CN113099356A - Method and device for self-adaptive sound field regulation and control - Google Patents
Method and device for self-adaptive sound field regulation and control Download PDFInfo
- Publication number
- CN113099356A CN113099356A CN202110311511.5A CN202110311511A CN113099356A CN 113099356 A CN113099356 A CN 113099356A CN 202110311511 A CN202110311511 A CN 202110311511A CN 113099356 A CN113099356 A CN 113099356A
- Authority
- CN
- China
- Prior art keywords
- sound field
- input
- neural network
- output sound
- output
- 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.)
- Granted
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
- H04R3/00—Circuits for transducers, loudspeakers or microphones
- H04R3/12—Circuits for transducers, loudspeakers or microphones for distributing signals to two or more loudspeakers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04R—LOUDSPEAKERS, MICROPHONES, GRAMOPHONE PICK-UPS OR LIKE ACOUSTIC ELECTROMECHANICAL TRANSDUCERS; DEAF-AID SETS; PUBLIC ADDRESS SYSTEMS
- H04R2430/00—Signal processing covered by H04R, not provided for in its groups
Abstract
The invention discloses a method and a device for self-adaptive sound field regulation and control, and belongs to the field of sound field regulation and control. The method utilizes the neural network to establish a complex corresponding relation between an output sound field and an input sound field, utilizes the input sound source to adaptively control the input sound field to train the neural network and obtain a specific target output sound field after a complex process. The device comprises an input sound source, a transmission medium, a sound field detection module and a neural network module, and further utilizes a reconfigurable acoustic element to replace an array sound source to realize the self-adaptive control from the input common Gaussian sound field to the specific target output sound field aiming at the Gaussian sound source. The invention breaks through the performance limitation of the traditional sound field regulation and control technology, provides a brand new thought for self-adaptive flexible regulation and control of the sound field through the combination of the array sound source, the reconfigurable acoustic element and the neural network, has wide potential application prospects in the fields of communication, sensing, measurement, imaging and the like based on the complex sound field regulation and control technology, and fills the blank in the related technical field of sound field regulation and control.
Description
Technical Field
The invention belongs to the field of sound field regulation and control, and particularly relates to a method and a device for self-adaptive sound field regulation and control.
Background
The ocean covers two thirds of the surface area of the earth, which is one of the leading areas of human exploration and research. The ocean not only plays an important role in international commerce and fisheries, but also contains climate information and abundant resources to be exploited. With the frequent and rapid development of global marine activities, the development of modern marine communication networks and technologies has become a focus problem in the academic and industrial fields of China. Underwater wireless communication is a key technology of marine communication, and can be generally divided into underwater electromagnetic wave communication and underwater acoustic communication. The marine environment is variable, electromagnetic wave (light wave, radio frequency) signals are influenced by various factors such as attenuation, scattering and absorption when being transmitted underwater due to high frequency, and the transmission distance is generally limited to a short distance (not more than 200 m). At present, sound waves become the best way to transmit information underwater for long distances. Due to the complex and variable marine environment, the sound field can be distorted and distorted to different degrees due to various complex disturbances in the transmission process, for example, suspended particles in water can scatter the sound field, bubbles in water can disturb the sound field, the transmission of the sound field can be disturbed by underwater turbulence and uneven gradient distribution, and the like, which can cause the output sound field to be greatly changed compared with the input sound field. The input sound field is a common gaussian sound field, and the amplitude and phase distribution of the output sound field after a complex process may become disordered, which may adversely affect the communication based on the sound field. In addition, the sound field can be widely applied to sensing, measurement, imaging and other applications, and the disturbance of the sound field caused by the complex marine environment can generate adverse effects on the applications. In view of this, how to flexibly regulate and control the sound field is a key technology which needs to be solved urgently in the application of the sound field.
It is noted that in communication, sensing, measurement and imaging applications based on sound fields, gaussian sound fields are commonly used in the conventional technology, and much more, the parameter dimensions such as frequency and amplitude of the gaussian sound fields are concerned, and the spatial distribution of the sound fields is rarely involved. In recent years, with the development of light wave space dimension resources, particularly, a vortex light field which has a spatial spiral phase structure distribution and carries orbital angular momentum is inspired, and a vortex sound field also receives wide attention. More generally, with the rapid development of the optical field regulation and control technology, the sound field regulation and control technology is also developed. The sound field regulation technology aims at tailoring the spatial amplitude and phase distribution of the sound field, such as regulating and generating a vortex sound field with a spiral phase structure and a structural sound field with general spatial amplitude and phase distribution. In many applications, the output sound field of a known input sound field becomes disordered after a complex process, wherein the complex process is unknown, and is like a black box, however, the output sound field is a specific sound field meeting certain application requirements, such as a converged gaussian sound field or a vortex sound field, having a specific spatial amplitude and phase distribution, which requires flexible regulation and control of the sound field. In particular, in order to cope with a complex and variable marine environment, an adaptive sound field regulation and control technology is needed, which has important research significance and practical value for flexibly regulating and controlling a sound field and practically applying to advanced communication, sensing, measurement and imaging technologies.
Disclosure of Invention
The invention provides a method and a device for regulating and controlling a self-adaptive sound field, which aim to establish a complex corresponding relation between an output sound field and an input sound field by utilizing a neural network, train the neural network by utilizing an array sound source to self-adaptively control the input sound field, acquire a specific target output sound field after a complex process, construct a device for regulating and controlling the self-adaptive sound field, and further replace the array sound source by utilizing a reconfigurable acoustic element to realize the self-adaptive control from an input common Gaussian sound field to the specific target output sound field, thereby breaking through the performance limitation of the traditional sound field regulation and control technology and filling the blank of the related technology.
In order to achieve the above object, according to an aspect of the present invention, there is provided a method for adaptive sound field regulation, wherein a plurality of pairs of known input sound fields and known output sound fields are used as training sample sets to train a neural network after the input sound fields are transmitted or transformed to obtain corresponding output sound fields, wherein the output sound fields are inputs of the neural network, and the input sound fields are outputs of the neural network;
after the neural network is trained, the target output sound field is used as the input of the trained neural network, the output of the neural network corresponds to the input sound field required for obtaining the target output sound field, and the required input sound field is regulated and controlled, so that the self-adaptive sound field regulation and control are realized, and the target output sound field is obtained.
Preferably, the training sample set of the neural network is a plurality of input sound fields generated by the array sound source and a plurality of output sound fields corresponding to the input sound fields after transmission or transformation, after the neural network is trained by the training sample set, the target output sound field is used as the neural network input, the required input sound field is output after the neural network learning, then the required input sound field is generated by utilizing the array sound source regulation and control, the input sound field is transmitted or transformed to obtain the corresponding output sound field, the output sound field is compared with the target output sound field, if the output sound field is inconsistent or the consistency degree is lower than a set value, the output sound field and the corresponding input sound field are used as the newly added supplementary sample on the basis of the training sample set, the neural network is retrained and the neural network parameters are adjusted, and then the target output is used as the input of the retrained neural network sound field, and repeating iteration until the actual output sound field obtained by the neural network learning is matched with the target output sound field.
Preferably, the input sound field is generated by an array sound source, the array sound source can be an annular array, a square array or other geometric arrangement arrays, and a plurality of known input sound fields with specific spatial variation amplitude and phase distribution can be generated according to needs by controlling the amplitude and phase of the sound field generated by each array unit of the array sound source.
Preferably, the transmission or transformation process may be a complex transmission process, or may be an input-to-output complex transformation process; the target output sound field can be a converged Gaussian sound field, a vortex sound field or any other structural sound field with specific spatial amplitude and phase distribution.
Preferably, the output sound field is obtained by transmitting or transforming the input sound field correspondingly, and each output sound field detects the amplitude and phase distribution of the sound field space point by point through a single sound field receiver space or detects the amplitude and phase distribution of the sound field space at one time through an array sound field receiver.
Preferably, the input sound source is a traditional single common gaussian sound source, which is located a certain distance in front of the array sound source, and still utilizes neural network learning to obtain an adaptive control input sound field a, which obtains a target output sound field a' after a complex process, at this time, an acoustic element is used to replace the array sound source, the transmission matrix of the acoustic element is H, and the input common gaussian sound source a0Sound field distribution a when a certain distance is transmitted to reach the position of the acoustic element0' can be obtained by acoustic transmission theory such that, in order to obtain a final target output sound field, the transmission matrix H of the acoustic elements of the substitute array sound source satisfies A0Thus, an acoustic element required for an alternative array sound source in order to obtain a target output sound field when the input sound source is a single ordinary gaussian sound source can be obtained.
According to another aspect of the present invention, there is provided an apparatus for adaptive sound field regulation, the apparatus comprising an input sound source, a transmission medium, a sound field detection module, and a neural network module; the input sound source is used for generating an arbitrary input sound field with spatial amplitude and phase distribution; the transmission medium is used for realizing the conversion from an input sound field to an output sound field; the sound field detection module is used for detecting the amplitude and phase distribution neural network module of the output sound field and is used for training the input sound field and the output sound field through a network as training sample sets, wherein the output sound field is input of the neural network, and the input sound field is output of the neural network; after the neural network is trained, the target output sound field is used as the input of the trained neural network, the neural network outputs the input sound field required for obtaining the target output sound field, and the required input sound field is regulated and controlled, so that the self-adaptive sound field regulation and control are realized.
Preferably, the training sample set of the neural network is a plurality of input sound fields generated by the array sound source and a plurality of output sound fields corresponding to the input sound fields after transmission or transformation, after the neural network is trained by the training sample set, the target output sound field is used as the neural network input, the required input sound field is output after the neural network learning, then the required input sound field is generated by utilizing the array sound source regulation and control, the input sound field is transmitted or transformed to obtain the corresponding output sound field, the output sound field is compared with the target output sound field, if the output sound field is inconsistent or the consistency degree is lower than a set value, the output sound field and the corresponding input sound field are used as the newly added supplementary sample on the basis of the training sample set, the neural network is retrained and the neural network parameters are adjusted, and then the target output is used as the input of the retrained neural network sound field, and repeating iteration until the actual output sound field obtained by the neural network learning is matched with the target output sound field.
Preferably, the transmission medium may be an air and underwater transmission medium of a real scene, a transmission medium of which transmission paths scatter some scatterers, a complex scattering medium, other media with unknown transmission characteristics and complex transmission matrices, a device with complex transformation matrices, or other unknown black boxes which are connected with an input sound field and an output sound field in a broad sense. .
Preferably, the sound field detection module is a single sound-electricity conversion element or an array sound-electricity conversion element, and can be a single microphone or an array microphone; detecting the spatial amplitude and phase distribution of a sound field by a single acoustoelectric conversion element through a spatial point-by-point scanning method; the array acoustoelectric conversion element can detect the spatial amplitude and phase distribution of the sound field at one time.
Preferably, the input sound source is an array sound source, which may be a ring array, a square array, or other geometrically arranged arrays, and each sound source may be a speaker, a hydrophone, or other general electroacoustic transducer. If the input sound source is a single Gaussian sound source, the output sound field of a Gaussian sound field generated by a common Gaussian sound source is very disordered after a complex input and output process, and a specific target output sound field can be obtained by adding an acoustic element to perform self-adaptive sound field regulation. The Gaussian sound source is used for generating a single sound field in Gaussian distribution; the acoustic elements are acoustic metamaterials, acoustic metasurfaces, near-type pure phase acoustic elements, amplitude and phase type acoustic elements, or transmission matrix programmable and reconfigurable acoustic elements.
Through the technical scheme, compared with the prior art, the invention has the following beneficial effects:
1. the invention discloses a method and a device for regulating and controlling a self-adaptive sound field, which can obtain a specific target output sound field after a complex process through self-adaptively controlling an input sound field.
2. The invention establishes the complex corresponding relation between the output sound field and the input sound field by utilizing the neural network, does not depend on a specific complex process, does not need to solve a definite input-output corresponding relation, only completes the training of the neural network by a plurality of pairs of self-adaptive control input sound fields and corresponding output sound fields, has the characteristics of flexibility and high efficiency, and provides a brand new thought for a novel target-oriented sound field regulation and control technology.
3. The array sound source is used for adaptively controlling the input sound field to train the neural network and finally obtaining the specific target output sound field after a complex process, the array sound source is flexible and convenient to use, a plurality of input sound fields required by the training of the neural network can be generated, and the input sound field can be adaptively controlled according to the specific target output sound field.
4. Aiming at a more common Gaussian sound source application scene, the reconfigurable acoustic element is used for replacing an array sound source to realize the self-adaptive control from the input common Gaussian sound field to the specific target output sound field, so that the method has important value for wider practical application of the sound field and is beneficial to the application and popularization of the self-adaptive sound field regulation and control technology.
5. The basic idea of the adaptive sound field regulation and control technology provided by the invention has wide universality, can be further applied to the adaptive light field regulation and control technology and the adaptive electromagnetic field regulation and control technology, and has important significance for expanding the application based on flexible light field regulation and control and flexible electromagnetic field regulation and control.
Drawings
Fig. 1 is a schematic flow chart of a method for adaptive sound field regulation according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of an apparatus for adaptive sound field regulation according to an embodiment of the present invention;
FIG. 3 is a supplementary schematic diagram of a neural network-based training process provided by an embodiment of the present invention;
fig. 4 is a schematic flow chart of adaptive sound field regulation based on a neural network according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
The invention provides a self-adaptive sound field regulation and control method, which comprises the steps of after an input sound field is transmitted or transformed to obtain a corresponding output sound field, training a neural network by taking a plurality of pairs of known input sound fields and output sound fields as training sample sets, wherein the output sound field is input into the neural network, and the input sound field is output from the neural network;
after the neural network is trained, the target output sound field is used as the input of the trained neural network, the output of the neural network corresponds to the input sound field required by obtaining the target output sound field, and the required input sound field is regulated and controlled, so that the self-adaptive sound field regulation and control are realized.
Specifically, a training sample set of a neural network is a plurality of input sound fields generated by an array sound source and a plurality of output sound fields corresponding to the input sound fields after transmission or transformation, after the neural network is trained by the training sample set, a target output sound field is used as the input of the neural network, the required input sound field is output after learning of the neural network, then the required input sound field is generated by utilizing the regulation and control of the array sound source, the input sound field is transmitted or transformed to obtain a corresponding output sound field, the output sound field is compared with the target output sound field, if the output sound field is inconsistent or the consistency degree is lower than a set value, the output sound field and the corresponding input sound field are used as a supplementary sample newly added on the basis of the training sample set, the neural network is retrained and the neural network parameters are adjusted, and then the target output is used as the input of the retrained neural network, and repeating iteration until the actual output sound field obtained by the neural network learning is matched with the target output sound field.
Specifically, an input sound field is generated by an array sound source, and the amplitude and the phase of the generated sound field can be regulated and controlled independently by each array unit of the array sound source, so that a structural sound field with spatially-changed amplitude and phase distribution is generated.
Specifically, the output sound field is obtained by transmitting or transforming the input sound field correspondingly, and each output sound field detects the spatial amplitude and phase distribution of the sound field point by point through a single sound field receiver space or detects the spatial amplitude and phase distribution of the sound field at one time through an array sound field receiver.
Specifically, when an input sound field is generated by a single Gaussian sound source, an output sound field is obtained by transmitting the input sound field through a free space to reach a preset acoustic element and then transmitting or transforming the input sound field, and the amplitude and the phase of the sound field generated by the acoustic element are regulated to replace an array sound source, so that a structural sound field with spatially-changed amplitude and phase distribution is generated.
The invention also provides a device for self-adapting sound field regulation, which comprises an input sound source, a transmission medium, a sound field detection module and a neural network module; the input sound source is used for generating an arbitrary input sound field with spatial amplitude and phase distribution; the transmission medium is used for realizing the conversion from an input sound field to an output sound field; the sound field detection module is used for detecting the amplitude and phase distribution neural network module of the output sound field and is used for training the input sound field and the output sound field through a network as training sample sets, wherein the output sound field is input of the neural network, and the input sound field is output of the neural network; after the neural network is trained, the target output sound field is used as the input of the trained neural network, the neural network outputs the input sound field required for obtaining the target output sound field, and the required input sound field is regulated and controlled, so that the self-adaptive sound field regulation and control are realized.
Specifically, a training sample set of a neural network is a plurality of input sound fields generated by an array sound source and a plurality of output sound fields corresponding to the input sound fields after transmission or transformation, after the neural network is trained by the training sample set, a target output sound field is used as the input of the neural network, the required input sound field is output after learning of the neural network, then the required input sound field is generated by utilizing the regulation and control of the array sound source, the input sound field is transmitted or transformed to obtain a corresponding output sound field, the output sound field is compared with the target output sound field, if the output sound field is inconsistent or the consistency degree is lower than a set value, the output sound field and the corresponding input sound field are used as a supplementary sample newly added on the basis of the training sample set, the neural network is retrained and the neural network parameters are adjusted, and then the target output is used as the input of the retrained neural network, and repeating iteration until the actual output sound field obtained by the neural network learning is matched with the target output sound field.
Specifically, the transmission medium is an air or underwater transmission medium of a real scene, or a medium whose transmission characteristics are unknown and which has a complex transmission matrix.
Specifically, the sound field detection module is a single acoustoelectric conversion element or an array acoustoelectric conversion element; detecting the spatial amplitude and phase distribution of a sound field by a single acoustoelectric conversion element through a spatial point-by-point scanning method; the array acoustoelectric conversion element can detect the spatial amplitude and phase distribution of the sound field at one time.
Specifically, the input sound source is an array sound source, which may be an annular array, a square array, or other geometrically arranged arrays, and each sound source may be a speaker, a hydrophone, or other general electroacoustic conversion elements. Sometimes, an input sound source is a single Gaussian sound source, a Gaussian sound field generated by a common Gaussian sound source is very disordered after a complex input and output process, and a specific target output sound field can be obtained by adding an acoustic element to perform self-adaptive sound field regulation. The Gaussian sound source is used for generating a single sound field in Gaussian distribution; the acoustic elements are acoustic metamaterials, acoustic metasurfaces, near-type pure phase acoustic elements, amplitude and phase type acoustic elements, or transmission matrix programmable and reconfigurable acoustic elements.
The following description is made with reference to the embodiments and the accompanying drawings.
Fig. 1 is a schematic flow chart of a method for adaptive sound field regulation and control according to the present invention, in which a neural network (e.g., a convolutional neural network) is used to establish a complex correspondence between an output sound field and an input sound field, the input sound field is generated by an array sound source, each array unit can arbitrarily and independently regulate the amplitude and phase of the generated sound field, so that the array sound source can generate a plurality of arbitrary input sound fields, a plurality of output sound fields are obtained by correspondence after passing through a transmission medium, the plurality of pairs of known input sound fields and output sound fields are used as a training sample set to train the neural network, wherein the output sound field is input to the neural network, and the input sound field is output to the neural network. After the neural network is trained, the target output sound field is used as the input of the neural network, the output of the neural network corresponds to the input sound field required for obtaining the target output sound field, and the required input sound field is generated through the regulation and control of the array sound source, so that the self-adaptive sound field regulation and control are realized, namely the input sound field is regulated and controlled in a self-adaptive mode to obtain the specific target output sound field after a complex process.
Fig. 2 is a schematic structural diagram of an adaptive sound field regulation and control apparatus provided by the present invention, which includes an array sound source, a transmission medium, a sound field detection module, and a neural network module. Each array unit of the array sound source can independently regulate and control the amplitude and the phase of a sound field generated at the position of the array sound source at will, and an input sound field with any spatial amplitude and phase distribution can be generated by controlling the array sound source; the input and output complex process realizes the conversion from an input sound field to an output sound field; the sound field detection module detects the output sound field, namely, the amplitude and phase distribution of the output sound field are obtained.
Fig. 3 is a supplementary schematic diagram of a neural network-based training process provided by the present invention. The output sound field is used as the input of the neural network, the output sound field is used as the output of the neural network, an input and output sample set training neural network is established, the neural network generally distributes an input layer, a hidden layer and an output layer, and the input and output corresponding relation is obtained after repeated iterative computation.
Fig. 4 is a schematic flow chart of adaptive sound field regulation based on a neural network according to the present invention. The neural network training module can be a convolution neural network, and n scattered sound fields (A) passing through a complex medium1’,A2’,…,An') n sound fields (A) generated by an array sound source as input information1,A2,…,An) As output information, a convolutional neural model is constructed. When given a target output sound field Sn+1' as the input of the convolutional neural network, the convolutional neural network can obtain the sound field distribution A of the input sound field through the corresponding relation obtained beforen+1Adjusting the amplitude phase of each unit of the array sound source according to the sound field distribution, and obtaining a corresponding output sound field An+1Comparing the output sound field with a target sound field, substituting the output sound field and the input sound field into a neural network as a newly added sample training set if the output sound field and the input sound field are inconsistent, adjusting parameters of the neural network, repeating the step to finally obtain an output sound field which is highly consistent with the target sound field, and simultaneously obtaining the input sound field generated by the corresponding array sound source to finish the training process of the neural network.
The present invention is not limited to the above embodiments, and those skilled in the art can implement the present invention in other various embodiments according to the disclosure of the present invention, so that all designs and concepts of the present invention can be changed or modified without departing from the scope of the present invention.
Claims (10)
1. A method of adaptive sound field regulation, the method comprising:
after a plurality of input sound fields are transmitted or transformed to obtain corresponding output sound fields, training a neural network by taking the known input sound fields and the known output sound fields as training sample sets, wherein the output sound field is input of the neural network, and the input sound field is output of the neural network;
after the neural network is trained, the target output sound field is used as the input of the trained neural network, the output of the neural network corresponds to the input sound field required for obtaining the target output sound field, and the required input sound field is regulated and controlled, so that the self-adaptive sound field regulation and control are realized, and the target output sound field is obtained.
2. The method according to claim 1, wherein the training sample set of the neural network is a plurality of input sound fields and a plurality of output sound fields corresponding to the input sound fields after transmission or transformation, the neural network is trained by the training sample set, a target output sound field is input as the neural network, the required input sound field is output after learning by the neural network, then the required input sound field is generated by adaptive control, the input sound field is transmitted or transformed to obtain a corresponding output sound field, the output sound field is compared with the target output sound field, if the output sound field and the corresponding input sound field are inconsistent or the degree of consistency is lower than a set value, the output sound field and the corresponding input sound field are used as a newly added supplementary sample on the basis of the training sample set, the neural network is retrained and neural network parameters are adjusted, and then the target output sound field is used as the input of the retrained neural network, and repeating iteration until the actual output sound field obtained by the neural network learning is matched with the target output sound field.
3. The method of claim 2, wherein the input sound field is generated by an array sound source, and each array unit of the array sound source can be arbitrarily and independently controlled in amplitude and phase to generate the sound field, so as to generate a structural sound field with spatially varying amplitude and phase distribution.
4. The method of claim 3, wherein the output sound fields are obtained by transmission or transformation of input sound fields, and each output sound field detects the amplitude and phase distribution of the sound field space by scanning the space of a single sound field receiver point by point or detects the amplitude and phase distribution of the sound field space by an array sound field receiver at one time.
5. The method according to claim 1, wherein the input sound field is generated by a single gaussian sound source, an acoustic element is preset behind the single gaussian sound source, the output sound field is obtained by transmitting the input sound field through a free space to the preset acoustic element and then transmitting or transforming, and the target adaptive control is realized by regulating the acoustic element to obtain a target output sound field a'; the transmission matrix H of the acoustic elements satisfies:
A0’H＝A
wherein A is0' is a Gaussian sound source A0The sound field distribution when reaching the position where the acoustic element is located, a being the input sound field.
6. A self-adaptive sound field regulation and control device is characterized by comprising an input sound source, a transmission medium, a sound field detection module and a neural network module;
the input sound source is used for generating an arbitrary input sound field with spatial amplitude and phase distribution;
the transmission medium is used for realizing the conversion from an input sound field to an output sound field;
the sound field detection module is used for detecting the amplitude and phase distribution of an output sound field;
the neural network module is used for training the input sound field and the output sound field through a network by taking the input sound field and the output sound field as training sample sets, wherein the output sound field is input into the neural network, and the input sound field is output from the neural network; after the neural network is trained, the target output sound field is used as the input of the trained neural network, the neural network outputs the input sound field required for obtaining the target output sound field, and the required input sound field is regulated and controlled, so that the self-adaptive sound field regulation and control are realized.
7. The apparatus of claim 6, wherein the training sample set of the neural network is a plurality of input sound fields and a plurality of output sound fields corresponding to the input sound fields after transmission or transformation, the neural network is trained by the training sample set, a target output sound field is input as the neural network, the neural network learns the target output sound field and outputs the required input sound field, the required input sound field is generated by adaptive control, the input sound field is transmitted or transformed to obtain a corresponding output sound field, the output sound field is compared with the target output sound field, if the output sound field and the target output sound field are inconsistent or the degree of consistency is lower than a set value, the output sound field and the corresponding input sound field are used as a supplementary sample newly added on the basis of the training sample set, the neural network is retrained and neural network parameters are adjusted, and then the target output sound field is used as the input of the retrained neural network, and repeating iteration until the actual output sound field obtained by the neural network learning is matched with the target output sound field.
8. The apparatus of claim 6, wherein the input sound field is generated by an array sound source, and each array unit of the array sound source can independently adjust and control the amplitude and phase of the sound field generated by the array sound source arbitrarily, so as to generate a structural sound field with spatially varying amplitude and phase distribution;
or the input sound field is generated by a single Gaussian sound source, an acoustic element is preset behind the single Gaussian sound source, the output sound field is obtained by transmitting or transforming the input sound field to the preset acoustic element through a free space, and target self-adaptive control is realized by regulating the acoustic element to obtain a target output sound field A'; the transmission matrix H of the acoustic elements satisfies:
A0’H＝A
wherein A is0' is a Gaussian sound source A0The sound field distribution when reaching the position where the acoustic element is located, a being the input sound field.
9. The apparatus of claim 6, wherein the transmission medium is air, water, or a medium with unknown transmission characteristics and a complex transmission matrix.
10. The apparatus of claim 6, wherein the sound field detection module is a single acousto-electric conversion element or an array acousto-electric conversion element; detecting the spatial amplitude and phase distribution of a sound field by a single acoustoelectric conversion element through a spatial point-by-point scanning method; the array acoustoelectric conversion element can detect the spatial amplitude and phase distribution of the sound field at one time.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110311511.5A CN113099356B (en) | 2021-03-24 | 2021-03-24 | Method and device for self-adaptive sound field regulation and control |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110311511.5A CN113099356B (en) | 2021-03-24 | 2021-03-24 | Method and device for self-adaptive sound field regulation and control |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113099356A true CN113099356A (en) | 2021-07-09 |
| CN113099356B CN113099356B (en) | 2021-11-19 |
Family
ID=76669207
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110311511.5A Active CN113099356B (en) | 2021-03-24 | 2021-03-24 | Method and device for self-adaptive sound field regulation and control |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113099356B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103438985A (en) * | 2013-08-14 | 2013-12-11 | 大连理工大学 | Sound field information collecting method for sound field synthesis |
| US20160293167A1 (en) * | 2013-10-10 | 2016-10-06 | Google Inc. | Speaker recognition using neural networks |
| CN109086824A (en) * | 2018-08-01 | 2018-12-25 | 哈尔滨工程大学 | A kind of sediment sonar image classification method based on convolutional neural networks |
| CN109993280A (en) * | 2019-03-27 | 2019-07-09 | 东南大学 | A kind of underwater sound source localization method based on deep learning |
-
2021
- 2021-03-24 CN CN202110311511.5A patent/CN113099356B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103438985A (en) * | 2013-08-14 | 2013-12-11 | 大连理工大学 | Sound field information collecting method for sound field synthesis |
| US20160293167A1 (en) * | 2013-10-10 | 2016-10-06 | Google Inc. | Speaker recognition using neural networks |
| CN109086824A (en) * | 2018-08-01 | 2018-12-25 | 哈尔滨工程大学 | A kind of sediment sonar image classification method based on convolutional neural networks |
| CN109993280A (en) * | 2019-03-27 | 2019-07-09 | 东南大学 | A kind of underwater sound source localization method based on deep learning |
Non-Patent Citations (4)
| Title |
|---|
| JIAN WANG ET AL: ""An application of VoIP communication on embedded system"", 《IEEE》 * |
| JIAN WANG ET AL: ""An Input Weights Dependent Complex-Valued Learning Algorithm Based on Wirtinger Calculus"", 《IEEE》 * |
| JIAN WANG ET AL: ""Experimental demonstration of a feedback-assisted orbital angular momentum-based underwater wireless optical link across the underwater-to-air interface"", 《IEEE》 * |
| 李思纯等: "基于矢量传感器和LMBP神经网络的水声目标识别", 《信号处理》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113099356B (en) | 2021-11-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN102928844B (en) | Underwater sub-wavelength resolution ratio three-dimensional imaging method | |
| CN113126028B (en) | Noise source positioning method based on multiple microphone arrays | |
| RU2694846C1 (en) | Method for formation of marine monitoring system with programmable neuron network control system | |
| CN104734795A (en) | Time-variant underwater acoustic channel capacity simulation model | |
| CN112257319B (en) | Acoustic metamaterial design method based on non-gradient topological optimization | |
| CN111983619A (en) | Underwater acoustic target forward scattering acoustic disturbance positioning method based on transfer learning | |
| CN108549052A (en) | A kind of humorous domain puppet sound intensity sound localization method of circle of time-frequency-spatial domain joint weighting | |
| CN113607447A (en) | Acoustic-optical combined fan fault positioning device and method | |
| CN107462317A (en) | A kind of water surface target radiation noise simulation method under complex scene | |
| CN113099356B (en) | Method and device for self-adaptive sound field regulation and control | |
| Alexandrou et al. | A methodology for acoustic seafloor classification | |
| CN103575377B (en) | Measure the method for difference frequency ripple spatial characteristics in parameter sound field | |
| Galvin et al. | Analysis of the performance of an underwater acoustic communications system and comparison with a stochastic model | |
| KR102199158B1 (en) | Methods and systems for estimating the location of sound sources using azimuth-frequency expression and convolution neural network model | |
| Møller et al. | Circular loudspeaker arrays with controllable directivity | |
| CN105372644B (en) | One kind is based on the modified Adaptive beamformer method and system of dynamic weight | |
| Zhu et al. | Internet of Underwater Things infrastructure: A shared underwater acoustic communication layer scheme for real-world underwater acoustic experiments | |
| Zhang et al. | Fast calculation of underwater acoustic horizontal range: A guarantee for B5G ocean mobile networks | |
| CN108650043B (en) | High-precision underwater acoustic communication channel modeling method | |
| CN115811304A (en) | Ocean impulse noise elimination method based on condition variational self-coding network | |
| CN113109795B (en) | Deep sea direct sound zone target depth estimation method based on deep neural network | |
| CN114925496A (en) | Marine environment noise forecasting method | |
| Elechi et al. | Analysis of a SONAR detecting system using multi-beamforming algorithm | |
| Gutierrez et al. | An eigenpath underwater acoustic communication channel simulation | |
| Sun et al. | Secondary channel estimation in spatial active noise control systems using a single moving higher order microphone |
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 |