CN110245608A - A kind of Underwater targets recognition based on semi-tensor product neural network - Google Patents
A kind of Underwater targets recognition based on semi-tensor product neural network Download PDFInfo
- Publication number
- CN110245608A CN110245608A CN201910513322.9A CN201910513322A CN110245608A CN 110245608 A CN110245608 A CN 110245608A CN 201910513322 A CN201910513322 A CN 201910513322A CN 110245608 A CN110245608 A CN 110245608A
- Authority
- CN
- China
- Prior art keywords
- semi
- neural network
- tensor product
- underwater
- product neural
- 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
Classifications
-
- 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
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/20—Scenes; Scene-specific elements in augmented reality scenes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
- G06F2218/04—Denoising
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Measurement Of Velocity Or Position Using Acoustic Or Ultrasonic Waves (AREA)
Abstract
The present invention provides a kind of Underwater targets recognitions based on semi-tensor product neural network, receive underwater sound signal by underwater sonar sensor, the time domain of acoustical signal and frequency domain information are presented in LOFAR map by Short Time Fourier Transform；It is constructed using LOFAR map sample as input feature vector matrix by data sample semi-tensor product neural network；The underwater sound signal received is divided into training set and verifying collection, input semi-tensor product neural network is trained and verifies；By choosing different hyper parameters, model training is carried out with double of tensor product neural network of training set, the test effect of contrast verification collection determines the high hyper parameter of test accuracy rate；Finally by the semi-tensor product neural network after the acoustical signal input model training of current collected submarine target, differentiation result is provided.The present invention can be improved Underwater Targets Recognition rate, expand application scenarios, suitable for identifying submarine target complicated ambient sea noise.
Description
Technical field
The invention belongs to field of signal processing, are related to neural network, Underwater Acoustic channels, semi-tensor product multiplication and underwater mesh
It identifies the methods of other.
Background technique
Target classification identification all has a very important significance all kinds of research fields, and traditional target classification identification is all
It is artificially to extract each category feature, then structural classification device carries out Classification and Identification.With the fast development of computer vision technique,
Target classification identification technology based on deep learning is studied extensively by everybody, also achieves the Classification and Identification effect for surmounting the mankind.
But it is mainly studied in the computer vision fields such as image and video at present, also concentrates on language for the research of acoustical signal
Sound signal processing and natural language processing, also rest in traditional method the Classification and Identification of all kinds of submarine targets.
Traditional acoustical signal recognizer generally have dynamic time warping technology, support vector machines, gauss hybrid models with
And hidden Markov model etc..Conventional method is extracted by characteristic of human nature and Environmental Noise Influence is larger, these identification models are all only
Be a kind of symbolism system, reduce the ability of modeling, the recognition performance of the acoustical signal in complex environment pair will substantially under
Drop, causes the Classification and Identification rate in ambient sea noise complicated and changeable not ideal enough.Artificial neural network is artificial in recent years
The research hotspot that smart field rises.As the research work of artificial neural network deepens continuously, have been achieved at present very big
Progress.
Summary of the invention
For overcome the deficiencies in the prior art, the present invention provides a kind of submarine target knowledge based on semi-tensor product neural network
This novel matrix operation of semi-tensor product multiplication is introduced neural network, establishes a kind of mind based on semi-tensor product by other method
Through network model, and it is applied to the processing of acoustical signal, can be improved Underwater Targets Recognition rate, expand application scenarios, is suitable for
Submarine target is identified in complicated ambient sea noise.
The technical solution adopted by the present invention to solve the technical problems by underwater sonar sensor the following steps are included: connect
Underwater sound signal is received, includes the acoustical signal of submarine target and the marine environment ambient noise without submarine target；By acoustical signal
Time domain and frequency domain information are presented in LOFAR map by Short Time Fourier Transform；It is special using LOFAR map sample as input
Matrix building is levied by data sample semi-tensor product neural network；The underwater sound signal received is divided into training set and verifying collection,
Input semi-tensor product neural network is trained and verifies；By choosing different hyper parameters, with double of tensor product mind of training set
Model training is carried out through network, the test effect of contrast verification collection determines the high hyper parameter of test accuracy rate；It finally will be current
Semi-tensor product neural network after the acoustical signal input model training of collected submarine target, provides differentiation result.
The underwater sound signalIn formula, h (t) indicates ocean Channel Impulse
Response, s (t) indicate Underwater Target Signal, and n (t) is ambient sea noise, and * indicates convolution algorithm, and t is time variable.
After the underwater sonar sensor receives underwater sound signal, design high-pass filter filters out low frequency range noise,
Then framing pretreatment is carried out.
A length of 1s when each frame signal in the framing pretreatment.
The Short Time Fourier Transform isX (t) representation signal in formula,
W (t) is window function, is played the role of the time limit, e-jωtPlay frequency limit.
In the semi-tensor product neural network,Y in formulaiIndicate semi-tensor product operation
Eigenmatrix afterwards, xiIndicate the eigenmatrix of input, WiFormula weight matrix, biIndicate bias term, fc() indicates semi-tensor product
Layer activation primitive,Indicate semi-tensor product operation；Input feature vector is after semi-tensor product carries out feature extraction, the characteristic pattern of output
It is passed to pond layer and carries out feature selecting and information filtering, pond process is zi=βipool(yi)+bi, in formula, pool ()
It indicates yiCarry out sampling operation, βiExpression multiplies biasing.
The underwater sound signal is divided into K subsample, and an individual subsample is kept as verifying model
Data, other K-1 sample are used to substitute into building semi-tensor product neural network and are trained；Cross validation repetition K times, each
Subsample verifying is primary, and average K result carrys out the performance indicator as classification of assessment device.
The beneficial effects of the present invention are: determining noise model is not assumed that because neural network model is trained by data,
Reduce and model bring error to ambient sea noise complicated and changeable, training arithmetic speed faster, and compares conventional machines
Learning algorithm has better robustness, while can also be improved the discrimination of underwater sound source target.This method can be adapted for
The ambient sea noise of all kinds of complexity, for submarine target Intellisense, seafari, underwater anomaly detection and wisdom
Ocean and ocean defence construction have far reaching significance.The present invention uses and successfully applies artificial intelligence technology in ocean letter
Breath perception and Underwater Target Detection have pushed the technology in the application and development in the fields such as wisdom ocean and ocean defence.
Detailed description of the invention
Fig. 1 is group method flow chart of the invention；
Fig. 2 is semi-tensor product neural network structure schematic diagram of the invention.
Specific embodiment
Present invention will be further explained below with reference to the attached drawings and examples, and the present invention includes but are not limited to following implementations
Example.
The present invention receives the acoustical signal of submarine target by underwater sonar sensor, acquires a large amount of naval vessels, fishing boat, speedboat
The acoustical signal of equal variety classes submarine target and all kinds of marine environment ambient noises, are believed by Short Time Fourier Transform generation sound
Number LOFAR spectrogram, by data sample be divided into training set and verifying collection input semi-tensor product neural network be trained.In conjunction with
Neural network parameter adjustment, so that the effect based on training set and verifying collection reaches best.Finally by the acoustical signal of submarine target
Input semi-tensor product neural network provides differentiation result.
As shown in Figure 1, the present invention the following steps are included:
Step 1: underwater sonar sensor receives signal
Underwater sonar sensor is placed in marine environment, the signal x (t) received are as follows:
In formula, h (t) indicates that ocean channel impulse response, s (t) indicate Underwater Target Signal, and n (t) makes an uproar for marine environment
Sound, * indicate convolution algorithm, and t is time variable.It is that receive be pure ambient sea noise without target.
Step 2: Signal Pretreatment
Since ocean low frequency environments noise level is higher, design high-pass filter, filter out low frequency range noise, then into
Row framing pretreatment, when each frame signal a length of 1s.Signal sub-frame processing can increase number of training and normalization sample letter
Number length is prepared to obtain the visualization spectrogram sample of unified size in next step.
Step 3: obtaining acoustical signal time-frequency visualizes LOFAR spectrogram
It takes and the time domain of acoustical signal and frequency domain information is presented on LOFAR map by way of Short Time Fourier Transform
In:
X (t) representation signal (indicating the reception sample of signal of 1s length herein) in formula, w (t) is window function, plays the time limit
Effect, e-jωtPlay frequency limit.
Step 4: building semi-tensor product neural network
The convolution algorithm in conventional convolution neural network is replaced with semi-tensor matrix multiplication product, by a nuclear matrix to every
Layer input directly carries out semi-tensor product multiplication.Matrix semi-tensor product matrix multiplication can realize two matrix multiples of Arbitrary Dimensions.
General Study is left semi-tensor product.For giving two matrix A ∈ Mm×nWith B ∈ Mp×q, then its semi-tensor is long-pending are as follows:
Wherein r=lcm (n, p) is the least common multiple of { n, p },For Kronecker product operation.
Semi-tensor lamination: the process of entire semi-tensor product can be indicated with following formula:
Y in formulaiIndicate the eigenmatrix after semi-tensor product operation, xiIndicate that the eigenmatrix of input (is obtained with third step
The LOFAR spectrogram sample obtained is as input feature vector matrix), WiFormula weight matrix, biIndicate bias term, fc() indicates semi-tensor
Lamination activation primitive,Indicate semi-tensor product operation.
Pond layer: for input feature vector after semi-tensor product carries out feature extraction, the characteristic pattern of output can be passed to pond
Layer carries out feature selecting and information filtering, that is, down-sampling.The process in pond can be indicated with following formula:
zi=βipool(yi)+bi
In formula, pool () indicates yi carrying out sampling operation, i.e., above-mentioned pondization operation；βiExpression multiplies biasing, biIndicate inclined
Set item.
Full articulamentum: full articulamentum usually builds the decline in convolutional neural networks hidden layer, only connects entirely to other
Connect layer transmitting signal.Full articulamentum is followed by output layer, and output layer uses logical function or normalization exponential function output category
Label.
Step 5: building training set sample set
At K subsample, an individual subsample is retained to be made the Segmentation of Data Set that underwater sonar sensor is acquired
For the data for verifying model, other K-1 sample is used to substitute into building semi-tensor product neural network and is trained.Cross validation
It repeats K times, each subsample verifying is primary, and average K result carrys out the performance indicator as classification of assessment device.
Step 6: adjusting hyper parameter training optimization neural network model
By choosing different hyper parameters, model training is carried out with training set, the test effect of contrast test collection determines
The high hyper parameter of test accuracy rate.The hyper parameter includes the learning rate of neural network, batch size, the number of iterations, half
Measure product core size and activation primitive.This step is used for optimization neural network model, improves the performance and effect of e-learning.
Step 7: realizing Underwater Targets Recognition classification
After testing data is become by pretreatment, the semi-tensor product neural network model that can have been optimized by training is provided
Classification results realize Underwater Targets Recognition classification.
The embodiment of the present invention for conventional method Underwater Targets Recognition rate is low and the weak problem of robustness, proposition based on
Steps are as follows for the Underwater targets recognition of semi-tensor product neural network:
Step 1: underwater sonar sensor receives signal
Underwater sonar sensor is placed in marine environment, the signal x (t) received are as follows:
In formula, h (t) indicates that ocean channel impulse response, n (t) are ambient sea noise, and * indicates convolution algorithm, when t is
Between variable.It is that receive be pure ambient sea noise without target.
Step 2: Signal Pretreatment
Since ocean low frequency environments noise level is higher, second order Butterworth filter is designed, bilinear transformation is passed through
Derive digital butterworth high pass filter.The voice data received is obtained by second order butterworth high pass filter
To filtered acoustical signal, filter out low frequency range noise, then carry out framing pretreatment, when each frame signal a length of 1s.
Step 3: obtaining acoustical signal time-frequency visualizes LOFAR spectrogram
It takes and the time domain of acoustical signal and frequency domain information is presented on LOFAR map by way of Short Time Fourier Transform
In:
X (t) representation signal in formula, w (t) are window function, are played the role of the time limit, e-jωtPlay frequency limit, window function is optional
With Hanning window, Fourier transformation points are selected as 1024, and window length is selected as 1024.
Step 4: building semi-tensor product neural network
The convolution algorithm in conventional convolution neural network is replaced with semi-tensor matrix multiplication product, by a nuclear matrix to every
Layer input directly carries out semi-tensor product multiplication.Matrix semi-tensor product matrix multiplication can realize two matrix multiples of Arbitrary Dimensions.
General Study is left semi-tensor product.Give two matrix A ∈ Mm×nWith B ∈ Mp×q, then its semi-tensor is long-pending are as follows:
Wherein t=lcm (n, p) is the least common multiple of { n, p },For Kronecker product operation.
Semi-tensor lamination: the process of entire semi-tensor product can be indicated with following formula:
Y in formulaiIndicate the eigenmatrix after semi-tensor product operation, xiIndicate the eigenmatrix of input, WiFormula weight square
Battle array, biIndicate bias term, fc() indicates semi-tensor lamination activation primitive,Indicate semi-tensor product operation.
Pond layer: for input feature vector after semi-tensor product carries out feature extraction, the characteristic pattern of output can be passed to pond
Layer carries out feature selecting and information filtering, that is, down-sampling.The process in pond can be indicated with following formula:
zi=βipool(yi)+bi
In formula, pool () is indicated yiDown-sampling operation is carried out, i.e., above-mentioned pondization operation；βiExpression multiplies biasing, biIt indicates
Bias term.
Full articulamentum: full articulamentum usually builds the decline in convolutional neural networks hidden layer, only connects entirely to other
Connect layer transmitting signal.Full articulamentum is followed by output layer, and output layer uses logical function or normalization exponential function output category
Label.
The convolutional layer that is of five storeys, 5 layers of pond layer, the semi-tensor product neural network of 1 layer of full articulamentum, output layer nerve are gathered around in building
First number is class categories number.
Step 5: building training set sample set
The tape label data acquired from underwater sonar sensor, then by Segmentation of Data Set at K sub- sample sets, one
Individual subsample is kept as the data of verifying model, other K-1 sample set is used to train.Cross validation repeats K
Secondary, each subsample verifying is primary, and average K result carrys out the performance indicator as classification of assessment device.
Step 6: adjusting hyper parameter training optimization neural network model
By adjusting, the learning rate of neural network, batch size, there are also semi-tensor product core sizes, activation letter for the number of iterations
The hyper parameters such as number, carry out optimization neural network model, improve the performance and effect of e-learning.Learning rate is chosen as 0.0001-
0.01, batch size is chosen as 10-30, and the number of iterations is selected as 2-5 times, and semi-tensor product core selects 3*3, non-linear excitation letter
Number selects Relu function, and pond layer choosing is the average pond of 2*2.
Step 7: realizing Underwater Targets Recognition classification
After testing data is become by pretreatment, the semi-tensor product neural network model that can have been optimized by training is provided
Classification results realize Underwater Targets Recognition classification.
Claims (7)
1. a kind of Underwater targets recognition based on semi-tensor product neural network, it is characterised in that the following steps are included: passing through
Underwater sonar sensor receives underwater sound signal, includes the acoustical signal of submarine target and the marine environment background without submarine target
Noise；The time domain of acoustical signal and frequency domain information are presented in LOFAR map by Short Time Fourier Transform；With LOFAR map
Sample is constructed as input feature vector matrix by data sample semi-tensor product neural network；The underwater sound signal received is divided into instruction
Practice collection and verifying collection, input semi-tensor product neural network is trained and verifies；By choosing different hyper parameters, training set is used
Double of tensor product neural network carries out model training, and the test effect of contrast verification collection determines the high hyper parameter of test accuracy rate；
Finally by the semi-tensor product neural network after the acoustical signal input model training of current collected submarine target, differentiation knot is provided
Fruit.
2. the Underwater targets recognition according to claim 1 based on semi-tensor product neural network, it is characterised in that: institute
The underwater sound signal statedIn formula, h (t) indicates ocean channel impulse response, s (t)
Indicate Underwater Target Signal, n (t) is ambient sea noise, and * indicates convolution algorithm, and t is time variable.
3. the Underwater targets recognition according to claim 1 based on semi-tensor product neural network, it is characterised in that: institute
After the underwater sonar sensor stated receives underwater sound signal, design high-pass filter filters out low frequency range noise, is then divided
Frame pretreatment.
4. the Underwater targets recognition according to claim 3 based on semi-tensor product neural network, it is characterised in that: institute
A length of 1s when each frame signal in the framing pretreatment stated.
5. the Underwater targets recognition according to claim 1 based on semi-tensor product neural network, it is characterised in that: institute
The Short Time Fourier Transform stated isX (t) representation signal in formula, w (t) are window letter
Number, plays the role of the time limit, e-jωtPlay frequency limit.
6. the Underwater targets recognition according to claim 1 based on semi-tensor product neural network, it is characterised in that: institute
In the semi-tensor product neural network stated,Y in formulaiIndicate the feature square after semi-tensor product operation
Battle array, xiIndicate the eigenmatrix of input, WiFormula weight matrix, biIndicate bias term, fc() indicates semi-tensor lamination activation primitive,Indicate semi-tensor product operation；For input feature vector after semi-tensor product carries out feature extraction, the characteristic pattern of output is passed to pond
Change layer and carry out feature selecting and information filtering, pond process is zi=βipool(yi)+bi, in formula, pool () is indicated yiIt carries out
Sampling operation, βiExpression multiplies biasing.
7. the Underwater targets recognition according to claim 1 based on semi-tensor product neural network, it is characterised in that:
The underwater sound signal is divided into K subsample, and an individual subsample is kept as the data of verifying model,
Other K-1 sample is used to substitute into building semi-tensor product neural network and is trained；Cross validation repeats K times, each subsample
Verifying is primary, and average K result carrys out the performance indicator as classification of assessment device.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910513322.9A CN110245608B (en) | 2019-06-14 | 2019-06-14 | Underwater target identification method based on half tensor product neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910513322.9A CN110245608B (en) | 2019-06-14 | 2019-06-14 | Underwater target identification method based on half tensor product neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110245608A true CN110245608A (en) | 2019-09-17 |
| CN110245608B CN110245608B (en) | 2022-05-17 |
Family
ID=67887015
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910513322.9A Active CN110245608B (en) | 2019-06-14 | 2019-06-14 | Underwater target identification method based on half tensor product neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110245608B (en) |
Cited By (19)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110807365A (en) * | 2019-09-29 | 2020-02-18 | 浙江大学 | Underwater target identification method based on fusion of GRU and one-dimensional CNN neural network |
| CN110837085A (en) * | 2019-11-14 | 2020-02-25 | 东南大学 | Fluctuation index calculation method for underwater target discrimination |
| CN111401548A (en) * | 2020-03-03 | 2020-07-10 | 西北工业大学 | L off line spectrum detection method based on deep learning |
| CN111505650A (en) * | 2020-04-28 | 2020-08-07 | 西北工业大学 | HPSS-based underwater target passive detection method |
| CN111523469A (en) * | 2020-04-23 | 2020-08-11 | 苏州浪潮智能科技有限公司 | Pedestrian re-identification method, system, equipment and computer readable storage medium |
| CN111624585A (en) * | 2020-05-21 | 2020-09-04 | 西北工业大学 | Underwater target passive detection method based on convolutional neural network |
| CN111931412A (en) * | 2020-06-19 | 2020-11-13 | 中国船舶重工集团公司第七一五研究所 | Underwater target noise LOFAR spectrogram simulation method based on generative countermeasure network |
| CN112163461A (en) * | 2020-09-07 | 2021-01-01 | 中国海洋大学 | Underwater target identification method based on multi-mode fusion |
| CN112364779A (en) * | 2020-11-12 | 2021-02-12 | 中国电子科技集团公司第五十四研究所 | Underwater sound target identification method based on signal processing and deep-shallow network multi-model fusion |
| CN112418181A (en) * | 2020-12-13 | 2021-02-26 | 西北工业大学 | Personnel overboard detection method based on convolutional neural network |
| CN112885362A (en) * | 2021-01-14 | 2021-06-01 | 珠海市岭南大数据研究院 | Target identification method, system, device and medium based on radiation noise |
| CN113109780A (en) * | 2021-03-02 | 2021-07-13 | 西安电子科技大学 | High-resolution range profile target identification method based on complex number dense connection neural network |
| CN113189595A (en) * | 2021-05-07 | 2021-07-30 | 山东大学 | Neural network-based two-way echo target positioning method, equipment and storage medium |
| CN113420668A (en) * | 2021-06-21 | 2021-09-21 | 西北工业大学 | Underwater target identification method based on two-dimensional multi-scale arrangement entropy |
| CN113420870A (en) * | 2021-07-04 | 2021-09-21 | 西北工业大学 | U-Net structure generation countermeasure network and method for underwater acoustic target recognition |
| CN113466839A (en) * | 2021-09-03 | 2021-10-01 | 北京星天科技有限公司 | Side-scan sonar sea bottom line detection method and device |
| CN115047408A (en) * | 2022-06-13 | 2022-09-13 | 天津大学 | Underwater multi-sound-source positioning method based on single-layer large convolution kernel neural network |
| CN116647376A (en) * | 2023-05-25 | 2023-08-25 | 中国人民解放军军事科学院国防科技创新研究院 | Voiceprint information-based underwater acoustic network node identity authentication method |
| CN116973901A (en) * | 2023-09-14 | 2023-10-31 | 海底鹰深海科技股份有限公司 | Algorithm application of time-frequency analysis in sonar signal processing |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2006039658A (en) * | 2004-07-22 | 2006-02-09 | Hitachi Software Eng Co Ltd | Image classification learning processing system and image identification processing system |
| CN102662167A (en) * | 2012-04-11 | 2012-09-12 | 西北工业大学 | Feature extraction method of radiated noise signal of underwater target |
| CN106919921A (en) * | 2017-03-06 | 2017-07-04 | 重庆邮电大学 | With reference to sub-space learning and the gait recognition method and system of tensor neutral net |
| CN107194404A (en) * | 2017-04-13 | 2017-09-22 | 哈尔滨工程大学 | Submarine target feature extracting method based on convolutional neural networks |
| CN108846323A (en) * | 2018-05-28 | 2018-11-20 | 哈尔滨工程大学 | A kind of convolutional neural networks optimization method towards Underwater Targets Recognition |
| 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 |
| US20190033447A1 (en) * | 2017-01-06 | 2019-01-31 | Massachusetts Institute Of Technology | Systems and methods for detecting objects in underwater environments |
| CN109448707A (en) * | 2018-12-18 | 2019-03-08 | 北京嘉楠捷思信息技术有限公司 | Voice recognition method and device, equipment and medium |
| CN109767785A (en) * | 2019-03-06 | 2019-05-17 | 河北工业大学 | Ambient noise method for identifying and classifying based on convolutional neural networks |
-
2019
- 2019-06-14 CN CN201910513322.9A patent/CN110245608B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2006039658A (en) * | 2004-07-22 | 2006-02-09 | Hitachi Software Eng Co Ltd | Image classification learning processing system and image identification processing system |
| CN102662167A (en) * | 2012-04-11 | 2012-09-12 | 西北工业大学 | Feature extraction method of radiated noise signal of underwater target |
| US20190033447A1 (en) * | 2017-01-06 | 2019-01-31 | Massachusetts Institute Of Technology | Systems and methods for detecting objects in underwater environments |
| CN106919921A (en) * | 2017-03-06 | 2017-07-04 | 重庆邮电大学 | With reference to sub-space learning and the gait recognition method and system of tensor neutral net |
| CN107194404A (en) * | 2017-04-13 | 2017-09-22 | 哈尔滨工程大学 | Submarine target feature extracting method based on convolutional neural networks |
| CN108846323A (en) * | 2018-05-28 | 2018-11-20 | 哈尔滨工程大学 | A kind of convolutional neural networks optimization method towards Underwater Targets Recognition |
| 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 |
| CN109448707A (en) * | 2018-12-18 | 2019-03-08 | 北京嘉楠捷思信息技术有限公司 | Voice recognition method and device, equipment and medium |
| CN109767785A (en) * | 2019-03-06 | 2019-05-17 | 河北工业大学 | Ambient noise method for identifying and classifying based on convolutional neural networks |
Non-Patent Citations (5)
| Title |
|---|
| DAIZHAN CHENG 等: "On Semi-tensor Product of Matrices and Its Applications", 《ACTA MATHEMATICAE APPLICATAE SINICA》 * |
| QIUYUAN HUANG 等: "Tensor Product Generation Networks for Deep NLP Modeling", 《PROCEEDINGS OF NAACL-HLT 2018》 * |
| XIAOQUAN KE 等: "Underwater Acoustic Target Recognition Based on Supervised Feature-Separation Algorithm", 《SENSORS》 * |
| 郭戈 等: "基于声呐图像的水下目标检测、识别与跟踪研究综述", 《控制与决策》 * |
| 顾正浩: "面向水下目标识别的卷积神经网络优化方法研究", 《中国优秀博硕士学位论文全文数据库(硕士) 信息科技辑》 * |
Cited By (32)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110807365A (en) * | 2019-09-29 | 2020-02-18 | 浙江大学 | Underwater target identification method based on fusion of GRU and one-dimensional CNN neural network |
| CN110807365B (en) * | 2019-09-29 | 2022-02-11 | 浙江大学 | Underwater target identification method based on fusion of GRU and one-dimensional CNN neural network |
| CN110837085A (en) * | 2019-11-14 | 2020-02-25 | 东南大学 | Fluctuation index calculation method for underwater target discrimination |
| CN110837085B (en) * | 2019-11-14 | 2022-06-14 | 东南大学 | Fluctuation index calculation method for underwater target discrimination |
| CN111401548A (en) * | 2020-03-03 | 2020-07-10 | 西北工业大学 | L off line spectrum detection method based on deep learning |
| CN111401548B (en) * | 2020-03-03 | 2022-03-22 | 西北工业大学 | Lofar line spectrum detection method based on deep learning |
| CN111523469A (en) * | 2020-04-23 | 2020-08-11 | 苏州浪潮智能科技有限公司 | Pedestrian re-identification method, system, equipment and computer readable storage medium |
| CN111505650B (en) * | 2020-04-28 | 2022-11-01 | 西北工业大学 | HPSS-based underwater target passive detection method |
| CN111505650A (en) * | 2020-04-28 | 2020-08-07 | 西北工业大学 | HPSS-based underwater target passive detection method |
| CN111624585A (en) * | 2020-05-21 | 2020-09-04 | 西北工业大学 | Underwater target passive detection method based on convolutional neural network |
| CN111931412A (en) * | 2020-06-19 | 2020-11-13 | 中国船舶重工集团公司第七一五研究所 | Underwater target noise LOFAR spectrogram simulation method based on generative countermeasure network |
| CN112163461A (en) * | 2020-09-07 | 2021-01-01 | 中国海洋大学 | Underwater target identification method based on multi-mode fusion |
| CN112163461B (en) * | 2020-09-07 | 2022-07-05 | 中国海洋大学 | Underwater target identification method based on multi-mode fusion |
| CN112364779A (en) * | 2020-11-12 | 2021-02-12 | 中国电子科技集团公司第五十四研究所 | Underwater sound target identification method based on signal processing and deep-shallow network multi-model fusion |
| CN112364779B (en) * | 2020-11-12 | 2022-10-21 | 中国电子科技集团公司第五十四研究所 | Underwater sound target identification method based on signal processing and deep-shallow network multi-model fusion |
| CN112418181B (en) * | 2020-12-13 | 2023-05-02 | 西北工业大学 | Personnel falling water detection method based on convolutional neural network |
| CN112418181A (en) * | 2020-12-13 | 2021-02-26 | 西北工业大学 | Personnel overboard detection method based on convolutional neural network |
| CN112885362A (en) * | 2021-01-14 | 2021-06-01 | 珠海市岭南大数据研究院 | Target identification method, system, device and medium based on radiation noise |
| CN112885362B (en) * | 2021-01-14 | 2024-04-09 | 珠海市岭南大数据研究院 | Target identification method, system, device and medium based on radiation noise |
| CN113109780A (en) * | 2021-03-02 | 2021-07-13 | 西安电子科技大学 | High-resolution range profile target identification method based on complex number dense connection neural network |
| CN113109780B (en) * | 2021-03-02 | 2022-08-05 | 西安电子科技大学 | High-resolution range profile target identification method based on complex number dense connection neural network |
| CN113189595A (en) * | 2021-05-07 | 2021-07-30 | 山东大学 | Neural network-based two-way echo target positioning method, equipment and storage medium |
| CN113420668A (en) * | 2021-06-21 | 2021-09-21 | 西北工业大学 | Underwater target identification method based on two-dimensional multi-scale arrangement entropy |
| CN113420668B (en) * | 2021-06-21 | 2024-01-12 | 西北工业大学 | Underwater target identification method based on two-dimensional multi-scale permutation entropy |
| CN113420870A (en) * | 2021-07-04 | 2021-09-21 | 西北工业大学 | U-Net structure generation countermeasure network and method for underwater acoustic target recognition |
| CN113420870B (en) * | 2021-07-04 | 2023-12-22 | 西北工业大学 | U-Net structure generation countermeasure network and method for underwater sound target recognition |
| CN113466839A (en) * | 2021-09-03 | 2021-10-01 | 北京星天科技有限公司 | Side-scan sonar sea bottom line detection method and device |
| CN115047408A (en) * | 2022-06-13 | 2022-09-13 | 天津大学 | Underwater multi-sound-source positioning method based on single-layer large convolution kernel neural network |
| CN115047408B (en) * | 2022-06-13 | 2023-08-15 | 天津大学 | Underwater multi-sound-source positioning method based on single-layer large convolution kernel neural network |
| CN116647376A (en) * | 2023-05-25 | 2023-08-25 | 中国人民解放军军事科学院国防科技创新研究院 | Voiceprint information-based underwater acoustic network node identity authentication method |
| CN116647376B (en) * | 2023-05-25 | 2024-01-26 | 中国人民解放军军事科学院国防科技创新研究院 | Voiceprint information-based underwater acoustic network node identity authentication method |
| CN116973901A (en) * | 2023-09-14 | 2023-10-31 | 海底鹰深海科技股份有限公司 | Algorithm application of time-frequency analysis in sonar signal processing |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110245608B (en) | 2022-05-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110245608A (en) | A kind of Underwater targets recognition based on semi-tensor product neural network | |
| CN110827837B (en) | Whale activity audio classification method based on deep learning | |
| CN107609488B (en) | Ship noise identification and classification method based on deep convolutional network | |
| CN107703486B (en) | Sound source positioning method based on convolutional neural network CNN | |
| CN108630209B (en) | Marine organism identification method based on feature fusion and deep confidence network | |
| CN113191178B (en) | Underwater sound target identification method based on auditory perception feature deep learning | |
| CN108922513A (en) | Speech differentiation method, apparatus, computer equipment and storage medium | |
| CN110148408A (en) | A kind of Chinese speech recognition method based on depth residual error | |
| Wei et al. | A method of underwater acoustic signal classification based on deep neural network | |
| CN111599376A (en) | Sound event detection method based on cavity convolution cyclic neural network | |
| CN113111786B (en) | Underwater target identification method based on small sample training diagram convolutional network | |
| CN115830436A (en) | Marine organism intelligent detection method based on deep learning | |
| Shen et al. | Improved auditory inspired convolutional neural networks for ship type classification | |
| CN115114949A (en) | Intelligent ship target identification method and system based on underwater acoustic signals | |
| CN117974736B (en) | Underwater sensor output signal noise reduction method and system based on machine learning | |
| CN112735466B (en) | Audio detection method and device | |
| CN110415685A (en) | A kind of audio recognition method | |
| CN118351881A (en) | Fusion feature classification and identification method based on noise reduction underwater sound signals | |
| CN112052880A (en) | Underwater sound target identification method based on weight updating support vector machine | |
| Zhang et al. | Underwater acoustic source separation with deep Bi-LSTM networks | |
| CN117711442A (en) | Infant crying classification method based on CNN-GRU fusion model | |
| CN117310668A (en) | Underwater sound target identification method integrating attention mechanism and depth residual error shrinkage network | |
| CN117275491A (en) | Sound classification method based on audio conversion and time diagram neural network | |
| CN116417011A (en) | Underwater sound target identification method based on feature fusion and residual CNN | |
| Wang et al. | Underwater acoustic target recognition combining multi-scale features and attention mechanism |
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 |