CN112737694A - Non-uniform quantization system of filter multi-carrier modulation optical communication system based on SOM - Google Patents

Non-uniform quantization system of filter multi-carrier modulation optical communication system based on SOM Download PDF

Info

Publication number
CN112737694A
CN112737694A CN202011580479.2A CN202011580479A CN112737694A CN 112737694 A CN112737694 A CN 112737694A CN 202011580479 A CN202011580479 A CN 202011580479A CN 112737694 A CN112737694 A CN 112737694A
Authority
CN
China
Prior art keywords
som
quantization
training
module
neuron
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
Application number
CN202011580479.2A
Other languages
Chinese (zh)
Other versions
CN112737694B (en
Inventor
林嘉芊
毕美华
卢旸
杨国伟
周雪芳
胡淼
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Hangzhou Dianzi University
Original Assignee
Hangzhou Dianzi University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Hangzhou Dianzi University filed Critical Hangzhou Dianzi University
Priority to CN202011580479.2A priority Critical patent/CN112737694B/en
Publication of CN112737694A publication Critical patent/CN112737694A/en
Application granted granted Critical
Publication of CN112737694B publication Critical patent/CN112737694B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • 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
    • H04B10/50—Transmitters
    • H04B10/501—Structural aspects
    • 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
    • H04B10/50—Transmitters
    • H04B10/516—Details of coding or modulation
    • 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
    • H04B10/60—Receivers
    • H04B10/66—Non-coherent receivers, e.g. using direct detection
    • 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
    • H04B10/60—Receivers
    • H04B10/66—Non-coherent receivers, e.g. using direct detection
    • H04B10/69—Electrical arrangements in the receiver
    • H04B10/691—Arrangements for optimizing the photodetector in the receiver

Landscapes

  • Physics & Mathematics (AREA)
  • Electromagnetism (AREA)
  • Engineering & Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Optical Communication System (AREA)

Abstract

The invention discloses a non-uniform quantization system of a filter multi-carrier modulation optical communication system based on SOM, which comprises an SOM training module and an SOM quantization module. In a pre-training mode, the SOM training module is used for training an SOM model of a filter multi-carrier modulation signal, the module initializes model parameters and a training set, iterates to convergence through operations such as normalization, competitive learning and weight adjustment, and inputs an obtained SOM model weight as a quantization level into the SOM quantization module; the quantization module does not operate. In a working mode, the SOM training module is used for learning signal characteristics on the basis of an original SOM model and adjusting quantization level in real time; and the SOM quantization module quantizes the input data sequence according to the quantization level. The SOM non-uniform quantization algorithm reduces the system quantization error, improves the system error rate performance and saves the cost.

Description

Non-uniform quantization system of filter multi-carrier modulation optical communication system based on SOM
Technical Field
The invention belongs to the technical field of optical communication, and particularly relates to a non-uniform quantization system of a filter multi-carrier modulation optical communication system based on SOM.
Background
In recent years, with the emergence of emerging technologies such as machine learning, social networking, cloud computing and the like, the demand of people on the size and the variety of access bandwidth is also increasing year by year, which requires that a future transmission network system can meet diversified requirements such as high capacity, asynchronous transmission and high spectrum efficiency to a certain extent. In order to implement 5G large capacity data access service, wireless forward access based on optical fiber link bearer is proposed and becomes a hotspot of current access network technology research, in a 5G access bearer scheme based on optical fiber network, an optical transmission system based on IMDD Multi-Carrier Modulation is widely concerned due to the comprehensive advantages of IMDD and Multi-Carrier technology, in the IMDD multi-carrier modulation optical transmission system, a low-resolution Digital-to-Analog converter (DAC) cannot meet the performance requirement of the system, and a high-resolution DAC greatly increases the power consumption and cost of the system. Therefore, in order to balance the performance and cost of the system, a non-uniform quantization method for a low-quantization bit number DAC is required to improve the transmission performance of the system and reduce the system cost.
Common quantization methods can be classified into uniform quantization and non-uniform quantization. Among them, uniform quantization is widely used due to its simple implementation, whereas in a filter multicarrier modulation communication system, signals are generally not uniformly distributed. The uniform quantization method cannot adaptively adjust the quantization level according to the distribution of the signal to be quantized, which may cause large quantization noise and decrease of system transmission performance. Therefore, a quantization method based on signal distribution is a very promising solution. In a conventional OFDM system, for example, Jizong Peng published in 2017 "SQNR Improvement Enabled by non-noise OFDM DAC Output Levels for IM-DD OFDM Systems" in IEEE Photonics Journal, based on the fact that OFDM signal distribution conforms to gaussian distribution, it is proposed to use a non-uniform quantization method based on gaussian distribution to improve the transmission performance of the IMDD-OFDM system. However, in the IMDD filter bank multi-carrier system, the waveform distribution of the signal after being processed by the filter is not a standard gaussian distribution, and a signal distribution curve has a peak, so that a certain optimization space exists in the non-uniform quantization scheme based on the assumption of the gaussian distribution. A non-uniform quantization method using non-parametric Histogram Estimation (NPHE) is proposed in "Performance Optimization by Nonparametric Histogram Estimation for Low Resolution in IMDD-OQAM-OFDM System" published in IEEE Photonics Journal. The method comprises the steps of firstly, carrying out estimation fitting on a probability density curve of an OQAM-OFDM signal by using an NPHE (nonlinear programming), and then, calculating the optimal quantization level combination under the minimum quantization noise through linear programming. Although the quantization performance of the method is improved to a certain extent compared with the quantization method based on Gaussian distribution. However, the method still has the problems that the number of samples used for fitting is huge, and the computational complexity and the storage complexity are increased sharply along with the increase of the quantization digit. Therefore, there is still room for improvement in the non-uniform quantization method based on signal distribution.
Disclosure of Invention
In view of the defects in the prior art, the present invention provides a non-uniform quantization system of a filter multi-carrier modulation optical communication system based on SOM to improve the quantization performance of DAC.
In order to achieve the purpose, the invention adopts the following technical scheme:
the non-uniform quantization system of the filter multi-carrier modulation optical communication system based on the SOM comprises an SOM training module and an SOM quantization module;
the SOM training module is used for carrying out data preprocessing on a data sequence sent by the filter multi-carrier modulation transmitter to obtain a training set, inputting the training set into an SOM model for training, and inputting the SOM model neuron weight obtained by training into the SOM quantization module as a quantization level;
and the SOM quantization module is used for quantizing the input data sequence according to the quantization level.
Preferably, the SOM training module has two working modes:
the first mode is a pre-training mode, and an SOM model is pre-trained by inputting a training set;
and the second mode is a working mode, the characteristics of the signals are learned in the process of quantizing by using the trained SOM model, and the weight is adjusted in real time to adjust the quantization level.
Preferably, the training steps in the two working modes are as follows:
step 1: initializing a training set, inputting a signal sent by a filter multi-carrier modulation transmitter into a data preprocessing module, and taking a processed data sequence as the input of an SOM model training module;
step 2: initializing an SOM training module, and setting a topological structure and training parameters of a model according to a training set and a target quantization digit;
and step 3: normalizing the data sequence;
and 4, step 4: calculating winning neurons, inputting training set data into a competition layer, calculating the similarity between each neuron and input data, wherein the neuron with the highest similarity is the winning neuron;
and 5: defining a win region according to the win neuron, and adjusting the weight of nodes in the win region;
step 6: and (5) circularly executing the step (3-5) until the SOM converges, and inputting the SOM neuron weight obtained by model convergence into the SOM quantization module as a quantization level.
Preferably, in step 1, based on the known characteristic that the real part signal distribution and the imaginary part signal distribution in the filter multicarrier signal are the same, the training set is formed by real part signal data of a signal transmitted by the filter multicarrier transmitter, and the training set represents a matrix X with a dimension of 1 × N1×N。
As a preferred scheme, in the step 2, the initialization parameters of the SOM model include two parts, namely model parameters and training parameters, the model parameters include a competitive layer topology structure of the SOM and a neuron weight W, and the training parameters include a learning rate α (t) and a win-win domain range function r (t);
the dimension of the competition layer depends on the dimension of the training set and the quantization bit number N, and the dimension of the competition layer is set to be N × m, wherein m is 2n；
In the initialization of the weight of the neuron, m sample data in a training set which is randomly extracted is used as an initial value w of the neuronj(j ═ 1,2, …, m); in the setting of the learning rate and the winning field, the values of the learning rate alpha (t) and the winning field r (t) are gradually reduced along with the increase of the training iteration number t;
the attenuation function is:
Figure BDA0002864304640000031
wherein T is the maximum iteration number which is the same as the number of data in the training set;
the learning rate function is:
α(t)＝α(0)/(1+t/T)
wherein, alpha (t) is the learning rate of the t-th round, and alpha (0) is the initial learning rate;
the winning node has the largest weight adjustment amount, the adjacent nodes have relatively small adjustment amounts, and the larger the distance from the winning node, the smaller the weight adjustment amount; r (t) is represented by:
r(t)＝r(0)/(1+t/T)
where r (t) is the winning domain range of the t-th round, and r (0) is the initial value of the winning domain range.
Preferably, in step 3, the normalization method is a Z-zero mean normalization method, and is expressed as (x- μ)/σ, where μ and σ are the mean and standard deviation of the input training set, respectively, and x is the input data sequence.
Preferably, in step 4, the similarity is an euclidean distance, and the formula is as follows:
Figure BDA0002864304640000041
where x is the input data value, wjThe weight of the jth neuron.
As a preferred scheme, in step 5, the weight value adjustment formula is:
wij(t+1)＝wij(t)+α(t)[xi-wij(t)]
i＝1,2,…,N,j∈Nj*(t)
wherein, wijThe weight of the jth neuron when the ith data is input is represented by j, and N is the winning neuronj*(t) represents the nodes that are encompassed by the current winning domain.
Preferably, in step 6, the convergence condition is that the number of iterations reaches a predetermined number of iterations T.
As a preferred scheme, the SOM quantization model is configured to calculate a similarity between input sample data and a weight, and output a quantization level with the maximum similarity as a quantization result;
the similarity dijFor the euclidean distance between the signal and the quantization level, the quantization process is expressed as:
Figure BDA0002864304640000042
wherein x isqiAnd selecting the quantization level with the minimum Euclidean distance as the quantization output value of the ith data.
Compared with the prior art, the invention has the following beneficial effects:
1. the non-uniform quantization system fully considers the characteristics of a high-speed filter multi-carrier system, and can effectively improve the quantization performance of the DAC, so that the performance of the filter multi-carrier modulation optical communication system is improved.
2. Compared with a nonlinear programming quantization method based on signal estimation, the non-uniform quantization system can obtain better signal estimation effect and quantization effect, and is simpler in algorithm and lower in complexity.
3. Compared with other non-uniform quantization methods based on clustering, the non-uniform quantization system can better adapt to signal distribution change.
Drawings
Fig. 1 is a schematic diagram of a non-uniform quantization system of a filter multi-carrier modulation optical communication system based on SOM according to an embodiment of the present invention;
fig. 2 is a schematic diagram of a filter multi-carrier modulation optical communication system according to an embodiment of the present invention.
Fig. 3 is a BER comparison graph of the high-speed optical generic filter multi-carrier system provided by the embodiment of the invention, which uses a 3-5-bit SOM-based non-uniform quantization method and uses a conventional uniform quantization method.
Detailed Description
The embodiments of the present invention are described below with reference to specific embodiments, and other advantages and effects of the present invention will be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention. It is to be noted that the features in the following embodiments and examples may be combined with each other without conflict.
The non-uniform quantization system of the filter multi-carrier modulation optical communication system based on the SOM comprises an SOM training model and an SOM quantization model, wherein the SOM training model preprocesses input data to generate a training set, the training set is trained to obtain the SOM model, and the weight of the obtained SOM model is used as a quantization level to be input into an SOM quantization module. The SOM quantization model quantizes the input signal, and the SOM training model learns the characteristics of the input signal and adjusts the quantization level in real time.
As shown in fig. 1, the non-uniform quantization system of the SOM-based filter multi-carrier modulation optical communication system according to the embodiment of the present invention includes a SOM training module M1, a SOM quantization module M2;
the SOM training module M1 is used for carrying out data preprocessing on a data sequence sent by the filter multi-carrier modulation transmitter to obtain a training set, and inputting the training set into an SOM model for training; training the obtained SOM model neuron weight as a quantization level to input into an SOM quantization module;
and the SOM quantization module M2 is configured to quantize the input data sequence according to the quantization level obtained by the SOM training module, and output a quantization result.
In the SOM training module M1, data preprocessing is performed on the data sequence sent by the filter multicarrier modulation transmitter to obtain a training set, and a quantization level obtained by training is input to the SOM quantization module. The training steps are as follows:
step 1: initializing a training set, inputting a signal sent by a filter multi-carrier modulation transmitter into a data preprocessing module, and taking a processed data sequence as the input of an SOM model training module; based on the characteristic that the real part signal distribution and the imaginary part signal distribution in the known filter bank multicarrier signal are the same, the training set is formed by real part signal data of a signal transmitted by the filter multicarrier transmitter. The training set may represent a matrix X with dimensions of 1N1×N。
Step 2: and initializing the SOM training module, and setting the topological structure and the training parameters of the model according to the training set and the target quantization digit. The initialization parameters of the SOM model comprise two parts, namely model parameters and training parameters, the model parameters comprise a competitive layer topological structure and neuron weight W of the SOM, and the training parameters comprise a learning rate alpha (t) and a win field range function r (t). The dimension of the competition layer depends on the dimension of the training set and the quantization bitN, the dimension of the competition layer is set to be N × m, wherein m is 2n. In the initialization of the weight of the neuron, m sample data in a training set which is randomly extracted is used as an initial value w of the neuronj(j ═ 1,2, …, m). In the setting of the learning rate and the winning area, the values of the learning rate α (t) and the winning area r (t) are gradually decreased as the training iteration number t increases. The most commonly used attenuation function is:
Figure BDA0002864304640000061
wherein T is the maximum iteration number and is the same as the number of data in the training set.
The learning rate function is:
α(t)＝α(0)/(1+t/T)
where α (t) is the learning rate of the t-th round, and α (0) is the initial learning rate.
The winning area range function R (t) is a area radius R set by taking a winning neuron as a center, and has the function of adjusting the weight of the neuron around the winning neuron, the winning node has the largest weight adjustment amount, the adjacent nodes have slightly smaller adjustment amounts, and the larger the distance from the winning node is, the smaller the weight adjustment amount is. r (t) can be expressed as:
r(t)＝r(0)/(1+t/T)
where r (t) is the winning domain range of the t-th round, and r (0) is the initial value of the winning domain range.
And step 3: and (5) carrying out data sequence normalization processing. The normalization method may select a 0-mean normalization method (Z-zero normalization), which may be expressed as (x- μ)/σ, where μ and σ are the mean and standard deviation, respectively, of the input training set.
And 4, step 4: and (4) calculating a winning neuron, inputting training set data into a competition layer, and calculating the similarity between each neuron and input data, wherein the neuron with the maximum similarity is the winning neuron. Wherein, the similarity is Euclidean distance, and the formula is as follows:
Figure BDA0002864304640000062
where x is the input data value, wjThe weight of the jth neuron.
And 5: and defining a win region according to the win neuron, and adjusting the weight of nodes in the win region. The weight value adjustment formula is as follows:
wij(t+1)＝wij(t)+α(t)[xi-wij(t)]
i＝1,2,…,N,j∈Nj*(t)
wherein, wijThe weight of the jth neuron when the ith data is input is represented by j, and N is the winning neuronj*(t) represents the nodes that are encompassed by the current winning domain.
Step 6: and (5) circularly executing the step (3-5) until the SOM converges, and inputting the SOM weight obtained by model convergence into the SOM quantization module as a quantization level. The convergence condition is that the number of iterations reaches a predetermined number of iterations T.
In the SOM quantization model M2, the similarity between the input sample data and the weight is calculated, and the full quantization level with the maximum similarity is output as the quantization result.
Degree of similarity dijIs the euclidean distance between the signal and the weight. The quantization process is represented as:
Figure BDA0002864304640000071
wherein x isqiAnd selecting the quantization level with the minimum Euclidean distance as the quantization output value of the ith data.
The quantization system of the embodiment fully considers the characteristics of the high-speed optical filter multi-carrier system, can effectively improve the quantization performance of the DAC, and further has a positive effect on the transmission performance of the system.
The filter multicarrier modulation optical communication system of the present example, as shown in fig. 2, includes: the optical transmission module inputs digital signals containing data information into the SOM-based quantization module to obtain analog signals, the analog signals are converted into high-speed optical signals through the optical modulator and sent to the optical fiber channel, the optical receiving module converts the optical signals into corresponding electric signals, and information data are obtained through demodulation.
The light emitting module includes: the system comprises a digital signal module, a baseband modulation module, an SOM-based non-uniform quantization module and an optical modulator;
a digital signal module: carrying out coding and mapping processing on an input data sequence and generating a high-speed digital electric signal to be transmitted;
a baseband modulation module: the non-uniform quantization module is connected with the digital signal module and outputs a digital signal to the SOM-based non-uniform quantization module;
and the non-uniform quantization module based on the SOM is connected with the baseband modulation model, splits the digital signal into a real part sample sequence and an imaginary part sample sequence, inputs the real part sample sequence into the SOM model for pre-training, iterates to obtain a quantization level and quantizes the input signal. Meanwhile, the module also learns the characteristics of the newly input signal and adjusts the quantization level of the signal in real time;
and the optical modulator is connected with the non-uniform quantization module based on the SOM and used for completing the electro-optical conversion of the signal which is taken as the quantized output through the optical modulator.
In an embodiment of the present invention, a light receiving module includes: the device comprises a photoelectric detector, a real-time oscilloscope, a baseband demodulation module and a data output unit; after the photoelectric detector samples the received electric signal through the real-time oscilloscope, the output signal of the real-time oscilloscope is processed and output through the baseband demodulation module, and the receiving of user data is achieved.
The embodiment of the invention verifies that the parameters of the SOM non-uniform quantization method are as follows: the signal is modulated by UFMC, and the length of input digital signal sequence is 524288 bit. The uniform quantization method inputs a digital signal sequence length of 524288.
As shown in fig. 3, which is a graph comparing BER performance of the system respectively subjected to uniform quantization of 3-5 bits and SOM non-uniform quantization in the high-speed optical general filter bank multi-carrier system, in the graph: the horizontal axis represents the received optical power, and the vertical axis represents the BER. It can be seen that the non-uniform quantization method of the present invention achieves better quantization performance than the conventional uniform quantization method with the same quantization bit number.
In summary, the influence factors of the small-amplitude signal in signal quantization can be better considered by adopting the non-uniform quantization algorithm based on the SOM. Compared with uniform quantization DAC, the cost is lower, and the quantization effect is better. Meanwhile, the SOM can adjust the quantization level in real time and can better adapt to the changed signal distribution. Therefore, the algorithm of the invention can be better applied to the requirements of the filter multi-carrier optical communication system.
The specific embodiments described herein are merely illustrative of the spirit of the invention. Various modifications or additions may be made to the described embodiments or alternatives may be employed by those skilled in the art without departing from the spirit or ambit of the invention as defined in the appended claims.

Claims (10)

1. The non-uniform quantization system of the filter multi-carrier modulation optical communication system based on the SOM is characterized by comprising an SOM training module and an SOM quantization module;
the SOM training module is used for carrying out data preprocessing on a data sequence sent by the filter multi-carrier modulation transmitter to obtain a training set, inputting the training set into an SOM model for training, and inputting the SOM model neuron weight obtained by training into the SOM quantization module as a quantization level;
and the SOM quantization module is used for quantizing the input data sequence according to the quantization level.
2. The non-uniform quantization system of an SOM-based filter multi-carrier modulated optical communication system of claim 1, wherein the SOM training module has two modes of operation:
the first mode is a pre-training mode, and an SOM model is pre-trained by inputting a training set;
and the second mode is a working mode, the characteristics of the signals are learned in the process of quantizing by using the SOM obtained by training, and the weight is adjusted in real time to adjust the quantization level.
3. The non-uniform quantization system of an SOM-based filter multi-carrier modulated optical communication system of claim 2, wherein the training steps in the two modes of operation are as follows:
step 1: initializing a training set, inputting a signal sent by a filter multi-carrier modulation transmitter into a data preprocessing module, and taking a processed data sequence as the input of an SOM model training module;
step 2: initializing an SOM training module, and setting a topological structure and training parameters of a model according to a training set and a target quantization digit;
and step 3: normalizing the data sequence;
and 4, step 4: calculating winning neurons, inputting training set data into a competition layer, calculating the similarity between each neuron and input data, wherein the neuron with the highest similarity is the winning neuron;
and 5: defining a win region according to the win neuron, and adjusting the weight of nodes in the win region;
step 6: and (5) circularly executing the step (3-5) until the SOM converges, and inputting the SOM neuron weight obtained by model convergence into the SOM quantization module as a quantization level.
4. The non-uniform quantization system of an SOM-based filtered multi-carrier modulated optical communication system of claim 3, wherein in step 1, the training set is formed by real signal data of the filtered multi-carrier transmitter transmission signal based on the known characteristics of real signal distribution and imaginary signal distribution in the filtered multi-carrier signal, the training set representing a matrix X with dimension of 1X N1×N。
5. The non-uniform quantization system of the SOM-based filter multi-carrier modulation optical communication system of claim 4, wherein in the step 2, the SOM model initialization parameters comprise two parts, namely model parameters and training parameters, the model parameters comprise a competition layer topology and neuron weights W of the SOM, and the training parameters comprise a learning rate α (t) and a dominance area range function r (t);
the dimension of the competition layer depends on the dimension of the training set and the quantization bit number N, and the dimension of the competition layer is set to be N × m, wherein m is 2n；
In the initialization of the weight of the neuron, m sample data in a training set which is randomly extracted is used as an initial value w of the neuronj(j ═ 1,2, …, m); in the setting of the learning rate and the winning field, the values of the learning rate alpha (t) and the winning field r (t) are gradually reduced along with the increase of the training iteration number t;
the attenuation function is:
Figure FDA0002864304630000021
wherein T is the maximum iteration number which is the same as the number of data in the training set;
the learning rate function is:
α(t)＝α(0)/(1+t/T)
wherein, alpha (t) is the learning rate of the t-th round, and alpha (0) is the initial learning rate;
the winning node has the largest weight adjustment amount, the adjacent nodes have relatively small adjustment amounts, and the larger the distance from the winning node, the smaller the weight adjustment amount; r (t) is represented by:
r(t)＝r(0)/(1+t/T)
where r (t) is the winning domain range of the t-th round, and r (0) is the initial value of the winning domain range.
6. The non-uniform quantization system of the SOM-based filter multi-carrier modulation optical communication system of claim 5, wherein in the step 3, the normalization method is a Z-zero mean normalization method, which is expressed as (x- μ)/σ, where μ and σ are the mean and standard deviation of the input training set, respectively, and x is the input data sequence.
7. The non-uniform quantization system of the SOM-based filter multi-carrier modulation optical communication system of claim 6, wherein in the step 4, the similarity is euclidean distance, which is expressed as:
Figure FDA0002864304630000022
where x is the input data value, wjThe weight of the jth neuron.
8. The non-uniform quantization system of the SOM-based filter multi-carrier modulation optical communication system of claim 7, wherein in the step 5, the weight adjustment formula is:
wij(t+1)＝wij(t)+α(t)[xi-wij(t)]
Figure FDA0002864304630000023
wherein, wijThe weight of the jth neuron when the ith data is input is shown as j, the winning neuron is shown as j,
Figure FDA0002864304630000024
representing the nodes that are encompassed by the current prevailing domain of interest.
9. The non-uniform quantization system of the SOM-based filter multi-carrier modulated optical communication system of claim 8, wherein the convergence condition in step 6 is that the number of iterations reaches a predetermined number of iterations T.
10. The non-uniform quantization system of the SOM-based filter multi-carrier modulation optical communication system of claim 9, wherein the SOM quantization model is configured to calculate similarities between input sample data and weights, and output a full quantization level with the largest similarity as a quantization result;
the similarity dijFor the euclidean distance between the signal and the quantization level, the quantization process is expressed as:
Figure FDA0002864304630000031
wherein x isqiAnd selecting the quantization level with the minimum Euclidean distance as the quantization output value of the ith data.
CN202011580479.2A 2020-12-28 2020-12-28 Non-uniform quantization system of filter multi-carrier modulation optical communication system based on SOM Active CN112737694B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011580479.2A CN112737694B (en) 2020-12-28 2020-12-28 Non-uniform quantization system of filter multi-carrier modulation optical communication system based on SOM

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011580479.2A CN112737694B (en) 2020-12-28 2020-12-28 Non-uniform quantization system of filter multi-carrier modulation optical communication system based on SOM

Publications (2)

Publication Number Publication Date
CN112737694A true CN112737694A (en) 2021-04-30
CN112737694B CN112737694B (en) 2021-11-16

Family

ID=75606528

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011580479.2A Active CN112737694B (en) 2020-12-28 2020-12-28 Non-uniform quantization system of filter multi-carrier modulation optical communication system based on SOM

Country Status (1)

Country Link
CN (1) CN112737694B (en)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105144598A (en) * 2012-12-03 2015-12-09 数字无线功率有限公司 Systems and methods for advanced iterative decoding and channel estimation of concatenated coding systems
US20190089442A1 (en) * 2017-09-15 2019-03-21 At&T Intellectual Property I, L.P. Two reference signal beam reporting and identification
CN110020695A (en) * 2019-04-19 2019-07-16 杭州电子科技大学 K-means non-uniform quantizing algorithm for filter bank multi-carrier modulation optical communication system
US10470077B1 (en) * 2018-06-08 2019-11-05 At&T Intellectual Property I, L.P. Messaging centers with rule based adaptive retry profiles for 5G or other next generation network
CN111756450A (en) * 2019-03-27 2020-10-09 中山大学 Vector mode multiplexing system based on discrete multi-carrier modulation technology

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105144598A (en) * 2012-12-03 2015-12-09 数字无线功率有限公司 Systems and methods for advanced iterative decoding and channel estimation of concatenated coding systems
US20190089442A1 (en) * 2017-09-15 2019-03-21 At&T Intellectual Property I, L.P. Two reference signal beam reporting and identification
US10470077B1 (en) * 2018-06-08 2019-11-05 At&T Intellectual Property I, L.P. Messaging centers with rule based adaptive retry profiles for 5G or other next generation network
CN111756450A (en) * 2019-03-27 2020-10-09 中山大学 Vector mode multiplexing system based on discrete multi-carrier modulation technology
CN110020695A (en) * 2019-04-19 2019-07-16 杭州电子科技大学 K-means non-uniform quantizing algorithm for filter bank multi-carrier modulation optical communication system

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
JIAQIAN LIN等: ""A study on performance improvement of IMDD-UFMC with modified"", 《OPTICS COMMUNICATIONS》 *

Also Published As

Publication number Publication date
CN112737694B (en) 2021-11-16

Similar Documents

Publication Publication Date Title
CN108462517B (en) MIMO link self-adaptive transmission method based on machine learning
CN110120926B (en) Communication signal modulation mode identification method based on evolution BP neural network
CN112637093B (en) Signal detection method based on model-driven deep learning
CN112910562B (en) Communication method based on probability shaping
WO2019071947A1 (en) Optical fiber nonlinearity equalization method for use in 64-qam coherent light communications system
CN111935042B (en) Probability shaping recognition system and method based on machine learning and receiving end
CN112235023B (en) MIMO-SCFDE self-adaptive transmission method based on model-driven deep learning
CN112865866B (en) Visible light PAM system nonlinear compensation method based on GSN
Balevi et al. Autoencoder-based error correction coding for one-bit quantization
CN109474388B (en) Low-complexity MIMO-NOMA system signal detection method based on improved gradient projection method
CN114567528B (en) Communication signal modulation mode open set recognition method and system based on deep learning
CN113473580B (en) User association joint power distribution method based on deep learning in heterogeneous network
CN110830124A (en) Spectrum sensing method based on quantum particle swarm optimization extreme learning machine
CN113038612B (en) Cognitive radio power control method based on deep learning
CN112153617B (en) Terminal equipment transmission power control method based on integrated neural network
CN111313971B (en) Lightgbm equalization system and method for improving IMDD short-distance optical communication system
CN112242969A (en) Novel single-bit OFDM receiver based on model-driven deep learning
CN110020695A (en) K-means non-uniform quantizing algorithm for filter bank multi-carrier modulation optical communication system
Wei et al. Automatic modulation recognition using neural architecture search
CN112737694B (en) Non-uniform quantization system of filter multi-carrier modulation optical communication system based on SOM
Zhang et al. High-order convolutional attention networks for automatic modulation classification in communication
CN111327558B (en) Method and system for GMM non-uniform quantization for filter multi-carrier modulation optical communication
CN111343115B (en) 5G communication modulation signal identification method and system
Naderializadeh Wireless link scheduling via graph representation learning: A comparative study of different supervision levels
CN113518457B (en) Power distribution strategy based on one-dimensional deep convolutional neural network

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