CN111895986A - MEMS gyroscope original output signal noise reduction method based on LSTM neural network - Google Patents
MEMS gyroscope original output signal noise reduction method based on LSTM neural network Download PDFInfo
- Publication number
- CN111895986A CN111895986A CN202010622709.0A CN202010622709A CN111895986A CN 111895986 A CN111895986 A CN 111895986A CN 202010622709 A CN202010622709 A CN 202010622709A CN 111895986 A CN111895986 A CN 111895986A
- Authority
- CN
- China
- Prior art keywords
- data
- neural network
- noise reduction
- model
- gyroscope
- 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.)
- Pending
Links
- 230000009467 reduction Effects 0.000 title claims abstract description 42
- 238000000034 method Methods 0.000 title claims abstract description 26
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 24
- 238000003062 neural network model Methods 0.000 claims abstract description 21
- 238000012549 training Methods 0.000 claims abstract description 21
- 238000012360 testing method Methods 0.000 claims abstract description 13
- 238000011156 evaluation Methods 0.000 claims abstract description 9
- 238000007781 pre-processing Methods 0.000 claims abstract description 5
- 230000006870 function Effects 0.000 claims description 19
- 238000012545 processing Methods 0.000 claims description 15
- 230000015654 memory Effects 0.000 claims description 13
- 238000010606 normalization Methods 0.000 claims description 12
- 230000005540 biological transmission Effects 0.000 claims description 9
- 238000004891 communication Methods 0.000 claims description 9
- 239000011159 matrix material Substances 0.000 claims description 9
- 238000012886 linear function Methods 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 4
- 238000006243 chemical reaction Methods 0.000 claims description 4
- 230000002441 reversible effect Effects 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 3
- 230000006386 memory function Effects 0.000 abstract description 2
- 230000008859 change Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 230000005284 excitation Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 230000003068 static effect Effects 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 238000012300 Sequence Analysis Methods 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 230000007787 long-term memory Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000010998 test method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01C—MEASURING DISTANCES, LEVELS OR BEARINGS; SURVEYING; NAVIGATION; GYROSCOPIC INSTRUMENTS; PHOTOGRAMMETRY OR VIDEOGRAMMETRY
- G01C19/00—Gyroscopes; Turn-sensitive devices using vibrating masses; Turn-sensitive devices without moving masses; Measuring angular rate using gyroscopic effects
- G01C19/56—Turn-sensitive devices using vibrating masses, e.g. vibratory angular rate sensors based on Coriolis forces
- G01C19/5776—Signal processing not specific to any of the devices covered by groups G01C19/5607 - G01C19/5719
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Signal Processing (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Gyroscopes (AREA)
Abstract
The invention discloses a method for reducing noise of an original output signal of an MEMS gyroscope based on an LSTM neural network, which comprises the following steps of; 1) acquiring original output data of the MEMS gyroscope; 2) preprocessing data; 3) constructing a noise reduction model, and constructing an LSTM neural network model based on a Keras framework; 4) noise reduction treatment, namely training a neural network model by using training set data, storing the model, and performing prediction treatment by using test set data as input of the trained model; 5) and evaluating the noise reduction result by evaluating the mean square error of the noise reduction evaluation index. The method adopts the memory function characteristic of the LSTM neural network model, is easier to obtain the potential rule of the data sequence output by the gyroscope, and has obvious noise reduction effect on the original output signal of the MEMS gyroscope.
Description
Technical Field
The invention belongs to the field of digital signal processing, relates to noise reduction of an output signal of an MEMS (micro-electromechanical system) gyroscope, and particularly relates to a noise reduction method of an original output signal of the MEMS gyroscope based on an LSTM (least Square-spline) neural network.
Background
At present, the method for reducing the noise of the output data of the MEMS gyroscope comprises hardware design and software programming, wherein the software design mainly establishes a model, and corrects errors through a proper filter. Common modeling methods are:
1. the sliding autoregressive modeling method is used for building a sliding autoregressive model based on time sequence analysis, but random drift errors cannot be accurately described if the order of the sliding autoregressive model is too high.
2. In the wavelet analysis-based modeling method, wavelet transformation energy conversion is performed with different resolution analysis according to signal frequency, but accurate modeling cannot be performed if wavelet basis functions are unreasonably selected.
3. According to the modeling method of the MEMS gyroscope data manual, the modeling accuracy is high but the modeling method is not suitable for all MEMS gyroscopes.
4. The Kalman filter becomes a common MEMS gyroscope noise reduction algorithm due to high resolving speed and good real-time performance, but the dispersion of the filter can be caused by the mismatching of the model, and the adaptability to external input is poor.
The random drift error of the MEMS gyroscope can change along with the change of the external environment, and an accurate error model is difficult to establish without a definite rule. Therefore, the identification and noise reduction method for the random error of the output data of the MEMS gyroscope is one of the subjects of attention of those skilled in the art.
Disclosure of Invention
In view of the above-mentioned drawbacks or shortcomings of the prior art, an object of the present invention is to provide a method for reducing noise of an original output signal of a MEMS gyroscope based on an LSTM neural network.
In order to realize the task, the invention adopts the following technical solution:
a method for reducing noise of an original output signal of a MEMS gyroscope based on an LSTM neural network is characterized by comprising the following steps:
step 1), acquiring original output data of the MEMS gyroscope
Firstly, an MEMS gyroscope data acquisition system mainly comprising an MEMS gyroscope module, upper computer software, a computer and a power supply is built, the system uploads acquired data information to the upper computer in a serial port communication mode, the upper computer software acquires MEMS gyroscope data mainly through an RS-422 serial port communication protocol, and balanced transmission adopts unidirectional/non-reversible transmission and a transmission line with an enabling end or without the enabling end; the MEMS gyroscope module is connected with a USB interface of a computer through an RS-422 serial port communication protocol-USB interface, and upper computer software acquires original output data of the MEMS gyroscope through an operating system call interface of the control computer;
step 2), data preprocessing
Reading the gyroscope output sample data obtained in the step 1), performing sample data format conversion, performing normalization processing on the gyroscope data by adopting a maximum and minimum normalization method, mapping the sample data to a range between [0 and 1], and further performing normalization on the sample data set according to a ratio of 7: 3, dividing training set data and test set data according to the proportion;
step 3): establishing a noise reduction model
Constructing an LSTM neural network model based on a Keras framework, wherein the LSTM neural network model comprises 1 layer of LSTM and 3 layers of Dense, and setting the structure and model parameters of the LSTM neural network model;
step 4): noise reduction processing
Carrying out model training on training set data through an LSTM neural network model, carrying out parameter adjustment through a loss function until the model is converged, and storing the model; performing noise reduction treatment by taking the test set data as the input of the model;
step 5): noise reduction result evaluation
And evaluating through the mean square error of the noise reduction evaluation index and the noise reduction result comparison graph.
According to the invention, the process of outputting the sample data to the gyroscope in the step 2) and carrying out the format conversion of the sample data is as follows; starting from the statistical view, processing the random error of the gyroscope into a group of random variables of an ordered time sequence, and converting the gyroscope data noise reduction problem into time sequence data prediction.
Further, the model parameter formula in step 3) includes:
forget door ft=σ(Wf[xt,ht-1]+bf) Wherein f ist∈[0,1]Selection weight of node pair representing time t to cell memory at time t-1, bfBias term for forgetting gate, ht-1Implicit state information, non-linear function, representing a t-1 node
Input door it=σ(Wi[xt,ht-1]+bi) Wherein it∈[0,1]The selection weight of the node at time t to the current node information, biFor the bias term of the input gate, a non-linear function
outputting the memory information Ct=ft·Ct-1+it·CinIn which C ist-1Memory information representing a t-1 node;
output gate ot=σ(Wo[xt,ht-1]+bo) Wherein o ist∈[0,1]Selection weight of node cell memory information representing time t, boIs the bias term of the output gate;
hidden layer state h at time tt=ot·tanh(Ct);
xtRepresenting the input vector, h, of the LSTM neural network node at time ttHidden layer state information representing LSTM neural network nodes at the time t;tan h is a hyperbolic tangent activation function; wfWeight matrix for forgetting gate, WiIs a weight matrix of input gates, WoA weight matrix representing the information to be updated.
The method for reducing the noise of the MEMS gyroscope original output signal based on the LSTM neural network adopts the memory function characteristic of the LSTM neural network model, is easier to obtain the potential rule of the gyroscope output data sequence, and has obvious noise reduction effect on the MEMS gyroscope original output signal.
Drawings
FIG. 1 is a flow chart of the method for reducing noise of the MEMS gyroscope raw output signal based on the LSTM neural network.
FIG. 2 is a block diagram of a MEMS gyroscope data acquisition flow;
FIG. 3 is a comparison graph of noise reduction results, wherein (a) is a comparison graph of noise reduction results of training set data, and (b) is a comparison graph of noise reduction results of test set data.
The present invention will be described in further detail with reference to the following drawings and examples.
Detailed Description
Referring to fig. 1, the embodiment provides a method for reducing noise of an original output signal of a MEMS gyroscope based on an LSTM neural network, which includes the following specific detailed implementation steps:
step 1) obtaining original output data of MEMS gyroscope
Preparing a test environment: in the embodiment, a zero-position signal of the MEMS, namely an output signal when the sensing angular velocity is in a zero state, is selected, and the selected test method is a static test. Meanwhile, in order to improve the reliability of data and reduce the influence of environment and temperature change on output data, the MEMS gyroscope module is kept to be tested at a standard room temperature of 25 ℃ (the MEMS gyroscope module is placed in a temperature control turntable with the temperature set to be 25 ℃), and data collection is started after the gyroscope is kept in a static state for 30 min.
Building a data acquisition system: the system mainly comprises an MEMS gyroscope module, a power supply, upper computer software, a computer and the like; the acquired data information is uploaded to an upper computer in a serial port communication mode, the upper computer software acquires MEMS gyroscope data mainly through an RS-422 serial port communication protocol, and balanced transmission adopts a transmission line which is unidirectional/non-reversible and has an enabling end or no enabling end. The MEMS gyroscope module is connected with a USB interface of a computer through an RS-422 serial port communication protocol-USB interface, and the upper computer software acquires MEMS gyroscope data by controlling the operating system of the computer to call the interface.
Collecting data: as shown in fig. 2, a serial port is set, a baud rate is set to 921600, a storage path is selected, a power supply is powered on, gyroscope data transmission is started, a PC receives the data, then, the data is checked and resolved, the data is stored in a cache area of a computer, upper computer software calls a computer operating system to obtain the data received by the cache area, and the MEMS gyroscope data is output through a resolution protocol.
Step 2) data preprocessing
In this embodiment, the data preprocessing mainly includes preparing data to enter a neural network layer, including data format change and data normalization processing of a gyroscope.
Reading MEMS gyroscope sample data, transforming the format of the sample data, converting the gyroscope sample data into a time series data format, and expressing sampling points in the sequence sample as follows: x is the number ofi=(x1,…,xi,…,xn) The method is suitable for entering a neural network layer to perform time series prediction.
Carrying out normalization processing and dividing training set data and test set data; the purpose of normalization processing is to eliminate the influence of dimensions, avoid training a model due to the influence of a large number of dimensions in the neural network training process, perform normalization processing by adopting a maximum and minimum normalization method, and map sample data between [0 and 1 ]. Dividing training set data and test set data, and preparing data to enter a neural network layer;
maximum and minimum normalization method formula:
and after normalization processing, the sample data set is divided into 7 parts: and 3, dividing training set data and test set data according to the proportion to prepare for entering a neural network layer.
Step 3) establishing a noise reduction model, and constructing an LSTM neural network model based on a keras frame
An LSTM neural network model is constructed, an LSTM (Long ShortMemoryNetwork) long and short memory neural network is a special recurrent neural network, is proposed for improving the problem of gradient disappearance (influence factors of past time on the current time exist among different hidden layers, but the influence is weakened along with the increase of time span), is good at exploring the nonlinear relation among time series data, and is proved to be more effective for the time series prediction problem; the MEMS gyroscope is used as an actually existing physical system, state continuity exists in output at adjacent time points, after gyroscope data is processed into time sequence data, certain correlation exists between the output at the current moment and the output at the adjacent moment, and the time sequence can be considered to reflect the internal law of the system, so that an LSTM neural network model is applied to noise reduction of an original output signal of the MEMS gyroscope, and further output of the system is predicted by a fitting sequence modeling method, so that a better noise reduction effect on the output data of the MEMS gyroscope is achieved.
The LSTM neural network model processes long-term memory by introducing a 'gate structure' to enable the model to have a long-term selective memory mode, the special structural design enables the model to be more suitable for predicting or processing time-based series data, and the influence of previous information on current information is controlled by the following three 'gate' structures, namely:
(1) a forget gate that determines which cell states should be forgotten;
(2) an input gate that determines which new states should be added;
(3) an output gate for determining an output based on the current state and the current input;
the formula for the "door structure" model includes:
forget door ft=σ(Wf[xt,ht-1]+bf) Wherein f ist∈[0,1]Selection weight of node pair representing time t to cell memory at time t-1, bfBias term for forgetting gate, ht-1Implicit state information, non-linear function, representing a t-1 node
Input door it=σ(Wi[xt,ht-1]+bi) Wherein i ist∈[0,1]The selection weight of the node at time t to the current node information, biFor the bias term of the input gate, a non-linear function
outputting the memory information Ct=ft·Ct-1+it·CinWherein, Ct-1Memory information representing a t-1 node;
output gate ot=σ(Wo[xt,ht-1]+bo) Wherein o ist∈[0,1]Selection weight of node cell memory information representing time t, boIs the bias term of the output gate;
hidden layer state h at time tt=ot·tanh(Ct);
Wherein x istRepresenting the input vector, h, of the LSTM neural network node at time ttHidden layer state information representing LSTM neural network nodes at the time t;tan h is a hyperbolic tangent activation function; wfWeight matrix for forgetting gate, WiIs a weight matrix of input gates, WoA weight matrix representing the information to be updated.
Configuring a training model, wherein the training model comprises a 1-layer LSTM neural network layer and a 3-layer Dense layer, setting the structure and the hyper-parameters of the LSTM neural network model, and mainly comprises input data dimensionality, output data dimensionality, the number of nodes of each layer, an excitation function, a learning rate, a loss function, an optimization function, batch processing quantity, iteration times and the like.
The input data dimension of the LSTM layer is 1, the output data dimension is 1, the activation function of the input layer is relu, the number of nodes of each layer is 128, and the learning rate is 0.001. The excitation function is a tanh function and a sigmoid function, the loss function loss is mse (mean square error), the optimization function is an Adam function, the batch processing quantity is 1, and the iteration number is 100;
the tanh function is formulated as:
the sigmoid function is formulated as:
step 4) noise reduction treatment
The main idea of noise reduction is to use a long-short term memory network LSTM neural network model to predict a time sequence, the LSTM model has strong learning capacity, the motion characteristics of gyroscope output data are learned from a gyroscope training data sequence, a loss function loss is minimized through an optimizer, and parameter adjustment is carried out to enable the model to be converged.
The LSTM neural network model training process is that 70% of gyroscope sample data is used as training set data to be input into an input layer of the LSTM neural network model, obtained weight and bias of the input layer are input into an output layer of the LSTM neural network model, the result passes through a 3-layer Dense layer through a full-connection network, the output result is a noise reduction result of the training model, and the model is stored.
And further, inputting the residual 30% of gyroscope sample data serving as data of a test set into the trained long-short term memory network LSTM model, and predicting further output of the gyroscope system to realize the noise reduction of the gyroscope data.
Step 5) noise reduction result evaluation
Evaluating through a noise reduction evaluation index; the objective evaluation uses Mean Squared Error (MSE) as an evaluation index, and the formula is as follows:
in the formula (f)iIs the predicted value, yiThe value is an observed value.
In this embodiment, the mean square error MSE after 100 iterations is reduced from 0.2046 to 0.0173, which is 90% reduction in mean square error.
FIG. 3 shows a comparison graph of noise reduction results, where (a) is a comparison graph of noise reduction results of training set data, and (b) is a comparison graph of noise reduction results of test set data.
Claims (3)
1. A method for reducing noise of an original output signal of a MEMS gyroscope based on an LSTM neural network is characterized by comprising the following steps:
step 1), acquiring original output data of the MEMS gyroscope
Firstly, an MEMS gyroscope data acquisition system mainly comprising an MEMS gyroscope module, upper computer software, a computer and a power supply is built, the system uploads acquired data information to the upper computer in a serial port communication mode, the upper computer software acquires MEMS gyroscope data mainly through an RS-422 serial port communication protocol, and balanced transmission adopts unidirectional/non-reversible transmission and a transmission line with an enabling end or without the enabling end; the MEMS gyroscope module is connected with a USB interface of a computer through an RS-422 serial port communication protocol-USB interface, and upper computer software acquires original output data of the MEMS gyroscope through an operating system call interface of the control computer;
step 2), data preprocessing
Reading the gyroscope output sample data obtained in the step 1), carrying out sample data format conversion, carrying out normalization processing on the gyroscope data by adopting a maximum and minimum normalization method, mapping the sample data to a range between [0 and 1], and further dividing a sample data set into training set data and test set data according to a ratio of 7: 3;
step 3): establishing a noise reduction model
Constructing an LSTM neural network model based on a Keras framework, wherein the LSTM neural network model comprises 1 layer of LSTM and 3 layers of Dense, and setting a structure and a model parameter formula of the LSTM neural network model;
step 4): noise reduction processing
Carrying out model training on training set data through an LSTM neural network model, carrying out parameter adjustment through a loss function until the model is converged, and storing the model; performing noise reduction treatment by taking the test set data as the input of the model;
step 5): noise reduction result evaluation
And evaluating through the mean square error of the noise reduction evaluation index and the noise reduction result comparison graph.
2. The method of claim 1, wherein in step 2), the gyroscope outputs sample data, and the sample data format is transformed; starting from the statistical view, processing the random error of the gyroscope into a group of random variables of an ordered time sequence, and converting the gyroscope data noise reduction problem into time sequence data prediction.
3. The method of claim 1, wherein the model parameter formula in step 3) comprises:
forget door ft=σ(Wf[xt,ht-1]+bf) Wherein f ist∈[0,1]Selection weight of node pair representing time t to cell memory at time t-1, bfBias term for forgetting gate, ht-1Implicit state information, non-linear function, representing a t-1 node
Input door it=σ(Wi[xt,ht-1]+bi) Wherein it∈[0,1]The selection weight of the node at time t to the current node information, biFor the bias term of the input gate, a non-linear function
outputting the memory information Ct=ft·Ct-1+it·CinIn which C ist-1Memory information representing a t-1 node;
output gate ot=σ(Wo[xt,ht-1]+bo) Wherein o ist∈[0,1]Selection weight of node cell memory information representing time t, boIs the bias term of the output gate;
hidden layer state h at time tt=ot·tanh(Ct);
xtRepresenting the input vector, h, of the LSTM neural network node at time ttHidden layer state information representing LSTM neural network nodes at the time t;tan h is a hyperbolic tangent activation function; wfWeight matrix for forgetting gate, WiIs a weight matrix of input gates, WoA weight matrix representing the information to be updated.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010622709.0A CN111895986A (en) | 2020-06-30 | 2020-06-30 | MEMS gyroscope original output signal noise reduction method based on LSTM neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010622709.0A CN111895986A (en) | 2020-06-30 | 2020-06-30 | MEMS gyroscope original output signal noise reduction method based on LSTM neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN111895986A true CN111895986A (en) | 2020-11-06 |
Family
ID=73191575
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010622709.0A Pending CN111895986A (en) | 2020-06-30 | 2020-06-30 | MEMS gyroscope original output signal noise reduction method based on LSTM neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111895986A (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113447021A (en) * | 2021-07-15 | 2021-09-28 | 北京理工大学 | MEMS inertial navigation system positioning enhancement method based on LSTM neural network model |
CN113838471A (en) * | 2021-08-10 | 2021-12-24 | 北京塞宾科技有限公司 | Noise reduction method and system based on neural network, electronic device and storage medium |
CN118035652A (en) * | 2024-04-10 | 2024-05-14 | 东北石油大学三亚海洋油气研究院 | Processing method and device for measurement while drilling data and electronic equipment |
CN118468229A (en) * | 2024-07-10 | 2024-08-09 | 北京理工大学 | Gyro array fusion system and method based on improved LMD and LSTM |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107239859A (en) * | 2017-06-05 | 2017-10-10 | 国网山东省电力公司电力科学研究院 | The heating load forecasting method of Recognition with Recurrent Neural Network is remembered based on series connection shot and long term |
CN108364028A (en) * | 2018-03-06 | 2018-08-03 | 中国科学院信息工程研究所 | A kind of internet site automatic classification method based on deep learning |
CN109389059A (en) * | 2018-09-26 | 2019-02-26 | 华南理工大学 | A kind of P300 detection method based on CNN-LSTM network |
US20190135616A1 (en) * | 2017-11-03 | 2019-05-09 | University Of Dayton | Deep learning software enhanced microelectromechanical systems (mems) based inertial measurement unit (imu) |
CN110728357A (en) * | 2019-09-19 | 2020-01-24 | 东南大学 | IMU data denoising method based on recurrent neural network |
CN111238462A (en) * | 2020-01-19 | 2020-06-05 | 湖北三江航天红峰控制有限公司 | LSTM fiber-optic gyroscope temperature compensation modeling method based on deep embedded clustering |
-
2020
- 2020-06-30 CN CN202010622709.0A patent/CN111895986A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107239859A (en) * | 2017-06-05 | 2017-10-10 | 国网山东省电力公司电力科学研究院 | The heating load forecasting method of Recognition with Recurrent Neural Network is remembered based on series connection shot and long term |
US20190135616A1 (en) * | 2017-11-03 | 2019-05-09 | University Of Dayton | Deep learning software enhanced microelectromechanical systems (mems) based inertial measurement unit (imu) |
CN108364028A (en) * | 2018-03-06 | 2018-08-03 | 中国科学院信息工程研究所 | A kind of internet site automatic classification method based on deep learning |
CN109389059A (en) * | 2018-09-26 | 2019-02-26 | 华南理工大学 | A kind of P300 detection method based on CNN-LSTM network |
CN110728357A (en) * | 2019-09-19 | 2020-01-24 | 东南大学 | IMU data denoising method based on recurrent neural network |
CN111238462A (en) * | 2020-01-19 | 2020-06-05 | 湖北三江航天红峰控制有限公司 | LSTM fiber-optic gyroscope temperature compensation modeling method based on deep embedded clustering |
Non-Patent Citations (3)
Title |
---|
JIANG CHANGHUI,ET.AL: "A MEMS IMU De-Noising Method Using Long Short Term Memory Recurrent Neural Networks (LSTM-RNN)", 《SENSORS》 * |
XUGUANG CAO: "The Realization Methods of Fuzzy Sensor Linguistic Symbol Description" * |
马艺翔: "深度学习在多因子选股交易中的应用研究", 《中国优秀硕士学位论文全文数据库经济与管理科学辑》 * |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113447021A (en) * | 2021-07-15 | 2021-09-28 | 北京理工大学 | MEMS inertial navigation system positioning enhancement method based on LSTM neural network model |
CN113838471A (en) * | 2021-08-10 | 2021-12-24 | 北京塞宾科技有限公司 | Noise reduction method and system based on neural network, electronic device and storage medium |
CN118035652A (en) * | 2024-04-10 | 2024-05-14 | 东北石油大学三亚海洋油气研究院 | Processing method and device for measurement while drilling data and electronic equipment |
CN118468229A (en) * | 2024-07-10 | 2024-08-09 | 北京理工大学 | Gyro array fusion system and method based on improved LMD and LSTM |
CN118468229B (en) * | 2024-07-10 | 2024-09-10 | 北京理工大学 | Gyro array fusion system and method based on improved LMD and LSTM |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111895986A (en) | MEMS gyroscope original output signal noise reduction method based on LSTM neural network | |
CN107688871B (en) | Water quality prediction method and device | |
CN109816095B (en) | Network flow prediction method based on improved gated cyclic neural network | |
CN108900346B (en) | Wireless network flow prediction method based on LSTM network | |
Andrieu et al. | Particle methods for change detection, system identification, and control | |
CN110502806B (en) | Wireless spectrum occupancy rate prediction method based on LSTM network | |
CN107544904B (en) | Software reliability prediction method based on deep CG-LSTM neural network | |
CN110083125B (en) | Machine tool thermal error modeling method based on deep learning | |
CN115759461A (en) | Internet of things-oriented multivariate time sequence prediction method and system | |
CN110991776A (en) | Method and system for realizing water level prediction based on GRU network | |
CN112580848A (en) | PT-LSTM-based time series prediction system and method | |
Tang et al. | Reinforcement learning compensated extended Kalman filter for attitude estimation | |
CN116432699A (en) | Method for predicting silicon content of blast furnace molten iron based on time sequence chart packing network | |
CN115293197A (en) | Borehole strain data anomaly detection method based on long-term and short-term memory network | |
CN116364203A (en) | Water quality prediction method, system and device based on deep learning | |
CN115482877A (en) | Fermentation process soft measurement modeling method based on time sequence diagram network | |
Lejarza et al. | Discovering governing equations via moving horizon learning: The case of reacting systems | |
Boso et al. | Information geometry of physics-informed statistical manifolds and its use in data assimilation | |
CN112132096B (en) | Behavior modal identification method of random configuration network for dynamically updating output weight | |
CN114239796A (en) | Power system state estimation method based on extended Kalman filtering | |
Shen et al. | Variational sequential optimal experimental design using reinforcement learning | |
Xu et al. | Residual autoencoder-LSTM for city region vehicle emission pollution prediction | |
CN116777452A (en) | Prepayment system and method for intelligent ammeter | |
CN114995248A (en) | Intelligent maintenance and environmental parameter big data internet of things system | |
CN115936236A (en) | Method, system, equipment and medium for predicting energy consumption of cigarette factory |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20201106 |