Frequency modulation digital audio broadcasting prediction method and device fusing entropy weight and neural network
By integrating entropy weights and neural networks, a prediction model for FM digital audio broadcasting is constructed, which solves the problems of poor applicability and low accuracy of radio wave propagation models, achieves high-precision prediction results, and supports the construction of wireless digital broadcasting coverage networks.
Patent Information
- Application Number
- CN202411293031.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-14
- Publication Date
- 2024-11-22
AI Technical Summary
Existing radio wave propagation models have poor applicability in different environments and low prediction accuracy, making it difficult to meet the high-precision coverage requirements of FM digital audio broadcasting systems.
By employing a method that integrates entropy weighting and neural networks, a prediction model for FM digital audio broadcasting is constructed. The radio wave propagation model is then optimized using an RNN neural network and the entropy method. The model is trained and validated in conjunction with ITU-R P.1546, ITU-R P.2001, and ITM models to improve prediction accuracy.
It achieves high-precision FM digital audio broadcast prediction in different environments, expands the applicability of the model, and supports the construction of a technologically advanced, fast-transmitting, widely covered, and secure wireless digital broadcast coverage network.
Smart Images

Figure CN119011055A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of wireless broadcasting technology, and in particular to a method and device for predicting FM digital audio broadcasting by integrating entropy weight and neural network. Background Art
[0002] At present, radio is an important channel for people to obtain news and entertainment information. It has the characteristics of public welfare, emergency and universality. It is also an important infrastructure to ensure that people can quickly and effectively receive authoritative emergency information in emergency situations. FM digital audio broadcasting (CDR) is a wide-coverage digital audio broadcasting system that runs in the traditional analog FM frequency band (88MHz~108MHz). It is designed to provide two types of services: digital audio broadcasting and data broadcasting. Compared with FM analog broadcasting systems, CDR has significant advantages in spectrum efficiency, scheduling flexibility and transmission reliability. At present, CDR is in a critical stage of full deployment and popularization. In order to accelerate the construction of wireless digital broadcasting coverage network, the State Administration of Radio and Television encourages wireless broadcasting stations in various places to build CDR transmission systems to build wireless digital broadcasting coverage networks with advanced technology, fast transmission, wide coverage, safety and reliability.
[0003] Wireless channels are the transmission pathways between transceiver devices in electronic information systems. Channel characteristics determine the performance of electronic information systems and play an important role in the entire life cycle of electronic information system demonstration, design, development, production, testing and verification, and operation and maintenance.
[0004] The setting of FM digital audio broadcasting system needs to consider environmental factors. This is because environmental factors have a significant impact on the propagation of radio signals. Specifically, the consideration of terrain factors in the setting of FM digital audio broadcasting system is mainly reflected in the following aspects:
[0005] The impact of terrain on signal propagation: Different terrains, such as mountains, buildings, fjords, etc., will hinder or reflect the propagation of radio signals, affecting the coverage and quality of the signal. Therefore, when selecting the transmission frequency and setting the antenna, these terrain factors need to be taken into account to ensure that the signal can effectively cover the target area.
[0006] Choose the appropriate transmission frequency: When choosing the transmission frequency, you should consider the influence of factors such as terrain, buildings, and electromagnetic interference. Lower frequencies can penetrate obstacles, but are susceptible to electromagnetic interference; higher frequencies can improve the speed and quality of signal transmission, but the signal penetration ability is poor. Therefore, it is necessary to make comprehensive considerations based on the actual situation and choose the appropriate transmission frequency.
[0007] Reasonable antenna setting: The location and direction of the antenna are crucial to improving signal coverage and reducing signal attenuation. When setting up the antenna, the influence of the terrain should be taken into consideration and a suitable location and direction should be selected to ensure that the signal can be effectively transmitted to the target area.
[0008] In order to better configure the FM digital audio broadcasting system, it is necessary to predict the signal propagation range and propagation loss before setting up, so as to reasonably set the antenna, signal transmission frequency and signal transmission power, etc. There are many radio wave propagation models, such as statistical models, experimental models, deterministic models, regression models, etc.
[0009] Statistical model: Through field measurement of the field strength in the mobile communication service area, the variation law of the median field strength with factors such as frequency, distance, antenna height, etc. is summarized by statistical methods, and expressed by formula or curve. This requires the collection of a large amount of data in the early stage, and the accuracy of the model is poor due to the influence of the environment.
[0010] Experimental model: Some radio wave propagation laws are obtained through experimental methods, but they are not expressed by formulas or curves like statistical models. Because the experiment cannot truly reflect the actual environment, the applicability of this type of model is poor and the model accuracy is poor.
[0011] Deterministic model: After appropriately idealizing the terrain, objects and other radio wave propagation environments, the deterministic method of electromagnetic field theory or geometric optics is used to obtain the law of field intensity changes. Due to the many environmental factors, this type of model has a large amount of calculations, and the accuracy deviation is large when corresponding to different environments.
[0012] Regression model: The path loss law is fitted by performing linear regression on the calculated or measured data of path loss changing with propagation distance according to the distance multiplication method. The model has poor applicability and poor accuracy.
[0013] The above models generally have the problems of poor applicability and large deviation in prediction accuracy in different environments. Summary of the invention
[0014] The present invention provides a method and device for predicting FM digital audio broadcasting by integrating entropy weight and neural network with high accuracy and wide application range in order to solve the technical problems existing in the known technology.
[0015] The technical solution adopted by the present invention to solve the technical problems existing in the known technology is:
[0016] A method for predicting FM digital audio broadcasting by integrating entropy weight and neural network comprises the following steps:
[0017] Step 1, determine the input feature parameters and target output parameters, and use a neural network to build a FM digital audio broadcast prediction model;
[0018] Step 2: Collect measured data of input feature parameters and output target parameters, and divide the collected data into a training set, a validation set, and a test set;
[0019] Step 3, assuming that the training set compiled by the collected measured data is the original training set; selecting the best from a plurality of different radio wave propagation models as the prediction sub-model, using the prediction sub-model based on the entropy weight method, and calculating the target output parameter data from the input feature parameter data in the original training set; compiling the input feature parameters and the corresponding calculated target output parameter data into an improved training set;
[0020] Step 4, setting the initial values of the hyperparameters used for iterative training of the FM digital audio broadcasting prediction model; using the improved training set, validation set and test set data, correspondingly training, optimizing and testing the FM digital audio broadcasting prediction model.
[0021] Furthermore, in step 1, the following input characteristic parameters are determined: longitude of the receiving point, latitude of the receiving point, distance between transceivers, terrain clearance angle, altitude of the transmitting antenna, altitude of the receiving antenna, terrain clearance and radio wave frequency; and the following target output parameters are determined: propagation loss.
[0022] Furthermore, in step 1, an FM digital audio broadcast prediction model is constructed using an RNN neural network, wherein the number of initial hidden layer neurons of the RNN neural network is set to 10, and the initial number of hidden layers is set to two.
[0023] Furthermore, in step 2, the collected feature data are randomly divided into a training set, a validation set, and a test set in a ratio of 6:2:2.
[0024] Further, step 3 includes the following sub-steps:
[0025] Step B1, preliminarily selecting an applicable radio wave propagation model according to the application scenario;
[0026] Step B2, selecting a better radio wave propagation model from the initially selected radio wave propagation models as a prediction sub-model;
[0027] Step B3, using the root mean square error between the predicted value and the measured value of the prediction sub-model as an evaluation index, and calculating the entropy value of each prediction sub-model based on the entropy method;
[0028] Step B4, based on the entropy value of each prediction sub-model, further calculating the weight of each prediction sub-model;
[0029] Step B5, performing weighted summation on the output of each prediction sub-model and its corresponding weight, and calculating the target output data as the target output parameter data for training the FM digital audio broadcasting prediction model.
[0030] Furthermore, in step B1, the applicable radio wave propagation models selected include ITU-R P.1546 model, ITU-RP.2001 model and ITM model.
[0031] Further, step B2 includes the following method steps:
[0032] If the prediction sub-model errors are distributed on the same side of the measured value, the prediction sub-model with the smallest error is selected as the better prediction sub-model;
[0033] If the prediction sub-model errors are distributed on the opposite side of the measured value, the two prediction sub-models with the smallest errors distributed on the opposite side are selected as the better prediction sub-models.
[0034] Furthermore, in step 4, the initial values of the hyperparameters for training the FM digital audio broadcasting prediction model are set as follows: the maximum number of iterations is 20 times, and the learning rate is 0.005.
[0035] Furthermore, step 4 includes the following sub-steps:
[0036] Step C1, training the FM digital audio broadcasting prediction model using the improved training set;
[0037] Step C2, input the validation set data into the constructed model, and compare and analyze the output results with the measured values;
[0038] Step C3, judging whether the model accuracy meets the expected requirements or whether the number of model iterations reaches the preset number according to the verification results;
[0039] Step C4, if the model accuracy does not meet the accuracy requirement or the number of iterations does not reach the set value, adjust the FM digital audio broadcast prediction model hyperparameters and retrain using the improved training set;
[0040] Step C5: If the model accuracy meets the accuracy requirement or the number of iterations reaches the set value, the test set data is input into the FM digital audio broadcasting prediction model to test its final performance and save the model.
[0041] The present invention also provides a device for an FM digital audio broadcast prediction method that integrates entropy weights and neural networks, comprising a memory and a processor, wherein the memory is used to store a computer program; the processor is used to execute the computer program and implement the steps of the FM digital audio broadcast prediction method that integrates entropy weights and neural networks as described above when executing the computer program.
[0042] The advantages and positive effects of the present invention are as follows: a method for predicting FM digital audio broadcasting that integrates entropy weight and neural network can be used to establish a high-precision prediction model for FM digital audio broadcasting that integrates entropy value method and neural network, and can be expanded to compile an appropriate amount of model training data to train the prediction model. The present invention has high accuracy and a wide range of applications, and can provide important support for building a wireless digital broadcasting coverage network with advanced technology, fast transmission, wide coverage, and safe and reliable. BRIEF DESCRIPTION OF THE DRAWINGS
[0043] Figure 1 The present invention is a flowchart of a method for predicting FM digital audio broadcasting by integrating entropy weight and neural network.
[0044] Figure 2 The present invention is a workflow diagram for preparing a FM digital audio broadcast prediction model training.
[0045] Figure 3 It is a workflow diagram for compiling an improved training set according to the present invention.
[0046] Figure 4 The present invention discloses a workflow diagram for training a FM digital audio broadcast prediction model. DETAILED DESCRIPTION
[0047] The present invention will be described in detail below with reference to the accompanying drawings and in combination with embodiments. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.
[0048] See also Figures 1 to 4 , a FM digital audio broadcast prediction method integrating entropy weight and neural network, comprising the following steps:
[0049] Step 1: Determine the input feature parameters and target output parameters, and use a neural network to build a FM digital audio broadcasting prediction model.
[0050] Step 2: Collect measured data of input feature parameters and output target parameters, and divide the collected data into training set, validation set and test set; apply to the training of FM digital audio broadcasting prediction model, optimization of hyperparameters and final performance verification.
[0051] Step 3, assume that the training set compiled by the collected measured data is the original training set; select the best one from multiple different radio wave propagation models as the prediction sub-model, and use the prediction sub-model based on the entropy weight method to calculate the target output parameter data from the input feature parameter data in the original training set; compile the input feature parameters and the corresponding calculated target output parameter data into an improved training set.
[0052] Step 4, setting the initial values of the hyperparameters used for iterative training of the FM digital audio broadcasting prediction model; using the improved training set, validation set and test set data, correspondingly training, optimizing and testing the FM digital audio broadcasting prediction model.
[0053] Preferably, in step 1, the following input characteristic parameters can be determined: longitude of the receiving point, latitude of the receiving point, distance between transceivers, terrain clearance angle, altitude of the transmitting antenna, altitude of the receiving antenna, terrain clearance and radio wave frequency; the following target output parameters can be determined: propagation loss.
[0054] Preferably, in step 1, a RNN neural network may be used to construct an FM digital audio broadcast prediction model, wherein the number of neurons in the initial hidden layer of the RNN neural network may be set to 10, and the number of initial hidden layers may be set to two.
[0055] RNN neural network, also known as recurrent neural network (RNN), is a type of artificial neural network with internal loop connections, used to process sequence data. Its biggest feature is the existence of loops in the network, which allows information to circulate in the network, realizing the storage and processing of sequence information.
[0056] Preferably, in step 2, the collected feature data can be randomly divided into a training set, a validation set and a test set in a ratio of 6:2:2.
[0057] Preferably, step 3 may include the following sub-steps:
[0058] Step B1: Preliminarily select an applicable radio wave propagation model based on the application scenario.
[0059] Step B2, selecting a better radio wave propagation model from the initially selected radio wave propagation models as a prediction sub-model.
[0060] Step B3, using the root mean square error between the predicted value and the measured value of the prediction sub-model as an evaluation index, and calculating the entropy value of each prediction sub-model based on the entropy method.
[0061] Step B4: further calculating the weight of each prediction sub-model based on the entropy value of each prediction sub-model.
[0062] Step B5, performing weighted summation on the output of each prediction sub-model and its corresponding weight, and calculating the target output data as the target output parameter data for training the FM digital audio broadcasting prediction model.
[0063] By using the prediction sub-model, sufficient training samples can be obtained to train the FM digital audio broadcasting prediction model.
[0064] Several commonly used radio wave propagation models, including ITU-R P.1546 model, ITU-R P.526 model, ITU-RP.2001 model, ITM model, Okumura-Hata model, etc.
[0065] The ITU-R P.1546 model is a method for predicting point-to-area radio propagation for terrestrial services provided by the International Telecommunication Union. The ITU-R P.1546 radio wave propagation model stipulates that the effect is not good when the transmitting antenna height exceeds 3000 meters.
[0066] When dealing with radio wave propagation prediction problems in complex mountainous areas, the ITU-RP.526 model provided by the International Telecommunication Union is closer to the measured value. The ITU-RP.526 propagation model is used to calculate the propagation loss of diffraction propagation. The polarization mode will affect the smooth spherical diffraction loss in the model, and the model will produce discontinuities at the edge of the line of sight.
[0067] The three methods for predicting the average annual transmission loss of tropospheric scatter in the International Telecommunication Union standards ITU-R P.452, ITU-R P.617 and ITU-R P.2001, the effective spectrum range of the radio wave propagation model ITU-R P.2001 is 30MHz~50GHz, and the propagation distance is 3km~1000km. In theory, it is suitable for the calculation of the coverage range of most civil aviation ground stations above 3km.
[0068] The ITM model, also known as the Longley-Rice model, is a propagation loss prediction model widely used in the field of wireless communications. It was proposed by Longley and Rice in the early 1970s. The model is based on electromagnetic theory and statistical analysis of terrain features and radio measurement results. It is mainly used to estimate the propagation loss of radio waves under various complex terrain and environmental conditions. The ITM model is a general model that is applicable to radio propagation problems with frequencies between 20MHz and 20GHz. It can predict the median attenuation of radio signals as a function of distance, as well as the variability of signals in time and space.
[0069] The Okumura-Hata model is mainly used in VHF and UHF bands, provides relatively comprehensive data, and is widely used in the planning and design of wireless communication systems. In addition, the model also takes into account factors such as antenna height and regional coverage type. It is most commonly used in models for predicting path loss in urban areas, covering a frequency range of 500 to 1500MHz, with antenna heights of up to 30 to 1000m, and is suitable for different propagation environments, such as cities, suburbs, and open areas.
[0070] Preferably, in step B1, the applicable radio wave propagation model selected may include ITU-R P.1546 model, ITU-RP.2001 model and ITM model.
[0071] From several commonly used models, the ITU-R P.1546 model, the ITU-R P.2001 model and the ITM model are preliminarily screened. The ITU-R P.1546 model, the ITU-R P.2001 model and the ITM model are prediction models with excellent performance and are widely used. The present invention selects these three models as prediction sub-models.
[0072] Assuming that the prediction sub-models include the ITU-R P.1546 model, the ITU-R P.2001 model and the ITM model, the method of using the entropy method to calculate the weight of each prediction sub-model may include the following method steps:
[0073] Data standardization: Before using the entropy method, the original data needs to be standardized. The purpose of standardization is to eliminate the impact of different dimensions on the evaluation results and make the indicators comparable.
[0074] Calculate the information entropy (H) of each prediction sub-model i ):
[0075]
[0076] Where i is the number of the prediction submodel; i = {1, 2, 3}, i = 1 corresponds to the ITU-R P.1546 model, i = 2 corresponds to the ITU-R P.2001 model, and i = 3 corresponds to the ITM model; p i is the probability of the i-th prediction sub-model, H i is the information entropy of the i-th prediction sub-model; p i The calculation formula is as follows:
[0077]
[0078] In the formula, e i is the error between the predicted value and the measured value of the ith prediction sub-model, and j is the number of prediction sub-models.
[0079] Determine the weight of each prediction sub-model (w i ):
[0080]
[0081] In the formula, w i is the weight of the i-th prediction sub-model.
[0082] Preferably, step B2 may include the following method steps:
[0083] If the prediction sub-model errors are distributed on the same side of the measured value, the prediction sub-model with the smallest error is selected as the better prediction sub-model.
[0084] If the prediction sub-model errors are distributed on the opposite side of the measured value, the two prediction sub-models with the smallest errors distributed on the opposite side are selected as the better prediction sub-models.
[0085] Preferably, in step 4, the initial values of the hyperparameters for training the FM digital audio broadcasting prediction model can be set as follows: the maximum number of iterations is 20 times, and the learning rate is 0.005.
[0086] Preferably, step 4 may include the following sub-steps:
[0087] Step C1, using the improved training set to train the FM digital audio broadcasting prediction model; during training, the learning rate descent algorithm is set to the stochastic gradient descent method with momentum.
[0088] Step C2: input the validation set data into the constructed model and compare and analyze the output results with the measured values.
[0089] Step C3, judging whether the model accuracy meets the expected requirements or whether the number of model iterations reaches the preset number according to the verification results.
[0090] Step C4: if the model accuracy does not meet the accuracy requirement or the number of iterations does not reach the set value, adjust the hyperparameters of the FM digital audio broadcasting prediction model and retrain using the improved training set.
[0091] Step C5: If the model accuracy meets the accuracy requirement or the number of iterations reaches the set value, the test set data is input into the FM digital audio broadcasting prediction model to test its final performance and save the model.
[0092] The present invention also provides a device for an FM digital audio broadcast prediction method that integrates entropy weights and neural networks, comprising a memory and a processor, wherein the memory is used to store a computer program; the processor is used to execute the computer program and implement the steps of the FM digital audio broadcast prediction method that integrates entropy weights and neural networks as described above when executing the computer program.
[0093] The above-mentioned ITU-R P.1546 model, ITU-R P.2001 model, ITM model, neural network, RNN neural network, entropy method, etc. can all adopt applicable functional modules, software systems and algorithms in the prior art; or adopt functional modules, software systems and algorithms in the prior art and construct them by conventional technical means.
[0094] The embodiments described above are only used to illustrate the technical ideas and features of the present invention, and their purpose is to enable those skilled in the art to understand the contents of the present invention and implement them accordingly. The patent scope of the present invention cannot be limited only by this embodiment, that is, any equivalent changes or modifications made to the spirit disclosed by the present invention still fall within the patent scope of the present invention.
Claims
1. A method for predicting FM digital audio broadcasting by integrating entropy weight and neural network, characterized in that: The steps include: Step 1, determine the input feature parameters and target output parameters, and use a neural network to build a FM digital audio broadcast prediction model; Step 2: Collect measured data of input feature parameters and output target parameters, and divide the collected data into a training set, a validation set, and a test set; Step 3, assuming that the training set compiled by the collected measured data is the original training set; selecting the best from a plurality of different radio wave propagation models as the prediction sub-model, using the prediction sub-model based on the entropy weight method, and calculating the target output parameter data from the input feature parameter data in the original training set; compiling the input feature parameters and the corresponding calculated target output parameter data into an improved training set; Step 4, setting the initial value of the hyperparameter for iterative training of the FM digital audio broadcasting prediction model; The FM digital audio broadcasting prediction model is trained, optimized and verified using the improved training set, validation set and test set data.
2. The FM digital audio broadcast prediction method integrating entropy weight and neural network according to claim 1 is characterized in that: In step 1, the following input characteristic parameters are determined: receiving point longitude, receiving point latitude, distance between transceivers, terrain clearance angle, transmitting antenna altitude, receiving antenna altitude, terrain clearance and radio wave frequency; the following target output parameters are determined: propagation loss.
3. The FM digital audio broadcast prediction method integrating entropy weight and neural network according to claim 1, characterized in that: In step 1, an FM digital audio broadcast prediction model is constructed using an RNN neural network, wherein the number of initial hidden layer neurons of the RNN neural network is set to 10, and the number of initial hidden layers is set to two.
4. The FM digital audio broadcast prediction method integrating entropy weight and neural network according to claim 1, characterized in that: In step 2, the collected feature data are randomly divided into training set, validation set and test set in the ratio of 6:2:
2.
5. The FM digital audio broadcast prediction method integrating entropy weight and neural network according to claim 1, characterized in that: Step 3 includes the following sub-steps: Step B1, preliminarily selecting an applicable radio wave propagation model according to the application scenario; Step B2, selecting a better radio wave propagation model from the initially selected radio wave propagation models as a prediction sub-model; Step B3, using the root mean square error between the predicted value and the measured value of the prediction sub-model as an evaluation index, and calculating the entropy value of each prediction sub-model based on the entropy method; Step B4, based on the entropy value of each prediction sub-model, further calculating the weight of each prediction sub-model; Step B5, performing weighted summation on the output of each prediction sub-model and its corresponding weight, and calculating the target output data as the target output parameter data for training the FM digital audio broadcasting prediction model.
6. The FM digital audio broadcast prediction method integrating entropy weight and neural network according to claim 5, characterized in that: In step B1, applicable radio wave propagation models are selected, including ITU-R P.1546 model, ITU-R P.2001 model and ITM model.
7. The method for predicting FM digital audio broadcasting by integrating entropy weight and neural network according to claim 5, characterized in that: Step B2 includes the following method steps: If the prediction sub-model errors are distributed on the same side of the measured value, the prediction sub-model with the smallest error is selected as the better prediction sub-model; If the prediction sub-model errors are distributed on the opposite side of the measured value, the two prediction sub-models with the smallest errors distributed on the opposite side are selected as the better prediction sub-models.
8. The method for predicting FM digital audio broadcasting by integrating entropy weight and neural network according to claim 1, characterized in that: In step 4, the initial values of the hyperparameters for training the FM digital audio broadcast prediction model are set as follows: the maximum number of iterations is 20, and the learning rate is 0.
005.
9. The method for predicting FM digital audio broadcasting by integrating entropy weight and neural network according to claim 1, characterized in that: Step 4 includes the following sub-steps: Step C1, training the FM digital audio broadcasting prediction model using the improved training set; Step C2, input the validation set data into the constructed model, and compare and analyze the output results with the measured values; Step C3, judging whether the model accuracy meets the expected requirements or whether the number of model iterations reaches the preset number according to the verification results; Step C4, if the model accuracy does not meet the accuracy requirement or the number of iterations does not reach the set value, adjust the FM digital audio broadcast prediction model hyperparameters and retrain using the improved training set; Step C5: If the model accuracy meets the accuracy requirement or the number of iterations reaches the set value, the test set data is input into the FM digital audio broadcasting prediction model to test its final performance and save the model.
10. A device for frequency modulation digital audio broadcast prediction method integrating entropy weight and neural network, comprising a memory and a processor, characterized in that: The memory is used to store computer programs; the processor is used to execute the computer program and implement the steps of the FM digital audio broadcast prediction method integrating entropy weight and neural network as described in any one of claims 1 to 9 when executing the computer program.
