CN109274624A - A kind of carrier frequency bias estimation based on convolutional neural networks - Google Patents
A kind of carrier frequency bias estimation based on convolutional neural networks Download PDFInfo
- Publication number
- CN109274624A CN109274624A CN201811320190.XA CN201811320190A CN109274624A CN 109274624 A CN109274624 A CN 109274624A CN 201811320190 A CN201811320190 A CN 201811320190A CN 109274624 A CN109274624 A CN 109274624A
- Authority
- CN
- China
- Prior art keywords
- signal
- convolutional neural
- neural networks
- frequency deviation
- sample
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L27/00—Modulated-carrier systems
- H04L27/0014—Carrier regulation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L27/00—Modulated-carrier systems
- H04L27/0014—Carrier regulation
- H04L2027/0024—Carrier regulation at the receiver end
- H04L2027/0026—Correction of carrier offset
Landscapes
- Engineering & Computer Science (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Digital Transmission Methods That Use Modulated Carrier Waves (AREA)
Abstract
The present invention relates to a kind of carrier frequency bias estimations based on convolutional neural networks, belong to signal processing technology field, and solving prior art frequency deviation can not accurately estimate and be difficult to adapt to all digital communication signal Modulation Types.A kind of carrier frequency bias estimation based on convolutional neural networks, comprising the following steps: obtain signal sample data;Convolutional neural networks are constructed based on the signal sample data;The convolutional neural networks of construction are trained, the convolutional neural networks model based on carrier wave frequency deviation is obtained;Signal carrier frequency deviation is estimated using the convolutional neural networks model based on carrier wave frequency deviation.The accurate estimation of carrier wave frequency deviation is realized, and is suitable for the signal of various different modulating patterns.
Description
Technical field
The present invention relates to signal processing technology field more particularly to a kind of Nonlinear Transformation in Frequency Offset Estimation based on convolutional neural networks
Method.
Background technique
In practical wireless communication systems, since sending and receiving end oscillator is unstable, precision is limited and Doppler frequency shift etc.
The influence of reason causes reception signal to generate certain frequency shift (FS), to influence receiver performance.Therefore, it is being concerned with
When demodulating, restoring to transmission unknown data, in order to effectively reduce influence of the frequency shift (FS) to system performance, needing to use has
The frequency excursion algorithm of effect accurately estimates frequency offseting value and compensates to it.The existing big spininess of carrier frequency bias estimation
Special signal type is designed, it is difficult to adapt to all digital communication signal Modulation Types.
Summary of the invention
In view of above-mentioned analysis, the present invention is intended to provide a kind of carrier frequency bias estimation based on convolutional neural networks,
The accurate estimation of carrier wave frequency deviation is realized, and is suitable for the signal of various different modulating patterns.
The present invention provides a kind of carrier frequency bias estimation based on convolutional neural networks, comprising the following steps:
Obtain signal sample data;
Convolutional neural networks are constructed based on the signal sample data;
The convolutional neural networks of construction are trained, the convolutional neural networks model based on carrier wave frequency deviation is obtained;
Signal carrier frequency deviation is estimated using the convolutional neural networks model based on carrier wave frequency deviation.
Above-mentioned technical proposal has the beneficial effect that on the one hand, signal sample data can be the signal of different modulating pattern
It is formed, therefore the neural network model of the convolutional neural networks based on signal sample data construction and training, is also applied for not
With Modulation Types signal, on the other hand, the sample by signal sample data as convolutional neural networks model, the convolution mind of formation
Through network model, the arrival precision of prediction of convolutional neural networks model can be made, to realize the accurate estimation of carrier wave frequency deviation.
Further, obtaining signal sample data includes: in the case where each presets frequency deviation, and emulation generates a complex radical and takes a message
Number, as a sample signal and digital quantity signal is converted to, the digital quantity signal is sampled, obtains complex base band sampling
Signal constitutes a matrix using the real and imaginary parts of the complex base band sampled signal as two row vectors of matrix;At M
Under default frequency deviation, sampling obtains M sample (xi,yi), as signal sample data;Wherein, xiIt is corresponding same for i-th of sample
The matrix that phase component and quadrature component are constituted, yiFor its corresponding frequency deviation value, M > 2.
The real and imaginary parts of complex base band sampled signal are used matrix by having the beneficial effect that for above-mentioned further technical solution
Form indicates, ensure that the globality of complex base band sampled signal under the default frequency deviation, by this matrix and its corresponding practical frequency deviation
Value combination constitutes a sample, convenient for convolutional neural networks training.
Further, the frequency deviation region of the complex baseband signal is [- fs/2,fs/ 2], fsFor signal sampling frequencies.
Above-mentioned further technical solution volume has the beneficial effect that the above method realizes the abundant sampling of complex baseband signal.
It further, is L to the sampling number of digital quantity signal sampling, L >=2, the sample matrix is the square that L row 2 arranges
Battle array.
Further, the convolutional neural networks, specifically include, an input layer, at least one convolutional layer, one it is non-thread
Property active coating, one batch of normalization layer, an output layer, the output layer is to return layer.
Further, construction convolutional neural networks detailed process includes, in input layer input sample matrix, passing through convolution
Layer, convolutional layer have multiple convolution kernels to carry out convolution, handle by batch normalization layer and nonlinear activation layer, then pass through dropout
It is handled, via recurrence layer outgoing carrier frequency deviation.
Further, convolutional neural networks are trained, obtain the convolutional neural networks model based on carrier wave frequency deviation, had
Body includes: that P are chosen from the M sample as training data, remaining is trained training data as test data
Initial model is obtained, whether expected precision is reached using test data verifying initial model, if so, the model is as final
Convolutional neural networks model training data is carried out if it is not, then carrying out tune ginseng to initial model using stochastic gradient descent method
Re -training, until model reaches expected precision, wherein 1 < P < M.
Above-mentioned having the beneficial effect that for further technical solution obtains initial model using training data, utilizes test data
Initial model is verified, and initial model is carried out continuing to adjust ginseng by stochastic gradient descent method, so that model accuracy is reached expected, mentions
The high training speed of convolutional neural networks.
Further, tune ginseng is carried out to initial model using stochastic gradient descent method, specifically included, each batch processing training
When, it calculates network error and carries out backpropagation in this, as error, according to the parameter value of the t-1 times iteration, objective function pair
The First-order Gradient and learning efficiency of parameter current update the parameter value of the t times iteration.
Further, signal carrier frequency deviation is estimated according to the convolutional Neural model based on carrier wave frequency deviation, including, it is right
Signal to be estimated is sampled, and complex base band sample sequence is obtained, and the real and imaginary parts for extracting the sample sequence constitute matrix, and
As the convolutional neural networks mode input based on carrier wave frequency deviation, the output valve of carrier wave frequency deviation is obtained.
Further, the real and imaginary parts for extracting the sample sequence constitute matrix, specifically include, if sample sequence z (n)
Sampling number be equal to L, extract the sample sequence z (n) real and imaginary parts constitute matrix;
If the sampling number of sample sequence z (n) is less than L, 0 is filled behind the sample sequence, is equal to its length
L constitutes new sample sequence, and the real and imaginary parts for extracting sample sequence constitute matrix;
If the sampling number of sample sequence z (n) is greater than L, sample sequence pi(m)=z ((i-1) L+m) extracts sampling
The real and imaginary parts of sequence constitute matrix, wherein m=0, and 1,2 ..., L-1, i=1,2 ..., I,Indicate little
In the maximum integer of N/L.
When having the beneficial effect that according to offset estimation of above-mentioned further technical solution, the sampling number of sample sequence with it is right
The relationship of sampling number, is adjusted sample sequence, so that enhancing in offset estimation when digital quantity signal samples
Input matrix also improves to the adaptability of convolutional neural networks model and trains obtained convolutional Neural model to signal carrier frequency
The versatility estimated partially.
Other features and advantages of the present invention will illustrate in the following description, also, partial become from specification
It obtains it is clear that understand through the implementation of the invention.The objectives and other advantages of the invention can be by written explanation
Specifically noted structure is achieved and obtained in book, claims and attached drawing.
Detailed description of the invention
Attached drawing is only used for showing the purpose of specific embodiment, and is not to be construed as limiting the invention, in entire attached drawing
In, identical reference symbol indicates identical component.
Fig. 1 is the flow diagram of the method for the embodiment of the present invention;
Fig. 2 is the convolutional neural networks exemplary diagram that the embodiment of the present invention provides.
Specific embodiment
Specifically describing the preferred embodiment of the present invention with reference to the accompanying drawing, wherein attached drawing constitutes the application a part, and
Together with embodiments of the present invention for illustrating the principle of the present invention.
A specific embodiment of the invention, discloses a kind of carrier frequency bias estimation based on convolutional neural networks.
As shown in Figure 1, comprising the following steps:
Step S1, signal sample data is obtained;
It specifically includes, in the case where each presets frequency deviation, emulation generates complex baseband signal, forms a sample, utilizes modulus
Sample signal is converted to digital quantity by analog quantity by converter, forms digital quantity signal, is then carried out to the digital quantity signal
Sampling, sampling number L obtains complex base band sampled signal, by the real part (in-phase component) and void of the complex base band sample signal
Two column vectors of portion's (quadrature component) as matrix are constituted the matrix of the column of L row 2 with this;Similarly, frequency deviation is preset at each
Under, the different complex baseband signal signal of another frequency deviation is generated, the matrix that L row 2 arranges is constituted;M sample (x is obtained with thisi,yi),
As signal sample data, wherein xiThe sample arranged for the L row 2 that the corresponding in-phase component of i-th of sample and quadrature component are constituted
Matrix, yiFor its corresponding frequency deviation value.Wherein, the frequency deviation region of above-mentioned complex baseband signal is [- fs/2,fs/ 2], fsIt is adopted for signal
Sample frequency.
It should be noted that the Modulation Types of complex baseband signal according to application need to select, may include AM, FM, PSK,
The various Modulation Types such as ASK, FSK, QAM;What since the signal that signal sample data can be different modulating pattern was formed, therefore
The neural network model of convolutional neural networks and training based on signal sample data construction, is also applied for different modulating pattern letter
Number.
Step S2, convolutional neural networks are constructed;
Specifically, convolutional neural networks input layer having a size of L row 2 arrange, centre comprising at least one convolutional layer, one it is non-thread
Property active coating, one batch of normalization layer, output layer is to return layer;The process of specific configuration convolutional neural networks is, in input layer
Input sample matrix, by convolutional layer, convolutional layer has multiple convolution kernels to carry out convolution, by batch normalization layer and nonlinear activation
Layer processing to prevent network model over-fitting, then passes through dropout and handles it, via recurrence layer outgoing carrier frequency deviation;
Fig. 2 is construction convolutional neural networks flow diagram, and L=1024 in figure, batch normalization indicate to criticize
Normalization, conv represent convolutional layer, the size of the digital representation convolution kernel before conv, of digital representation convolution kernel later
Number;Regression Layer is indicated to return layer, is criticized and normalize layer between convolutional layer and nonlinear activation layer, all activated
Function all uses ReLU (line rectification function), and for adjusting the output of convolutional layer and batch normalization layer, finally output is carrier wave
Frequency deviation;
Step S3, convolutional neural networks are trained, obtain the convolutional neural networks model based on carrier wave frequency deviation;
Initial model is obtained using training data, verifies initial model using test data, and pass through stochastic gradient descent
Method carries out continuing to adjust ginseng to initial model, and model accuracy is made to reach expected;
Specifically, from the M sample (xi,yi) as training data, remaining is used as tests middle selection P a (1 < P < M)
Data are trained the convolutional neural networks, the specific training process of convolutional neural networks are as follows: from the M sample
It chooses P and is used as training data, remaining is trained training data as test data, initial model is obtained, using test
Whether data verification initial model reaches expected precision, if so, the model is final convolutional neural networks model, if
It is no, then tune ginseng is carried out to initial model using stochastic gradient descent method, re -training is carried out to training data, until model reaches
It is expected that precision;
Wherein, stochastic gradient descent method is the rudimentary algorithm of neural metwork training, i.e., calculates net when each batch processing is trained
Network error (error i.e. between reality output and ideal output) and the backpropagation as error, after according to First-order Gradient information
Parameter is updated, more new strategy is represented by wt←wt-1- η g, specifically, according to the parameter value of the t-1 times iteration, mesh
Scalar functions update the parameter value of the t times iteration, wherein w to the First-order Gradient and learning efficiency of parameter currenttRepeatedly for the t times
The parameter value in generation, wt-1For the parameter value of the t-1 times iteration, η is learning efficiency, and g is a ladder of the objective function to parameter current
Degree;
Step S4, signal carrier frequency deviation is estimated according to the convolutional Neural model based on carrier wave frequency deviation;
It specifically includes, signal to be estimated is acquired, be converted to signal to be estimated by analog quantity by analog-digital converter
Digital quantity, sampling obtains signal sample data to be estimated, and thus obtains complex base band sample sequence z (n), n=0,1,
2 ..., N-1, N are its sampling number;Since the sampling number of complex base band sample sequence z (n) may be inconsistent with L, then
Point or less three kinds of situations, to signal to be estimated carry out offset estimation;
If signal length (the complex base band sample sequence sampling number of signal to be estimated) to be estimated is equal to the letter of model supports
Number input length (sampling number L) N=L, then directly extract z (n) real and imaginary parts, constitute 2 row L arrange matrix, as institute
The input of the convolutional neural networks model based on carrier wave frequency deviation is stated, output result is to estimate resulting frequency deviation value;
If the signal that signal length to be estimated is less than model supports inputs length, i.e. N < L, then by signal sampling to be identified
Sequence z (n) mends N-L 0, i.e. z (n)=0, n=N ..., L-1 below, extracts the real and imaginary parts of z (n), constitutes what L row 2 arranged
Matrix, as the input of the convolutional neural networks model based on carrier wave frequency deviation, exporting result is to estimate resulting frequency deviation
Value;
If the signal that signal length to be estimated is greater than model supports inputs length, i.e. N > L then extracts each signal segment pi
(m), wherein pi(m)=z ((i-1) L+m), m=0,1,2 ..., L-1, i=1,2 ..., I, whereinIt is indicated
Maximum integer no more than N/L;P is extracted respectivelyi(m) real and imaginary parts constitute the matrix that L row 2 arranges, as described based on load
The input of the inclined convolutional neural networks model of wave frequency, output result are oi, i=1,2 ..., I;Final frequency deviation value is oiBe averaged
Value.
In conclusion the present invention discloses a kind of carrier frequency bias estimation based on convolutional neural networks, pass through emulation
Form generates sample data, on the one hand, can be with the sample of various Modulation Types, on the other hand, also in order to obtaining sample data
Accurate frequency deviation, thus convenient to constructing the convolutional neural networks model with higher estimated accuracy;Meanwhile sample data uses
The matrix form that L row 2 arranges, ensure that the globality of complex base band sampled signal under default frequency deviation, by this matrix and its corresponding reality
Frequency deviation value combination in border constitutes a sample, convenient for convolutional neural networks training;When constructing convolutional neural networks, pass through dropout
It is handled, prevents convolutional neural networks over-fitting, and in training pattern, using stochastic gradient descent method, improve instruction
Practice precision of the model in actual estimated;In practical offset estimation, believed according to the sampling number of sample sequence with to digital quantity
Number sampling when sampling number relationship, sample sequence is adjusted, enhances input matrix to convolutional neural networks model
Adaptability also improves trained obtained convolutional Neural model to the versatility of signal carrier offset estimation.
The present invention is trained convolutional neural networks by the signal sample data under various frequency deviations, and recycling trains
Convolutional neural networks to receive signal carrier wave frequency deviation estimate, to realize the accurate estimation of carrier wave frequency deviation, and can fit
The signal for answering various different modulating patterns, has the characteristics that universality;And the accurate estimation of carrier wave frequency deviation, be conducive to reduce frequency
Deviate the influence to system receptivity.
The foregoing is only a preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto,
In the technical scope disclosed by the present invention, any changes or substitutions that can be easily thought of by anyone skilled in the art,
It should be covered by the protection scope of the present invention.
Claims (10)
1. a kind of carrier frequency bias estimation based on convolutional neural networks, which comprises the following steps:
Obtain signal sample data;
Convolutional neural networks are constructed based on the signal sample data;
The convolutional neural networks of construction are trained, the convolutional neural networks model based on carrier wave frequency deviation is obtained;
Signal carrier frequency deviation is estimated using the convolutional neural networks model based on carrier wave frequency deviation.
2. the method according to claim 1, wherein acquisition signal sample data includes:
Under each default frequency deviation, emulation generates a complex baseband signal, as a sample signal and is converted to number
Measure signal;
The digital quantity signal is sampled, complex base band sampled signal is obtained;
Using the real and imaginary parts of the complex base band sampled signal as two row vectors of matrix, a matrix is constituted;
Under M default frequency deviations, sampling obtains M sample (xi,yi), as signal sample data;Wherein, xiFor i-th of sample
The matrix that corresponding in-phase component and quadrature component are constituted, yiFor its corresponding frequency deviation value, M > 2.
3. according to the method described in claim 2, it is characterized in that, the frequency deviation region of the complex baseband signal is [- fs/2,fs/
2], fsFor signal sampling frequencies.
4. according to the method described in claim 2, it is characterized in that, being L, L to the sampling number of digital quantity signal sampling
>=2, the sample matrix is the matrix that L row 2 arranges.
5. one inputs the method according to claim 1, wherein the convolutional neural networks, specifically include
Layer, at least one convolutional layer, a nonlinear activation layer, one batch of normalization layer, an output layer, the output layer is to return
Layer.
6. according to the method described in claim 2, it is characterized in that, construction convolutional neural networks detailed process includes inputting
Layer input sample matrix, by convolutional layer, convolutional layer has multiple convolution kernels to carry out convolution, by batch normalization layer and non-linear swashs
Layer processing living, then it is handled by dropout, via recurrence layer outgoing carrier frequency deviation.
7. according to the method described in claim 2, obtaining based on carrier wave it is characterized in that, be trained to convolutional neural networks
The convolutional neural networks model of frequency deviation, specifically includes:
P are chosen from the M sample and is used as training data, remaining is trained training data as test data
To initial model, whether expected precision is reached using test data verifying initial model;
If so, the model is final convolutional neural networks model, if it is not, then using stochastic gradient descent method to introductory die
Type carries out tune ginseng, carries out re -training to training data, until model reaches expected precision, wherein 1 < P < M.
8. the method according to the description of claim 7 is characterized in that being adjusted using stochastic gradient descent method to initial model
Ginseng, specifically includes, and when each batch processing is trained, calculates network error and simultaneously carries out backpropagation in this, as error, according to t-1
First-order Gradient and learning efficiency of the parameter value, objective function of iteration to parameter current, the parameter value of update t iteration.
9. according to the method described in claim 4, it is characterized in that, according to the convolutional Neural model based on carrier wave frequency deviation to signal
Carrier wave frequency deviation estimated, including,
Signal to be estimated is sampled, complex base band sample sequence is obtained, the real and imaginary parts for extracting the sample sequence are constituted
Matrix, and as the convolutional neural networks mode input based on carrier wave frequency deviation, obtain the output valve of carrier wave frequency deviation.
10. according to the method described in claim 9, it is characterized in that, the real and imaginary parts for extracting the sample sequence constitute square
Battle array, specifically includes, if the sampling number of sample sequence z (n) is equal to L, extracts the real and imaginary parts structure of the sample sequence z (n)
At matrix;
If the sampling number of sample sequence z (n) is less than L, 0 is filled behind the sample sequence, its length is made to be equal to L, structure
The sample sequence of Cheng Xin, the real and imaginary parts for extracting sample sequence constitute matrix;
If the sampling number of sample sequence z (n) is greater than L, sample sequence pi(m)=z ((i-1) L+m) extracts sample sequence
Real and imaginary parts constitute matrix, wherein m=0, and 1,2 ..., L-1, i=1,2 ..., I,It indicates no more than N/L's
Maximum integer.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811320190.XA CN109274624B (en) | 2018-11-07 | 2018-11-07 | Carrier frequency offset estimation method based on convolutional neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811320190.XA CN109274624B (en) | 2018-11-07 | 2018-11-07 | Carrier frequency offset estimation method based on convolutional neural network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109274624A true CN109274624A (en) | 2019-01-25 |
CN109274624B CN109274624B (en) | 2021-04-27 |
Family
ID=65192276
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811320190.XA Active CN109274624B (en) | 2018-11-07 | 2018-11-07 | Carrier frequency offset estimation method based on convolutional neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109274624B (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110048980A (en) * | 2019-04-19 | 2019-07-23 | 中国电子科技集团公司第三十六研究所 | A kind of blind demodulation method of digital communication and device |
CN110061939A (en) * | 2019-04-19 | 2019-07-26 | 中国电子科技集团公司第三十六研究所 | A kind of symbol rate estimation method and apparatus |
CN110781755A (en) * | 2019-09-28 | 2020-02-11 | 深圳中物智建科技有限公司 | Intelligent identification and metering method and system for construction steel bars |
CN110995327A (en) * | 2019-12-17 | 2020-04-10 | 电子科技大学 | Mixed beam forming optimization method and system for multi-carrier MIMO system |
CN112187266A (en) * | 2020-09-29 | 2021-01-05 | 中国电子科技集团公司第三十六研究所 | Nonlinear correction method and device for analog-to-digital converter and electronic equipment |
CN112910810A (en) * | 2021-02-01 | 2021-06-04 | 中国人民解放军国防科技大学 | Synchronization method, apparatus and medium for MIMO OFDM system |
CN113609982A (en) * | 2021-08-05 | 2021-11-05 | 东南大学 | High-discrimination radio frequency fingerprint learning and extracting method based on intelligent carrier synchronization |
CN113726416A (en) * | 2021-09-01 | 2021-11-30 | 北京邮电大学 | Satellite communication carrier synchronization method and device and communication equipment |
CN114598886A (en) * | 2022-05-09 | 2022-06-07 | 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) | Image coding method, decoding method and related device |
CN115103434A (en) * | 2022-06-23 | 2022-09-23 | 东南大学 | Machine learning-based 5G NR downlink timing synchronization method |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1585289A (en) * | 2004-06-02 | 2005-02-23 | 中兴通讯股份有限公司 | Method for compensating frequency offset in wireless mobile communication system |
CN103188198A (en) * | 2013-03-25 | 2013-07-03 | 河海大学常州校区 | OFDM (Orthogonal Frequency Division Multiplexing) symbol timing and frequency offset estimation method based on particle swarm optimization (PSO) algorithm |
US9190053B2 (en) * | 2013-03-25 | 2015-11-17 | The Governing Council Of The Univeristy Of Toronto | System and method for applying a convolutional neural network to speech recognition |
CN106875002A (en) * | 2017-02-20 | 2017-06-20 | 中国石油大学(华东) | Complex value neural network training method based on gradient descent method Yu generalized inverse |
CN108509911A (en) * | 2018-04-03 | 2018-09-07 | 电子科技大学 | Interference signal recognition methods based on convolutional neural networks |
CN108616471A (en) * | 2018-05-07 | 2018-10-02 | 中国电子科技集团公司第三十六研究所 | A kind of signal modulate method and apparatus based on convolutional neural networks |
-
2018
- 2018-11-07 CN CN201811320190.XA patent/CN109274624B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1585289A (en) * | 2004-06-02 | 2005-02-23 | 中兴通讯股份有限公司 | Method for compensating frequency offset in wireless mobile communication system |
CN103188198A (en) * | 2013-03-25 | 2013-07-03 | 河海大学常州校区 | OFDM (Orthogonal Frequency Division Multiplexing) symbol timing and frequency offset estimation method based on particle swarm optimization (PSO) algorithm |
US9190053B2 (en) * | 2013-03-25 | 2015-11-17 | The Governing Council Of The Univeristy Of Toronto | System and method for applying a convolutional neural network to speech recognition |
CN106875002A (en) * | 2017-02-20 | 2017-06-20 | 中国石油大学(华东) | Complex value neural network training method based on gradient descent method Yu generalized inverse |
CN108509911A (en) * | 2018-04-03 | 2018-09-07 | 电子科技大学 | Interference signal recognition methods based on convolutional neural networks |
CN108616471A (en) * | 2018-05-07 | 2018-10-02 | 中国电子科技集团公司第三十六研究所 | A kind of signal modulate method and apparatus based on convolutional neural networks |
Non-Patent Citations (1)
Title |
---|
王雪安: ""基于循环谱的调制方式盲识别与频偏估计"", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110061939A (en) * | 2019-04-19 | 2019-07-26 | 中国电子科技集团公司第三十六研究所 | A kind of symbol rate estimation method and apparatus |
CN110048980A (en) * | 2019-04-19 | 2019-07-23 | 中国电子科技集团公司第三十六研究所 | A kind of blind demodulation method of digital communication and device |
CN110781755A (en) * | 2019-09-28 | 2020-02-11 | 深圳中物智建科技有限公司 | Intelligent identification and metering method and system for construction steel bars |
CN110995327A (en) * | 2019-12-17 | 2020-04-10 | 电子科技大学 | Mixed beam forming optimization method and system for multi-carrier MIMO system |
CN110995327B (en) * | 2019-12-17 | 2021-05-04 | 电子科技大学 | Mixed beam forming optimization method and system for multi-carrier MIMO system |
CN112187266A (en) * | 2020-09-29 | 2021-01-05 | 中国电子科技集团公司第三十六研究所 | Nonlinear correction method and device for analog-to-digital converter and electronic equipment |
CN112187266B (en) * | 2020-09-29 | 2024-03-08 | 中国电子科技集团公司第三十六研究所 | Nonlinear correction method and device of analog-to-digital converter and electronic equipment |
CN112910810B (en) * | 2021-02-01 | 2022-11-29 | 中国人民解放军国防科技大学 | Synchronization method, apparatus and medium for mimo-ofdm system |
CN112910810A (en) * | 2021-02-01 | 2021-06-04 | 中国人民解放军国防科技大学 | Synchronization method, apparatus and medium for MIMO OFDM system |
CN113609982A (en) * | 2021-08-05 | 2021-11-05 | 东南大学 | High-discrimination radio frequency fingerprint learning and extracting method based on intelligent carrier synchronization |
CN113609982B (en) * | 2021-08-05 | 2024-03-19 | 东南大学 | High-discrimination radio frequency fingerprint learning and extracting method based on intelligent carrier synchronization |
CN113726416A (en) * | 2021-09-01 | 2021-11-30 | 北京邮电大学 | Satellite communication carrier synchronization method and device and communication equipment |
CN113726416B (en) * | 2021-09-01 | 2022-10-11 | 北京邮电大学 | Satellite communication carrier synchronization method and device and communication equipment |
CN114598886B (en) * | 2022-05-09 | 2022-09-13 | 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) | Image coding method, decoding method and related devices |
CN114598886A (en) * | 2022-05-09 | 2022-06-07 | 哈尔滨工业大学(深圳)(哈尔滨工业大学深圳科技创新研究院) | Image coding method, decoding method and related device |
CN115103434A (en) * | 2022-06-23 | 2022-09-23 | 东南大学 | Machine learning-based 5G NR downlink timing synchronization method |
Also Published As
Publication number | Publication date |
---|---|
CN109274624B (en) | 2021-04-27 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109274624A (en) | A kind of carrier frequency bias estimation based on convolutional neural networks | |
O'Shea et al. | Physical layer communications system design over-the-air using adversarial networks | |
Barriac et al. | Distributed beamforming for information transfer in sensor networks | |
CN106059972B (en) | A kind of Modulation Identification method under MIMO correlated channels based on machine learning algorithm | |
CN103630894B (en) | The control method of broadband multi-channel coherent radar imaging system | |
CN101010922A (en) | Method and apparatus for calculating log-likelihood ratio for decoding in a receiver for a mobile communication system | |
Zhou et al. | RCNet: Incorporating structural information into deep RNN for online MIMO-OFDM symbol detection with limited training | |
CN109314539B (en) | Method, computer-readable storage medium, and apparatus for demodulating received signal | |
Zhao et al. | Echo state network for fast channel prediction in ricean fading scenarios | |
CN112153616A (en) | Power control method in millimeter wave communication system based on deep learning | |
CN113630130A (en) | End-to-end digital communication demodulation method | |
Ding et al. | Digital-twin-enabled city-model-aware deep learning for dynamic channel estimation in urban vehicular environments | |
CN115085774B (en) | Common sense fusion hybrid beamforming method based on Keramelteon | |
CN110417516A (en) | The method of adjustment and device of radio modem neural network based | |
Guo et al. | Deep learning for joint channel estimation and feedback in massive MIMO systems | |
CN105917624A (en) | Communication apparatus, demodulation apparatus, carrier reproduction apparatus, phase error compensation apparatus, phase error compensation method, and storage medium storing phase error compensation program | |
CN109495150A (en) | A kind of millimetre-wave attenuator codebook design method | |
Wang et al. | Methods of generating multiple uncorrelated Rayleigh fading processes | |
Chen et al. | Joint initial access and localization in millimeter wave vehicular networks: a hybrid model/data driven approach | |
CN112217579B (en) | Vector signal generation and device with adjustable error vector amplitude | |
CN108566255A (en) | Time correlation mimo system channel prediction method based on multi-task learning | |
Skima et al. | A multi-criteria comparative analysis of different Rayleigh fading channel simulators | |
CN107276657A (en) | Millimeter wave mixed-beam forming method based on parallel algorithm | |
Khanh et al. | Deep learning for uplink spectral efficiency in cell-free massive MIMO systems | |
Gorday et al. | GFSK demodulation using learned sequence correlation |
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 |