CN110362789B - GPR model-based adaptive sound masking system and method - Google Patents
GPR model-based adaptive sound masking system and method Download PDFInfo
- Publication number
- CN110362789B CN110362789B CN201910656438.8A CN201910656438A CN110362789B CN 110362789 B CN110362789 B CN 110362789B CN 201910656438 A CN201910656438 A CN 201910656438A CN 110362789 B CN110362789 B CN 110362789B
- Authority
- CN
- China
- Prior art keywords
- sound
- masking
- satisfaction
- database
- adaptive
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000000873 masking effect Effects 0.000 title claims abstract description 163
- 238000000034 method Methods 0.000 title claims abstract description 58
- 230000003044 adaptive effect Effects 0.000 title claims description 29
- 238000012216 screening Methods 0.000 claims abstract description 40
- 230000008569 process Effects 0.000 claims abstract description 18
- 230000006870 function Effects 0.000 claims abstract description 13
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 9
- 241000238631 Hexapoda Species 0.000 claims abstract description 7
- 230000007613 environmental effect Effects 0.000 claims description 17
- 230000007246 mechanism Effects 0.000 claims description 15
- 230000015654 memory Effects 0.000 claims description 13
- 239000011159 matrix material Substances 0.000 claims description 10
- 238000011156 evaluation Methods 0.000 claims description 9
- 238000005070 sampling Methods 0.000 claims description 8
- 230000006054 immunological memory Effects 0.000 claims description 7
- 238000012847 principal component analysis method Methods 0.000 claims description 5
- 238000012549 training Methods 0.000 claims description 5
- 238000010276 construction Methods 0.000 claims description 4
- 238000012360 testing method Methods 0.000 claims description 4
- 238000000513 principal component analysis Methods 0.000 claims 1
- 238000005516 engineering process Methods 0.000 abstract description 29
- 239000003795 chemical substances by application Substances 0.000 abstract description 10
- 238000013473 artificial intelligence Methods 0.000 abstract description 9
- 238000002474 experimental method Methods 0.000 description 7
- 230000009467 reduction Effects 0.000 description 5
- 230000003213 activating effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000009413 insulation Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 210000001015 abdomen Anatomy 0.000 description 1
- 238000010521 absorption reaction Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000008076 immune mechanism Effects 0.000 description 1
- 210000000987 immune system Anatomy 0.000 description 1
- 230000036039 immunity Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Computational Mathematics (AREA)
- Mathematical Optimization (AREA)
- Theoretical Computer Science (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Analysis (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Operations Research (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- Computing Systems (AREA)
- Soundproofing, Sound Blocking, And Sound Damping (AREA)
Abstract
The invention relates to a GPR model-based self-adaptive sound masking system and a GPR model-based self-adaptive sound masking method, wherein the system comprises a masking sound database composed of various intermittent insect sound, bird sound, water flow sound and other natural sounds, a sound satisfaction agent model modeled by a Gaussian process regression algorithm and a self-adaptive masking sound screening subsystem built based on the database and the agent model, and the system has the functions of screening optimal masking sound in the database based on a unified sound satisfaction agent model aiming at different medium-low frequency background noise, and obtaining corresponding combined sound and satisfaction values thereof. Compared with the prior art, the invention applies the artificial intelligence method to the sound masking technology for the first time, builds a complete self-adaptive sound masking system, improves the feasibility and universality of the sound masking technology, saves the screening time of masking sound, and reduces the cost of relying on manual adjustment in the traditional sound masking technology.
Description
Technical Field
The invention relates to the technical field of environmental noise control, in particular to a GPR model-based adaptive sound masking system and method.
Background
With the rapid development of urban areas, the problem of environmental noise pollution is increasingly concerned by the public. The traditional noise reduction technology is concentrated on noise reduction of a sound source, sound absorption and insulation of a propagation path process and sound insulation treatment of the environment of a noise receiver, so that the noise level of the urban environment is reduced to a certain extent, and the living environment of residents is improved. However, due to the limitations of the traditional noise reduction measures in the treatment of traffic noise and fixed equipment noise and the characteristic that medium-low frequency noise is not easy to attenuate, urban environment noise is often mainly characterized by medium-low frequency, and particularly, the problem that sound level reaches the standard but still disturbs people in dense cities is solved. The noise in the abdomen of residential communities along urban arterial roads and in urban public green areas is tracked and monitored, and the sound level of the areas is mostly maintained at a steady-state level of 55-60dBA, however, the public places are in such an environment, the satisfaction degree is still lower, and the satisfaction degree of areas with higher sound levels is lower.
In recent years, studies have found that the public satisfaction with the acoustic environment is not only dependent on the sound pressure level, but that pleasant sounds can improve the human satisfaction with the acoustic environment even if the sound pressure level is high. In practice, this is the application of psycho-acoustics in ambient acoustics. It has been found that water sounds (e.g., fountain, running water, etc.) and bird song, etc., help to improve the annoyance caused by traffic noise. Therefore, researchers begin to consider using sound masking technology to improve public satisfaction with the sound environment in which they are located, and this method has good application prospects and market demands. At present, the environmental noise control technology related to the sound masking method still stays in the starting stage, and related patents of the sound masking method of the system are not reported yet. The only relatively complete sound masking method at present is a sound masking-based method for improving sound environment satisfaction degree studied by the environmental science institute in 2018, and the main functions of the method are to manually select natural sounds with the same effective characteristics aiming at individual effective characteristics of environmental background noise, and then manually iterate and adjust corresponding effective characteristics of the natural sounds to obtain the optimal masking sound. The method does not determine the unified effective characteristics of each background sound, does not establish a unified satisfaction model depending on effective characteristic parameters, and also stays in the stage of manual test for selecting and adjusting the optimal masking sound. In fact, due to the numerous characteristic parameters affecting satisfaction, the acoustic satisfaction model is a complex high-dimensional nonlinear system, and researchers have difficulty in identifying the intrinsic physical mechanism of the complex system, so that the biggest difficulty of the traditional acoustic masking technology is in building the acoustic satisfaction mechanism model.
Currently, artificial intelligence technology has been widely applied to various fields of the internet, industry, agriculture, traffic, etc., with excellent results. In recent years, there has been a growing interest in researchers in the environmental field. In the field of environmental noise control, researchers have been trying to apply artificial intelligence techniques to traditional noise control methods. In fact, the big data technology in the artificial intelligence technology can well solve the problem that the mechanism characteristic in the traditional noise technology is difficult to obtain. Therefore, the artificial intelligence technology is applied to the sound masking method, and an adaptive sound masking system with universality for improving the satisfaction of the sound environment is established by utilizing a database, a proxy model and an immune memory mechanism in the artificial intelligence method, so that the sound masking technology which is difficult to realize in the traditional method can be constructed into a complete and feasible sound masking system, the feasibility and the practicability of the sound masking technology are improved, the labor cost of the sound masking technology is avoided, and the time consumption is reduced.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a GPR model-based adaptive sound masking system and a GPR model-based adaptive sound masking method, which are combined with a database technology and a proxy model method, so that the feasibility and universality of the sound masking technology can be improved, and the labor and time cost for implementing the technology can be reduced.
The invention has the innovation points that the effective characteristics of the satisfaction degree of the acoustic environment are determined, a sufficient amount of natural sound samples are collected to construct a masking sound database, and a memory mechanism in an immune system is introduced into a masking sound screening subsystem by using a proxy model to approximate the universal acoustic satisfaction degree model. It can search the masking sound database for the optimal masking sound based directly on the satisfaction agent model, and at the same time, generate the memory cell of the masking response behavior, so that the system can mask the repeated background sound directly next time. The invention effectively improves the feasibility and practicality of the sound masking technology, and greatly saves a great deal of cost and time consumption generated by manual iteration of masking sound.
The aim of the invention can be achieved by the following technical scheme:
an adaptive sound masking system based on a GPR model is established by a masking sound database, a sound satisfaction agent model and an adaptive masking sound screening subsystem, wherein the masking sound database is formed by a large number of intermittent insect sound, bird sound, water flow sound and other various natural sounds, the sound satisfaction agent model is formed by modeling through a Gaussian process regression algorithm, and the adaptive masking sound screening subsystem is established based on the masking sound database and the sound satisfaction agent model and is introduced into an immune memory mechanism.
The invention also provides a construction method for the adaptive sound masking system based on the GPR model, which comprises the following steps:
step 1: collecting background sound samples and natural sound samples and establishing a masking sound database by using collected sample data;
step 2: regression combining the acoustic satisfaction proxy model using the GPR model;
step 3: an adaptive masking sound screening subsystem is built on a masking sound database based on a sound satisfaction proxy model.
Further, the step 1 includes the following sub-steps:
step 11: collecting environmental background sound samples of different scenes, seasons and time periods, obtaining basic characteristics of background noise through subjective evaluation experiments, and obtaining effective characteristic parameters affecting the satisfaction degree of the acoustic environment by using a principal component analysis method;
step 12: and acquiring intermittent insect sound, bird sound, water flow sound and other natural sound data according to the effective characteristic parameters, and establishing a masking sound database by utilizing the acquired natural sound data and the corresponding effective characteristic parameter values.
Further, the effective characteristic parameters in the step 11 refer to main characteristics affecting the satisfaction degree of the background sound, and are obtained by analyzing the main components of the basic physical characteristics of the noise, such as the sound level of A, the sound pressure level of the characteristic frequency, the loudness, the roughness, the sharpness, the shaking degree, the tone schedule and the like.
Further, the step 2 includes the following sub-steps:
step 21: collecting a sufficient amount of masking sound samples in a masking sound database by adopting a uniform sampling method;
step 22: combining the collected environmental background sound sample with the masking sound sample to obtain a combined sound sample, and obtaining an effective characteristic parameter value of the combined sound;
step 23: carrying out satisfaction subjective evaluation test on all the combined sounds in the combined sound sample and obtaining corresponding satisfaction observation values;
step 24: taking all combined sounds in the combined sound sample and the corresponding satisfaction observation values as training samples;
step 25: consider a combined acoustic satisfaction proxy model and model the proxy model using a gaussian process regression algorithm.
Further, the proxy model in the step 24 has the following description formula:
in which acoustic samples are combinedRepresents a d-dimensional input variable constituted by the combined sound effective feature quantity,representing a potential (true) satisfaction function, y (·) representing the value corresponding to the combined sound sample +.>The random deviation variable epsilon indicates that the real acoustic satisfaction value is subject to different satisfaction observation functionsInfluence of subjective differences among individuals.
Further, the prediction formula of the gaussian process regression algorithm in the step 25 is:
in the method, in the process of the invention,respectively represent any combination sound +.>A predicted value and a predicted variance of satisfaction of (c),represent satisfaction observation vector, K *,mn 、K mn And K ** All represent a functional covariance matrix,>representing a bias covariance matrix, I mn Representing an mn-level identity matrix.
Further, the step 3 includes the following sub-steps:
step 31: identifying a new incoming background sound and simultaneously activating the memory cells;
step 32: if the new background sound identification result is the repeated background sound, the optimal masking sound is directly matched and memorized in a masking sound database of the self-adaptive sound masking system;
step 33: if the new background sound identification result is brand-new background sound, calculating the combined sound satisfaction value of the background sound and the masking sound database one by utilizing a prediction formula of the combined sound satisfaction, screening the database to obtain the optimal masking sound corresponding to the maximum satisfaction value, and generating a memory cell at the same time;
step 04: outputting the optimal masking sound and performing control intervention on the background sound.
Compared with the prior art, the invention has the following advantages:
1. sound masking techniques based on masking effects: the invention introduces the thought of masking effect in psychoacoustics into the noise control field, and provides an acoustic masking technology for improving subjective feeling of people. When the environmental sound level is difficult to reduce and the satisfaction degree of the acoustic environment of the crowd is low, the satisfaction degree of the crowd on the acoustic environment is improved by adding effective masking sound without considering how to strive to reduce noise, and the satisfaction degree of the public on the acoustic environment is improved by using the acoustic masking technology from the viewpoint of subjective perception of the crowd by considering the psychoacoustic masking benefit.
2. Sound masking systems with universality: the invention establishes a complete self-adaptive sound masking system by utilizing the database, the agent model, the artificial immunity and other artificial intelligent technologies, effectively improves the feasibility and universality of the sound masking technology, and improves the application prospect of the sound masking technology. The application of the artificial intelligence method constructs the sound masking technology which is difficult to realize by the traditional method into a complete and feasible sound masking system, wherein the application of the database lays a data foundation for automatic screening of masking sound; the application of the Gaussian process regression algorithm realizes the screening of the masking sound system with high precision and high efficiency.
3. Combined sound satisfaction model: the invention unifies the effective characteristics of background sound, establishes a masking sound database according to the effective characteristics, further provides a combined sound satisfaction model based on effective characteristic parameters, and can greatly simplify the modeling complexity of a sound masking system, wherein the effective characteristic database firstly determines the effective characteristics influencing the sound environment according to environmental background noise samples with representative meanings of different environments, seasons and time periods; and secondly, a sufficient number of masking sound databases applicable to different middle-low frequency background sounds are established based on the effective characteristics. The construction of the database provides a data basis for the establishment of the acoustic satisfaction proxy model and the automatic screening of masking sounds.
4. GPR (Gaussian process regression) model regression acoustic satisfaction model: the method is based on the thought of the proxy model, utilizes the GPR model regression combination acoustic satisfaction model, and effectively identifies a high-latitude nonlinear complex system which is difficult to obtain by a mechanism model. And a GPR algorithm regression combined sound satisfaction agent model is adopted to deduce a prediction formula of the combined sound satisfaction, so that a high-precision screening mechanism is provided for the masking sound screening subsystem. Moreover, the influence of individual subjective deviation on the acoustic satisfaction value is considered in the GPR model, and the accuracy and reliability of satisfaction prediction are further improved.
5. Acoustic masking system based on immune mechanism: the invention designs a masking sound screening subsystem based on a masking sound database and a combined sound satisfaction agent model, introduces an immune memory mechanism into the screening subsystem, can enhance the screening adaptability of the subsystem, improves the masking sound matching efficiency, saves screening time, and on one hand, realizes the optimal matching of background sound by taking a combined sound satisfaction prediction formula as a screening mechanism and automatically screening in the masking sound database; on the other hand, the immune memory mechanism is introduced into the screening subsystem, so that the direct matching of repeated background sounds is realized, the self-adaptability of the screening subsystem of the masking sound is further enhanced, and the screening efficiency of the sound masking system is improved.
Drawings
FIG. 1 is a flow chart of a method of constructing an adaptive sound masking system based on a GPR model of the present invention;
FIG. 2 is a schematic block diagram of an adaptive sound masking system of the present invention;
fig. 3 is a masking sound screening flow chart of the adaptive sound masking system of the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
Studies have shown that sound masking techniques can improve sound environment satisfaction when the ambient background noise is below 70dB (a), and that noise reduction measures should be first taken to handle when the sound level is above 70dB (a). The invention provides an intelligent sound masking technology for improving the satisfaction of sound environment aiming at the background noise lower than 70dB (A). The basic principle is based on the concept of masking effect in psychoacoustics, and proper masking sound is selected to adjust the physical and psychoacoustics characteristics of background sound, so that the satisfaction degree of people on the sound environment is improved.
Since different background sounds have different physical and perceived characteristics, the characteristics of the respective sounds are also different. And determining unified characteristics of the background noise affecting satisfaction, and adding masking sound in a targeted manner based on the characteristics to adjust the characteristic value of the background sound so as to play a role in adjusting the satisfaction of the acoustic environment. In order to efficiently screen out the optimal masking sound corresponding to the background sound under the condition of reducing the cost as much as possible, the invention designs a self-adaptive sound masking system with high precision and universality, as shown in figure 1, firstly, effective characteristics of background noise are determined through subjective evaluation experiments and a principal component analysis method, and a masking sound database is established according to the effective characteristics; then modeling a sound satisfaction proxy model by using a GPR algorithm; finally, an adaptive masking sound screening subsystem is constructed on the masking sound database based on the GPR model predictive formula. The specific method comprises the following steps:
first, the effective characteristics of the ambient background noise are determined. Studies have shown that the main factors determining the satisfaction of the acoustic environment are sound pressure level and loudness, and that it is necessary to determine the influence of other parameters on the satisfaction of the acoustic environment, considering that the addition of masking sound can improve the satisfaction of the acoustic environment with a proper increase in loudness. Specifically, m environmental background noise samples with representative meanings in different environments, seasons and time periods are collected for subjective evaluation experiments. The experiment adopts a multi-path loudspeaker for playback, restores the sound field of the actual environment as much as possible, adopts a multi-level odd-number category semantic subdivision method to obtain the satisfaction evaluation result of the crowd on the noise sample, and determines the influence on the fullness of soundBasic physical features of meaning: a sound levelCharacteristic frequency sound pressure level +.>Loudness->Roughness R, sharpness S, jitter F, and pitch K. The main characteristic parameters are subjected to dimension reduction by using a principal component analysis method, and d (less than or equal to 7) effective characteristic parameters affecting the satisfaction degree of the acoustic environment are obtained: />
Then, a masking sound database is constructed. Effective characteristic parameter according to d dimensionA large amount of natural sound data (or downloading various natural sound databases on the internet) such as intermittent insect sound, bird sound and water flow sound are collected. Further, a masking sound database is constructed by expanding the data volume by adjusting the values of each effective feature in the natural sound sampleThe database is composed of the effective parameters of the natural sound, and contains natural sound data quantity N.
Thereafter, the acoustic satisfaction model was combined using GPR model regression. Sampling n masking sound samples in a masking database using uniform samplingAnd m background sound samples->Combining to obtain mn combined sounds +.>Obtaining the satisfaction degree observation value y of the combined sound through subjective evaluation experiments 1 ,y 2 ,…y mn . Since these observations are affected by subjective differences of different individuals, it is assumed that y 1 ,y 2 ,…y mn Is subject to a mean value of 0 and variance of +.>Is influenced by the random deviation epsilon. True mechanism model due to acoustic satisfaction +.>It is difficult to obtain, based on the thought of proxy models, to approximate potential (true) satisfaction models using data modeling methods. Using mn combined acoustic samplesAnd training a proxy model of satisfaction. By analysis of mn training samplesCan be aware of the potential satisfaction model +.>Is a complex function with high latitude and nonlinearity, and therefore, a GPR algorithm is adopted to model a proxy model. Can deduce any new combined sound sampleIs satisfied with a mean value of +.>Variance is->Is a gaussian distribution of (c):
finally, an adaptive masking sound screening subsystem is established. For any newly arrived background sound, using the combined sound prediction model of formula (1)In a masking database->Automatically screening to calculate maximum value +.>This is the background sound +.>Is>Furthermore, an immune memory mechanism is introduced into the screening subsystem, and memory cells are generated while masking sound matching is successful every time, so that the memory cells are automatically activated when background sound is repeated next time, and immediate direct matching is achieved.
As shown in fig. 2, which is a schematic organization chart of the adaptive sound masking system of the present invention, the specific principle steps of the system are as follows:
1. collecting background sound samples and natural sound samples, and establishing a masking sound database
1) Collecting background sound samplesCollecting m environmental background sound samples (each sample has the same duration) with representative significance in different scenes, different seasons and different periods of time through subjectivityEvaluation test analysis shows the characteristics of background noise: a sound level +.>Characteristic frequency sound pressure level +.>Loudness +.>Roughness R, sharpness S, jitter F, and pitch K. Further, main characteristic affecting satisfaction degree of different background sounds is obtained by adopting a principal component analysis method, d (less than or equal to 7) main characteristic parameters are determined, and the main characteristic parameters are recorded as follows: />Defining the d main characteristic quantities as effective characteristic parameters affecting the satisfaction degree of the acoustic environment;
2) Construction of masking sound databaseAnd acquiring a large amount of natural sound data such as intermittent insect sound, bird sound and water flow sound according to the effective characteristic parameters (or expanding the sample size by adjusting the values of effective characteristics in the natural sound sample). Establishing a masking sound database by using the collected N pieces of natural sound data and the corresponding effective characteristic values thereof>
2. Approximation of combined acoustic satisfaction model using proxy model
1) And (3) uniformly sampling: masking sound database by adopting uniform sampling methodN masking sound samples are collected>
2) Acquisition ofCombined sound sampleRespectively combining the collected m background sounds with n masking sound samples to obtain mn combined sound samples and corresponding mn groups of effective characteristic parameter values
3) Subjective evaluation experiment: for mn combined soundsSubjective evaluation experiments of satisfaction are carried out to obtain mn satisfaction observation values y 1 ,y 2 ,…y mn Each observed value y k (k=1, …, mn) represents a true satisfaction value +.>And subjective individual deviation value ε k A kind of electronic device.
4) Training a combined sound satisfaction proxy model: consider a proxy modelWherein the method comprises the steps ofIs a d-dimensional input variable composed of effective feature quantities of the combined sound, < >>Representing the potential (true) satisfaction function, y (·) represents the satisfaction observation function corresponding to the combined sound sample, which is affected by a random deviation epsilon. Here epsilon means that the sound satisfaction value is affected by subjective differences of different individuals, assuming epsilon satisfies a gaussian distribution:
5) GPR modeling: due to the combined acoustic satisfaction modelIs a continuous nonlinear complex system with high dimension, a proxy model is built by using a GPR model, and mn potential (real) satisfaction values are assumed to be +.>Obeying a gaussian process: />Wherein (1)>Representation->Mean function of>Representing a sampleAnd->Covariance function between i, j=1 …, mn. To simplify the model, consider the zero-mean Gaussian process prior, let +.>Therefore, mn satisfaction function values->The gaussian process prior is satisfied:furthermore, mn observations y= [ Y ] 1 ,…,y mn ] T Satisfy Gaussian distribution->Finally, the process is carried out,it is possible to deduce any new combined sound sample +.>Is a predictive model of (1) its satisfaction value +.>Satisfy mean +.>Variance is->Is a gaussian distribution of (c):
in the method, in the process of the invention,respectively represent any combination sound +.>A predicted value and a predicted variance of satisfaction of (c),represent satisfaction observation vector, K *,mn 、K mn And K ** All represent a functional covariance matrix,>representing a bias covariance matrix, I mn Representing an mn-level identity matrix.
Here, covariance matrix K *,mn 、K mn And K ** From covariance function valuesThe specific expression is as follows:
here, the square-index function is used as the covariance function, i.eThe calculation formula of (2) is as follows:
3. establishing an adaptive masking sound screening subsystem based on an immune memory mechanism
Adaptive masking sound screening: identifying new coming background soundsActivating memory cells, judging->Whether or not the background sound is brand new: if the background sound is repeated, directly matching the memory masking sound; if the background sound is brand new, respectively calculating +.>And N masking sounds in the database +.>Is>Searching for maximum satisfaction value +.>The corresponding masking sound is +.>Is>And producing memory cells. The masking sound screening algorithm flow chart is shown in fig. 3.
Further summarised, a noise control method employing an adaptive sound masking system based on a GPR model is derived, the method comprising the steps of:
step 01: processing the new background sound to obtain effective parameter values;
step 02: inputting the background sound into an adaptive masking sound screening subsystem, identifying the background sound and activating the memory cells at the same time;
step 03: if the background sound identification result is the repeated background sound, the optimal masking sound is directly matched and memorized in a masking sound database;
step 04: if the background sound identification result is brand-new background sound, calculating the combined sound satisfaction values of the background sound and N masking sounds in a database one by utilizing a prediction formula of the combined sound satisfaction, screening the masking sound database to obtain the optimal masking sound corresponding to the maximum satisfaction value, and generating a memory cell at the same time;
step 05: outputting the optimal masking sound and performing control intervention on the background sound.
Example 1:
at the crossing with busy traffic and along the urban arterial road, the effective noise parameters of the low-medium-high frequency band traffic noise are determined by data processingThen, screening is performed by using an adaptive masking system, and the screening is performed in a masking database +.>Is matched with the optimal masking sound +.>Simultaneously obtain the corresponding optimal combination sound +.>In this environment, an optimal masking sound is used +.>For the traffic noise->Intervention is performed to make the crowd in the environment feel the combined sound of traffic sound and masking sound +.>Subjective satisfaction of people in the environment is improved, and annoyance feeling of the people caused by traffic noise is reduced.
Example 2:
in the open-air platform of rail transit, wheel track noise generated for train entering and exitingScreening in an adaptive masking sound system to match the optimal masking sound>And the combination sound thereof>At this station, the masking sound is used>For->Performing intervention to obtain optimal combined sound +.>The frequency spectrum of the environmental sound is more balanced, the interference of the crowd is reduced, and the public is enabled to feel as if the public is in the natural state by matching with the corresponding scenic poster or video put in the station, so that the trouble caused by noise is relieved.
The invention improves the living environment from the perspective of subjective feeling of people based on the psychological acoustic correlation law, improves the satisfaction degree of people to the acoustic environment, and accords with the human-based concept; by utilizing the artificial intelligence correlation technology, a proxy model of subjective feeling of noise is provided, an intelligent self-adaptive sound masking system is constructed, an intelligent sound masking technology is realized, and the artificial intelligence technology is fused in the field of noise control.
While the invention has been described with reference to certain preferred embodiments, it will be understood by those skilled in the art that various changes and substitutions of equivalents may be made and equivalents will be apparent to those skilled in the art without departing from the scope of the invention. Therefore, the protection scope of the invention is subject to the protection scope of the claims.
Claims (3)
1. The construction method of the adaptive sound masking system based on the GPR model is characterized in that the system is constructed by a masking sound database, a sound satisfaction degree proxy model and an adaptive masking sound screening subsystem, wherein the masking sound database is composed of intermittent insect sound, bird sound, water flow sound and other various natural sounds, the sound satisfaction degree proxy model is composed of Gaussian process regression algorithm modeling, and the adaptive masking sound screening subsystem is constructed based on the masking sound database and the sound satisfaction degree proxy model and introduces an immune memory mechanism;
the method comprises the following steps:
step 1: collecting background sound samples and natural sound samples and establishing a masking sound database by using collected sample data;
step 2: regression combining the acoustic satisfaction proxy model using the GPR model;
step 3: constructing an adaptive masking sound screening subsystem on a masking sound database based on the sound satisfaction proxy model;
the step 2 comprises the following sub-steps:
step 21: collecting masking sound samples in a masking sound database by adopting a uniform sampling method;
step 22: combining the environmental background sound sample and the masking sound sample to obtain a combined sound sample and a plurality of groups of corresponding effective characteristic parameter values;
the specific process of combining the masking sound sample with the environmental background sound sample after the masking sound sample is collected in the step 21 and the step 22 is as follows: sampling n masking sound samples in a masking database using uniform samplingAnd m background sound samplesCombining to obtain mn combined sounds +.>
Step 23: carrying out satisfaction subjective evaluation test on all the combined sounds in the combined sound sample and obtaining corresponding satisfaction observation values;
step 24: taking all the effective characteristic values of the combined sound in the combined sound sample and the corresponding satisfaction observation values as training samples;
step 25: modeling a sound satisfaction proxy model by using a Gaussian process regression algorithm;
the proxy model has the description formula:
in which acoustic samples are combinedRepresents a d-dimensional input variable constituted by the combined sound effective feature quantity,representing a true satisfaction function, +.>Representation corresponds to a combined sound sample +>The random deviation epsilon indicates that the real sound satisfaction value is influenced by subjective differences of different individuals;
the gaussian process regression model in the step 25 derives a satisfaction prediction formula as follows:
in the method, in the process of the invention,respectively represent any combination sound +.>A predicted value and a predicted variance of satisfaction of (c),represent satisfaction observation vector, K *,mn 、K mn And K ** All represent a functional covariance matrix,>representing a bias covariance matrix, I mn Representing an mn-level identity matrix;
the step 3 comprises the following sub-steps:
step 31: the self-adaptive masking sound screening subsystem identifies any new background sound and activates the memory cells at the same time;
step 32: if the new background sound identification result is the repeated background sound, the optimal masking sound is directly matched and memorized in a masking sound database of the self-adaptive masking sound system;
step 33: if the new background sound identification result is brand-new background sound, calculating the combined sound satisfaction value of the background sound and the masking sound database one by utilizing a prediction formula of the combined sound satisfaction, screening the database to obtain the optimal masking sound corresponding to the maximum satisfaction value, and generating a memory cell at the same time;
step 34: outputting the optimal masking sound and performing control intervention on the new background sound.
2. The method for constructing an adaptive sound masking system based on a GPR model according to claim 1, wherein said step 1 comprises the following sub-steps:
step 11: collecting environmental background sound samples of different scenes, seasons and time periods, and obtaining a plurality of effective characteristic parameters affecting the satisfaction degree of the sound environment by using a principal component analysis method;
step 12: intermittent insect sound, bird sound, water flow sound and other natural sound data are collected based on the effective characteristic parameters, and a masking sound database is established by utilizing the collected data and the corresponding effective characteristic parameters.
3. The method for constructing an adaptive sound masking system based on a GPR model according to claim 2, wherein the effective characteristic parameter in step 11 is a main characteristic affecting the satisfaction of background sound, specifically by applying to the physical characteristics of noise: and (3) carrying out principal component analysis on the A sound level, the characteristic frequency sound pressure level, the loudness, the roughness, the sharpness, the shaking degree and the tone degree.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910656438.8A CN110362789B (en) | 2019-07-19 | 2019-07-19 | GPR model-based adaptive sound masking system and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910656438.8A CN110362789B (en) | 2019-07-19 | 2019-07-19 | GPR model-based adaptive sound masking system and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110362789A CN110362789A (en) | 2019-10-22 |
| CN110362789B true CN110362789B (en) | 2024-02-02 |
Family
ID=68221159
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910656438.8A Active CN110362789B (en) | 2019-07-19 | 2019-07-19 | GPR model-based adaptive sound masking system and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110362789B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112932489A (en) * | 2020-12-30 | 2021-06-11 | 广东电网有限责任公司电力科学研究院 | Transformer substation noise subjective annoyance degree evaluation model establishing method and model establishing system |
| WO2023211385A1 (en) * | 2022-04-27 | 2023-11-02 | Nanyang Technological University | Soundscape augmentation system and method of forming the same |
| EP4365890A1 (en) | 2022-11-07 | 2024-05-08 | Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. | Adaptive harmonic speech masking sound generation apparatus and method |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106933146A (en) * | 2017-03-14 | 2017-07-07 | 吉林大学 | Electrocar pedestrian's caution sound method for designing, caution sound control system and method |
| CN108357445A (en) * | 2018-03-20 | 2018-08-03 | 吉林大学 | Car masking sound quality self-adapting control system and method |
| CN109238448A (en) * | 2018-09-17 | 2019-01-18 | 上海市环境科学研究院 | A method of acoustic environment satisfaction is improved based on sound masking |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20030103632A1 (en) * | 2001-12-03 | 2003-06-05 | Rafik Goubran | Adaptive sound masking system and method |
-
2019
- 2019-07-19 CN CN201910656438.8A patent/CN110362789B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106933146A (en) * | 2017-03-14 | 2017-07-07 | 吉林大学 | Electrocar pedestrian's caution sound method for designing, caution sound control system and method |
| CN108357445A (en) * | 2018-03-20 | 2018-08-03 | 吉林大学 | Car masking sound quality self-adapting control system and method |
| CN109238448A (en) * | 2018-09-17 | 2019-01-18 | 上海市环境科学研究院 | A method of acoustic environment satisfaction is improved based on sound masking |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110362789A (en) | 2019-10-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110362789B (en) | GPR model-based adaptive sound masking system and method | |
| Romero et al. | Modelling the soundscape quality of urban waterfronts by artificial neural networks | |
| Kang et al. | A model for implementing soundscape maps in smart cities | |
| CN107293286B (en) | Voice sample collection method based on network dubbing game | |
| CN105788592A (en) | Audio classification method and apparatus thereof | |
| CN109346063B (en) | Voice data enhancement method | |
| Kang | From understanding to designing soundscapes | |
| CN109238448B (en) | Method for improving urban sound environment satisfaction based on sound masking | |
| CN103471709A (en) | Method for predicting noise quality of noise inside passenger vehicle | |
| CN110246481B (en) | Automobile active noise reduction method for predicting engine speed | |
| CN109256118B (en) | End-to-end Chinese dialect identification system and method based on generative auditory model | |
| CN109767778A (en) | Bi-LSTM and WaveNet fused voice conversion method | |
| CN108206027A (en) | A kind of audio quality evaluation method and system | |
| CN107507625A (en) | Sound source distance determines method and device | |
| Gloaguen et al. | Road traffic sound level estimation from realistic urban sound mixtures by non-negative matrix factorization | |
| Yue et al. | A visualized soundscape prediction model for design processes in urban parks | |
| CN113327624B (en) | Method for intelligent monitoring of environmental noise by adopting end-to-end time domain sound source separation system | |
| Cai et al. | Music creation and emotional recognition using neural network analysis | |
| CN117672251A (en) | Urban sound scene intelligent digital twin system based on edge calculation | |
| CN110333484A (en) | The room area grade localization method with analysis is known based on environmental background phonoreception | |
| CN113409798A (en) | Method, device and equipment for generating noise-containing voice data in vehicle | |
| CN116312640A (en) | Self-adaptive environment voice emotion recognition method and device | |
| CN115828722A (en) | Rainfall-driven deep learning flood forecasting method based on production convergence process | |
| Hamadicharef et al. | Intelligent and perceptual-based approach to musical instruments sound design | |
| CN118523320B (en) | Method, device, equipment, storage medium and product for protecting electricity of acousto-optic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |