OFDM communication parameter self-adaptive selection method and system based on layering moduleTechnical Field
The invention belongs to the technical field of communication, and particularly relates to an OFDM communication parameter self-adaptive selection method and system based on a layering module
Background
Orthogonal Frequency Division Multiplexing (OFDM) is a widely applied multi-carrier modulation technology in high-speed wired/wireless communication, such as wireless 5G, high-speed two-wire network and the like, and has the characteristics of high spectrum utilization rate, multipath interference resistance and the like. However, OFDM has some drawbacks in practical applications. One of the problems is the high peak-to-average power ratio (PAPR), which leads to non-linear distortion of the power amplifier. Generally, industrial automation has low power consumption, low cost and high reliability requirements for communication equipment, and a higher PAPR causes a problem of power loss of the system, which not only shortens the service life of equipment related devices, but also reduces the working efficiency of a power amplifier. Therefore, the research on the method for reducing the peak-to-average ratio of the OFDM system signal has strong practical significance.
Currently, the Partial Transmit Sequence (PTS) method is a widely used method for reducing the peak-to-average ratio problem. The key to the PTS method is the division of the number of data blocks of the incoming data, generally denoted V. However, the PTS method cannot adaptively complete the selection of the processing parameter V for different OFDM models. Furthermore, the need for an additional Inverse Fast Fourier Transform (IFFT) and the selection of the optimal phase factor from the set of phase factors introduces a high computational complexity. The above problems limit the processing effect of the PTS method.
Disclosure of Invention
In order to solve the above-mentioned defects, the present invention proposes a hierarchical module-based adaptive selection method and system for OFDM communication parameters.
In order to solve the technical problems, the invention adopts the following technical scheme:
An OFDM communication parameter self-adaptive selection method based on a layering module comprises the following steps:
Presetting modeling parameters;
Constructing a fuzzy neural network;
obtaining training sample data, wherein the sample data is used as fuzzy neural network training data;
Training the fuzzy neural network;
and acquiring input data, and responding to output data of the input data, wherein the input data is the parameter data.
The method is further improved as follows: the obtaining input data, responding to the output data of the input data, then comprises performing block and weighting processing on the output data.
The method is further improved as follows: the preset modeling parameters comprise input variables and output variables, the input variables comprise the number N of subcarriers, the power Po of signals and a power threshold Pt, the output variables comprise a blocking processing parameter V, and the power threshold Pt is set according to the linear working range of the power amplifier.
The method is further improved as follows: the construction of the fuzzy neural network comprises the following steps: an input layer, a blurring layer, a fuzzy inference layer, a normalization layer and an output layer, wherein,
Confirming the number of the nerve cells of the input layer;
The blurring layer carries out data blurring processing after responding to the input layer;
the fuzzy reasoning layer is used for matching the data after the fuzzification processing with the fuzzy rule of the fuzzy reasoning layer;
The normalization layer performs data normalization processing after responding to the fuzzy layer;
the output layer defuzzifies the normalized data.
The method is further improved as follows: the method for confirming the number of the neurons of the input layer comprises the following steps of:
Where xin is the number of neuron points, in represents input, and n is the dimension of the input.
The method is further improved as follows: the method for confirming the number of the neurons of the input layer comprises the following steps of: the data blurring processing after the blurring layer responds to the input layer comprises the following steps:
the membership function of the blurring process is as follows:
Where i represents the ith input variable, j represents the jth fuzzy hierarchy, σij is the standard deviation of the membership function, and cij is the center of the membership function.
The method is further improved as follows: the fuzzy inference layer is matched with the fuzzy rule of the fuzzy inference layer through the data after fuzzification processing, and the method comprises the following steps:
the fitness function of the fuzzy rule is:
Where j is {1,2,., mi }, k is {1,2,., m },M is the number of fuzzy inference layer rules, k represents the kth fuzzy rule, and n is the dimension of an input variable;
the normalization layer performs data normalization processing after the response of the fuzzy layer, and comprises the following steps:
The normalization function is:
The output layer defuzzifies the normalized data, which comprises the following steps:
The defuzzification function is:
wherein yo is the output of the fuzzy neural network, r is the number of variables required to be output by the network, and omegaok is the weight parameter corresponding to the kth rule.
The method is further improved as follows: the method for obtaining training sample data, wherein the sample data is used as fuzzy neural network training data and comprises the following steps:
The training sample data capacity function of the fuzzy neural network is as follows:
Wherein, P is the training sample data capacity of the fuzzy neural network, L is the number of fuzzy layers in the fuzzy neural network, N is the number of input layers, Ri is the number of fuzzy grades contained in the ith input xi, Ncp is the parameter contained in each fuzzy grade membership function, and M is the number of output variables of the fuzzy neural network.
The method is further improved as follows: the training of the fuzzy neural network comprises the following steps:
Setting the iteration times L and the expected training errors Ee, training the fuzzy neural network through a gradient descent method, and continuously optimizing the front piece parameter sigmaij,cij and the weight parameter omegaok of the membership function to finish the optimal fuzzy neural network.
OFDM communication parameter self-adaptation selection system based on layering module includes:
the preset module is used for presetting modeling parameters;
The fuzzy neural network module is used for constructing a fuzzy neural network model according to preset modeling parameters, and is also used for acquiring input data and outputting data responding to the input data;
the sample module is used for obtaining training sample data;
the training module is used for training the fuzzy neural network model according to training sample data;
A partial transmission sequence module for performing block and weighting processing on data output in response to the input data;
and the orthogonal frequency division multiplexing module is used for modulating the data responded by the partial transmission sequence module.
By adopting the technical scheme, the invention has the following technical progress:
the invention combines the learning ability of the neural network and the reasoning ability of the fuzzy control method, and realizes the self-adaptive selection of the processing parameters of the PTS method.
The invention considers the signal characteristic of the OFDM system and the characteristic of the power amplifier, selects three parameters with larger influence on the processing parameters as the input of the fuzzy neural network, namely the number N of subcarriers, the power Po of the signal and the power threshold Pt, and is applicable to different OFDM models.
The FNN-APTS method provided by the invention only needs to process the signals with the power larger than the threshold value, so that the number of segmented signals is greatly reduced, the calculation complexity of the system is reduced, and the bit error rate performance is improved.
Drawings
FIG. 1 is an overall flow chart of the method of the present invention;
FIG. 2 is a schematic diagram of a selected Gaussian membership function according to the present invention;
FIG. 3 is a block diagram of the fuzzy neural network architecture of the present invention;
FIG. 4 is a graph of the results of the N membership function of the input variable after training in accordance with the present invention;
FIG. 5 is a graph of the membership function results of the trained input variable Po according to the present invention;
FIG. 6 is a graph of the membership function results of the trained input variable Pt according to the present invention;
Detailed Description
The invention is described in further detail below with reference to the accompanying drawings:
The invention provides an OFDM communication parameter self-adaptive selection method based on a layering module, as shown in fig. 1, comprising the following steps:
Presetting modeling parameters;
Constructing a fuzzy neural network;
Obtaining training sample data, wherein the sample data is used as fuzzy neural network training data;
Training the fuzzy neural network;
Acquiring input data, and responding to output data of the input data, wherein the input data is the parameter data;
Specifically, the invention combines the learning ability of the neural network and the reasoning ability of the fuzzy control method, and realizes the self-adaptive selection of the processing parameters of the PTS method; only signals with power larger than a threshold value need to be processed, so that the number of segmented signals is greatly reduced, the calculation complexity of the system is reduced, and the bit error rate performance is improved.
Further, obtaining input data, and responding to output data of the input data includes performing block and weighting processing on the output data.
Further, the preset modeling parameters comprise input variables and output variables, the input variables comprise the number N of subcarriers, the power Po of signals and a power threshold Pt, the output variables comprise a blocking processing parameter V, and the power threshold Pt is set according to the linear working range of the power amplifier; in order to better reflect the network training effect, in this embodiment, the signal characteristics and the power amplifier characteristics of the OFDM system are comprehensively considered, and three parameters with great influence on the processing parameters are selected as inputs of the fuzzy neural network, namely, the number N of subcarriers, the power Po of the signal, and the power threshold Pt.
Further, as shown in fig. 3, constructing the fuzzy neural network includes: an input layer, a blurring layer, a fuzzy inference layer, a normalization layer and an output layer, wherein,
Confirming the number of the nerve cells of the input layer;
The blurring layer carries out data blurring processing after responding to the input layer;
the fuzzy reasoning layer is used for matching the data after the fuzzification processing with the fuzzy rule of the fuzzy reasoning layer;
The normalization layer performs data normalization processing after responding to the fuzzy layer;
the output layer defuzzifies the normalized data.
Still further, the confirming the number of the neurons of the input layer includes obtaining the number of the neurons by the following method:
Wherein xin is the number of the neurons, in represents input, and n is the dimension of the input quantity; in this embodiment, the first layer is an input of the entire network, and the node of the input layer of the fuzzy neural network is the entry of fuzzy information, and the number of the nodes (NN) of the neurons of the input layer of the network depends on the dimension of the input information, i.e., (NN)1 =n=3.
Still further, the data obfuscation after the obfuscation layer responds to the input layer includes the following methods:
as shown in fig. 2, the membership function of the blurring process is:
Wherein i represents an ith input variable, j represents a jth fuzzy grade, sigmaij is a standard deviation of a membership function, and cij is a center of the membership function; the layer network in this embodiment can divide the input information into 5 fuzzy sets of large (VB), large (BI), medium (MI), small (SM) and small (VS), so that mi =5 is the number of Neuronal Nodes (NN) in the layer
Still further, the fuzzy inference layer is matched with the fuzzy rule of the fuzzy inference layer through the data after the fuzzification processing, and the method comprises the following steps:
the fitness function of the fuzzy rule is:
Where j is {1,2,., mi }, k is {1,2,., m },M is the number of fuzzy inference layer rules, k represents the kth fuzzy rule, and n is the dimension of an input variable; the rule base of this layer in this embodiment contains all combinations of fuzzy sets. The rule base contains k=125 rules reflecting the relationship between the ambiguous input and output. Number of neuronal nodes of the layer
The normalization processing of the data after the response of the normalization layer to the fuzzy layer comprises the following steps:
The normalization function is:
the number of neuronal nodes included in this layer is the same as in the fuzzy inference layer in this embodiment.
The output layer defuzzifies the normalized data, which comprises the following steps:
The defuzzification function is:
Wherein yo is the output of the fuzzy neural network, r is the number of variables required to be output by the network, and omegaok is the weight parameter corresponding to the kth rule; in this embodiment, the layer includes a weight parameter ωij corresponding to each branch output, which is called a back-part parameter, and the layer includes only one output data, i.e., a processing parameter V, so that the number of neuron nodes in the layer is (NN)5 =1.
Further, training sample data is obtained, and the sample data is used as fuzzy neural network training data, and comprises the following steps:
The training sample data capacity function of the fuzzy neural network is as follows:
Wherein, P is the training sample data capacity of the fuzzy neural network, L is the number of fuzzy layers in the fuzzy neural network, N is the number of input layers, Ri is the number of fuzzy grades contained in the ith input xi, Ncp is the parameter contained in each fuzzy grade membership function, and M is the number of output variables of the fuzzy neural network; according to the function, the training sample data capacity P of the fuzzy neural network can be obtained, and the data capacity P is 125-155, and is as follows:
A training sample set is then obtained, the sample set comprising three inputs, namely the number of sub-carriers N, the signal power Po and the power threshold Pt, comprising an output data, the processing parameter V.
Further, training the fuzzy neural network includes the following method:
Setting the iteration times L and the expected training errors Ee, training the fuzzy neural network by a gradient descent method, and continuously optimizing the front piece parameter sigmaij,cij and the weight parameter omegaok of the membership function to finish the optimal fuzzy neural network; in this embodiment, the iteration number L may be set to 100 times, and the expected training error Ee may be set to 0.01, so as to train and generate the fuzzy neural network FNN-APTS, and the trained membership function result graphs of the input variables are shown in fig. 4, fig. 5 and fig. 6.
The system based on the OFDM communication parameter self-adaptive selection method based on the layering module comprises the following steps:
the preset module is used for presetting modeling parameters;
The fuzzy neural network module is used for constructing a fuzzy neural network model according to preset modeling parameters, and is also used for acquiring input data, responding to output data of the input data, wherein the input data is the parameter data;
The sample module is used for obtaining training sample data which is used as fuzzy neural network training data;
The training module is used for training the fuzzy neural network model;
the partial transmission sequence module is used for carrying out block and weighting processing on the output data;
and the orthogonal frequency division multiplexing module is used for modulating the data responded by the partial transmission sequence module.
In summary, the invention comprehensively considers the signal characteristics and the power amplifier characteristics of the OFDM system, selects the parameters with larger influence on the processing parameters as the input variables of the fuzzy neural network, and constructs the fuzzy neural network. In the off-line stage, the iteration times L and the expected training error Ee are set, the membership function is continuously optimized, and the fuzzy neural network is trained and generated by judging whether the set error condition is reached. And in an online stage, according to the input parameters of different OFDM systems, the processing parameters V are obtained in a self-adaptive mode. The invention only needs to process the signal with the power larger than the threshold value, and improves the calculation complexity and the bit error rate performance of the system while ensuring the PAPR performance.
The above examples are only illustrative of the preferred embodiments of the present invention and are not intended to limit the scope of the present invention, and various modifications and improvements made by those skilled in the art to the technical solution of the present invention should fall within the scope of protection defined by the claims of the present invention without departing from the spirit of the design of the present invention.