CN109962747A - The photon neural network self-adaptive processing device of wideband electromagnetic signal - Google Patents
The photon neural network self-adaptive processing device of wideband electromagnetic signal Download PDFInfo
- Publication number
- CN109962747A CN109962747A CN201910187390.0A CN201910187390A CN109962747A CN 109962747 A CN109962747 A CN 109962747A CN 201910187390 A CN201910187390 A CN 201910187390A CN 109962747 A CN109962747 A CN 109962747A
- Authority
- CN
- China
- Prior art keywords
- signal
- wavelength
- unit
- optical filter
- road
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
- G06N3/067—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using optical means
- G06N3/0675—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using optical means using electro-optical, acousto-optical or opto-electronic means
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B10/00—Transmission systems employing electromagnetic waves other than radio-waves, e.g. infrared, visible or ultraviolet light, or employing corpuscular radiation, e.g. quantum communication
- H04B10/25—Arrangements specific to fibre transmission
- H04B10/2575—Radio-over-fibre, e.g. radio frequency signal modulated onto an optical carrier
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04J—MULTIPLEX COMMUNICATION
- H04J14/00—Optical multiplex systems
- H04J14/02—Wavelength-division multiplex systems
- H04J14/0227—Operation, administration, maintenance or provisioning [OAMP] of WDM networks, e.g. media access, routing or wavelength allocation
- H04J14/0254—Optical medium access
- H04J14/0261—Optical medium access at the optical multiplex section layer
- H04J14/0265—Multiplex arrangements in bidirectional systems, e.g. interleaved allocation of wavelengths or allocation of wavelength groups
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04B—TRANSMISSION
- H04B2210/00—Indexing scheme relating to optical transmission systems
- H04B2210/006—Devices for generating or processing an RF signal by optical means
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Neurology (AREA)
- Electromagnetism (AREA)
- Optical Communication System (AREA)
Abstract
The present invention relates to field of signal processing, disclose the photon neural network self-adaptive processing device of wideband electromagnetic signal.Structure is as follows: electrooptic switching element connection optics serioparallel exchange unit obtains the input of the road N signal parallel, the optics serioparallel exchange unit is sequentially connected feature extraction unit, signal is combined into all the way by the first wavelength-division Multiplexing Unit, it is connected to the first optical filter group array element output road L signal, L is the node number of corresponding hidden layer, first optical filter connects L photon non-linear neural member, non-linear neural member is consecutively connected to the second wavelength-division multiplex unit, second optical filter group array element, the second optical filter group array element, which connects N number of photoelectric conversion unit, is superimposed the road N linearly, it is superimposed to be signally attached to weight control unit, weight control unit is connected to the first optical filter group array element.Above scheme achieves the advantage that ultra wide band electrically magnetic signal is received and handled;Flexible adaptive ability;Low energy consumption;Processing in real time.
Description
Technical field
The present invention relates to the photon neural network self-adaptive processing of field of signal processing, especially wideband electromagnetic signal dresses
It sets.
Background technique
Since electromagnetic wave is found by the mankind, research and application to it are just in accelerated development situation always.Current society
In meeting, each field such as wireless communication, radar, remote sensing, broadcast, mapping, electronic countermeasure all be unable to do without electromagnetic signal.From national economy
From the perspective of safety, electromagnetic signal has become the strategic resources for having to firmly grasp.
In order to preferably utilize electromagnetic spectrum resource, people plan signal frequency range purposes.However, electromagnetic signal
Opening cause it to be very easy to misapply, abuse, even illegal use, some intentional electromagnetic interferences or unintentionally electricity
Magnetic leakage is it is also possible to impact the key facilities such as communication system with system, and there is an urgent need to some wide spectrums and high performance
Spectral monitoring equipment carries out monitoring, control, investigation and the processing of frequency spectrum resource.
In general, target electromagnetic signal is converted to the signal for being easy to the Unified Form of subsequent processing by reception system, must
Must have corresponding ability, such as: target electromagnetic signal distributions are in certain frequency bandwidth, it is desirable that the system that receives has equal
Or bigger bandwidth of operation;Target electromagnetic signal may be very weak, it is also possible to and it is relatively strong, in order to can correctly monitor, it is desirable that connect
Receipts system has certain sensitivity and dynamic range.
Currently, electromagnetic signal, which is received, has been obtained extensive research with processing, microwave/radio frequency processing body is based on tradition
The reception system that system is established generallys use superhet search pattern, and combines flexible Digital Signal Processing (DSP).But it is super
Multistage frequency conversion technology (being three-level frequency conversion in typical case) in heterodyne system makes the reception of system, and structure is complicated, and then brings
The problems such as volume is big, power consumption is high;In addition, the common bandwidth of operation of microwave receiver is larger, and DSP processing is then limited to modulus and turns
The performance of (AD) and the calculating speed of itself are changed, bandwidth, the mismatch of rate are often resulted in, thus at the place of big bandwidth signal
Bottleneck is encountered in reason.
Electromagnetic signal coherent reception method and device of new generation based on microwave photon technology overcome multiple frequency conversion and band
Multinomial achievement especially has been obtained in wide problem under the support of US Department of Defense Advanced Research Projects Agency (DARPA).But
It is still to depend on traditional AD and DSP in signal processing link, does not dramatically increase the processing capacity of electromagnetic signal.
In recent years, the development of neural network considerably increases the flexibility and high efficiency of electromagnetic signal processing, and essence is
Using a kind of nonlinear treatment process, optimal response adaptively is carried out to target, in identification, sorting, memory, association
It is shown excellent performance, is widely used, and bring the advantage of low-power consumption.But for electricity of the invention
For magnetic signal, tradition does not break through the bottleneck in its bandwidth and rate still based on the neural network of electronic device building.With
Upper scheme has their own advantages, and is also respectively suitable for different application fields.But prior art or the nothing in receptivity
Method adapts to the requirement of Future broadband or is unable to satisfy quick, flexible requirement on signal processing.
Summary of the invention
The technical problems to be solved by the present invention are: in view of the above problems, providing the light of wideband electromagnetic signal
Sub-neural network self-adaptive processing device.
The technical solution adopted by the invention is as follows: the photon neural network self-adaptive processing device of wideband electromagnetic signal, packet
Include electrooptic switching element, the optics serioparallel exchange unit of wavelength resolution, feature extraction unit, two wavelength-division multiplex units, two
Optical filter group array element, photon non-linear neural member, N number of photoelectric conversion unit and weight control unit, the electro-optic conversion
Unit connection optics serioparallel exchange unit obtains the input of the road N signal parallel, and the optics serioparallel exchange unit is sequentially connected feature
Signal is combined into all the way by extraction unit, the first wavelength-division Multiplexing Unit, is connected to the first optical filter group array element output road L letter
Number, the L is the node number of corresponding hidden layer, and first optical filter connects L photon non-linear neural member, described non-thread
Nerve member is consecutively connected to the second wavelength-division multiplex unit, the second optical filter group array element, the second optical filter group battle array
Unit, which connects N number of photoelectric conversion unit, is superimposed the road N linearly, superimposed to be signally attached to weight control unit, described
Weight control unit is connected to the first optical filter group array element.
Further, the optics serioparallel exchange unit of the wavelength resolution includes 1:N coupler, N-1 wavelength convert list
Optical signal is divided into the road N by member, N-1 bit delay cell, the 1:N coupler, respectively λ0、λ1、λ2、…λN-1, wherein λ0It is straight-through
To one of bit delay cell, the other road N-1 signal λ1、λ2、…λN-2、λN-1Pass through a wavelength conversion unit respectively
Separate different parallel signals on wavelength, the λ after wavelength conversion unit1、λ2、…λN-2A bit is led to respectively again to prolong
Slow unit, the λ0、λ1、λ2、…λN-2、λN-1Delay be respectively N-1bit, N-2bit, N-3bit ... 1bit, 0bit.
Further, the optics serioparallel exchange unit of the wavelength resolution includes signal light, detection light, high non-linearity Jie
Matter, filter, the detection light generate the light-pulse generator of wavelength-division, and the light-pulse generator and signal light of the wavelength-division are sent into Gao Fei simultaneously
Linear medium, the high non-linearity medium are connected to filter.
Further, the high non-linearity medium uses nonlinear optical fiber.
Further, the high non-linearity medium uses semiconductor optical amplifier.
Further, the first optical filter group array element, photon non-linear neural member, the second wavelength-division multiplex unit,
One of realization structure of the photon neural network of second optical filter group array element, photoelectric conversion unit composition are as follows: described
First optical filter group array element uses 1:L optocoupler seaming element and matrix size (to indicate its filter for the optical filter group battle array of L × N
Wave device is arranged as L row N column), the photon non-linear neural member uses balance detection device assembly I and electro-optic conversion component, institute
Stating the second optical filter group array element uses 1:N coupler component and matrix size (to indicate it for the optical filter group battle array of N × L
Filter is arranged as L row N column), the photoelectric conversion unit uses balance detection component II, and the wavelength-division with N number of wavelength is multiple
The road L signal is separated by 1:L optocoupler seaming element with optical signal, wherein L is the number of nodes of a hidden layer, and the road L signal enters matrix
Balance detection device assembly I is sequentially entered after the optical filter group battle array that scale is L × N again and electro-optic conversion component, the second wavelength-division are answered
With unit, the second wavelength-division multiplex unit is connected to 1:N optocoupler seaming element and separates the road N signal, and the road N signal enters matrix size
To enter balance detection component II after the optical filter group battle array of N × L.
Compared with prior art, by adopting the above technical scheme have the beneficial effect that 1) realize ultra wide band electrically magnetic signal receive
With processing;2) flexible adaptive ability;3) low energy consumption;4) processing in real time.
Detailed description of the invention
Fig. 1 is the structural schematic diagram of the photon neural network self-adaptive processing device of wideband electromagnetic signal of the present invention.
Fig. 2 is the structural schematic diagram of one embodiment of the optics serioparallel exchange unit of wavelength resolution of the present invention.
Fig. 3 is the schematic diagram of one embodiment of the optics serioparallel exchange unit of wavelength resolution of the present invention.
Fig. 4 is the structural schematic diagram of another embodiment of the optics serioparallel exchange unit of wavelength resolution of the present invention.
Fig. 5 is the process of the multi-wavelength conversion of another embodiment of the optics serioparallel exchange unit of wavelength resolution of the present invention
Schematic diagram.
Fig. 6 is the structural schematic diagram of one embodiment of photon neural network of the present invention.
Fig. 7 is the corresponding equivalent topologies structure of Fig. 6.
Fig. 8 be baud rate be 100MHz 6 kinds of modulated signals time waveform figure, wherein 8 (a), 8 (b), 8 (c), 8 (d),
8 (e), 8 (f) respectively correspond 2PSK signal, 4PSK signal, 2FSK signal, 4FSK signal, 2PSK signal, 4PSK signal.
Fig. 9 is the training process schematic diagram of 2ASK signal identification.
Figure 10 is the accuracy rate schematic diagram of 6 kinds of Modulation Signals Recognitions.
Specific embodiment
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.
As shown in Figure 1, the photon neural network self-adaptive processing device of wideband electromagnetic signal, including electrooptic switching element
(101), 103, two wavelength-division multiplex units 104 of the optics serioparallel exchange unit 102, feature extraction unit of wavelength resolution1~
1042, two optical filter group array elements 1051~1052, photon non-linear neural member 1061~106L, N number of photoelectric conversion unit
1071~107NWith weight control unit 108, firstly, by 101 receive microwave signal of electrooptic switching element and be transformed on light wave,
Optical signal completes the parallel input of the road N signal by the optics serioparallel exchange unit 102 of wavelength resolution, and every road input signal has
Different optical wavelength, and be input in light feature extraction list 103 simultaneously, the extraction of essential feature is completed, it is a certain such as to extract completion
Parameter information needed for a function, such as spectrum density maximum value loss severity absolute value standard, standard nonlinear component deviation, mark
Quasi-component deviation etc..Each characteristic parameter is loaded on a wavelength, passes through a wavelength-division multiplex unit 1041It is combined into all the way.It connects
, according to the number of set hidden node, the optical filter group of a 1:2L (indicating that there is 1 input, 2L output)
Array element 1051First input signal is divided into the road L (corresponding hidden layer is L node) in interior radical, it is every to have the filtering of N group light all the way
Device is filtered to the long signal of Different lightwave and (is equivalent to weighed value adjusting), is 2 signal streams to be weighted per output all the way.It adjusts
Signal after whole enters L photon non-linear neural member 1061~106LThe simulation for realizing nonlinear activation function, wherein each light
Sub- neuron has different wavelength.Finally, passing through a wavelength division multiplexer 1042, 1:2N (indicate that there is 1 input,
2N output) optical filter group array element 1052With N number of photoelectric conversion unit 1071~107NRealize the linear folded of the road N signal
Add, to realize the building of an intact nervous network.The receiver can by training sequence and weight control unit 108 come
The connection weight of two optical filter group array elements is adjusted, to realize the reconstruct of different processing functions.
One embodiment of the optics serioparallel exchange unit of wavelength resolution: Fig. 2 is the device of optics serioparallel exchange unit, packet
1:N coupler, N-1 wavelength conversion unit, N-1 bit delay cell are included, optical signal is divided into the road N by the 1:N coupler,
λ respectively0、λ1、λ2、…λN-1, wherein λ0Through to one of bit delay cell, the other road N-1 signal λ1、λ2、…
λN-2、λN-1Separate different parallel signals on wavelength by a wavelength conversion unit respectively, by wavelength conversion unit
λ afterwards1、λ2、…λN-2Lead to a bit delay cell, the λ respectively again0、λ1、λ2、…λN-2、λN-1Delay be respectively N-
1bit,N-2bit,N-3bit,…1bit,0bit.Such as the principle that Fig. 3 is optics serioparallel exchange.The electromagnetic signal of M-bit all the way
Being modulated to wavelength is λ0Light carrier after, by optical fiber transmit reach receiver after both carry out serioparallel exchange, according to processing essence
The requirement of degree, such as above structure, it is determined that the number N of parallel signal, wherein wavelength X all the way0Signal is straight-through, and in addition the road N-1 is believed
Number respectively by corresponding wavelength conversion unit, separates different parallel signals on wavelength, be denoted as λ respectively1、λ2、…λN-1。
Different wave length signal passes through different delays, such as λ respectively0Delay is N-1bit, λ1Delay is N-2bit, λN-1Delay is
0bit.Parallel signal data format after delay is as shown in figure 3, with λ0Data start bit start, successively record the road N signal
Data, so that it may realize the conversion of optics string and data flow, while realize wavelength convert.
Another embodiment of the optics serioparallel exchange unit of wavelength resolution: as shown in figure 4, the optics of the wavelength resolution
Serioparallel exchange unit include signal light, detection light, high non-linearity medium, filter, detect light generation can use it is super continuous
Generation+spectrum carving of spectrum+light path escaping scheme (only giving a feasible scheme here), the purpose is to realize a wave
(wavelength is represented by λ ' to the light-pulse generator divided1、λ’2…λ’N-1), the light-pulse generator and signal light of the wavelength-division are sent into Gao Fei simultaneously
Linear medium is to generate four-wave mixing effect, and wherein the power of signal light is greater than detection light.The high non-linearity medium is using non-
Linear optical fiber can also use semiconductor optical amplifier.According to formula ωi=2 ωp-ωsIt is found that the difference generated by four-wave mixing
At the ideler frequency light of wavelength can completely replica signal light information, and utilize an optical filter in output end, and rationally control
The central wavelength of filter, so that it may obtain that N number of amplitude is equal, pulse signal stream of wavelength resolution, as shown in Figure 5.
The backbone serioparallel exchange unit that above structure is realized is at low cost, controls simple and flexible.
One embodiment of photon neural network: photon neural network is by the first optical filter group array element, the light in Fig. 1
Sub- non-linear neural member, the second wavelength-division multiplex unit, the second optical filter group array element, photoelectric conversion unit composition, such as Fig. 6 institute
Show in the present embodiment, the first optical filter group array element uses 1:L optocoupler seaming element and matrix size to filter for the light of L × N
Wave device group battle array (indicate its filter row be classified as L row N column), the photon non-linear neural member using balance detection device assembly I and
Electro-optic conversion component, the second optical filter group array element use 1:N coupler component and matrix size to filter for the light of N × L
Wave device group battle array (indicates that its filter row is classified as N row L column), and the photoelectric conversion unit uses balance detection component II, has N number of
The wavelength-division-multiplexed optical signal of wavelength separates the road L signal by 1:L optocoupler seaming element, and wherein L is the number of nodes of a hidden layer, the road L
Signal sequentially enters balance detection device assembly I and electro-optic conversion group after entering the optical filter group battle array that matrix size is L × N again
Part, the second wavelength-division multiplex unit, the second wavelength-division multiplex unit are connected to 1:N optocoupler seaming element and separate the road N signal, the road N letter
Number enter matrix size be N × L optical filter group battle array after enter balance detection component II.
The structural topology figure of above-mentioned neural network is as shown in fig. 7, the neural network is divided into three layers, an input layer, node
Number is N, is determined by number of wavelengths;One hidden layer, number of nodes L;One output layer, number of nodes N are determined by input layer.Photon
The specific implementation process of neural network are as follows: branch, branch are carried out by the photo-coupler of 1:L first by the optical signal of wavelength-division multiplex
Signal afterwards enters L × N filtering unit group battle array of photon neural network simultaneously.Since the optical signal of input has N number of wavelength,
N number of optical filter is needed to control each wavelength signals per filter unit all the way, the realization of filter is here with micro-loop
For (but it is not only limited in micro-loop).The different signal of N number of wavelength is after N number of filter is filtered, into balance detection
The upper arm of component I, and unfiltered part (through connect signal) enters the lower arm of balance detection component I, therefore can realize subtracting for signal
Method operation thus completes the weighting operations of a hidden node to simulate positive and negative weighted effect.When complete the road L it is above-mentioned
When operation, the weighting by N number of input layer to L hidden node, V as shown with 7 are representedij, wherein i is 1~L, j 1
~N.Electric signal is modulated on the different light carrier of wavelength by the signal after weighting again by electro-optic conversion component, is adjusted at this time
Device processed can simulate nonlinear activation function due to its nonlinear transmission characteristic in neural network.Therefore, hidden node ui's
Output can indicate are as follows:
Signal after electro-optic conversion is combined into all the way again by Interleave muiltiplexing component element, has L wave in signal stream at this time
It is long.As procedure described above, the photo-coupler of optical signals 1:N is divided into the identical signal in the road N, then the filter assembly by a N × L
It carries out hidden layer with the balance detection component II of N × 1 to operate to the linear weighted function of output layer, W as shown in Figure 7ji.Output layer at this time
Node expression be
Formula (2) illustrates that a complete Neural Network Based Nonlinear functional simulation function passes through in photon neural network
Fuel factor or electrode power-up mode come change the relevant parameter of micro-loop can be to the adjustment of weight coefficient, to complete broadband
The power function realized required for receiver.
It is applied below with one for being identified as above-mentioned broadband receiver of modulation format to be illustrated, it should be noted that
For those of ordinary skill in the art, without creative efforts, it can also be obtained according to the above process
The measurement of the information such as other function, such as phase, direction, frequency and pulsewidth.
The present embodiment establishes above-mentioned analogue system using matlab, and is believed using neural fusion 6 kinds of modulation
Number identification, comprising: 2ASK, 4ASK, 2FSK, 4FSK, 2PSK and 4PSK.
Fig. 8 is the time waveform figure that baud rate is modulated signal in the 6 of 100MHz, wherein Fig. 8 (a) 2PSK signal, Fig. 8
It (b) is 4PSK signal, Fig. 8 (c) is 2FSK signal, and Fig. 8 (d) is 4FSK signal, and Fig. 8 (e) is 2PSK signal, and Fig. 8 (f) is 4PSK
Signal.According to the research of E.E.Azzouz and A.K.Nandi, the present invention has carried out five extraordinary extractions to this signal, point
5 inputs for not corresponding to neural network, spectrum density, the non-weak signal section of zero center including zero center normalization instantaneous amplitude are instantaneous
The standard deviation of the non-weak signal end instantaneous phase nonlinear component of standard deviation, the zero center of phase nonlinear component absolute value, zero
Center normalizes the standard deviation of non-weak signal section instantaneous frequency absolute value and the mark of zero center normalization instantaneous amplitude absolute value
Quasi- deviation.
For the identification using neural fusion to above-mentioned signal, it is 5 that input neuron number, which is arranged, in the present invention, non-thread
Nerve member number of nodes is 15, and output node number is 6, number of training 2000.Fig. 9 gives in 2ASK signal identification,
The training process of neural network.With 10-3For the target of training error, it can be seen that the neural network only needs 26 steps that can realize
Convergence.The training of other modulated signals can obtain similar result without creative efforts.
The accuracy rate for each Modulation Signals Recognition that Figure 10 is signal-to-noise ratio when being 5dB.Wherein, 2ASK signal is 94%, 4AKS letter
It number be 98%, 2FSK signal be 100%, 4FSK signal be 82%, 2PSK signal be 92%, 4PSK signal is 86%.
By in the above simulation result it is observed that the present invention is successfully realized using photon neural network is to rate
The high-precisions of 6 kinds of modulated signals of 100MHz identifies, it was demonstrated that wideband electromagnetic signal is received to be had with the method and device that handles
Effect property.The program is in frequency spectrum resource monitoring, control and investigation field is of great significance and application value.
The invention is not limited to specific embodiments above-mentioned.The present invention, which expands to, any in the present specification to be disclosed
New feature or any new combination, and disclose any new method or process the step of or any new combination.If this
Field technical staff is altered or modified not departing from the unsubstantiality that spirit of the invention is done, should belong to power of the present invention
The claimed range of benefit.
Claims (6)
1. the photon neural network self-adaptive processing device of wideband electromagnetic signal, which is characterized in that including electrooptic switching element, wave
The long optics serioparallel exchange unit differentiated, feature extraction unit, two wavelength-division multiplex units, two optical filter group array elements,
Photon non-linear neural member, N number of photoelectric conversion unit and weight control unit, the electrooptic switching element connect optics string and turn
It changes unit and obtains the input of the road N signal parallel, the optics serioparallel exchange unit is sequentially connected feature extraction unit, the first wavelength-division is answered
Signal is combined into all the way with unit, is connected to the first optical filter group array element output road L signal, the L is corresponding hidden layer
Node number, first optical filter connects that L photon non-linear neural is first, and the non-linear neural member is consecutively connected to
Second wavelength-division multiplex unit, the second optical filter group array element, the second optical filter group array element connect N number of photoelectric conversion
Unit is superimposed the road N linearly, superimposed to be signally attached to weight control unit, and the weight control unit is connected to the
One optical filter group array element.
2. the photon neural network self-adaptive processing device of wideband electromagnetic signal as described in claim 1, which is characterized in that institute
The optics serioparallel exchange unit for stating wavelength resolution includes 1:N coupler, N-1 wavelength conversion unit, N-1 bit delay cell,
Optical signal is divided into the road N by the 1:N coupler, respectively λ0、λ1、λ2、…λN-1, wherein λ0Postpone through to one of bit
Unit, the road N-1 signal λ in addition1、λ2、…λN-2、λN-1Different parallel signals is set to exist by a wavelength conversion unit respectively
It is separated on wavelength, the λ after wavelength conversion unit1、λ2、…λN-2Lead to a bit delay cell, the λ respectively again0、λ1、
λ2、…λN-2、λN-1Delay be respectively N-1bit, N-2bit, N-3bit ... 1bit, 0bit.
3. the photon neural network self-adaptive processing device of wideband electromagnetic signal as described in claim 1, which is characterized in that institute
The optics serioparallel exchange unit for stating wavelength resolution includes signal light, detection light, high non-linearity medium, filter, the detection light
The light-pulse generator of wavelength-division is generated, the light-pulse generator and signal light of the wavelength-division are sent into high non-linearity medium simultaneously, and the height is non-thread
Property medium is connected to filter.
4. the photon neural network self-adaptive processing device of wideband electromagnetic signal as claimed in claim 3, which is characterized in that institute
High non-linearity medium is stated using nonlinear optical fiber.
5. the photon neural network self-adaptive processing device of wideband electromagnetic signal as claimed in claim 3, which is characterized in that institute
High non-linearity medium is stated using semiconductor optical amplifier.
6. the photon neural network self-adaptive processing device of wideband electromagnetic signal as described in any one in claim 1-5, special
Sign is that the first optical filter group array element, photon non-linear neural member, the second wavelength-division multiplex unit, the second light filter
One of realization structure of the photon neural network of device group array element, photoelectric conversion unit composition are as follows: the first light filtering
Device group array element uses 1:L optocoupler seaming element and matrix size for the optical filter group battle array of L × N, the photon non-linear neural
Member uses balance detection device assembly I and electro-optic conversion component, and the second optical filter group array element uses 1:N coupler component
The optical filter group battle array for being N × L with matrix size, the photoelectric conversion unit use balance detection component II, have N number of wavelength
Wavelength-division-multiplexed optical signal the road L signal is separated by 1:L optocoupler seaming element, wherein L be a hidden layer number of nodes, the road L signal
Balance detection device assembly I and electro-optic conversion component, the are sequentially entered after into the optical filter group battle array that matrix size is L × N again
Two wavelength-division multiplex units, the second wavelength-division multiplex unit are connected to 1:N optocoupler seaming element and separate the road N signal, and the road N signal enters
Enter balance detection component II after the optical filter group battle array that matrix size is N × L.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910187390.0A CN109962747B (en) | 2019-03-13 | 2019-03-13 | Photon neural network self-adaptive processing device for broadband electromagnetic signals |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910187390.0A CN109962747B (en) | 2019-03-13 | 2019-03-13 | Photon neural network self-adaptive processing device for broadband electromagnetic signals |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109962747A true CN109962747A (en) | 2019-07-02 |
CN109962747B CN109962747B (en) | 2020-05-26 |
Family
ID=67024304
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910187390.0A Active CN109962747B (en) | 2019-03-13 | 2019-03-13 | Photon neural network self-adaptive processing device for broadband electromagnetic signals |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109962747B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112232129A (en) * | 2020-09-17 | 2021-01-15 | 厦门熙重电子科技有限公司 | Electromagnetic information leakage signal simulation system and method based on generation countermeasure network |
CN112398543A (en) * | 2019-08-19 | 2021-02-23 | 上海诺基亚贝尔股份有限公司 | Method, apparatus, system, device and computer readable medium for optical communication |
CN113822425A (en) * | 2021-11-02 | 2021-12-21 | 广西大学 | Integrated optical neural network system capable of resisting transmission loss interference |
CN116681117A (en) * | 2023-05-24 | 2023-09-01 | 浙江大学 | Photoelectric computer based on optical neuron wavelength multiplexing module |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20030135971A1 (en) * | 1997-11-12 | 2003-07-24 | Michael Liberman | Bundle draw based processing of nanofibers and method of making |
CN103178901A (en) * | 2013-02-05 | 2013-06-26 | 西安邮电大学 | Optical fiber simulated Raman scattering effect based optical neuron and establishment method thereof |
CN105790880A (en) * | 2016-03-01 | 2016-07-20 | 中国电子科技集团公司第五十四研究所 | Radio frequency exchange method based on microwave photon technology |
CN109039472A (en) * | 2018-07-11 | 2018-12-18 | 南通智大信息技术有限公司 | A kind of data center's optic communication dispersive estimates and management method based on deep learning |
CN109379138A (en) * | 2018-08-24 | 2019-02-22 | 中国电子科技集团公司第二十九研究所 | High Linear microwave photon link implementation method and structure based on photon neural network |
-
2019
- 2019-03-13 CN CN201910187390.0A patent/CN109962747B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20030135971A1 (en) * | 1997-11-12 | 2003-07-24 | Michael Liberman | Bundle draw based processing of nanofibers and method of making |
CN103178901A (en) * | 2013-02-05 | 2013-06-26 | 西安邮电大学 | Optical fiber simulated Raman scattering effect based optical neuron and establishment method thereof |
CN105790880A (en) * | 2016-03-01 | 2016-07-20 | 中国电子科技集团公司第五十四研究所 | Radio frequency exchange method based on microwave photon technology |
CN109039472A (en) * | 2018-07-11 | 2018-12-18 | 南通智大信息技术有限公司 | A kind of data center's optic communication dispersive estimates and management method based on deep learning |
CN109379138A (en) * | 2018-08-24 | 2019-02-22 | 中国电子科技集团公司第二十九研究所 | High Linear microwave photon link implementation method and structure based on photon neural network |
Non-Patent Citations (1)
Title |
---|
王睿等: "光学神经拟态计算研究进展", 《激光与光电子学进展》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112398543A (en) * | 2019-08-19 | 2021-02-23 | 上海诺基亚贝尔股份有限公司 | Method, apparatus, system, device and computer readable medium for optical communication |
CN112398543B (en) * | 2019-08-19 | 2023-08-25 | 上海诺基亚贝尔股份有限公司 | Methods, apparatus, systems, devices, and computer readable media for optical communications |
CN112232129A (en) * | 2020-09-17 | 2021-01-15 | 厦门熙重电子科技有限公司 | Electromagnetic information leakage signal simulation system and method based on generation countermeasure network |
CN113822425A (en) * | 2021-11-02 | 2021-12-21 | 广西大学 | Integrated optical neural network system capable of resisting transmission loss interference |
CN113822425B (en) * | 2021-11-02 | 2023-07-18 | 广西大学 | Transmission loss interference resistant integrated optical neural network system |
CN116681117A (en) * | 2023-05-24 | 2023-09-01 | 浙江大学 | Photoelectric computer based on optical neuron wavelength multiplexing module |
Also Published As
Publication number | Publication date |
---|---|
CN109962747B (en) | 2020-05-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109962747A (en) | The photon neural network self-adaptive processing device of wideband electromagnetic signal | |
CN101651472B (en) | Carrier wave communication device of power line and carrier wave communication method thereof | |
Guesmi et al. | Experimental demonstration of simultaneous modulation format/symbol rate identification and optical performance monitoring for coherent optical systems | |
CN105099553B (en) | A kind of visible light communication method of reseptance and its system based on neuroid | |
CN106452593B (en) | The construction method and device and nonlinear noise suppressing method and system of filter | |
CN101777953A (en) | Light double-sideband modulation device and method for transmitting two paths of signals | |
CN105547654A (en) | Method and system for measurement of optical device on the basis of optical double-side band modulation | |
CN104618023B (en) | Optical link linearization method | |
CN109510633A (en) | Power equalization method and modulator for specific harmonic elimination multilevel radio frequency pulse width modulation | |
Yu et al. | Multi-task learning convolutional neural network and optical spectrums enabled optical performance monitoring | |
CN106788591A (en) | Photovoltaic parallel in system based on power line carrier communication | |
CN108566244A (en) | The optical device spectral response measurement method and device of multi-channel parallel | |
CN106506088A (en) | A kind of carrier aggregation based on single sideband modulation and the method and system for depolymerizing | |
CN101909029A (en) | Method and system for controlling bias points of DQPSK (Differential Quadrature Reference Phase Shift Keying) modulator | |
US20090214213A1 (en) | Modulation device for generating optical signal with quadruple frequency and method thereof | |
CN101369845B (en) | Chromatic dispersion monitoring method, system and apparatus | |
CN114268433B (en) | Nonlinear compensation method of high-speed continuous variable quantum key distribution system | |
CN107517174B (en) | A kind of phase-shift keying (PSK) demodulator circuit based on analog switch | |
CN109001554A (en) | Immunity experiment multifrequency signal generator | |
CN205864375U (en) | A kind of High Linear high efficiency power amplifier | |
CN112787815B (en) | Continuous variable quantum key communication method and system based on attack perception and defense | |
CN103856432B (en) | Micro-strip resonance coherent demodulator for AMPSK modulating signals | |
CN113472713A (en) | High-order modulation signal demodulation method based on neural network and receiver | |
CN204065629U (en) | The tunable whole-optical Raman wavelength converter that a kind of Output optical power is equal | |
Wang et al. | High speed signal transmission technology based on partial power channel for capacitive power transfer system |
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 |