CN112257484A - Multi-sound-source direction finding method and system based on deep learning - Google Patents

Multi-sound-source direction finding method and system based on deep learning Download PDF

Info

Publication number
CN112257484A
CN112257484A CN201910661146.3A CN201910661146A CN112257484A CN 112257484 A CN112257484 A CN 112257484A CN 201910661146 A CN201910661146 A CN 201910661146A CN 112257484 A CN112257484 A CN 112257484A
Authority
CN
China
Prior art keywords
neural network
deep neural
sound
fourier transform
sound source
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
Application number
CN201910661146.3A
Other languages
Chinese (zh)
Other versions
CN112257484B (en
Inventor
徐及
黄兆琼
颜永红
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Institute of Acoustics CAS
Original Assignee
Institute of Acoustics CAS
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Institute of Acoustics CAS filed Critical Institute of Acoustics CAS
Priority to CN201910661146.3A priority Critical patent/CN112257484B/en
Publication of CN112257484A publication Critical patent/CN112257484A/en
Application granted granted Critical
Publication of CN112257484B publication Critical patent/CN112257484B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/08—Feature extraction
    • G06F2218/10—Feature extraction by analysing the shape of a waveform, e.g. extracting parameters relating to peaks
    • 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
    • G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • General Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Optimization (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Pure & Applied Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computational Mathematics (AREA)
  • Mathematical Analysis (AREA)
  • Biomedical Technology (AREA)
  • Signal Processing (AREA)
  • Algebra (AREA)
  • Databases & Information Systems (AREA)
  • Circuit For Audible Band Transducer (AREA)
  • Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)

Abstract

The invention relates to a deep learning-based multi-sound-source direction finding method and system, wherein the method comprises the following steps: converting the acoustic source signals received by the array into digital acoustic signals; performing Fourier transform on the digital sound signal to obtain a signal after Fourier transform; inputting the signals after Fourier transform into a deep neural network, and training the deep neural network by using a back propagation method until the deep neural network is converged; and finding the azimuth corresponding to the peak value from the posterior probability output by the converged deep neural network, wherein the azimuth is the azimuth of the sound source. The method of the invention utilizes the deep neural network, finds the optimal solution of the sound source direction of arrival by optimizing the criterion function, obtains the direction of the target directly from the signals received by the array, and realizes simple and efficient multi-sound source direction finding.

Description

Multi-sound-source direction finding method and system based on deep learning
Technical Field
The invention relates to the technical field of sound source direction finding, in particular to a multi-sound-source direction finding method and system based on deep learning.
Background
The sound source direction finding technology can indicate the spatial direction of a sound source target, and provides important spatial information for subsequent information acquisition and processing.
The traditional method mainly utilizes the modern digital signal processing technology to estimate the azimuth information of the sound source, gives the direction of arrival of the sound source through a lattice point matching search or analysis mode, and has higher calculation complexity.
In recent years, deep neural networks have been widely used to solve the conventional signal processing problem, however, most of the previous studies of applying deep learning to direction finding are based on shallow neural networks or conventional neural networks. However, the deep neural network has stronger capability of finding the optimal solution than the shallow neural network, and the deep neural network has certain advantages when being used for finding the optimal solution of the arrival direction of the sound source.
Disclosure of Invention
The invention aims to solve the problem of multi-sound-source direction finding in the actual environment, provides a multi-sound-source direction finding method and system based on deep learning, and realizes robust and efficient multi-sound-source direction finding by utilizing a deep neural network.
In order to achieve the above object, the present invention provides a deep learning-based multi-sound-source direction finding method, including:
converting the acoustic source signals received by the array into digital acoustic signals; performing Fourier transform on the digital sound signal to obtain a signal after Fourier transform;
inputting the signals after Fourier transform into a deep neural network, and training the deep neural network by using a back propagation method until the deep neural network is converged;
and finding the azimuth corresponding to the peak value from the posterior probability output by the converged deep neural network, wherein the azimuth is the azimuth of the sound source.
As an improvement of the above method, said converting the acoustic source signal received by the array into a digital acoustic signal; performing Fourier transform on the digital sound signal to obtain a signal after Fourier transform; the method specifically comprises the following steps:
the array comprises K microphones; fourier transform is carried out on the signals y (t) received by the array, and real parts and imaginary parts of Fourier transform coefficients at each frequency are connected in series with a vector X:
Y(fi)＝fft(y(t))，
Figure BDA0002138587670000021
wherein, Y (f)i) Is a frequency fiThe Fourier transform coefficients of (A); i is a frequency index, and F is the number of points of Fourier transform; real (Y (f)i) Is Y (f)i) Real part of (a), imag (Y (f)i) Is Y (f)i) An imaginary part of (d); y (f)i)＝[Y1(fi)，Y2(fi)，…，YL(fi)]And L is the number of fast beats.
As an improvement of the above method, the signals after fourier transform are input into a deep neural network, and the deep neural network is trained by using a back propagation method until the deep neural network converges; the method specifically comprises the following steps:
inputting the vector X into a deep neural network, wherein a training criterion function gamma of the deep neural network is as follows:
Figure BDA0002138587670000022
wherein, gamma isf，p＝H(θp，fi)[HH(θp，fi)H(θp，fi)]HH(θp，fi) Defining a steering vector
Figure BDA0002138587670000023
Figure BDA0002138587670000024
Theta P is the direction of arrival of the P sound source, and P is more than or equal to 1 and less than or equal to P; p is the number of sound sources;τkis the time delay between the kth microphone and the first microphone; k is more than or equal to 1 and less than or equal to K; gamma-shapedf，pYl(fi) Is a K x 1 dimensional observation vector projected onto a steering vector H (theta)p，fi) In space of (a);
when the training criterion is minimized, the deep neural network converges.
As an improvement of the above method, the direction of arrival of the sound source is: thetapAngle interval p Δ θ
Figure BDA0002138587670000025
Figure BDA0002138587670000026
As an improvement of the above method, the finding the azimuth corresponding to the peak from the posterior probability output by the converged deep neural network is that the azimuth of the sound source appears, and specifically:
after the deep neural network converges, the output vector is: z is ═ z1，z2，…，zP]T，zp∈[0，1]The posterior probability of the occurrence of the p sound source;
from z1，z2，…，zPExtracting a plurality of maxima in the posterior probability;
calculating a threshold value delta:
δ＝Oavg+μ(Omax-Oavg)，
wherein, OavgAnd OmaxRepresents the average value and the maximum value of the posterior probability, and mu is a parameter;
for maxima greater than the threshold δ, the corresponding bearing is the bearing where the sound source may appear.
The invention also provides a deep learning-based multi-sound-source direction-finding system, which comprises:
the signal conversion module is used for converting the sound source signals received by the array into digital sound signals; performing Fourier transform on the digital sound signal to obtain a signal after Fourier transform;
the deep neural network training module is used for inputting the signals subjected to Fourier transform into a deep neural network and training the deep neural network by using a back propagation method until the deep neural network converges;
and the sound source direction finding module is used for finding the direction corresponding to the peak value from the posterior probability output by the converged deep neural network, wherein the direction is the direction in which the sound source appears.
As an improvement of the above system, the specific implementation process of the signal conversion module is as follows:
fourier transform is carried out on the signals y (t) received by the array, and real parts and imaginary parts of Fourier transform coefficients at each frequency are connected in series with a vector X:
Y(fi)＝fft(y(t))，
Figure BDA0002138587670000031
wherein the array comprises K microphones; y (f)i) Is a frequency fiThe Fourier transform coefficients of (A); i is a frequency index, and F is the number of points of Fourier transform; real (Y (f)i) Is Y (f)i) Real part of (a), imag (Y (f)i) Is Y (f)i) An imaginary part of (d); y (f)i)＝[Y1(fi)，Y2(fi)，…，YL(fi)]And L is the number of fast beats.
As an improvement of the above system, the implementation process of the deep neural network training module is as follows:
inputting the vector X into a deep neural network, wherein a training criterion function gamma of the deep neural network is as follows:
Figure BDA0002138587670000032
wherein, gamma isf，p＝H(θp，fi)[HH(θp，fi)H(θp，fi)]HH(θp，fi) Defining a steering vector
Figure BDA0002138587670000033
Figure BDA0002138587670000034
θpP is more than or equal to 1 and less than or equal to P as the arrival direction of the pth sound source; p is the number of sound sources; thetapAngle interval p Δ θ
Figure BDA0002138587670000035
τkIs the time delay between the kth microphone and the first microphone; k is more than or equal to 1 and less than or equal to K; gamma-shapedf，pYl(fi) Is a K x 1 dimensional observation vector projected onto a steering vector H (theta)p，fi) In space of (a);
when the training criterion is minimized, the deep neural network converges.
As an improvement of the above system, the sound source direction finding module is implemented by the following specific processes:
after the deep neural network converges, the output vector is: z is ═ z1，z2，…，zP]T，zp∈[0，1]The posterior probability of the occurrence of the p sound source;
from z1，z2，…，zPExtracting a plurality of maxima in the posterior probability;
calculating a threshold value delta:
δ＝Oavg+μ(Omax-Oavg)，
wherein, OavgAnd OmaxRepresents the average value and the maximum value of the posterior probability, and mu is a parameter;
for maxima greater than the threshold δ, the corresponding bearing is the bearing where the sound source may appear.
The invention has the advantages that:
the invention provides an underwater multi-sound-source direction finding method based on deep learning, which utilizes a deep neural network to search an optimal solution of the sound source direction of arrival through an optimization criterion function.
Drawings
Fig. 1 is a flowchart of an underwater multi-sound-source direction finding method based on deep learning according to embodiment 1 of the present invention.
Detailed Description
The invention will now be further described with reference to the accompanying drawings and specific embodiments.
Example 1:
as shown in fig. 1, embodiment 1 of the present invention provides a deep learning-based multi-sound-source direction finding method, including the following steps
Step 1) converting sound source signals received by an array into digital sound signals;
the array includes K sensors (microphones).
Step 2) carrying out Fourier transform on the digital sound signal;
step 3) inputting the signals after Fourier transform into a deep neural network;
fourier transform is carried out on the signals y (t) received by the array, and the real part and the imaginary part of a Fourier transform coefficient on each frequency are connected in series to serve as the input X of the neural network:
Y(fi)＝fft(y(t))，
Figure BDA0002138587670000041
wherein, Y (f)i) Is a frequency fiThe Fourier transform coefficients of (A); i is a frequency index, and F is the number of points of Fourier transform; real (Y (f)i) Is Y (f)i) Real part of (a), imag (Y (f)i) Is Y (f)i) An imaginary part of (d);
step 4), training a neural network by using a back propagation method until convergence;
inputting the vector X into a deep neural network, wherein a training criterion function gamma of the deep neural network is as follows:
Figure BDA0002138587670000051
wherein, gamma isf，p＝H(θp，fi)[HH(θp，fi)H(θp，fi)]HH(θp，fi) Defining a steering vector
Figure BDA0002138587670000052
Figure BDA0002138587670000053
θpP is more than or equal to 1 and less than or equal to P as the arrival direction of the pth sound source; p is the number of sound sources; thetapAngle interval p Δ θ
Figure BDA0002138587670000054
τkIs the time delay between the kth microphone and the first microphone; k is more than or equal to 1 and less than or equal to K; gamma-shapedf，pYl(fi) Is a K x 1 dimensional observation vector projected onto a steering vector H (theta)p，fi) In space of (a);
when the training criterion is minimized, the deep neural network converges.
And 5) analyzing the converged neural network output posterior probability, and finding the azimuth corresponding to the peak value, namely the azimuth of the sound source.
After the neural network converges, extracting an extreme value in the posterior probability as the position where the sound source appears, and when the posterior probability output by the neural network is higher than a threshold:
δ＝Oavg+μ(Omax-Oavg)，
wherein, OavgAnd OmaxThe mean and maximum values of the posterior probabilities are represented, where μ is generated empirically. And extracting the azimuth corresponding to the probability higher than the threshold delta as the azimuth of the sound source possibly appearing.
The posterior probability of the output of the neural network reflects the probability of the sound source appearing at this orientation when the minimum is reached by the criterion function in step 4) above. After the neural network converges, the occurrence orientation of the sound source can be obtained by analyzing the peak value of the posterior probability.
Example 2:
embodiment 2 of the present invention provides a deep learning-based multi-sound-source direction finding system, which includes:
the signal conversion module is used for converting the sound source signals received by the array into digital sound signals; performing Fourier transform on the digital sound signal to obtain a signal after Fourier transform;
the specific implementation process of the signal conversion module is as follows:
fourier transform is carried out on the signals y (t) received by the array, and real parts and imaginary parts of Fourier transform coefficients at each frequency are connected in series with a vector X:
Y(fi)＝fft(y(t))，
Figure BDA0002138587670000061
wherein the array comprises K microphones; y (f)i) Is a frequency fiThe Fourier transform coefficients of (A); i is a frequency index, and F is the number of points of Fourier transform; real (Y (f)i) Is Y (f)i) Real part of (a), imag (Y (f)i) Is Y (f)i) An imaginary part of (d); y (f)i)＝[Y1(fi)，Y2(fi)，…，YL(fi)]And L is the number of fast beats.
The deep neural network training module is used for inputting the signals subjected to Fourier transform into a deep neural network and training the deep neural network by using a back propagation method until the deep neural network converges;
the specific implementation process of the deep neural network training module is as follows:
inputting the vector X into a deep neural network, wherein a training criterion function gamma of the deep neural network is as follows:
Figure BDA0002138587670000062
wherein, gamma isf，p＝H(θp，fi)[HH(θp，fi)H(θp，fi)]HH(θp，fi) Defining a steering vector
Figure BDA0002138587670000063
Figure BDA0002138587670000064
θpP is more than or equal to 1 and less than or equal to P as the arrival direction of the pth sound source; p is the number of sound sources; thetapAngle interval p Δ θ
Figure BDA0002138587670000065
τkIs the time delay between the kth microphone and the first microphone; k is more than or equal to 1 and less than or equal to K; gamma-shapedf，pYl(fi) Is a K x 1 dimensional observation vector projected onto a steering vector H (theta)p，fi) In space of (a);
when the training criterion is minimized, the deep neural network converges.
And the sound source direction finding module is used for finding the direction corresponding to the peak value from the posterior probability output by the converged deep neural network, wherein the direction is the direction in which the sound source appears.
The concrete implementation process of the sound source direction finding module is as follows:
after the deep neural network converges, the output vector is: z is ═ z1，z2，…，zP]T，zp∈[0，1]The posterior probability of the occurrence of the p sound source;
from z1，z2，…，zPExtracting a plurality of maxima in the posterior probability;
calculating a threshold value delta:
δ＝Oavg+μ(Omax-Oavg)，
wherein, OavgAnd OmaxRepresents the average value and the maximum value of the posterior probability, and mu is a parameter;
for maxima greater than the threshold δ, the corresponding bearing is the bearing where the sound source may appear.
Finally, it should be noted that the above embodiments are only used for illustrating the technical solutions of the present invention and are not limited. Although the present invention has been described in detail with reference to the embodiments, it will be understood by those skilled in the art that various changes may be made and equivalents may be substituted without departing from the spirit and scope of the invention as defined in the appended claims.

Claims (9)

1. A deep learning based multi-source direction finding method, the method comprising:
converting the acoustic source signals received by the array into digital acoustic signals; performing Fourier transform on the digital sound signal to obtain a signal after Fourier transform;
inputting the signals after Fourier transform into a deep neural network, and training the deep neural network by using a back propagation method until the deep neural network is converged;
and finding the azimuth corresponding to the peak value from the posterior probability output by the converged deep neural network, wherein the azimuth is the azimuth of the sound source.
2. The deep learning based multi-source direction finding method according to claim 1, wherein the converting the acoustic source signals received by the array into digital acoustic signals; performing Fourier transform on the digital sound signal to obtain a signal after Fourier transform; the method specifically comprises the following steps:
fourier transform is carried out on the signals y (t) received by the array, and real parts and imaginary parts of Fourier transform coefficients at each frequency are connected in series with a vector X:
Y(fi)＝fft(y(t))，
Figure FDA0002138587660000011
wherein the array comprises K microphones; y (f)i) Is a frequency fiThe Fourier transform coefficients of (A); i is a frequency index, and F is the number of points of Fourier transform; real (Y (f)i) Is Y (f)i) Real part of (a), imag (Y (f)i) Is Y (f)i) An imaginary part of (d); y (f)i)＝[Y1(fi)，Y2(fi)，…，YL(fi)]And L is the number of fast beats.
3. The deep learning-based multi-sound-source direction finding method according to claim 2, wherein the Fourier transformed signal is input into a deep neural network, and the deep neural network is trained by using a back propagation method until the deep neural network converges; the method specifically comprises the following steps:
inputting the vector X into a deep neural network, wherein a training criterion function gamma of the deep neural network is as follows:
Figure FDA0002138587660000012
wherein, gamma isf，p＝H(θp，fi)[HH(θp，fi)H(θp，fi)]HH(θp，fi) Defining a steering vector
Figure FDA0002138587660000013
Figure FDA0002138587660000014
θpP is more than or equal to 1 and less than or equal to P as the arrival direction of the pth sound source; p is the number of sound sources; tau iskIs the time delay between the kth microphone and the first microphone; k is more than or equal to 1 and less than or equal to K; gamma-shapedf，pYl(fi) Is a K x 1 dimensional observation vector projected onto a steering vector H (theta)p，fi) In space of (a);
when the training criterion is minimized, the deep neural network converges.
4. Root of herbaceous plantThe deep learning-based multi-sound-source direction finding method according to claim 3, wherein the directions of arrival of the sound sources are: thetapAngle interval p Δ θ
Figure FDA0002138587660000021
5. The deep learning-based multi-sound-source direction finding method according to claim 3 or 4, wherein the position corresponding to the peak is found from the posterior probability output by the converged deep neural network, and the position is a position where a sound source appears, and specifically:
after the deep neural network converges, the output vector is: z is ═ z1，z2，…，zP]T，zp∈[0，1]The posterior probability of the occurrence of the p sound source;
from z1，z2，…，zPExtracting a plurality of maxima in the posterior probability;
calculating a threshold value delta:
δ＝Oavg+μ(Omax-Oavg)，
wherein, OavgAnd OmaxRepresents the average value and the maximum value of the posterior probability, and mu is a parameter;
for maxima greater than the threshold δ, the corresponding bearing is the bearing where the sound source may appear.
6. A deep learning based multi-source direction finding system, the system comprising:
the signal conversion module is used for converting the sound source signals received by the array into digital sound signals; performing Fourier transform on the digital sound signal to obtain a signal after Fourier transform;
the deep neural network training module is used for inputting the signals subjected to Fourier transform into a deep neural network and training the deep neural network by using a back propagation method until the deep neural network converges;
and the sound source direction finding module is used for finding the direction corresponding to the peak value from the posterior probability output by the converged deep neural network, wherein the direction is the direction in which the sound source appears.
7. The deep learning-based multi-sound-source direction finding system according to claim 6, wherein the signal conversion module is implemented by the following steps:
fourier transform is carried out on the signals y (t) received by the array, and real parts and imaginary parts of Fourier transform coefficients at each frequency are connected in series with a vector X:
Y(fi)＝fft(y(t))，
Figure FDA0002138587660000022
wherein the array comprises K microphones; y (f)i) Is a frequency fiThe Fourier transform coefficients of (A); i is a frequency index, and F is the number of points of Fourier transform; real (Y (f)i) Is Y (f)i) Real part of (a), imag (Y (f)i) Is Y (f)i) An imaginary part of (d); y (f)i)＝[Y1(fi)，Y2(fi)，…，YL(fi)]And L is the number of fast beats.
8. The deep learning-based multi-sound-source direction finding system according to claim 7, wherein the deep neural network training module is implemented by the following steps:
inputting the vector X into a deep neural network, wherein a training criterion function gamma of the deep neural network is as follows:
Figure FDA0002138587660000031
wherein, gamma isf，p＝H(θp，fi)[HH(θp，fi)H(θp，fi)]HH(θp，fi) Defining a steering vector
Figure FDA0002138587660000032
Figure FDA0002138587660000033
θpP is more than or equal to 1 and less than or equal to P as the arrival direction of the pth sound source; p is the number of sound sources; thetapAngle interval p Δ θ
Figure FDA0002138587660000034
τkIs the time delay between the kth microphone and the first microphone; k is more than or equal to 1 and less than or equal to K; gamma-shapedf，pYl(fi) Is a K x 1 dimensional observation vector projected onto a steering vector H (theta)p，fi) In space of (a);
when the training criterion is minimized, the deep neural network converges.
9. The deep learning-based multi-sound-source direction-finding system according to claim 8, wherein the sound source direction-finding module is implemented by the following steps:
after the deep neural network converges, the output vector is: z is ═ z1，z2，…，zP]T，zp∈[0，1]The posterior probability of the occurrence of the p sound source;
from z1，z2，…，zPExtracting a plurality of maxima in the posterior probability;
calculating a threshold value delta:
δ＝Oavg+μ(Omax-Oavg)，
wherein, OavgAnd OmaxRepresents the average value and the maximum value of the posterior probability, and mu is a parameter;
for maxima greater than the threshold δ, the corresponding bearing is the bearing where the sound source may appear.
CN201910661146.3A 2019-07-22 2019-07-22 Multi-sound source direction finding method and system based on deep learning Active CN112257484B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910661146.3A CN112257484B (en) 2019-07-22 2019-07-22 Multi-sound source direction finding method and system based on deep learning

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910661146.3A CN112257484B (en) 2019-07-22 2019-07-22 Multi-sound source direction finding method and system based on deep learning

Publications (2)

Publication Number Publication Date
CN112257484A true CN112257484A (en) 2021-01-22
CN112257484B CN112257484B (en) 2024-03-15

Family

ID=74224714

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910661146.3A Active CN112257484B (en) 2019-07-22 2019-07-22 Multi-sound source direction finding method and system based on deep learning

Country Status (1)

Country Link
CN (1) CN112257484B (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113835065A (en) * 2021-09-01 2021-12-24 深圳壹秘科技有限公司 Sound source direction determining method, device, equipment and medium based on deep learning
CN115902774A (en) * 2022-10-13 2023-04-04 广州成至智能机器科技有限公司 Unmanned aerial vehicle sound source positioning method and device, unmanned aerial vehicle and storage medium

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105204001A (en) * 2015-10-12 2015-12-30 Tcl集团股份有限公司 Sound source positioning method and system
CN105676168A (en) * 2015-12-02 2016-06-15 江苏科技大学 Acoustic vector array DOA estimation method
CN109031200A (en) * 2018-05-24 2018-12-18 华南理工大学 A kind of sound source dimensional orientation detection method based on deep learning
CN109839612A (en) * 2018-08-31 2019-06-04 大象声科（深圳）科技有限公司 Sounnd source direction estimation method based on time-frequency masking and deep neural network
CN109975762A (en) * 2017-12-28 2019-07-05 中国科学院声学研究所 A kind of underwater sound source localization method

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105204001A (en) * 2015-10-12 2015-12-30 Tcl集团股份有限公司 Sound source positioning method and system
CN105676168A (en) * 2015-12-02 2016-06-15 江苏科技大学 Acoustic vector array DOA estimation method
CN109975762A (en) * 2017-12-28 2019-07-05 中国科学院声学研究所 A kind of underwater sound source localization method
CN109031200A (en) * 2018-05-24 2018-12-18 华南理工大学 A kind of sound source dimensional orientation detection method based on deep learning
CN109839612A (en) * 2018-08-31 2019-06-04 大象声科（深圳）科技有限公司 Sounnd source direction estimation method based on time-frequency masking and deep neural network

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
ZHAOQIONG HUANG 等: "Source localization using deep neural networks in a shallow water environment", 2018 ACOUSTICAL SOCIETY OF AMERICA, pages 2922 - 2932 *

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113835065A (en) * 2021-09-01 2021-12-24 深圳壹秘科技有限公司 Sound source direction determining method, device, equipment and medium based on deep learning
CN113835065B (en) * 2021-09-01 2024-05-17 深圳壹秘科技有限公司 Sound source direction determining method, device, equipment and medium based on deep learning
CN115902774A (en) * 2022-10-13 2023-04-04 广州成至智能机器科技有限公司 Unmanned aerial vehicle sound source positioning method and device, unmanned aerial vehicle and storage medium
CN115902774B (en) * 2022-10-13 2023-11-07 广州成至智能机器科技有限公司 Unmanned aerial vehicle sound source positioning method and device, unmanned aerial vehicle and storage medium

Also Published As

Publication number Publication date
CN112257484B (en) 2024-03-15

Similar Documents

Publication Publication Date Title
CN109800700B (en) Underwater acoustic signal target classification and identification method based on deep learning
CN110610718B (en) Method and device for extracting expected sound source voice signal
CN112526451B (en) Compressed beam forming and system based on microphone array imaging
CN109427328B (en) Multichannel voice recognition method based on filter network acoustic model
CN109975762B (en) Underwater sound source positioning method
CN112257484B (en) Multi-sound source direction finding method and system based on deep learning
CN1251194A (en) Recognition system
CN110534126B (en) Sound source positioning and voice enhancement method and system based on fixed beam forming
CN112037813B (en) Voice extraction method for high-power target signal
CN109541526A (en) A kind of ring array direction estimation method using matrixing
CN111352075B (en) Underwater multi-sound-source positioning method and system based on deep learning
CN109581291B (en) Direct positioning method based on artificial bee colony
CN111446998B (en) Direction-of-arrival estimation method based on deep learning
CN106033120B (en) A kind of asynchronous multi-frame joint detection method of multistation radar
CN110927668A (en) Sound source positioning optimization method of cube microphone array based on particle swarm
CN108269581B (en) Double-microphone time delay difference estimation method based on frequency domain coherent function
CN116631438A (en) Width learning and secondary correlation sound source positioning method based on minimum p norm
CN108269583B (en) Voice separation method based on time delay histogram
Fu et al. Iterative sound source localization for unknown number of sources
CN111308423B (en) Robust sound source positioning system and method thereof
CN110673132B (en) Real-time filtering method for trace point sequence for multi-frame joint detection and tracking
CN110632579B (en) Iterative beam forming method using subarray beam domain characteristics
CN113223552A (en) Speech enhancement method, speech enhancement device, speech enhancement apparatus, storage medium, and program
CN111948608B (en) Underwater sound multipath signal arrival time difference estimation method based on sparse modeling
CN113628634B (en) Real-time voice separation method and device guided by directional information

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