CN111814578A - Method for extracting frequency of ultralow frequency Doppler signal - Google Patents
Method for extracting frequency of ultralow frequency Doppler signal Download PDFInfo
- Publication number
- CN111814578A CN111814578A CN202010540223.2A CN202010540223A CN111814578A CN 111814578 A CN111814578 A CN 111814578A CN 202010540223 A CN202010540223 A CN 202010540223A CN 111814578 A CN111814578 A CN 111814578A
- Authority
- CN
- China
- Prior art keywords
- frequency
- signal
- time
- doppler
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- 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
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/02—Systems using the reflection of electromagnetic waves other than radio waves
- G01S17/50—Systems of measurement based on relative movement of target
- G01S17/58—Velocity or trajectory determination systems; Sense-of-movement determination systems
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Electromagnetism (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Computer Networks & Wireless Communication (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a frequency extraction method of an ultra-low frequency Doppler signal, which comprises the steps of firstly determining required target parameters according to the requirements of users, then collecting and integrating relevant Doppler signal data of a specific target, and separating a time sequence and a signal value sequence of the collected data. The model selection and construction is to select and construct parameters of the neural network model according to the requirements and targets of the user, complete model training and evaluation, and evaluate the obtained model to obtain a reliable model. And finally, stopping the DC and analyzing the frequency domain, and specifically extracting the required Doppler frequency by utilizing a Fourier transform or time-frequency analysis method. The invention uses the neural network model to carry out accurate function generation on the ultra-low frequency (millihertz) Doppler signals of limited sampling, and generates new Doppler signal data which is easy to carry out frequency domain analysis in real time, thereby well solving the problem that the frequency of the ultra-low frequency signals of limited sampling is difficult to extract from the frequency domain.
Description
Technical Field
The invention belongs to the technical field of optical measurement, and particularly relates to a frequency extraction method of an ultralow frequency Doppler signal.
Background
In recent years, with the development of precision equipment manufacturing, there has been an increasing demand for a wide-range, rapid, and high-precision mechanical quantity measurement technique. The research of corresponding multi-scenario, high-precision speed measurement technology is becoming more and more important for the requirements of science, industry and public safety. For example, radar speed measurement in the traffic safety field of the public security industry, blind area monitoring and lane change assistance of an automatic driving automobile, high-precision height-fixing and anti-collision early warning of an unmanned aerial vehicle through a radar technology, measurement and control and maintenance of satellite attitude and speed in a satellite formation flying deep space detection task, and the like. At present, the mainstream high-precision speed measurement method is mainly a speed measurement technology based on the laser doppler effect. The laser doppler velocity measurement technology has made great progress in recent years of development and application because of its advantages of non-contact measurement, fast response speed, high spatial resolution, high measurement accuracy, large measurement range, and the like. After more than forty years of development, the single-frequency laser Doppler velocimeter has become mature, the variety is diversified, and the single-frequency laser Doppler velocimeter is widely applied to various fields. However, the system structure of the single-frequency laser doppler velocimeter is complex, sensitive to the measurement environment, poor in anti-interference performance, severe in signal direct current drift phenomenon, high in requirement for the bandwidth of the signal demodulation system, and only capable of measuring a low speed in the actual measurement environment, and the application of the single-frequency laser doppler velocimeter is limited due to the defects.
The laser Doppler velocimeter based on microwave modulation is superior to a single-frequency laser Doppler velocimeter, physical quantity information to be measured is converted into a frequency modulation signal by utilizing a carrier technology, so that the measurement range is improved, the defect that the single-frequency laser Doppler velocimeter is easily influenced by a measurement environment can be overcome, the anti-interference capability is high, and the signal-to-noise ratio is high. The microwave modulation technology has important scientific and economic value for speed measurement. For the traditional laser Doppler velocity measurement technology, when the speed of the object motion is 100m/s, the frequency range of the Doppler signal is as high as 107Magnitude, the general circuit is difficult to process; the dual-frequency laser system-based optical carrier microwave velocity measurement technology is to detect the motion velocity of an object by using a microwave signal modulated on an optical carrier, and when the frequency of an intensity modulation signal is 109A Doppler shift of 10 for a movement speed of 100m/s in Hz2Order of magnitude, readily available, sophisticated signal processing in the microwave frequency domainAnd (6) circuit acquisition. In recent years, speed measurement research using the dual-frequency laser technology has achieved many results, however, for practical low-speed measurement, extraction of the ultra-low frequency signal frequency requires a very long time of sampling to improve the frequency resolution, and thus is difficult to achieve through frequency domain analysis.
Disclosure of Invention
The invention aims to provide a frequency extraction method of an ultralow frequency Doppler signal, which utilizes an ultralow frequency signal with a limited sampling number to perform machine learning fitting, then generates a signal with a high sampling number, and performs frequency domain analysis to extract the Doppler frequency.
The technical scheme adopted by the invention for solving the technical problems is an ultralow frequency extraction method based on a neural network, the method fits a function suitable for frequency domain analysis based on the neural network, and then carries out subsequent processing, and the method comprises the following steps:
step 1: targeting
Determining required target parameters according to the requirements of users;
step 2: data acquisition and integration
Collecting and integrating the relevant Doppler signal data of a specific target;
and step 3: data cleansing
Separating a time sequence and a signal value sequence of the acquired data, namely separating a time sequence t and a signal value sequence s from the acquired ultralow frequency signal with limited sampling number;
and 4, step 4: selecting and constructing models
Selecting and constructing parameters of a neural network model according to the requirements and targets of users, wherein the constructed neural network model takes a time sequence t as the input of model training, u as the label value of the model training, Doppler signal data is a binary group (t, u) respectively representing the time sequence and the signal value sequence,
Signal＝(t,u),Signalidenoted as the ith pair of time and signal value in the doppler signal,
ΔSignalirepresenting predicted signal and true signal dataDeviation therebetween (i th time);
and 5: model training and assessment
Evaluating the obtained model to obtain a reliable model;
step 6: DC blocking and frequency domain analysis
Specifically extracting the required Doppler frequency by using a Fourier transform or time-frequency analysis method, performing DC blocking processing on the generated new signal value sequence to obtain a pure alternating current signal, performing Fourier transform on the pure alternating current signal to obtain a frequency domain signal, and finally accurately extracting the Doppler frequency by using a peak value extraction algorithm or a time-frequency analysis method.
The method for constructing the neural network model comprises the following steps:
step 4-1: setting a neural network model parameter, setting the number of input and output layers as a neuron, setting the number of hidden layers and the number of neurons in each layer, setting the iteration times Epoch and Performance parameter Performance, and utilizing the Delta SignaliWhether the termination condition of the threshold control model training is reached or not, then, carrying out supervised training on the model by taking the time sequence t as the input of the model training and u as the label value of the model training;
step 4-2: and generating a new time sequence t ', which is far longer than t, and sending the time sequence t ' into the generated neural network model to obtain a new signal value sequence u '.
Furthermore, the increment of the time sequence t 'generated each time is the same as the increment of the time sequence t of the acquired real signal, but the total length of the time sequence t' is far greater than the length of the time sequence t of the real signal; meanwhile, the method removes the Doppler signals by using the characteristic that strong direct current values exist in the generated Doppler signals and using a DC blocking algorithm and a time-frequency analysis method, thereby obtaining frequency domain signals easy for peak value extraction.
Furthermore, the method has a user adjusting function, and a user can determine the prediction resolution of the neural network model by adjusting the iteration number Epoch and the Performance parameter Performance, so that the frequency extraction precision is influenced.
Furthermore, the user can customize the number of hidden layers and the number of neurons in each layer according to actual requirements, and select the number of layers and the number of neurons with the shortest training time and the optimal precision.
Compared with the prior art, the invention has the following beneficial technical effects:
1. the invention uses the neural network model to carry out accurate function generation on the ultra-low frequency (millihertz) Doppler signals of limited sampling, and generates new Doppler signal data which is easy to carry out frequency domain analysis in real time, thereby well solving the problem that the frequency of the ultra-low frequency signals of limited sampling is difficult to extract from the frequency domain.
2. The model generated by the invention can be matched with any complex Doppler frequency shift signal, and the Doppler frequency at any moment can be well extracted by using a time-frequency analysis method.
3. The method has the advantages of simple model training process and low calculation cost, and can ensure credible frequency domain extraction results and prevent disorder.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Fig. 2 is a schematic diagram of an ultra-low frequency doppler signal and a neural network prediction signal thereof.
Fig. 3 shows the prediction signal after the dc blocking process.
Fig. 4 is a diagram of spectral analysis of a predicted signal.
Detailed Description
The invention is described in further detail below with reference to the drawings.
Fig. 1 is a flow chart of the method of the present invention, and the method for extracting frequency of ultra-low frequency doppler signal provided by the present invention comprises the following 6 steps:
the method comprises the following steps: targeting
Determining required target parameters according to the requirements of users;
step two: data acquisition and integration
Collecting and integrating the relevant Doppler signal data of a specific target;
step three: data cleansing
Separating the time sequence and the signal value sequence of the acquired data;
step four: selecting and constructing models
Selecting and constructing parameters of a neural network model according to the requirements and targets of a user;
step five: model training and assessment
Evaluating the obtained model to obtain a reliable model;
step six: DC blocking and frequency domain analysis
And specifically extracting the required Doppler frequency by utilizing a Fourier transform or time-frequency analysis method.
The steps are further developed:
step 1: determining a target, and determining required target parameters according to the requirements of a user;
step 2: data acquisition and integration, wherein relevant Doppler signal data of a specific target are acquired and integrated;
and step 3: data cleaning, namely separating a time sequence and a signal value sequence from the acquired data, namely separating a time sequence t and a signal value sequence s from the acquired ultralow frequency signal with limited sampling number;
and 4, step 4: selecting and constructing a model, selecting and constructing parameters of the neural network model according to the requirements and targets of users, wherein the constructed neural network model takes a time sequence t as the input of model training, u as the label value of the model training, Doppler signal data is a binary group (t, u) respectively representing the time sequence and the signal value sequence,
Signal＝(t,u),Signalidenoted as the ith pair of time and signal value in the doppler signal,
ΔSignalirepresents the deviation between the predicted signal and the true signal data (i-th time);
and 5: model training and evaluation, namely evaluating the obtained model to obtain a reliable model;
step 6: and (3) performing blocking and frequency domain analysis, specifically extracting the required Doppler frequency by using a Fourier transform or time frequency analysis method, performing blocking processing on the generated new signal value sequence to obtain a pure alternating current signal, performing Fourier transform on the pure alternating current signal to obtain a frequency domain signal, and finally accurately extracting the Doppler frequency by using a peak value extraction algorithm or a time frequency analysis method.
The construction of the neural network model comprises the following steps:
step 4-1: setting a neural network model parameter, setting the number of input and output layers as a neuron, setting the number of hidden layers and the number of neurons in each layer, setting the iteration times Epoch and Performance parameter Performance, and utilizing the Delta SignaliWhether the termination condition of the threshold control model training is reached or not, then, carrying out supervised training on the model by taking the time sequence t as the input of the model training and u as the label value of the model training;
step 4-2: and generating a new time sequence t ', which is far longer than t, and sending the time sequence t ' into the generated neural network model to obtain a new signal value sequence u '.
Each time the time series t' is generated is the same as the time series t increment of the real signal acquired.
The user can determine the prediction resolution of the neural network model by adjusting the iteration number Epoch and the Performance parameter Performance, so that the frequency extraction precision is influenced.
The user can customize the number of hidden layers and the number of neurons in each layer according to actual requirements.
The user can select the number of layers and the number with the shortest training time and the best precision.
In specific implementation, a neural network model can be constructed by utilizing a Python software third-party deep learning tool package, the time sequence t is used as the input of model training, and u is used as the label value of the model training.
The doppler signal data is a binary set (t, u) representing a sequence of time and a sequence of signal values, respectively,
Signal＝(t,u),Signalirecording as the ith pair of time and signal value in the Doppler signal;
ΔSignalirepresents the deviation between the predicted signal and the true signal data (i-th time);
when the neural network model is constructed, the number of hidden layers and the number of neurons in each layer are specifically set according to actual conditions.
And during the DC blocking and frequency domain analysis, carrying out DC blocking treatment on u 'to obtain a pure AC signal u ", and then carrying out Fourier transform on the pure AC signal u' to obtain a frequency domain signal f. And finally, accurately extracting the Doppler frequency by using a peak extraction algorithm or a time-frequency analysis method.
To facilitate the realization of the idea of the invention by a person skilled in the art, a specific embodiment is now provided, as follows:
1. and acquiring an ultralow frequency Doppler signal which is measured when the frequency difference of the dual-frequency laser is about 1GHz and the standard guide rail moves at the speed of 5mm/s, namely the true value in the figure 2.
2. Setting neural network model parameters, wherein input and output layers are all 1 neuron, the number of hidden layers is 2, the number of neurons in each layer is 10, the iteration number Epoch is 1000, and the Performance parameter Performance is 10-7. The model prediction generated ultra low frequency doppler signals up to 1000s (adjustable) with a frequency resolution of 0.001Hz (adjustable), i.e. the neural network fit values in fig. 2.
3. The prediction signal is subjected to a differentiation process to obtain a differentiated signal, as shown in fig. 3.
4. The differential signal is fourier transformed to obtain a clear signal frequency domain distribution diagram, and as shown in fig. 4, a frequency value of 0.036Hz, corresponding to a movement speed of 5mm/s, can be obtained by using peak value extraction.
The above description is only a preferred embodiment of the present invention, but the scope of the present invention is not limited thereto. Changes and substitutions that can be easily made within the technical scope of the invention disclosed should be covered by the technical scope of the invention disclosed. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (6)
1. The method for extracting the frequency of the ultra-low frequency Doppler signal is characterized by comprising the following steps of:
step 1: targeting
Determining required target parameters according to the requirements of users;
step 2: data acquisition and integration
Collecting and integrating the relevant Doppler signal data of a specific target;
and step 3: data cleansing
Separating a time sequence and a signal value sequence of the acquired data, namely separating a time sequence t and a signal value sequence s from the acquired ultralow frequency signal with limited sampling number;
and 4, step 4: selecting and constructing models
Selecting and constructing parameters of a neural network model according to the requirements and targets of users, wherein the constructed neural network model takes a time sequence t as the input of model training, u as the label value of the model training, Doppler signal data is a binary group (t, u) respectively representing the time sequence and the signal value sequence,
Signal＝(t,u),Signalidenoted as the ith pair of time and signal value in the doppler signal,
ΔSignalirepresents the deviation between the predicted signal and the true signal data (i-th time);
and 5: model training and assessment
Evaluating the obtained model to obtain a reliable model;
step 6: DC blocking and frequency domain analysis
Specifically extracting the required Doppler frequency by using a Fourier transform or time-frequency analysis method, performing DC blocking processing on the generated new signal value sequence to obtain a pure alternating current signal, performing Fourier transform on the pure alternating current signal to obtain a frequency domain signal, and finally accurately extracting the Doppler frequency by using a peak value extraction algorithm or a time-frequency analysis method.
2. The method for extracting ultra low frequency doppler signal frequency according to claim 1, wherein the step 4 of constructing the neural network model comprises the steps of:
step 4-1: setting the neural network model parameters, setting the input and output layers as a neuron, setting the number of hidden layers and the neuron number of each layer,setting iteration times Epoch and Performance parameter Performance, and utilizing delta SignaliWhether the termination condition of the threshold control model training is reached or not, then, carrying out supervised training on the model by taking the time sequence t as the input of the model training and u as the label value of the model training;
step 4-2: and generating a new time sequence t ', which is far longer than t, and sending the time sequence t ' into the generated neural network model to obtain a new signal value sequence u '.
3. The method of claim 2, wherein the time series t' generated at each time is the same as the time series t increment of the real signal acquired.
4. The method according to claim 2, wherein the user can determine the predicted resolution of the neural network model by adjusting the iteration number Epoch and the Performance parameter Performance, so as to affect the frequency extraction accuracy.
5. The method of claim 2, wherein the number of hidden layers and the number of neurons in each layer can be customized according to actual requirements.
6. The method of claim 2, wherein the user can select the number of layers and the number of layers with the shortest training time and the best accuracy.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010540223.2A CN111814578B (en) | 2020-06-15 | 2020-06-15 | Method for extracting frequency of ultralow frequency Doppler signal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010540223.2A CN111814578B (en) | 2020-06-15 | 2020-06-15 | Method for extracting frequency of ultralow frequency Doppler signal |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111814578A true CN111814578A (en) | 2020-10-23 |
| CN111814578B CN111814578B (en) | 2021-03-05 |
Family
ID=72846104
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010540223.2A Active CN111814578B (en) | 2020-06-15 | 2020-06-15 | Method for extracting frequency of ultralow frequency Doppler signal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111814578B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113218520A (en) * | 2021-04-30 | 2021-08-06 | 南京森林警察学院 | Optimized neural network extraction method for laser pulse width |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6581046B1 (en) * | 1997-10-10 | 2003-06-17 | Yeda Research And Development Co. Ltd. | Neuronal phase-locked loops |
| US20130041859A1 (en) * | 2011-08-11 | 2013-02-14 | Greenray Industries, Inc. | Neural network frequency control |
| CN104568444A (en) * | 2015-01-28 | 2015-04-29 | 北京邮电大学 | Method for extracting fault characteristic frequencies of train rolling bearings with variable rotational speeds |
| CN106054159A (en) * | 2016-05-12 | 2016-10-26 | 北京航空航天大学 | Instantaneous frequency extraction method of Doppler signals |
| CN106067004A (en) * | 2016-05-30 | 2016-11-02 | 西安电子科技大学 | The recognition methods of digital modulation signals under a kind of impulsive noise |
| CN106872171A (en) * | 2017-04-10 | 2017-06-20 | 中国科学技术大学 | A kind of adaptive learning bearing calibration of Doppler's acoustic signal |
| CN108663576A (en) * | 2018-05-08 | 2018-10-16 | 集美大学 | Weak electromagnetic red signal detection method under a kind of complex environment |
| CN109188470A (en) * | 2018-09-11 | 2019-01-11 | 西安交通大学 | A kind of GNSS cheating interference detection method based on convolutional neural networks |
| CN111220958A (en) * | 2019-12-10 | 2020-06-02 | 西安宁远电子电工技术有限公司 | Radar target Doppler image classification and identification method based on one-dimensional convolutional neural network |
-
2020
- 2020-06-15 CN CN202010540223.2A patent/CN111814578B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6581046B1 (en) * | 1997-10-10 | 2003-06-17 | Yeda Research And Development Co. Ltd. | Neuronal phase-locked loops |
| US20130041859A1 (en) * | 2011-08-11 | 2013-02-14 | Greenray Industries, Inc. | Neural network frequency control |
| CN104568444A (en) * | 2015-01-28 | 2015-04-29 | 北京邮电大学 | Method for extracting fault characteristic frequencies of train rolling bearings with variable rotational speeds |
| CN106054159A (en) * | 2016-05-12 | 2016-10-26 | 北京航空航天大学 | Instantaneous frequency extraction method of Doppler signals |
| CN106067004A (en) * | 2016-05-30 | 2016-11-02 | 西安电子科技大学 | The recognition methods of digital modulation signals under a kind of impulsive noise |
| CN106872171A (en) * | 2017-04-10 | 2017-06-20 | 中国科学技术大学 | A kind of adaptive learning bearing calibration of Doppler's acoustic signal |
| CN108663576A (en) * | 2018-05-08 | 2018-10-16 | 集美大学 | Weak electromagnetic red signal detection method under a kind of complex environment |
| CN109188470A (en) * | 2018-09-11 | 2019-01-11 | 西安交通大学 | A kind of GNSS cheating interference detection method based on convolutional neural networks |
| CN111220958A (en) * | 2019-12-10 | 2020-06-02 | 西安宁远电子电工技术有限公司 | Radar target Doppler image classification and identification method based on one-dimensional convolutional neural network |
Non-Patent Citations (2)
| Title |
|---|
| 刘留 等: "基于神经网络的时变无线信道仿真", 《北京交通大学学报》 * |
| 王燕 等: "高精度多普勒信号模拟方法研究", 《舰船科学技术》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113218520A (en) * | 2021-04-30 | 2021-08-06 | 南京森林警察学院 | Optimized neural network extraction method for laser pulse width |
| CN113218520B (en) * | 2021-04-30 | 2021-11-09 | 南京森林警察学院 | Optimized neural network extraction method for laser pulse width |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111814578B (en) | 2021-03-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111461038B (en) | Pedestrian re-identification method based on layered multi-mode attention mechanism | |
| CN106338406A (en) | On-line monitoring and fault early-warning system and method for traction electric transmission system of train | |
| CN111649817B (en) | Distributed optical fiber vibration sensor system and mode identification method thereof | |
| CN109994203A (en) | A kind of epilepsy detection method based on EEG signal depth multi-angle of view feature learning | |
| KR101628154B1 (en) | Multiple target tracking method using received signal strengths | |
| CN108197743A (en) | A kind of prediction model flexible measurement method based on deep learning | |
| CN105389917A (en) | Rapid early-warning method based on phase-sensitive optical time-domain reflectometer | |
| CN113391282B (en) | Human body posture recognition method based on radar multi-dimensional feature fusion | |
| CN103558519A (en) | GIS partial discharge ultrasonic signal identification method | |
| CN114818916B (en) | Road target classification method based on millimeter wave radar multi-frame point cloud sequence | |
| Chen et al. | Attention-based dual-stream vision transformer for radar gait recognition | |
| CN101464172A (en) | Soft measuring method for power boiler breeze concentration mass flow | |
| CN118130984B (en) | Cable partial discharge fault real-time monitoring method based on data driving | |
| CN112462355A (en) | Sea target intelligent detection method based on time-frequency three-feature extraction | |
| CN111814578B (en) | Method for extracting frequency of ultralow frequency Doppler signal | |
| CN109357747A (en) | A kind of identification of online train and speed estimation method based on fiber-optic vibration signal | |
| CN111122162A (en) | Industrial system fault detection method based on Euclidean distance multi-scale fuzzy sample entropy | |
| Li et al. | Quickly build a high-precision classifier for Φ-OTDR sensing system based on transfer learning and support vector machine | |
| CN112926767A (en) | Annular fog flow gas phase apparent flow velocity prediction method based on particle swarm BP neural network | |
| CN112068120A (en) | micro-Doppler time-frequency plane individual soldier squad identification method based on two-dimensional Fourier transform | |
| Shen et al. | SSCT-Net: A semisupervised circular teacher network for defect detection with limited labeled multiview MFL samples | |
| CN113436442B (en) | Vehicle speed estimation method using multiple geomagnetic sensors | |
| CN110210326A (en) | A kind of identification of online train and speed estimation method based on fiber-optic vibration signal | |
| CN116466408B (en) | Artificial neural network superbedrock identification method based on aeromagnetic data | |
| CN115031794B (en) | Novel gas-solid two-phase flow measuring method based on multi-feature graph convolution |
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 |