CN108304855B - Multi-submarine characteristic signal blind source separation method in marine environment - Google Patents

Multi-submarine characteristic signal blind source separation method in marine environment Download PDF

Info

Publication number
CN108304855B
CN108304855B CN201711267293.XA CN201711267293A CN108304855B CN 108304855 B CN108304855 B CN 108304855B CN 201711267293 A CN201711267293 A CN 201711267293A CN 108304855 B CN108304855 B CN 108304855B
Authority
CN
China
Prior art keywords
clustering
cluster
mask
signal
clustering algorithm
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
Application number
CN201711267293.XA
Other languages
Chinese (zh)
Other versions
CN108304855A (en
Inventor
孙海信
王连生
成垦
沈艺珊
蔚然
王远旭
陈海兰
孙伟涛
苗永春
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.)
Xiamen University
Original Assignee
Xiamen University
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 Xiamen University filed Critical Xiamen University
Priority to CN201711267293.XA priority Critical patent/CN108304855B/en
Publication of CN108304855A publication Critical patent/CN108304855A/en
Application granted granted Critical
Publication of CN108304855B publication Critical patent/CN108304855B/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
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G06F18/232—Non-hierarchical techniques
    • G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • 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

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Obtaining Desirable Characteristics In Audible-Bandwidth Transducers (AREA)
  • Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)

Abstract

The invention discloses a multi-submarine characteristic signal blind source separation method in a marine environment, which comprises the following steps: collecting a signal sample; clustering the mixed signal samples by adopting a clustering algorithm based on a Hermitian angle sample vector and a reference vector in a time-frequency domain to obtain a plurality of clustering clusters, and taking a membership function in the clustering algorithm as a mask; performing cluster verification based on the estimated cluster number; clustering the masks into Q mask cluster based on a K-means clustering algorithm; by mask estimation, signal separation is performed. The method can solve the problem of blind source separation under the underdetermined condition, does not need to estimate any mixed signal matrix or source position in mask estimation, simplifies the complexity of signal separation and improves the operation efficiency.

Description

Multi-submarine characteristic signal blind source separation method in marine environment
Technical Field
The invention relates to the field of underwater acoustic signal processing, in particular to a method and a system for separating blind sources of multi-submarine characteristic signals in a marine environment.
Background
The current blind separation algorithm has many factor limitations, firstly, the blind separation algorithm is mainly used in a transient mixed form of signals, and the blind separation algorithms in a signal convolution mixed form are relatively few and can be roughly divided into the following categories: 1) the network solves the problem by using a feedback circuit based on a signal separation network that eliminates cross-correlation between signal components. The algorithm has a great defect in practical application, and has a poor separation effect even under the condition that the scale difference between signals is large or a mixed matrix is pathological, so that the signals cannot be separated. 2) The output is transformed by a nonlinear transfer function, such as a entropy maximization method based on the principle of information transmission maximization. The algorithm has a low convergence rate, and the inversion of the separation matrix brings unstable values. 3) The nonlinear Principal Component Analysis (PCA) algorithm, which is a generalization of the linear principal component analysis method. In addition, the above algorithm needs to estimate any mixed signal matrix or source position in mask estimation, so that the signal separation process is relatively complex and the operation efficiency is not high.
Disclosure of Invention
The invention aims to provide a multi-submarine characteristic signal blind source separation method in a marine environment, which can solve the problem of blind source separation under an underdetermined condition, simplifies the complexity of signal separation and improves the operation efficiency.
In order to achieve the purpose, the invention adopts the following technical scheme:
a multi-submarine characteristic signal blind source separation method in a marine environment comprises the following steps:
s1, collecting a mixed signal sample;
s2, clustering the mixed signal samples by adopting a clustering algorithm to obtain a plurality of clustering clusters based on Hermitian angle sample vectors and reference vectors in a time-frequency domain, and taking membership functions in the clustering algorithm as masks;
s3, performing cluster verification based on the estimated cluster number;
s4, clustering the masks into Q mask cluster based on K-means clustering algorithm, and recording as cqQ is 1.. Q, and satisfies DqQ1.. the sum of Q is minimal, DqIs the total distance of the mask from the cluster centroid in the qth cluster, i.e.:
Figure GDA0002470723420000021
wherein the content of the first and second substances,
Figure GDA0002470723420000022
is the ith mask, C, of the k frequency pointqIs the qth cluster cqThe center of mass of the magnetic field sensor,
Figure GDA0002470723420000023
is that
Figure GDA0002470723420000024
And cluster centroid CqPearson's correlation coefficient between, kstAnd kendAre the start and end frequency points of the adjacent frequency point groups used for clustering, i.e. the total number of frequency points is kend-kst+1，
Figure GDA0002470723420000025
As a distance measure, such highly correlated masks will come from one cluster;
s5, signal separation is performed by mask estimation.
Preferably, the clustering algorithm in step S2 adopts a K-means clustering or fuzzy C-means clustering algorithm.
After adopting the technical scheme, compared with the background technology, the invention has the following advantages:
the method is based on the sample vector and the reference vector of the Hermitian angle, the mixed signal samples are gathered by using a clustering algorithm, the unknown signals are separated, the problem of blind source separation under the underdetermined condition can be solved, any mixed signal matrix or source position does not need to be estimated in mask estimation, the complexity of signal separation is simplified, and the operation efficiency is improved.
Drawings
FIG. 1 is a schematic flow diagram of the present invention;
FIG. 2 is a schematic diagram of the arrangement of source signals and receivers;
FIG. 3 is a schematic diagram of a channel model;
FIG. 4a shows the original acoustic signal waveform of a blue whale;
FIG. 4b shows the original acoustic signal waveform of a whale;
FIG. 5a shows the original sound spectrogram of beluga;
FIG. 5b shows the original sound spectrum of Centaurium;
FIG. 6a shows a waveform of a mixed sound captured by the receiver R1;
FIG. 6b shows the waveform of the mixed sound captured by the receiver R2;
FIG. 7a is a waveform of an isolated sound of a blue whale;
FIG. 7b is a waveform of an isolated whale sound;
FIG. 8a is a graph of the isolated sound spectrum of a blue whale;
fig. 8b is a graph of the isolated sound spectrum of whale standing.
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.
Example one
Referring to fig. 1, the invention discloses a method for separating blind sources of multi-submarine characteristic signals in a marine environment, comprising the following steps:
and S1, collecting mixed signal samples.
And S2, clustering the mixed signal samples by adopting a clustering algorithm to obtain a plurality of clustering clusters based on a Hermitian (Hermitian matrix) angle sample vector and a reference vector in a time-frequency domain, and taking a membership function in the clustering algorithm as a mask. The clustering algorithm adopts a K-mean clustering algorithm or a fuzzy C-mean clustering algorithm.
And S3, performing cluster verification based on the estimated cluster number.
S4, clustering the masks into Q mask cluster based on K-means clustering algorithm, and recording as cqQ is 1.. Q, and satisfies DqQ1.. the sum of Q is minimal, DqIs masked in the qth clusterThe total distance from the cluster centroid, i.e.:
Figure GDA0002470723420000041
wherein the content of the first and second substances,
Figure GDA0002470723420000042
is the ith mask, C, of the k frequency pointqIs the qth cluster cqThe center of mass of the magnetic field sensor,
Figure GDA0002470723420000043
is that
Figure GDA0002470723420000044
And cluster centroid CqPearson's correlation coefficient between, kstAnd kendAre the start and end frequency points of the adjacent frequency point groups used for clustering, i.e. the total number of frequency points is kend-kst+1，
Figure GDA0002470723420000045
As a distance measure, such highly correlated masks will come from one cluster.
S5, signal separation is performed by mask estimation.
Evaluation of experiments
One, channel design
Two signal receiving ends (R1 and R2) and two sound signal emitting ends (S1 and S2) are simulated underwater at any positions. Wherein R1 is 3.5km from S1, R2 is 4.0km from S2, R1 is 4.5km from S2, and R2 is 4.7km from S1, the arrangement is shown in FIG. 2.
R1 and S1 form channel h11, R1 and S2 form channel h12, R2 and S1 form channel h21, R2 and S2 form channel h22, and four channel models are shown in fig. 3.
Second, simulation experiment
The simulation experiment is carried out on a Matlab2014b platform, the processor of the computer is AMD dual-core A6-4400M2.7GHz, and the memory is 4G.
Sounds used in the simulation experiment are the sounds of blue whales and standing whales respectively, and a sound separation experiment for different types of whales and a sound separation experiment for two-head blue whales are performed respectively.
FIG. 4a shows the original acoustic signal waveform of a whale, and FIG. 4b shows the original acoustic signal waveform of a whale; FIG. 5a shows the original sound spectrogram of beluga, and FIG. 5b shows the original sound spectrogram of Ontario whale; fig. 6a shows a mixed sound waveform captured by the receiving terminal R1, and fig. 6b shows a mixed sound waveform captured by the receiving terminal R2.
Third, analysis of separation results
The blind source separation method according to the present invention separates the mixed audio signal into waveforms as shown in fig. 7a and 7b, where fig. 7a is the separated sound waveform of blue whale and fig. 7b is the separated sound waveform of whale. Comparing fig. 7a with fig. 4a, and comparing fig. 7b with fig. 4b, it can be seen that the blind source separation method of the present invention can better separate the mixed signals.
For further examination, the separated signal was subjected to Short Time Fourier Transform (STFT) to obtain spectral images as shown in fig. 8a and 8b, where fig. 8a is a spectrum of a separated blue whale sound and fig. 8b is a spectrum of a separated whale standing sound. Comparing fig. 8a with fig. 5a, the characteristic energies at the bottom of the two are consistent; comparing fig. 8b with fig. 5b, the characteristic energies at the bottom of both are consistent.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims (2)

1. A multi-submarine characteristic signal blind source separation method in a marine environment is characterized by comprising the following steps:
s1, collecting a mixed signal sample;
s2, clustering the mixed signal samples by adopting a clustering algorithm to obtain a plurality of clustering clusters based on Hermitian angle sample vectors and reference vectors in a time-frequency domain, and taking membership functions in the clustering algorithm as masks;
s3, performing cluster verification based on the estimated cluster number;
s4, clustering the masks into Q mask cluster based on K-means clustering algorithm, and recording as cqQ is 1.. Q, and satisfies DqQ1.. the sum of Q is minimal, DqIs the total distance of the mask from the cluster centroid in the qth cluster, i.e.:
Figure FDA0002454399080000011
wherein the content of the first and second substances,
Figure FDA0002454399080000012
is the ith mask, C, of the k frequency pointqIs the qth cluster cqThe center of mass of the magnetic field sensor,
Figure FDA0002454399080000013
is that
Figure FDA0002454399080000014
And cluster centroid CqPearson's correlation coefficient between, kstAnd kendAre the start and end frequency points of the adjacent frequency point groups used for clustering, i.e. the total number of frequency points is kend-kst+1，
Figure FDA0002454399080000015
As a distance measure, such highly correlated masks will come from one cluster;
s5, signal separation is performed by mask estimation.
2. The method for blind source separation of multi-submarine signatures according to claim 1, wherein: the clustering algorithm in step S2 adopts a K-means clustering or fuzzy C-means clustering algorithm.
CN201711267293.XA 2017-12-05 2017-12-05 Multi-submarine characteristic signal blind source separation method in marine environment Active CN108304855B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201711267293.XA CN108304855B (en) 2017-12-05 2017-12-05 Multi-submarine characteristic signal blind source separation method in marine environment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201711267293.XA CN108304855B (en) 2017-12-05 2017-12-05 Multi-submarine characteristic signal blind source separation method in marine environment

Publications (2)

Publication Number Publication Date
CN108304855A CN108304855A (en) 2018-07-20
CN108304855B true CN108304855B (en) 2020-06-23

Family

ID=62870258

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201711267293.XA Active CN108304855B (en) 2017-12-05 2017-12-05 Multi-submarine characteristic signal blind source separation method in marine environment

Country Status (1)

Country Link
CN (1) CN108304855B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112133321B (en) * 2020-09-23 2021-05-14 青岛科技大学 Underwater acoustic signal Gaussian/non-Gaussian noise suppression method based on blind source separation

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105282067A (en) * 2015-09-16 2016-01-27 长安大学 Complex field blind source separation method

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105282067A (en) * 2015-09-16 2016-01-27 长安大学 Complex field blind source separation method

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
基于盲源分离的心肺音信号分离方法研究与应用;郭鹿鸣;《中国优秀硕士学位论文全文数据库 (基础科学辑)》;20170131;全文 *
欠定盲源分离算法及其应用研究;张良俊;《中国优秀硕士学位论文全文数据库 (基础科学辑)》;20150320;全文 *

Also Published As

Publication number Publication date
CN108304855A (en) 2018-07-20

Similar Documents

Publication Publication Date Title
Wang et al. A region-growing permutation alignment approach in frequency-domain blind source separation of speech mixtures
Kounades-Bastian et al. A variational EM algorithm for the separation of time-varying convolutive audio mixtures
US8364483B2 (en) Method for separating source signals and apparatus thereof
Yang et al. Under-determined convolutive blind source separation combining density-based clustering and sparse reconstruction in time-frequency domain
Bao et al. A compressed sensing approach to blind separation of speech mixture based on a two-layer sparsity model
Kidron et al. Cross-modal localization via sparsity
Traa et al. Blind multi-channel source separation by circular-linear statistical modeling of phase differences
Li et al. An EM algorithm for audio source separation based on the convolutive transfer function
CN109997186B (en) Apparatus and method for classifying acoustic environments
CN108304855B (en) Multi-submarine characteristic signal blind source separation method in marine environment
Serviere et al. Permutation correction in the frequency domain in blind separation of speech mixtures
CN110265060B (en) Speaker number automatic detection method based on density clustering
Jafari et al. Sparse coding for convolutive blind audio source separation
Kim et al. Efficient neighborhood-based topic modeling for collaborative audio enhancement on massive crowdsourced recordings
Mitianoudis et al. Using beamforming in the audio source separation problem
Mitianoudis Audio source separation using independent component analysis
CN110956978B (en) Sparse blind separation method based on underdetermined convolution aliasing model
Li et al. Expectation‐maximisation for speech source separation using convolutive transfer function
Fakhry et al. Underdetermined source detection and separation using a normalized multichannel spatial dictionary
Dias et al. Multichannel source separation using time-deconvolutive CNMF
Al-Saegh Independent Component Analysis for Separation of Speech Mixtures: A Comparison Among Thirty Algorithms.
Saab et al. Underdetermined sparse blind source separation with delays
Chen et al. LocSelect: Target Speaker Localization with an Auditory Selective Hearing Mechanism
Haddad et al. Blind and semi-blind anechoic mixing system identification using multichannel matching pursuit
Watanabe et al. DNN-based frequency component prediction for frequency-domain audio source separation

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