CN116708104A - Modulation format identification method of underwater visible light communication system based on reserve pool calculation - Google Patents
Modulation format identification method of underwater visible light communication system based on reserve pool calculation Download PDFInfo
- Publication number
- CN116708104A CN116708104A CN202310450808.9A CN202310450808A CN116708104A CN 116708104 A CN116708104 A CN 116708104A CN 202310450808 A CN202310450808 A CN 202310450808A CN 116708104 A CN116708104 A CN 116708104A
- Authority
- CN
- China
- Prior art keywords
- modulation format
- visible light
- light communication
- communication system
- folding
- 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
- 238000004891 communication Methods 0.000 title claims abstract description 40
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000004364 calculation method Methods 0.000 title claims abstract description 31
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 62
- 230000009466 transformation Effects 0.000 claims abstract description 16
- 230000004913 activation Effects 0.000 claims description 15
- 238000009826 distribution Methods 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 5
- 230000002776 aggregation Effects 0.000 claims 1
- 238000004220 aggregation Methods 0.000 claims 1
- 238000007781 pre-processing Methods 0.000 abstract description 2
- 239000010410 layer Substances 0.000 description 38
- 230000006870 function Effects 0.000 description 18
- 238000013528 artificial neural network Methods 0.000 description 15
- 238000012549 training Methods 0.000 description 9
- 238000005516 engineering process Methods 0.000 description 8
- 238000000605 extraction Methods 0.000 description 7
- 238000010586 diagram Methods 0.000 description 6
- 238000013461 design Methods 0.000 description 5
- 238000002474 experimental method Methods 0.000 description 5
- 238000000354 decomposition reaction Methods 0.000 description 4
- 238000013135 deep learning Methods 0.000 description 4
- 230000005540 biological transmission Effects 0.000 description 3
- 230000008859 change Effects 0.000 description 3
- 238000007635 classification algorithm Methods 0.000 description 3
- 238000013459 approach Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000001965 increasing effect Effects 0.000 description 2
- 239000011229 interlayer Substances 0.000 description 2
- 230000000306 recurrent effect Effects 0.000 description 2
- 229920006395 saturated elastomer Polymers 0.000 description 2
- 230000004931 aggregating effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 239000000969 carrier Substances 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000009827 uniform distribution Methods 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L27/00—Modulated-carrier systems
- H04L27/0012—Modulated-carrier systems arrangements for identifying the type of modulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- 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
- G06N3/047—Probabilistic or stochastic 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
- G06N3/048—Activation functions
-
- 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
-
- 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
- H04B10/80—Optical aspects relating to the use of optical transmission for specific applications, not provided for in groups H04B10/03 - H04B10/70, e.g. optical power feeding or optical transmission through water
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Probability & Statistics with Applications (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Electromagnetism (AREA)
- Optical Communication System (AREA)
Abstract
The invention discloses a modulation format identification method of an underwater visible light communication system based on reserve pool calculation; the method comprises the steps of utilizing a received signal of a system receiving end, firstly utilizing a coordinate transformation algorithm and a folding algorithm to perform data preprocessing and feature enhancement on the received signal, and then performing classification judgment of modulation format recognition based on a reserve pool computing network; the method can realize low calculation cost, high real-time performance, high accuracy and strong robustness in the modulation format identification task of the underwater visible light communication system.
Description
Technical Field
The invention relates to the technical field of electronic information, in particular to an underwater visible light communication system modulation format identification method based on reservoir calculation.
Background
Underwater visible light communication (Underwater Visible Light Communication, UVLC) is a novel wireless communication technology that utilizes visible light as a carrier to transmit information through an underwater channel. Compared with the traditional underwater communication technology, the underwater visible light communication has the advantages of higher communication rate, stronger electromagnetic interference resistance, better communication stability and the like, so that the method is widely applied to underwater resource development, ocean science and other special scenes [1,2]. In the foreseeable future, underwater visible light communication should have stronger dynamic performance to meet the demands of various frequency bands and improve spectrum utilization. As one of key technologies for directly affecting the transmission quality and rate of underwater visible light communication data, modulation format identification (Modulation Format Recognition, MFR) has the capability of autonomously detecting the modulation format of a received signal, without any prior information from a transmitting end, so that a receiver can dynamically adjust a demodulation mode, and the dynamic performance and frequency utilization rate of a communication system are enhanced.
However, in the task of modulation format identification, it is a difficult task to classify the modulation formats directly due to lack of a priori information from the transmitting end. In the long-term evolution, researchers have proposed various methods to improve the performance of the modulation format recognition task [3]. These methods can be broadly divided into two categories: probability-based (LB) methods and feature-based (FB) methods. In the probability-based method, researchers solve the problem of modulation format recognition by using methods such as probability and hypothesis test parameters, and the error probability is reduced to the minimum through proper hypothesis and proper threshold value, so as to obtain the optimal solution [4] in the Bayesian sense. Although this approach can provide an optimal solution, it is computationally complex and difficult to implement in practical applications. In contrast, feature-based methods extract salient features from received data and then use these features for modulation format identification. The method is easy to implement and can obtain near-optimal modulation format recognition accuracy [5] through proper feature extraction. Of course, how to design efficient feature extraction algorithms is also considered an important topic.
With the rapid development of artificial intelligence algorithms for several years, neural network-based algorithms have achieved significant success in a number of scientific fields. Algorithms based on neural networks are also applied to modulation format recognition tasks, and researchers obtain better performances in the modulation format recognition tasks through various deep learning algorithms such as various convolutional neural networks, cyclic neural networks, graph neural networks and the like [6,7]. However, due to the large data driving performance of the neural network, in an actual application scene, a user often has difficulty in acquiring enough mass data to sufficiently train the neural network. Furthermore, the vast computational resources and computational time overhead of neural network training are often unacceptable in systems with high real-time requirements. Therefore, how to design a new neural network structure, simplify the network framework, reduce the training overhead and improve the algorithm performance is a subject worthy of further discussion.
[ reference ]
1.S.Arnon,"Underwater optical wireless communication network,"Opt.Eng 49(1),015001(2010).
2.H.Kaushal and G.Kaddoum,"Underwater Optical Wireless Communication,"IEEE Access 4,1518–1547(2016).
3.O.A.Dobre,A.Abdi,Y.Bar-Ness,and W.Su,"Survey of automatic modulation classification techniques:classical approaches and new trends,"IET Commun.1(2),137(2007).
4.W.Wei and J.M.Mendel,"Maximum-likelihood classification for digital amplitude-phase modulations,"IEEE TRANSACTIONS ON COMMUNICATIONS 48(2),(2000).
5.A.K.Jain,R.P.W.Duin,and J.Mao,"Statistical pattern recognition:a review,"IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 22(1),(2000).
6.F.N.Khan,K.Zhong,W.H.Al-Arashi,C.Yu,C.Lu,and A.P.T.Lau,"Modulation Format Identification in Coherent Receivers Using Deep Machine Learning,"IEEE PHOTONICS TECHNOLOGY LETTERS 28(17),(2016).
7.D.Wang,M.Zhang,Z.Li,J.Li,M.Fu,Y.Cui,and X.Chen,"Modulation Format Recognition and OSNR Estimation Using CNN-Based Deep Learning,"IEEE Photon.Technol.Lett.29(19),1667–1670(2017).
Disclosure of Invention
In view of the above-mentioned shortcomings of the prior art, an object of the present invention is to provide a method for identifying modulation format of an underwater visible light communication system based on reservoir calculation (reservoir computing, RC) in combination with coordinate transformation and folding algorithm; the method can realize low calculation cost, high real-time performance, high accuracy and strong robustness in the modulation format identification task of the underwater visible light communication system.
In the invention, for the received signal of the receiving end of the underwater visible light communication, a digital signal processing technology is firstly utilized to obtain a corresponding IQ signal (I: in-phase, Q: quality), then a coordinate transformation algorithm and a folding algorithm are utilized to carry out efficient feature extraction, the accuracy of the subsequent modulation format classification is improved, and finally a reserve pool calculation is utilized to carry out efficient and simple classification, so that a corresponding modulation format classification result is obtained. The scheme of the invention is specifically described as follows.
A modulation format identification method of an underwater visible light communication system based on reservoir calculation comprises the following steps:
step 1): in an underwater visible light communication system, data sent by an LED at a transmitting end are received at a receiving end after passing through an underwater channel, and complex-value receiving signals are obtained after digital signal processing;
step 2): decomposing the received complex value signals under a rectangular coordinate system and a polar coordinate system respectively through a coordinate transformation algorithm to obtain an IQ component under the rectangular coordinate system and a polar angle polar diameter component under the polar coordinate system respectively, and then aggregating data under different coordinate systems;
step 3): carrying out folding operation on the feature graphs corresponding to the data under the rectangular coordinate system and the polar coordinate system by utilizing symmetry according to a symmetry axis through a folding algorithm, so as to reduce the range of the feature graphs and realize the removal of redundant information and the salient of significant features;
step 4): sending the data of the folded feature map obtained in the step 3) into a reserve pool computing RC network, generating an output weight of an optimal solution in the least square sense through a ridge regression method, linearly combining node states of the middle layer, obtaining probability that input data belong to each modulation format through a softmax function, and selecting a category corresponding to the maximum probability to obtain a classification decision of the final modulation format.
In the invention, in the step 3), for the two-dimensional signal under the polar coordinate system, the folding algorithm is carried out according to different angles in consideration of the symmetry of the distribution of the feature map; recording the angle at the current polar coordinateIn the range of->To->Where n is the current folding order, the unfolded time order is 0, i represents the ith data point, and the symmetry axis of the current signal distribution isThe symmetry axis is always the average value of the current angle range, and then the folding operation folds according to the symmetry axis:
in view of symmetry, each folding operation is equivalent to folding to the left or right of the symmetry axis, the equivalent folding operation to the other direction being expressed as:
in the invention, in the step 3), the folding order is 3-4 times; in the step 4), the number of nodes in the middle layer in the RC network is 400-800.
In the invention, in step 4), a nonlinear activation function is adopted in the RC network by the reserve pool calculation, wherein the nonlinear function is tanh, sigmoid or ReLU.
Compared with the prior art, the invention has the beneficial effects that:
1. the reserve pool computing network algorithm is applied to a modulation format recognition task of the underwater visible light communication system, so that higher accuracy is realized compared with a traditional classification algorithm, training expenditure and time cost are greatly reduced compared with a neural network method based on deep learning, and better accuracy, instantaneity and robustness are realized;
2. by introducing a coordinate transformation algorithm, different characteristics of PSK and QAM signals are highlighted, so that efficient characteristic enhancement is carried out on input data, local salient characteristics are highlighted, characteristic redundancy is reduced, and global characteristics are combined with local characteristics.
3. By introducing the folding algorithm, the redundancy of input data is effectively reduced through folding operation of different orders on the original feature map, and the global feature and the local feature are combined, so that the calculated amount and time expenditure of the algorithm are reduced, and the accuracy of the algorithm is improved.
Drawings
Fig. 1 is a schematic diagram of a modulation format recognition technique of an underwater visible light communication system based on reservoir calculation according to the present invention.
Fig. 2 is a system diagram of the underwater visible light communication system modulation format recognition technology based on reservoir calculation of the present invention.
Fig. 3 shows the transformation results of the coordinate transformation of the present invention in rectangular coordinate system and polar coordinate system for six modulation format data.
Fig. 4 is a transformation result of the folding algorithm of the present invention for performing different order folding operations on six modulation format data.
Fig. 5 is a study of the accuracy of the identification of the debug format according to the present invention along with the change of the emission voltage of the LED at the emission end.
FIG. 6 is a graph showing the algorithm accuracy of the present invention as a function of non-linear activation, reservoir interlayer size, and leakage coefficient, among other hyper-parameters.
Detailed Description
The technical scheme of the invention is described in detail below with reference to the accompanying drawings and examples.
Aiming at the modulation format recognition task of the underwater visible light communication, the invention mainly considers how to design a light and efficient algorithm, relieves huge calculation overhead and time consumption of a common neural network algorithm, and meets the requirement of a high-real-time system. Meanwhile, considering the complexity of the underwater channel, how to design the feature extraction algorithm in a targeted way, and perform data preprocessing and feature enhancement on the input data, so that the follow-up algorithm can be light in weight and high in accuracy.
In the invention, aiming at a modulation format recognition task in an underwater visible light communication system, namely, a received signal at a system receiving end is utilized, the pretreatment and the characteristic enhancement are carried out firstly, and then the classification judgment of the modulation format recognition is carried out. In an actual scene, due to the complexity of the underwater channel, how to effectively extract the characteristics of the received signals, and complete the modulation format classification judgment with the smallest time overhead and calculation resources, has very wide application value. The invention provides a modulation format identification technical scheme based on a feature extraction algorithm and reserve pool calculation (RC). The specific flow comprises the following steps: for the receiving signals of the receiving end of the underwater visible light communication, a digital signal processing technology is utilized to obtain corresponding IQ signals (I: in-phase, Q: quality), then a coordinate transformation algorithm and a folding algorithm are utilized to conduct efficient feature extraction, the accuracy of the subsequent modulation format classification is improved, and finally a reserve pool is utilized to conduct efficient and simple classification, so that corresponding modulation format classification results are obtained. By utilizing the technical scheme of modulation format identification provided by the patent, experiments are carried out on six modulation formats of OOK,4QAM,8QAM-DIA,8QAM-CIR,16APSK and 16QAM in an underwater visible light communication system, and the experimental results prove the effectiveness and high efficiency of the technical scheme provided by the invention.
1. Coordinate transformation algorithm
In an underwater visible light communication system, a plurality of signals received by a receiving end are recorded as(where N represents the length of each set of received signals), to obtain the corresponding IQ signal, consider:
In phase＝Imag[y(i)-
Quadrature＝Real[y(i)-#(1)
by separating the real part and the imaginary part of the complex signal, we convert the received one-dimensional complex signal into two-dimensional IQ signal, and in the traditional modulation format recognition algorithm, the subsequent classification algorithm directly takes the IQ signal as input and directly classifies the IQ signal to complete modulation format recognition. This patent considers carrying out further feature extraction to traditional IQ signal, makes it easier for subsequent modulation format classification decision.
The conventional IQ signal can be regarded as a component of the complex signal in the two-dimensional rectangular coordinate system, and the decomposition mode is suitable for modulation formats like QAM signals which are uniformly distributed in the x-axis and the y-axis of the two-dimensional rectangular coordinate system, but is not suitable for modulation formats like PSK which have ring symmetry characteristics. For this purpose, the invention contemplates the introduction of decomposing complex signals in two-dimensional polar coordinates
θ＝Atan2(Imag[y(i)-,Real[y(i)-)#(2)
(wherein ρ represents the radial length in polar coordinates, θ represents the angle in polar coordinates, θ is expressed as 0 ° To 360 degrees ° Or-180 ° To 180 ° All things being equal, the end result is not essentially different
This ring symmetry, such as PSK, is well characterized in polar coordinates, making its features more prominent. Meanwhile, the representation of the ring symmetry also reveals richer characteristics of rectangular symmetrical signals such as QAM (just like the representation of PSK signals by rectangular decomposition), and the characteristics of the received complex signals are more fully represented by rectangular decomposition under a rectangular coordinate system and ring decomposition under a polar coordinate system, which is beneficial to the accuracy of the identification and classification of the subsequent modulation formats.
2. Folding algorithm
Conventional modulation formatsThe recognition algorithm considers the signal distribution in the range of the complete constellation (for example, for IQ signals in rectangular coordinates, I is directly considered min ≤I≤I max ，Q min ≤Q≤Q max Complete constellation within the initial range). The problem with this approach is that the range of the feature map represented by the original constellation is too large and the key features such as the distance, angle, etc. of most of the constellation points to the origin are not emphasized, which increases the difficulty of the subsequent classification algorithm to learn the features and thus to identify the modulation format. Therefore, the patent considers that a folding algorithm is introduced, and the range of the feature map is further narrowed by folding the initial feature map, so that the key features are more prominent, and some redundant repeated features are eliminated. Meanwhile, the features reflected by the feature graphs after different folding times (called folding orders) are focused, and the purposes of eliminating feature redundancy and retaining outstanding features are respectively considered in different degrees.
Specifically, for a two-dimensional signal under a polar coordinate system, in consideration of symmetry of distribution of a feature map, we perform folding algorithms according to different angles respectively. Recording the angle at the current polar coordinateIn the range of->To->(where n is the current folding order, the unfolded time order is 0, i represents the ith data point), then the symmetry is used, the symmetry axis of the current signal distribution isThe folding operation then proceeds according to this symmetry axis:
of course, in view of symmetry, each fold operation is equivalent to folding to the left or right of the symmetry axis, and an equivalent fold operation to the other direction can be expressed as:
in the present invention, the angle in the initial polar coordinate isThe range is represented as-180 ° To +180 ° When the original characteristic diagram is folded, the original characteristic diagram is folded according to symmetry axis>Four folds were performed. Higher-order folding operation can be performed in principle, but in practical application, the folded characteristic diagram is too small, and part of useful information is lost, so that the accuracy of the subsequent modulation format recognition is affected. Meanwhile, the folding operation with fewer orders cannot sufficiently remove redundant information in the original feature map, and the local feature information cannot be emphasized. In practice, it is therefore appropriate to set the folding order to 3 or 4, and in different application scenarios, the specific folding order should be determined according to a trade-off between removing redundant information in the original feature map and highlighting local information of the folded feature map.
3. Pool calculation
In recent years, with the wide application of deep learning technology, the modulation format recognition technology based on deep learning utilizes the signal of the receiving end to directly send data into the neural network for end-to-end training, and finally directly output classification judgment. However, due to the complexity of the transmission channel, in order to fully learn the characteristics of the received signals under various modulation formats, the neural network needs to be designed into a deep complex architecture, which results in huge calculation cost and time overhead, and the neural network runs in some hardware platforms with limited calculation resources, so that the requirement of adaptability to real-time is often met. Therefore, the invention provides a classifying algorithm which is simple in structure and efficient in calculation and is based on reservoir calculation for identifying the modulation format.
Reservoir Computing (RC) is a special recurrent neural network architecture, whose network architecture comprises three parts: an input layer, an intermediate layer, and an output layer. The input layer comprises a plurality of nodes, each node corresponds to one input data, the middle layer belongs to one type of recurrent neural network, and the final output layer is an adder with weight. Accordingly, the pool calculation algorithm contains three important weights: the weight of the input layer (denoted as W in ) Connection weight between intermediate layer nodes (denoted as W res ) And the output weight (denoted as W) between the intermediate layer and the output layer out )。
The biggest characteristic of the classical pool computing network architecture is its input layer weight W in And weight W of intermediate layer res Are randomly generated and remain unchanged during the training process. For input sequence data { x } i I=1, 2, …, K }, and the intermediate layer has N nodes, and the state of each node is denoted as u (t), then the state updating mode of the intermediate layer node is as follows:
u(t)＝f(W in ·x(t)+W res ·u(t-1))#(5)
(wherein f represents a nonlinear activation function, enhancing the nonlinear fitting capability of the network)
In practice, to enhance the dynamic performance of reservoir calculations, a leakage constant may be introduced to dynamically update the middle layer state:
u(t)＝(1-α)·u(t-1)+αf(W in ·x(t)+W res ·u(t-1))#(6)
(where α is the leakage constant, the variation of which can affect the dynamic performance of reservoir calculations)
Finally, the input data enters the network through the input layer, then the linear separability is theoretically realized in the high-dimensional space through the dimension increasing operation of the middle layer, and the output layer trains an adder to carry out linear classification, so that the final classification result is obtained. Specifically, output { y } i I=1, 2, …, L } is derived from a linear weighted combination of intermediate layer node states u (t):
y(t)＝W out ·u(t)#(7)
wherein the output layer weight W out Is to minimize the least mean square error of the prediction result and the actual class:
(where ε is the regularization coefficient to prevent this optimization problem from overfitting)
Finally, the training modes of the output weight are many, for example, the training methods based on SVM, MLP and the like can be used for effective training, but in order to make the characteristics of simple and efficient structure of the reserve pool calculation self more outstanding, the scheme of the invention adopts a ridge regression method of the optimal solution in the least square sense, and the closed solution can be expressed as follows:
thus, for the entire pool computing network architecture, the weight of the input layer, W in And weight W of intermediate layer res Are randomly generated, do not need to be trained, and only need to be trained by the weight W of the output layer out And the method can be directly obtained through ridge regression closed-form solution, so that the portability and instantaneity of a modulation format recognition algorithm are fully ensured.
In addition, the invention further provides a deep research on the design mode of the pool computing network structure. Weight W for input layer in And weight W of intermediate layer res The random weight is generated by adopting random normal distribution or random uniform distribution, and better random weight can be obtained through multiple attempts in practice. Experiments prove that the performance of the pool calculation algorithm can be improved obviously when the nonlinear activation function is used compared with the non-linear activation function which is not used. For a plurality of nonlinear activation functions, three nonlinear activation functions of tan, sigmoid and ReLU are selected for comparison analysis:
tanh(x)＝(e x -e -x )/(e x +e -x )#(10)
sigmoid(x)＝1/(1+e -x )#(11)
ReLU(x)＝max(0,x)#(12)
experimental results show that the ReLU activation function performs slightly worse than the other two, due to its truncation to negative values. the tanh and sigmoid nonlinear activation functions have similar performance, and because the tanh and sigmoid nonlinear activation functions have similar shapes, the middle area changes quickly, and the two sides tend to be saturated, so that the nonlinear fitting capability in the algorithm is similar.
Experiments show that the algorithm performance can be improved to a certain extent along with the change of the scale of the middle layer, but the weak performance improvement can be slowed down and finally tends to be saturated, meanwhile, the calculation time consumption index level of the algorithm is increased along with the expansion of the scale of the middle layer, the scale of the middle layer of the storage pool is selected to be the most suitable about 500 in the weight of algorithm precision and time consumption cost, and the optimal accuracy rate at the moment reaches more than 90% under the condition of various emission voltages.
In order to make the purposes, technical schemes and advantages of the invention clearer, the following describes the embodiments of the invention in detail by combining the drawings and experimental results, the modulation recognition task of the underwater visible light communication based on the module comprises the following specific steps:
step 101: in an underwater visible light communication system based on QAM-CAP/APSK-CAP modulation, original data is firstly divided into six modulation formats, and the six modulation formats are up-sampled and then divided into IQ two paths of signals for multi-carrier transmission, wherein the number of each group of sub-carriers is 1024, and 128 data are transmitted by each sub-carrier. The data is sent by an LED at the transmitting end, received at the receiving end after passing through an underwater channel of 1.2m, and a complex value receiving signal is obtained after digital signal processing;
step 102: and decomposing the received complex value signals under a rectangular coordinate system and a polar coordinate system respectively through a coordinate transformation algorithm to obtain an IQ component under the rectangular coordinate system and a polar angle polar diameter component under the polar coordinate system respectively. Then, data under different coordinate systems are aggregated;
step 103: and carrying out folding operation on the feature graphs corresponding to the data under different coordinate systems through a folding algorithm, so as to remove redundant information and highlight salient features. The optimal folding order in the experiment is about 3 or 4 times;
step 104: 128 data of each group of two paths are flattened to obtain input data with the length of 256. Each set of input data is in turn input to the pool computing network. The input weight between the input layer and the middle layer is randomly generated, and a random matrix with 256 multiplied by 500 is input assuming that the node number of the middle layer is 500;
step 105: by randomly generating 500×500 intermediate layer connection weights, selecting an appropriate activation function (such as tanh), and updating the state of the internal node at each time step in combination with the input data transformed by the input weights.
Step 106: generating an output weight of an optimal solution in the least square sense by a ridge regression method, linearly combining node states of the middle layer, obtaining the probability that input data belong to each modulation format by a softmax function, and selecting a category corresponding to the maximum probability, namely the classification judgment of the final modulation format.
According to experimental results, fig. 3 and fig. 4 show transformation result diagrams of coordinate transformation and folding algorithm on complex value information received in an underwater visible light communication system, and the two feature enhancement algorithms are remarkably prominent in local features, and feature redundancy is reduced, so that the high efficiency and accuracy of the algorithm are improved.
Under the condition of calculation and super-selection of a proper reserve pool, fig. 5 is a study of the identification accuracy of the debugging format according to the invention along with the change of the emitting voltage of the emitting end LED. When the voltage of the LED emitting terminal is changed from 0.1V to 1.3V, the optimal accuracy is almost over 90 percent, and the highest accuracy is close to 100 percent. Meanwhile, the inset of fig. 5 shows the algorithm accuracy as a function of reservoir interlayer size for a transmit voltage of 0.3V. In the experiment, the size of the storage pool is about 500, the accuracy of more than 98% can be achieved, meanwhile, the algorithm calculation can be completed within a few seconds, and the calculation cost and the time consumption of the modulation format recognition algorithm are greatly reduced.
Finally, fig. 6 shows the transformation study of the modulation format recognition accuracy of the algorithm along with the super parameters of the nonlinear activation function, the scale of the middle layer of the reserve tank, the leakage coefficient and the like. The influence of the selection of different super parameters on the accuracy of the algorithm can be up to 10%, so that the proper super parameter selection can further enhance the performance of the algorithm in the actual modulation format recognition task.
In the invention, parameters such as a generation mode of random weight, selection of a nonlinear activation function, scale of intermediate layer nodes, number of folding operations and the like can be adjusted according to practical application so as to obtain optimal performance.
In summary, the invention proposes to combine the reserve pool calculation and the feature enhancement algorithm to perform the modulation format recognition task in the underwater visible light communication system. Through the reserve pool computing network, the computing overhead and the training time consumption of the algorithm can be greatly reduced, and the instantaneity of the algorithm is improved. Meanwhile, by introducing a coordinate transformation and folding algorithm, the original input data can be subjected to efficient data enhancement, the local significant features are highlighted, the feature redundancy is reduced, and the global features and the local features are combined, so that the algorithm performance and the robustness of the invention are further improved.
The division of the steps in the present invention is only for the purpose of clearly explaining the principle, and some steps may be combined or split when implemented, so long as the similar implementation principle and logic relationship are included within the protection scope of the present patent.
It will be understood by those of ordinary skill in the art that the foregoing embodiments are specific examples of carrying out the invention and that various changes in form and details may be made therein without departing from the spirit and scope of the invention. For example, the order of the folding operation in the folding algorithm, the generation mode of random weights in the storage pool calculation network structure, the scale of the middle layer, the selection of the nonlinear activation function and the like can be adjusted according to actual needs so as to obtain the optimal effect.
Claims (5)
1. The method for identifying the modulation format of the underwater visible light communication system based on reservoir calculation is characterized by comprising the following steps of:
step 1): in an underwater visible light communication system, data sent by an LED at a transmitting end are received at a receiving end after passing through an underwater channel, and complex-value receiving signals are obtained after digital signal processing;
step 2): decomposing the received complex value signals under a rectangular coordinate system and a polar coordinate system respectively through a coordinate transformation algorithm to obtain an IQ component under the rectangular coordinate system and a polar angle polar path component under the polar coordinate system respectively, and then splicing data under the rectangular coordinate system and the polar coordinate system to realize aggregation of data characteristics under different coordinate systems;
step 3): carrying out folding operation on the feature graphs corresponding to the data under the rectangular coordinate system and the polar coordinate system by utilizing symmetry according to a symmetry axis through a folding algorithm, so as to reduce the range of the feature graphs and realize the removal of redundant information and the salient of significant features;
step 4): sending the data of the folded feature map obtained in the step 3) into a reserve pool computing RC network, generating an output weight of an optimal solution in the least square sense through a ridge regression method, linearly combining node states of the middle layer, obtaining probability that input data belong to each modulation format through a softmax function, and selecting a category corresponding to the maximum probability to obtain a classification decision of the modulation format of the final underwater visible light communication system.
2. The method for identifying modulation format of underwater visible light communication system according to claim 1, wherein in step 3), folding algorithms are performed according to different angles for two-dimensional signals under polar coordinate system in consideration of symmetry of distribution of feature map; recording the angle at the current polar coordinateIn the range of->To->Wherein n is the current folding order, the unfolded time order is 0, i represents the ith data point, the symmetry axis of the current signal distribution is +.>The symmetry axis is always the average value of the current angle range, and then the folding operation folds according to the symmetry axis:
in view of symmetry, each folding operation is equivalent to folding to the left or right of the symmetry axis, the equivalent folding operation to the other direction being expressed as:
3. the method for identifying modulation format of underwater visible light communication system according to claim 1, wherein in step 3), the folding order is 3-4 times, and in step 4), the number of nodes in the middle layer is 400-800 in the RC network calculated by the reserve pool.
4. The method for identifying modulation format of an underwater visible light communication system according to claim 1, wherein in step 4), a nonlinear activation function is adopted in the reservoir calculation RC network, and the nonlinear function is tanh, sigmoid or ReLU.
5. The method for identifying the modulation format of the underwater visible light communication system according to claim 1, wherein the modulation format of the underwater visible light communication system comprises OOK,4QAM,8QAM-DIA,8QAM-CIR,16APSK and 16QAM.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310450808.9A CN116708104A (en) | 2023-04-25 | 2023-04-25 | Modulation format identification method of underwater visible light communication system based on reserve pool calculation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310450808.9A CN116708104A (en) | 2023-04-25 | 2023-04-25 | Modulation format identification method of underwater visible light communication system based on reserve pool calculation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116708104A true CN116708104A (en) | 2023-09-05 |
Family
ID=87840018
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310450808.9A Pending CN116708104A (en) | 2023-04-25 | 2023-04-25 | Modulation format identification method of underwater visible light communication system based on reserve pool calculation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116708104A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117807529A (en) * | 2024-02-29 | 2024-04-02 | 南京工业大学 | Modulation mode identification method and system for output signals of signal generator |
-
2023
- 2023-04-25 CN CN202310450808.9A patent/CN116708104A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117807529A (en) * | 2024-02-29 | 2024-04-02 | 南京工业大学 | Modulation mode identification method and system for output signals of signal generator |
| CN117807529B (en) * | 2024-02-29 | 2024-05-07 | 南京工业大学 | Modulation mode identification method and system for output signals of signal generator |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Peng et al. | Modulation classification using convolutional neural network based deep learning model | |
| Jdid et al. | Machine learning based automatic modulation recognition for wireless communications: A comprehensive survey | |
| Meng et al. | Automatic modulation classification: A deep learning enabled approach | |
| CN108234370B (en) | Communication signal modulation mode identification method based on convolutional neural network | |
| Weng et al. | Deep cascading network architecture for robust automatic modulation classification | |
| CN110113288B (en) | Design and demodulation method of OFDM demodulator based on machine learning | |
| Zhang et al. | Lightweight automatic modulation classification via progressive differentiable architecture search | |
| CN116708104A (en) | Modulation format identification method of underwater visible light communication system based on reserve pool calculation | |
| Wei et al. | Automatic modulation recognition using neural architecture search | |
| Huang et al. | Identifying physical-layer attacks for IoT security: An automatic modulation classification approach using multi-module fusion neural network | |
| Al‐Makhlasawy et al. | Modulation classification in the presence of adjacent channel interference using convolutional neural networks | |
| Ke et al. | Application of adversarial examples in communication modulation classification | |
| Ali et al. | Algorithm for automatic recognition of PSK and QAM with unique classifier based on features and threshold levels | |
| Bouchenak et al. | A semi-supervised modulation identification in MIMO systems: A deep learning strategy | |
| CN116628566A (en) | Communication signal modulation classification method based on aggregated residual transformation network | |
| Chen et al. | Automatic modulation classification using multi-scale convolutional neural network | |
| Kumar et al. | Automatic modulation classification for adaptive OFDM systems using convolutional neural networks with residual learning | |
| Wu et al. | Deep multi-scale representation learning with attention for automatic modulation classification | |
| Zhang et al. | Heterogeneous deep model fusion for automatic modulation classification | |
| Zhou et al. | Automatic modulation classification with genetic backpropagation neural network | |
| Wang et al. | Enhancing the generalization ability of deep learning model for radio signal modulation recognition | |
| Hussein et al. | Automatic Modulation Classification: Convolutional Deep Learning Neural Networks Approaches | |
| Sun et al. | Survey of Research on Application of Deep Learning in Modulation Recognition | |
| Xu et al. | PSK/QAM modulation recognition by convolutional neural network | |
| Shi et al. | STTMC: A Few-shot Spatial Temporal Transductive Modulation Classifier |
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 |