according to claim 1, characterized in that, The disturbance includes amplitude attenuation and signal-to-noise ratio degradation. The training process of the dynamic interaction model includes: Collect N environmental parameter combination samples. Each sample includes the specific values ​​of M environmental parameters and the actual disturbance amount of each frequency band in the corresponding scenario. The environmental parameter combination samples are divided into a training set and a validation set; A neural network model is used, with environmental parameters from the training set as input, the disturbance amount in each frequency band as output, and a preset loss function to train the neural network model; The trained model is validated using a validation set. When the validation accuracy is greater than a preset threshold, it is determined to be the dynamic interaction model.

3. The audio control method for stage performance according to claim 2, characterized in that, The loss function for training the dynamic interaction model is: ； in, ; ； ； ； in, For the total loss, The amplitude attenuation L1 loss for each frequency band, The MSE loss is the amount of signal-to-noise ratio degradation in each frequency band. This is a penalty term for prediction bias in high-priority frequency bands. , , These are the weighting coefficients, and K represents the number of frequency bands, and k represents the frequency band index. This represents the weighting coefficient for the k-th frequency band. The weighting coefficient for higher-priority frequency bands is greater than that for lower-priority frequency bands. This represents the actual signal-to-noise ratio degradation in the k-th frequency band for the i-th sample. The model predicts the signal-to-noise ratio degradation for the k-th frequency band in the i-th sample; This represents the actual amplitude attenuation of the i-th sample in the k-th frequency band. This is the predicted amplitude attenuation value of the model for the k-th frequency band of the i-th sample; The deviation threshold, Let be the overall prediction bias of the i-th sample in the m-th high-priority frequency band, where m is the index of the high-priority frequency band. This represents the actual amplitude attenuation of the m-th high-priority frequency band for the i-th sample. This is the predicted amplitude attenuation value of the model for the m-th high-priority frequency band of the i-th sample; This represents the actual signal-to-noise ratio degradation of the m-th high-priority frequency band for the i-th sample. The predicted signal-to-noise ratio degradation of the m-th high-priority frequency band for the i-th sample; This is the penalty coefficient.

4. The audio control method for stage performance according to claim 1, characterized in that, After the step of compensating each frequency band of the stage performance audio signal to be output based on the compensation gain to obtain the compensated audio signal, the method further includes: Collect air humidity and air temperature; Calculate the time compensation value of each speaker in the multi-speaker system based on the air humidity and air temperature; The compensated audio signal is controlled to be played on each speaker based on the time compensation value of each speaker.

5. The audio control method for stage performance according to claim 4, characterized in that, The step of calculating the time compensation value of each speaker in the multi-speaker system based on the air humidity and air temperature includes: Calculate the current speed of sound based on the air humidity and air temperature; Calculate the time compensation value of each speaker in a multi-speaker system based on the speed of sound when the air is dry and the current speed of sound.

6. The audio control method for stage performance according to claim 5, characterized in that, The step of calculating the current speed of sound based on the air humidity and air temperature includes: According to the formula Calculate the current speed of sound; where v is the current speed of sound, T is the air temperature in degrees Celsius, and h is the relative humidity percentage.

7. The audio control method for stage performance according to claim 5, characterized in that, The step of calculating the time compensation value of each speaker in the multi-speaker system based on the air humidity and air temperature includes: According to the formula Calculate the time compensation value for each speaker in the multi-speaker system; where Δt is the time compensation value of the target speaker in the multi-speaker system, d is the distance between the target speaker and the audience reference point, and v is the current speed of sound. The speed of sound when the air is dry.

8. An audio control system for stage performances, characterized in that, The system includes: The data acquisition module is used to acquire real-time environmental parameters of the open-air stage, including at least two of the following: rainfall, wind speed, ambient noise, temperature and humidity, and audience noise. The input module is used to input the real-time environmental parameters into a preset dynamic interaction model and output the disturbance amount of each frequency band; the dynamic interaction model is a nonlinear model obtained by training through multi-factor combination samples, and the multi-factor combination includes the actual disturbance amount of each frequency band under different combinations of environmental parameters. The calculation module is used to calculate the compensation gain of each frequency band based on the disturbance amount of each frequency band; The compensation module is used to compensate each frequency band of the stage performance audio signal to be output based on the compensation gain, so as to obtain a compensated audio signal.

9. The audio control system for stage performance according to claim 8, characterized in that, The disturbance includes amplitude attenuation and signal-to-noise ratio degradation. The training process of the dynamic interaction model includes: Collect N environmental parameter combination samples. Each sample includes the specific values ​​of M environmental parameters and the actual disturbance amount of each frequency band in the corresponding scenario. The environmental parameter combination samples are divided into a training set and a validation set; A neural network model is used, with environmental parameters from the training set as input, the disturbance amount in each frequency band as output, and a preset loss function to train the neural network model; The trained model is validated using a validation set. When the validation accuracy is greater than a preset threshold, it is determined to be the dynamic interaction model.

10. The audio control system for stage performance according to claim 9, characterized in that, The The loss function for training the dynamic interaction model is: ； in, ; ； ； ； in, For the total loss, The amplitude attenuation L1 loss for each frequency band, The MSE loss is the amount of signal-to-noise ratio degradation in each frequency band. This is a penalty term for prediction bias in high-priority frequency bands. , , These are the weighting coefficients, and K represents the number of frequency bands, and k represents the frequency band index. This represents the weighting coefficient for the k-th frequency band. The weighting coefficient for higher-priority frequency bands is greater than that for lower-priority frequency bands. This represents the actual signal-to-noise ratio degradation in the k-th frequency band for the i-th sample. The model predicts the signal-to-noise ratio degradation for the k-th frequency band in the i-th sample; This represents the actual amplitude attenuation of the i-th sample in the k-th frequency band. This is the predicted amplitude attenuation value of the model for the k-th frequency band of the i-th sample; The deviation threshold, Let be the overall prediction bias of the i-th sample in the m-th high-priority frequency band, where m is the index of the high-priority frequency band. This represents the actual amplitude attenuation of the m-th high-priority frequency band for the i-th sample. This is the predicted amplitude attenuation value of the model for the m-th high-priority frequency band of the i-th sample; This represents the actual signal-to-noise ratio degradation of the m-th high-priority frequency band for the i-th sample. The predicted signal-to-noise ratio degradation of the m-th high-priority frequency band for the i-th sample; This is the penalty coefficient.

Images