Communication signal interference optimization system and method based on machine learning

Abstract

The invention discloses a communication signal interference optimization system and method based on machine learning, and the system comprises a channel data collection module which collects channel data in a communication environment, builds a channel data group, and transmits the channel data group to a data processing and feature extraction module; the data processing and feature extraction module is used for receiving the channel data set in real time and carrying out preprocessing and multi-feature engineering processing on the channel data set to obtain a channel feature set and a signal quality data set; the machine learning model module is used for constructing a multi-level model and training the constructed multi-level model; the self-adaptive channel selection optimization module is used for performing interference prediction analysis by using the training data transmitted by the machine learning model module and triggering a multi-channel switching mechanism; and the iterative optimization module is used for acquiring interference data in real time through the signal interference detection equipment group after the initial signal selection, evaluating and optimizing the current channel signal quality, and generating an iterative mechanism.

Description

Technical Field

[0001] The present invention relates to the field of wireless communications, and in particular to a communication signal interference optimization system and method based on machine learning. Background Art

[0002] In the field of modern wireless communications, with the continuous growth of user demand and the popularization of communication equipment, the management and optimization of wireless communication channels have become increasingly important. Especially in the case of limited spectrum resources and complex communication environment, how to effectively reduce signal interference and improve communication quality has become one of the core issues in wireless communication research. Based on this background, using machine learning technology to optimize interference management of wireless communication signals has become a forward-looking solution.

[0003] At present, in existing wireless communication systems, channel selection usually relies on preset rules or simple interference detection mechanisms, which show obvious limitations in dynamically changing communication environments. On the one hand, traditional methods cannot adapt to complex environmental changes in real time, resulting in low accuracy of channel selection; on the other hand, fixed channel switching thresholds often cannot cope with sudden channel quality degradation or deterioration, thereby increasing the risk of communication interruption. More importantly, with the surge in the number of wireless devices and users, interference between channels has become more unpredictable, and existing interference management systems have difficulty coping with this complexity. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a communication signal interference optimization system and method based on machine learning, which solves the problems mentioned in the background technology.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0006] A communication signal interference optimization system based on machine learning, comprising:

[0007] The channel data acquisition module collects channel data in the communication environment through a channel detection device group installed in the communication environment, constructs a channel data group, and transmits the channel data group to the data processing and feature extraction module;

[0008] The data processing and feature extraction module is used to receive the channel data group in real time, and perform preprocessing and multiple feature engineering on the channel data group to obtain the channel feature set and signal quality data set;

[0009] The machine learning model module is used to build a multi-level model and train the built multi-level model;

[0010] Adaptive channel selection optimization module, which uses the training data transmitted by the machine learning model module to perform interference prediction analysis and trigger multiple channel switching mechanisms;

[0011] The iterative optimization module is used to collect interference data in real time through the signal interference detection device group after the adaptive channel selection optimization module performs the initial signal selection, evaluate and optimize the current channel signal quality, and generate an iterative mechanism.

[0012] Preferably, the channel data acquisition module includes a data acquisition unit and a data transmission unit;

[0013] The data collection unit collects channel data in the communication environment in real time through a channel detection device group installed in the communication environment;

[0014] The channel detection equipment group includes a signal detection equipment group and a signal quality detection equipment group;

[0015] The signal detection equipment group includes a limit noise analyzer, a multipath measurement instrument, a spectrum analyzer, a vector network analyzer, a path loss test equipment, an antenna array system and an impulse response measurement instrument;

[0016] The signal quality detection equipment group includes bit error rate test equipment, radio frequency test equipment and network performance monitoring tools;

[0017] The channel data includes signal data and signal quality data; the signal data includes transmitter phase infection TPI, multipath reflection angle MRA, dynamic frequency offset DFS, electromagnetic reflection coefficient ERC, path loss factor PLF, spatial channel correlation coefficient SCC, delay spread time DLT and frequency phase modulation FPM; the signal quality data includes the number of error bits Cw, the total number of transmitted bits Zc, signal power Xg, interference power Pi, noise power Pn, packet loss number Db and total number of sent packets Zf;

[0018] The data transmission unit is used to transmit the acquired channel data group and signal quality data to the data processing and feature extraction module through wireless transmission.

[0019] Preferably, the channel data acquisition module introduces the signal data into nonlinear transformation and interaction terms to generate a channel data group UCSDF;

[0020] The channel data set UCSDF is as follows:

[0021]

[0022] Preferably, the data processing and feature extraction module includes a data processing unit and a feature extraction unit;

[0023] The data processing unit is used to receive the collected channel data group and signal quality data in real time, and pre-process the channel data group and signal quality data, wherein the pre-processing includes filling in missing values ​​and removing outliers, and then performing normalization processing;

[0024] The feature extraction unit includes a multiple feature extraction unit and a channel quality feature extraction unit;

[0025] The multiple feature extraction unit is used to extract features from the channel data group to obtain a channel feature set; the multiple feature extraction process includes polynomial features, logarithmic transformation, sine and cosine transformation and feature interaction terms;

[0026] The communication quality feature extraction unit is used to perform feature extraction based on the collected signal quality data, thereby obtaining a signal quality data set, wherein the signal quality data includes a bit error rate BER, a signal to interference plus noise ratio SINR, and a packet loss rate PLR.

[0027] Preferably, the channel feature set AUCSD is:

[0028]

[0029] The bit error rate BER is calculated and extracted by the number of error bits Cw and the total number of transmitted bits Zc in the signal quality data. The calculation formula is:

[0030] The signal-to-interference-plus-noise ratio SINR is calculated and extracted by the signal power Xg, interference power Pi, and noise power Pn in the signal quality data, and the calculation formula is:

[0031] The packet loss rate PLR ​​is calculated and extracted by the number of lost packets Db and the total number of sent packets Zf in the signal quality data. The calculation formula is:

[0032] Preferably, the machine learning model module includes a machine learning model unit and a model training unit;

[0033] The machine learning model unit constructs a multi-level model by selecting a multi-level model architecture combining a deep neural network DNN, a spatiotemporal convolutional network STCN and integrated learning, predicts channel interference, extracts channel nonlinear characteristics through the deep neural network DNN, and the spatiotemporal correlation of the spatiotemporal convolutional network STCN channel, and then integrates multiple models through integrated learning to optimize the stability of the multi-level model;

[0034] The model training unit uses a large number of historical channel data sets for preprocessing to obtain the historical channel feature set AUCSD l, train a multi-level model, obtain a multi-stage prediction model, and output the channel interference prediction value ICP;

[0035] The channel interference prediction value ICP is output by the following multi-stage prediction model ICPmodel;

[0036] ICP(c i )=f DNN+STCN+Ensemble (AUCSD l );

[0037] In the formula, f DNN+STCN+Ensemble represents the combined function of the deep neural network DNN, the spatiotemporal convolutional network STCN and the integrated learning model, c i represents the i-th channel.

[0038] Preferably, the adaptive channel selection optimization module sorts the channels through the channel interference prediction value output by the machine learning model module, obtains the channel with the least interference, sets the channel threshold and compares and analyzes the channel interference prediction value of the current channel, and triggers a multi-channel switching mechanism.

[0039] Preferably, the adaptive channel selection optimization module includes an interference prediction and analysis unit and a dynamic adjustment and multiple switching determination unit;

[0040] The interference prediction and analysis unit predicts the channel feature sets of all current channels by using the trained multi-stage prediction model, obtains the channel interference prediction value of each channel, sorts all channels according to the channel interference prediction value, and selects the interference threshold channel C*;

[0041] The interference valley channel C* is screened by the following sorting method;

[0042]

[0043] In the formula, fhistory(c i ) represents channel c i The history of the function; α represents the channel interference prediction value ICP (c i );β represents the channel history representation function fhistory(c i ); γ represents the channel path loss factor PLF(c i ) weight value; c n represents n channels;

[0044] The dynamic adjustment and multiple switching determination unit is used to collect the channel feature set of the current communication environment in real time after selecting the current interference valley channel C*, and input the multi-stage prediction model, output the channel interference prediction value ICP of the current interference valley channel C*, and preset the channel threshold ICPthresh based on the signal interference standard, compare and evaluate the channel threshold ICPthresh with the current channel interference prediction value ICP, and trigger the multiple channel switching mechanism Switch;

[0045] The triggering multi-channel switching mechanism Switch is triggered by the following evaluation content;

[0046]

[0047] When Switch is True, the channel switching is triggered and switched to the interference valley channel C*; True means switching, False means not switching, c d Indicates the current channel.

[0048] Preferably, after the initial channel selection, the iterative optimization module collects interference data in real time through a signal interference detection device group, performs comprehensive calculations to obtain comprehensive communication quality indicators, presets signal quality thresholds, performs comparative evaluation and analysis on the signal quality of the current channel, and executes an optimization mechanism based on the evaluation results to further optimize the channel selection strategy, dynamically adjust the channel threshold, and generate an iterative mechanism.

[0049] Preferably, the iterative optimization module includes a channel quality analysis unit, a dynamic threshold adjustment unit and an iterative optimization unit;

[0050] The channel quality analysis unit includes a channel quality calculation unit and a channel quality assessment unit;

[0051] The channel quality calculation unit is used to perform comprehensive calculation to obtain a comprehensive communication quality indicator CQV based on the extracted quality data set and the spatial channel correlation coefficient SCC in the channel feature set AUCSD;

[0052] The channel quality evaluation unit is used to preset a signal quality threshold X based on the average value of historical normal communication quality, and then compare and evaluate it with the obtained comprehensive communication quality indicator CQV to analyze the communication quality of the current channel;

[0053] The dynamic threshold adjustment unit is used to combine the channel feature set AUCSD of the current channel with the channel interference prediction value ICP of the current channel, update the multi-stage prediction model ICPmodel through an adaptive weighting mechanism, further optimize the channel selection strategy, and dynamically adjust the channel threshold ICPthresh when identifying that the communication quality of the current channel is abnormal;

[0054] The iterative optimization unit is used to execute the iterative strategy after updating the multi-stage prediction model and dynamically adjusting the channel threshold.

[0055] Preferably, the comprehensive communication quality indicator CQV is calculated and obtained by the following algorithm formula;

[0056]

[0057] Wherein, h represents the demand function, and its specific form is set according to the design requirements of the system; log represents the logarithmic function; BER is the signal quality data including the bit error rate; SINR is the signal-to-interference-plus-noise ratio; and PLR is the packet loss rate.

[0058] Preferably, the specific evaluation contents of the comparative evaluation are as follows:

[0059] When the comprehensive communication quality index CQV ≥ the signal quality threshold X, it indicates that the current channel communication quality is abnormal;

[0060] When the comprehensive communication quality indicator CQV is less than the signal quality threshold X, it indicates that the current channel communication quality is normal.

[0061] Preferably, the channel threshold ICPthresh is adjusted by the following dynamic adjustment method;

[0062] ΔICPthresh=μ×(CQV-CQVtarget);

[0063] Where μ represents the adjustment coefficient, CQVtarget represents the target communication quality value, and △ICPthresh represents the dynamically adjusted channel threshold.

[0064] Preferably, the iteration strategy is as follows:

[0065] When the comprehensive communication quality index CQV ≥ the signal quality threshold X, continue to iterate;

[0066] When the comprehensive communication quality indicator CQV is less than the signal quality threshold X, the iteration is stopped.

[0067] The present invention also provides a communication signal interference optimization method based on machine learning, comprising the following steps:

[0068] S1. Collecting channel data in the communication environment through a channel detection device group installed in the communication environment, constructing a channel data group, and transmitting the channel data group wirelessly;

[0069] S2, receiving a channel data group in real time, and performing preprocessing and multiple feature engineering on the channel data group to obtain a channel feature set and a signal quality data set;

[0070] S3. Construct a multi-level model by setting a model architecture; use historical communication data to train the multi-level model, further generate a multi-stage prediction model, and input the feature data group into the multi-stage prediction model to output a channel interference prediction value;

[0071] S4. Sort the channels according to the output channel interference prediction values, obtain the channel with the least interference, set a channel threshold, compare and analyze the set channel threshold with the channel interference prediction value of the current channel, and trigger a multi-channel switching mechanism;

[0072] S5. After the initial channel selection, interference data is collected in real time through the signal interference detection equipment group, and a comprehensive calculation is performed to obtain a comprehensive communication quality index; a preset signal quality threshold X is compared and evaluated with the comprehensive communication quality index to analyze the signal quality of the current channel; based on the evaluation results, an optimization mechanism is executed to further optimize the channel selection strategy, dynamically adjust the channel threshold, and generate an iterative mechanism.

[0073] Preferably, the channel data in step S1 includes signal data and signal quality data;

[0074] The signal data includes transmitter phase infection TPI, multipath reflection angle MRA, dynamic frequency shift DFS, electromagnetic reflection coefficient ERC, path loss factor PLF, spatial channel correlation coefficient SCC, delay spread time DLT and frequency phase modulation FPM;

[0075] The signal quality data includes the number of error bits Cw, the total number of transmitted bits Zc, the signal power Xg, the interference power Pi, the noise power Pn, the number of lost packets Db and the total number of sent packets Zf.

[0076] Preferably, in step S1, a channel data group UCSDF is generated by introducing signal data into nonlinear transformation and interaction terms;

[0077] The channel data set UCSDF is as follows:

[0078]

[0079] Preferably, the preprocessing in step S2 is preprocessing of the channel data group and the signal quality data, and the preprocessing includes filling in missing values ​​and removing outliers, and then normalizing; the multiple feature engineering processing includes polynomial features, logarithmic transformation, sine and cosine transformation and feature interaction terms.

[0080] Preferably, in step S3, a multi-level model is constructed by selecting a multi-level model architecture combining a deep neural network DNN, a spatiotemporal convolutional network STCN and integrated learning to predict channel interference, and the nonlinear characteristics of the channel are extracted through the deep neural network DNN, and the spatiotemporal correlation of the spatiotemporal convolutional network STCN channel, and then multi-model integration is performed through integrated learning to optimize the stability of the multi-level model;

[0081] Use the historical channel data set for preprocessing to obtain the historical channel feature set AUCSD l , train a multi-level model, obtain a multi-stage prediction model, and output the channel interference prediction value ICP;

[0082] The channel interference prediction value ICP is output by the following multi-stage prediction model ICPmodel;

[0083] ICP(c i ) = f DNN+STCN+Ensemble (AUCSD l );

[0084] In the formula, f DNN+STCN+Ensemble represents the combined function of the deep neural network DNN, the spatiotemporal convolutional network STCN and the integrated learning model, c i represents the i-th channel.

[0085] Preferably, in step S4, the channel feature sets of all current channels are predicted by using the trained multi-stage prediction model to obtain the channel interference prediction value of each channel, and all channels are sorted according to the channel interference prediction value to select the interference threshold channel C*;

[0086] The interference valley channel C* is screened by the following sorting method;

[0087]

[0088] Where, ICP is the channel interference prediction value; fhistory(c i ) represents channel c i The history of the function; α represents the channel interference prediction value ICP (c i );β represents the channel history representation function fhistory(c i ); γ represents the channel path loss factor PLF(c i ) weight value; c n represents n channels;

[0089] After selecting the current interference valley channel C*, the channel feature set of the current communication environment is collected in real time and input into the multi-stage prediction model to output the channel interference prediction value ICP of the current interference valley channel C*, and the channel threshold ICPthresh is preset based on the signal interference standard, and the channel threshold ICPthresh is compared and evaluated with the current channel interference prediction value ICP, and the multiple channel switching mechanism Switch is triggered;

[0090] The triggering multi-channel switching mechanism Switch is triggered by the following evaluation content;

[0091]

[0092] When Switch is True, the channel switching is triggered and switched to the interference valley channel C*; True means switching, False means not switching, c d Indicates the current channel.

[0093] Compared with the prior art, the present invention has the following beneficial effects:

[0094] (1) By introducing the channel data acquisition module and the machine learning model module, the system can obtain the channel data in the communication environment in real time. The channel data is preprocessed and feature-engineered by the data processing and feature extraction module to generate an enhanced data group UCSDF, which is then trained and predicted by a multi-level machine learning model to generate a channel interference prediction value ICP. Compared with the traditional fixed threshold and simple interference detection mechanism, the system can dynamically respond to complex and changing communication environments and select the channel C* with the least interference in real time, thereby reducing the number of channel switching and improving the stability and reliability of the communication link. This adaptive channel selection mechanism significantly improves the accuracy of channel selection and reduces the occurrence of communication interruptions and bit error rates.

[0095] (2) By integrating the deep neural network DNN, the spatiotemporal convolutional network STCN and the integrated learning model, it can capture the nonlinear characteristics and spatiotemporal correlation in the channel data, and perform multi-model integrated optimization to generate a more stable and accurate channel interference prediction model ICPmodel. The adaptive channel selection optimization module obtains the channel with the least interference by sorting the channel interference prediction value ICP value, and sets the dynamic threshold ICPthresh, which is compared and analyzed with the interference prediction value of the current channel to trigger the multiple channel switching mechanism Switch. This adaptive interference management and channel switching strategy enables the system to respond quickly to complex and changeable interference environments, ensure the stability of communication quality, and significantly enhance the system's anti-interference ability.

[0096] (3) After the initial channel selection, the iterative optimization module collects the interference data of the current channel in real time through the signal interference detection equipment group, and comprehensively calculates the comprehensive communication quality index CQV. When the comprehensive communication quality index CQV does not meet the standard, the system will update the multi-stage prediction model ICPmodel through an adaptive weighting mechanism, and dynamically adjust the channel threshold ICPthresh to optimize the channel selection strategy. After updating the model and threshold, the iterative optimization unit executes the iterative strategy to ensure that the system can continuously optimize and achieve the best communication quality and resource utilization. This dynamic optimization mechanism not only improves the system's utilization efficiency of spectrum resources, but also effectively avoids the degradation of communication quality and waste of resources due to interference, thereby achieving more efficient spectrum management and improvement of communication service quality. BRIEF DESCRIPTION OF THE DRAWINGS

[0097] Figure 1 This is a principle block diagram of a communication signal interference optimization system based on machine learning in the present invention;

[0098] Figure 2 The present invention is a schematic flow chart of the steps of a communication signal interference optimization method based on machine learning. DETAILED DESCRIPTION

[0099] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0100] Example 1

[0101] See also Figure 1 , the present invention provides a communication signal interference optimization system based on machine learning, including a channel data acquisition module, a data processing and feature extraction module, a machine learning model module, an adaptive channel selection optimization module and an iterative optimization module;

[0102] The channel data acquisition module collects the channel data in the environment by installing a channel detection device group in the communication environment, and constructs a channel data group UCSDF, and transmits the channel data group UCSDF to the data processing and feature extraction module through wireless transmission;

[0103] The data processing and feature extraction module is used to receive the channel data group UCSDF in real time, and perform preprocessing and multiple feature engineering on the channel data group UCSDF to obtain the feature data group AUCSD;

[0104] The machine learning model module constructs a multi-level model by setting the model architecture, and uses historical communication data to train the multi-level model to generate a multi-stage prediction model ICPmodel. The feature data group AUCSD is input into the multi-stage prediction model ICPmodel, and the channel interference prediction value ICP is output;

[0105] The adaptive channel selection optimization module sorts the channels by the output channel interference prediction value ICP, obtains the channel C* with the least interference, sets the channel threshold ICPthresh and the channel interference prediction value ICP of the current channel, performs comparative analysis, and triggers the multi-channel switching mechanism Switch;

[0106] After the initial channel selection, the iterative optimization module collects interference data in real time through the signal interference detection equipment group, performs comprehensive calculations to obtain the comprehensive communication quality index CQV, presets the signal quality threshold X, performs comparative evaluation and analysis on the signal quality of the current channel, and executes the optimization mechanism based on the evaluation results to further optimize the channel selection strategy, dynamically adjust the channel threshold ICPthresh, and generate an iterative mechanism.

[0107] In this embodiment, the system acquires multi-dimensional channel data in the communication environment in real time by deploying a channel detection device group through the channel data acquisition module, constructs a channel data group UCSDF, and provides a solid foundation for subsequent data processing and model training. The data processing and feature extraction module generates an enhanced data group AUCSD through preprocessing and feature engineering of the channel data to ensure the quality of the data and the comprehensiveness of the features. On this basis, the machine learning model module constructs a multi-level model ICPmodel, and trains the model through historical data to output the channel interference prediction value ICP. The smooth execution of this process ensures that the system can accurately predict the interference of the channel and provide a scientific basis for channel selection. The adaptive channel selection optimization module sorts the channels according to the ICP value, selects the channel C* with the least interference, and triggers the multiple channel switching mechanism Switch by setting the dynamic threshold ICPthresh, realizing intelligent channel switching and optimization in a complex and changeable communication environment. The iterative optimization module further collects and calculates the communication quality index CQV in real time, and compares it with the preset signal quality threshold X, executes the adaptive optimization mechanism, and dynamically adjusts the channel selection strategy and threshold. Compared with traditional fixed thresholds and simple detection methods, this full-process adaptive optimization mechanism greatly improves the system's adaptability and anti-interference ability in complex communication environments, reduces the blindness and frequency of channel switching, and optimizes the system's stability and communication quality.

[0108] Example 2

[0109] This embodiment is explained in Example 1, please refer to Figure 1 ,Specifically: the channel data acquisition module includes a data acquisition unit and a data transmission unit;

[0110] The data collection unit collects channel data in the communication environment in real time by installing a channel detection device group in the communication environment, wherein the channel detection device group includes a signal detection device group and a signal quality detection device group;

[0111] The signal detection equipment group includes limit noise analyzers, multipath meters, spectrum analyzers, vector network analyzers, path loss test equipment, antenna array systems and impulse response meters;

[0112] The signal quality detection equipment group includes bit error rate test equipment, radio frequency test equipment and network performance monitoring tools;

[0113] Channel data includes signal data and signal quality data;

[0114] Signal data includes transmitter phase infection TPI, multipath reflection angle MRA, dynamic frequency offset DFS, electromagnetic reflection coefficient ERC, path loss factor PLF, spatial channel correlation coefficient SCC, delay spread time DLT and frequency phase modulation FPM;

[0115] Signal quality data includes the number of error bits Cw, the total number of transmitted bits Zc, the signal power Xg, the interference power Pi, the noise power Pn, the number of packet losses Db and the total number of packets sent Zf;

[0116] Then the signal data is introduced into nonlinear transformation and interaction terms to generate the channel data group UCSDF;

[0117] The channel data group UCSDF is as follows;

[0118]

[0119] By performing nonlinear transformation on the original parameters, complex features that are difficult to reflect with simple linear relationships can be captured;

[0120] Quadratic TPI 2 The squared term of the transmitter phase interference is calculated to capture the impact of TPI on system performance in the nonlinear range,

[0121] log(DFS) logarithmic transformation, which performs logarithmic transformation on the dynamic frequency shift to reduce the impact of extreme values ​​and emphasize relative changes;

[0122] Sin (MRA) sine transform, which performs logarithmic transformation on dynamic frequency deviation to reduce the impact of extreme values ​​and emphasize relative changes;

[0123] Interaction terms reflect the joint effects between two or more parameters, and these terms can reveal the hidden relationships between the original parameters;

[0124] The ERC×PLF parameter product is used to calculate the product of the electromagnetic reflection coefficient and the path loss factor, and to explore the comprehensive impact of their correlation on signal attenuation;

[0125] DLT 2 +FPM combination term, which directly adds the square term of the delay spread time to the frequency phase modulation to explore the combined effect of the two;

[0126] Cos(DFS×MRA) combines sine-cosine transform to perform cosine transform on the product of dynamic frequency offset and multipath reflection angle to capture their potential complex impact on channel correlation;

[0127] The data transmission unit is used to transmit the acquired channel data group UCSDF and signal quality data to the data processing and feature extraction module through the WIFI protocol.

[0128] In this embodiment, through the implementation of the channel data acquisition module, the system can acquire and transmit multi-dimensional channel data and signal quality data in the communication environment in real time. By introducing nonlinear transformations and interaction terms, the system successfully generates the channel data group UCSDF. These enhanced data can capture complex signal characteristics, reveal hidden channel relationships, and further provide a scientific basis for optimizing communication quality. Compared with traditional fixed threshold monitoring and simple interference detection methods, the system significantly improves the detection accuracy and comprehensiveness of channel status through advanced detection equipment and complex feature processing procedures. In addition, the data transmission unit efficiently transmits data through the WIFI protocol, ensuring the timely processing and analysis of channel data.

[0129] Example 3

[0130] This embodiment is explained in Example 2. Please refer to Figure 1 ,Specifically: the data processing and feature extraction module includes a data processing unit and a feature extraction unit;

[0131] The data processing unit is used to receive the collected channel data group UCSDF and signal quality data in real time, and pre-process the channel data group UCSDF and signal quality data, wherein the pre-processing includes filling in missing values ​​and removing outliers, and then performing normalization processing;

[0132] The feature extraction unit includes a multiple feature extraction unit and a channel quality feature extraction unit;

[0133] The multiple feature extraction unit is used to extract features from the channel data group UCSDF to obtain the channel feature set AUCSD. The multiple feature extraction process includes polynomial features, logarithmic transformation, sine and cosine transformation and feature interaction terms.

[0134] The channel feature set AUCSD is;

[0135]

[0136] The communication quality feature extraction unit is used to extract features based on the collected signal quality data, so as to obtain a signal quality data set, where the signal quality data includes a bit error rate BER, a signal to interference plus noise ratio SINR, and a packet loss rate PLR;

[0137] The bit error rate BER is calculated and extracted by the number of error bits Cw in the signal quality data and the total number of transmitted bits Zc. The calculation formula is:

[0138] The signal-to-interference-plus-noise ratio SINR is calculated and extracted by the signal power Xg, interference power Pi, and noise power Pn in the signal quality data. The calculation formula is:

[0139] The packet loss rate PLR ​​is calculated and extracted by the number of lost packets Db and the total number of sent packets Zf in the signal quality data. The calculation formula is:

[0140] In this embodiment, the data processing and feature extraction module of the system ensures that high-quality and highly available feature sets are extracted from the collected data through close collaboration between the data processing unit and the feature extraction unit. The data processing unit effectively preprocesses the channel data group UCSDF and signal quality data, including processing missing values, removing outliers, and normalizing, thereby ensuring the consistency and reliability of the data. The feature extraction unit generates an enhanced channel feature set AUCSD and signal quality data set through multiple feature extraction projects and channel quality feature extraction. These data sets cover complex nonlinear features, interaction terms, and quality indicators such as bit error rate BER, signal interference plus noise ratio SINR, and packet loss rate PLR. Compared with traditional simple feature extraction methods, the system greatly enriches the dimension and depth of features through multiple feature extraction technologies such as polynomial features, logarithmic transformation, and sine-cosine transformation, which can more accurately reflect the channel status and communication quality. This improvement significantly improves the prediction accuracy and stability of subsequent machine learning models, thereby achieving more accurate channel selection and better communication services.

[0141] Example 4

[0142] This embodiment is explained in Example 3, please refer to Figure 1 ,Specifically: the machine learning model module includes a machine learning model unit and a model training unit;

[0143] The machine learning model unit constructs a multi-level model by selecting a multi-level model architecture that combines deep neural network DNN, spatiotemporal convolution network STCN and ensemble learning to predict channel interference. It extracts channel nonlinear characteristics through deep neural network DNN and spatiotemporal correlation of spatiotemporal convolution network STCN channels, and then integrates multiple models through ensemble learning to optimize the stability of the multi-level model.

[0144] The model training unit uses a large number of historical channel data sets UCSDF l Perform preprocessing to obtain the historical channel feature set AUCSD l , train the multi-level model, obtain the multi-stage prediction model ICPmodel, and output the channel interference prediction value ICP;

[0145] The channel interference prediction value ICP is output through the following multi-stage prediction model ICPmodel;

[0146] ICP(c i )=f DNN+STCN+Ensemble (AUCSD l );

[0147] Where f DNN+STCN+Ensemble represents the combined function of the deep neural network DNN, the spatiotemporal convolutional network STCN and the integrated learning model, c i represents the i-th channel.

[0148] In this embodiment, the machine learning model module of the system achieves accurate prediction of channel interference by combining a multi-level model architecture of deep neural network DNN, spatiotemporal convolutional network STCN and ensemble learning. Through the deep neural network DNN, the system can extract nonlinear features in the channel data and capture complex changes in channel states; the spatiotemporal convolutional network STCN further explores the spatiotemporal correlation of the channel to ensure the responsiveness to the dynamic channel environment; and the ensemble learning model optimizes the stability and accuracy of the overall model through multi-model integration. The model training unit uses a large number of historical channel data sets UCSDF l Get the historical channel feature set AUCSD l, trained a multi-level model, generated a multi-stage prediction model ICPmodel, and output the channel interference prediction value ICP to guide channel selection. This multi-level model architecture significantly improves the prediction ability of complex channel interference, ensures the generalization performance of the model in different scenarios, and thus achieves significant improvement in channel selection and interference management. The system not only achieves more accurate channel interference prediction, but also effectively improves the overall performance and reliability of the communication system by optimizing the channel selection strategy.

[0149] Example 5

[0150] This embodiment is explained in Example 4. Please refer to Figure 1 ,Specifically: the adaptive channel selection optimization module includes an ,interference prediction and analysis unit and a dynamic adjustment and multiple ,switching determination unit;

[0151] The interference prediction and analysis unit predicts the channel feature set AUCSD of all current channels by using the trained multi-stage prediction model ICPmodel, obtains the channel interference prediction value ICP of each channel, sorts all channels according to the channel interference prediction value ICP, and selects the interference threshold channel C*;

[0152] The interference valley channel C* is screened by the following sorting method;

[0153]

[0154] In the formula, fhistory(c i ) represents channel c i The history representation function, α, β and γ represent the channel interference prediction value ICP(c i ), channel history representation function fhistory(c i ) and channel path loss factor PLF(c i ), c n Indicates n channels.

[0155] The dynamic adjustment and multiple switching determination unit is used to collect the channel feature set AUCSD of the current communication environment in real time after selecting the current interference valley channel C*, and input the multi-stage prediction model ICPmodel, output the channel interference prediction value ICP of the current interference valley channel C*, and compare and evaluate the preset channel threshold ICPthresh with the interference prediction value ICP of the current channel based on the signal interference standard, and trigger the multiple channel switching mechanism Switch;

[0156] The multi-channel switching mechanism Switch is triggered by the following evaluation content;

[0157]

[0158] When Switch is True, the channel switching is triggered and switched to the interference valley channel C*; True means switching, False means not switching, c d Indicates the current channel.

[0159] In this embodiment, the adaptive channel selection optimization module of the system realizes intelligent optimization of channel selection through the collaborative work of the interference prediction and analysis unit and the dynamic adjustment and multiple switching determination unit. The interference prediction and analysis unit uses the trained multi-stage prediction model ICPmodel to predict the channel feature set AUCSD of all current channels and obtain the channel interference prediction value ICP. The system sorts the channels according to the interference prediction value ICP value, selects the interference valley channel C*, and ensures that the optimal channel is selected for communication in a complex communication environment. The dynamic adjustment and multiple switching determination unit further collects the current communication environment data in real time, continuously evaluates the channel quality, and triggers the channel switching mechanism Switch when necessary to ensure the continuity and stability of communication. Compared with the traditional fixed threshold and simple switching mechanism, this module realizes the accuracy and flexibility of channel selection through real-time prediction, dynamic adjustment and intelligent switching, greatly improves the system's anti-interference ability and communication quality, significantly reduces the number of unnecessary channel switching, and optimizes the overall performance and resource utilization efficiency of the communication system.

[0160] Example 6

[0161] This embodiment is explained in Example 5. Please refer to Figure 1 ,Specifically: the iterative optimization module includes a channel quality analysis unit, a dynamic threshold adjustment unit and an iterative optimization unit;

[0162] The channel quality analysis unit includes a channel quality calculation unit and a channel quality evaluation unit;

[0163] The channel quality calculation unit is used to perform comprehensive calculation to obtain a comprehensive communication quality index CQV based on the extracted quality data set and the spatial channel correlation coefficient SCC in the channel feature set AUCSD;

[0164] The comprehensive communication quality index CQV is calculated and obtained by the following algorithm formula;

[0165]

[0166] In the formula, h represents the demand function, the specific form of which depends on the design requirements of the system, and log represents the logarithmic function;

[0167] The channel quality evaluation unit is used to preset the signal quality threshold X based on the average value of the historical normal communication quality, and then compare and evaluate it with the obtained comprehensive communication quality index CQV to analyze the communication quality of the current channel. The specific evaluation content is as follows;

[0168] When the comprehensive communication quality index CQV ≥ the signal quality threshold X, it indicates that the current channel communication quality is abnormal;

[0169] When the comprehensive communication quality indicator CQV is less than the signal quality threshold X, it indicates that the current channel communication quality is normal.

[0170] The dynamic threshold adjustment unit is used to combine the channel feature set AUCSD of the current channel with the channel interference prediction value ICP of the channel, update the multi-stage prediction model ICPmodel through an adaptive weighting mechanism, further optimize the channel selection strategy, and dynamically adjust the channel threshold ICPthresh when the current channel communication quality is identified to be abnormal;

[0171] The channel threshold ICPthresh is adjusted through the following dynamic adjustment methods:

[0172] ΔICPthresh=μ×(CQV-CQVtarget);

[0173] Where μ represents the adjustment coefficient, CQVtarget represents the target communication quality value, and △ICPthresh represents the dynamically adjusted channel threshold.

[0174] The iterative optimization unit is used to update the multi-stage prediction model ICPmodel and dynamically adjust the channel threshold ICPthresh, and then execute the iterative strategy. The specific contents are as follows;

[0175] When the comprehensive communication quality index CQV ≥ the signal quality threshold X, continue to iterate;

[0176] When the comprehensive communication quality indicator CQV is less than the signal quality threshold X, the iteration is stopped.

[0177] In this embodiment, the iterative optimization module of the system continuously improves the accuracy of channel selection and the communication quality of the system through the close cooperation of the channel quality analysis unit, the dynamic threshold adjustment unit and the iterative optimization unit. The channel quality analysis unit obtains the comprehensive communication quality index CQV through comprehensive calculation, accurately evaluates the communication quality of the current channel, and compares it with the historical communication quality to ensure real-time identification of abnormal channel quality. When the dynamic threshold adjustment unit detects abnormal communication quality, it combines the current channel feature set AUCSD and the interference prediction value ICP, updates the model through an adaptive weighting mechanism and dynamically adjusts the channel threshold ICPthresh, thereby optimizing the channel selection strategy. The iterative optimization unit executes the iterative strategy on this basis to ensure that the system is continuously adjusted and optimized until the target communication quality is achieved. Compared with the traditional static threshold setting and single optimization method, this module realizes the adaptive optimization capability of the system through dynamic adjustment and multiple iterations, significantly improves the stability and adaptability of the communication system, reduces the communication interruption and performance degradation caused by channel quality fluctuations, and finally achieves a comprehensive improvement in communication quality and resource utilization efficiency.

[0178] Example 7

[0179] See also Figure 1 and Figure 2 , a communication signal interference optimization method based on machine learning, comprising the following steps:

[0180] S1, by installing a channel detection device group in the communication environment, collecting channel data in the environment, and constructing a channel data group UCSDF, and transmitting the channel data group UCSDF through wireless transmission;

[0181] S2, receiving the channel data group UCSDF in real time, and performing preprocessing and multiple feature engineering on the channel data group UCSDF to obtain the feature data group AUCSD;

[0182] S3, set the model architecture, build a multi-level model, use historical communication data to train the multi-level model, generate a multi-stage prediction model ICPmodel, and input the feature data group AUCSD into the multi-stage prediction model ICPmodel, and output the channel interference prediction value ICP;

[0183] S4. Sort the channels by the output channel interference prediction value ICP, obtain the channel C* with the least interference, set the channel threshold ICPthresh and the channel interference prediction value ICP of the current channel, perform comparative analysis, and trigger the multi-channel switching mechanism Switch;

[0184] S5. After the initial channel selection, the signal interference detection equipment group is used to collect interference data in real time, and a comprehensive calculation is performed to obtain the comprehensive communication quality index CQV, and the signal quality threshold X is preset to perform comparative evaluation and analysis on the signal quality of the current channel. Based on the evaluation results, an optimization mechanism is executed to further optimize the channel selection strategy, dynamically adjust the channel threshold ICPthresh, and generate an iterative mechanism.

[0185] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and variations can be made to the embodiments without departing from the principles and spirit of the invention.

Claims

1. A communication signal interference optimization system based on machine learning, characterized in that: include: The channel data acquisition module collects channel data in the communication environment through a channel detection device group installed in the communication environment, constructs a channel data group, and transmits the channel data group to the data processing and feature extraction module; The data processing and feature extraction module is used to receive the channel data group in real time, and perform preprocessing and multiple feature engineering on the channel data group to obtain the channel feature set and signal quality data set; The machine learning model module is used to build a multi-level model and train the built multi-level model; Adaptive channel selection optimization module, which uses the training data transmitted by the machine learning model module to perform interference prediction analysis and trigger multiple channel switching mechanisms; The iterative optimization module is used to collect interference data in real time through the signal interference detection device group after the adaptive channel selection optimization module performs the initial signal selection, evaluate and optimize the current channel signal quality, and generate an iterative mechanism.

2. A communication signal interference optimization system based on machine learning according to claim 1, characterized in that: The channel data acquisition module includes a data acquisition unit and a data transmission unit; The data collection unit collects channel data in the communication environment in real time through a channel detection device group installed in the communication environment; The channel detection equipment group includes a signal detection equipment group and a signal quality detection equipment group; The signal detection equipment group includes a limit noise analyzer, a multipath measurement instrument, a spectrum analyzer, a vector network analyzer, a path loss test equipment, an antenna array system and an impulse response measurement instrument; The signal quality detection equipment group includes bit error rate test equipment, radio frequency test equipment and network performance monitoring tools; The channel data includes signal data and signal quality data; the signal data includes transmitter phase infection TPI, multipath reflection angle MRA, dynamic frequency offset DFS, electromagnetic reflection coefficient ERC, path loss factor PLF, spatial channel correlation coefficient SCC, delay spread time DLT and frequency phase modulation FPM; the signal quality data includes the number of error bits Cw, the total number of transmitted bits Zc, signal power Xg, interference power Pi, noise power Pn, packet loss number Db and total number of sent packets Zf; The data transmission unit is used to transmit the acquired channel data group and signal quality data to the data processing and feature extraction module through wireless transmission.

3. A communication signal interference optimization system based on machine learning according to claim 2, characterized in that: The channel data acquisition module introduces signal data into nonlinear transformation and interaction terms to generate a channel data group UCSDF; The channel data set UCSDF is as follows:

4. The communication signal interference optimization system based on machine learning according to claim 2, characterized in that: The data processing and feature extraction module includes a data processing unit and a feature extraction unit; The data processing unit is used to receive the collected channel data group and signal quality data in real time, and pre-process the channel data group and signal quality data, wherein the pre-processing includes filling in missing values ​​and removing outliers, and then performing normalization processing; The feature extraction unit includes a multiple feature extraction unit and a channel quality feature extraction unit; The multiple feature extraction unit is used to extract features from the channel data group to obtain a channel feature set; the multiple feature extraction process includes polynomial features, logarithmic transformation, sine and cosine transformation and feature interaction terms; The communication quality feature extraction unit is used to perform feature extraction based on the collected signal quality data, thereby obtaining a signal quality data set, wherein the signal quality data includes a bit error rate BER, a signal to interference plus noise ratio SINR, and a packet loss rate PLR.

5. The communication signal interference optimization system based on machine learning according to claim 4 is characterized in that: The channel feature set AUCSD is: The bit error rate BER is calculated and extracted by the number of error bits Cw and the total number of transmitted bits Zc in the signal quality data. The calculation formula is: The signal-to-interference-plus-noise ratio SINR is calculated and extracted by the signal power Xg, interference power Pi, and noise power Pn in the signal quality data, and the calculation formula is: The packet loss rate PLR ​​is calculated and extracted by the number of lost packets Db and the total number of sent packets Zf in the signal quality data. The calculation formula is:

6. A communication signal interference optimization system based on machine learning according to claim 5, characterized in that: The machine learning model module includes a machine learning model unit and a model training unit; The machine learning model unit constructs a multi-level model by selecting a multi-level model architecture combining a deep neural network DNN, a spatiotemporal convolutional network STCN and integrated learning, predicts channel interference, extracts channel nonlinear characteristics through the deep neural network DNN, and the spatiotemporal correlation of the spatiotemporal convolutional network STCN channel, and then integrates multiple models through integrated learning to optimize the stability of the multi-level model; The model training unit uses a large number of historical channel data sets for preprocessing to obtain the historical channel feature set AUCSD l , train a multi-level model, obtain a multi-stage prediction model, and output the channel interference prediction value ICP; The channel interference prediction value ICP is output by the following multi-stage prediction model ICPmodel; ICP(c i )=f DNN+STCN+Ensemble (AUCSD l ); In the formula, f DNN+STCN+Ensemble represents the combined function of the deep neural network DNN, the spatiotemporal convolutional network STCN and the integrated learning model, c i represents the i-th channel.

7. The communication signal interference optimization system based on machine learning according to claim 1, characterized in that: The adaptive channel selection optimization module sorts the channels according to the channel interference prediction value output by the machine learning model module, obtains the channel with the least interference, sets the channel threshold and compares and analyzes the channel interference prediction value of the current channel, and triggers the multi-channel switching mechanism.

8. The communication signal interference optimization system based on machine learning according to claim 7, characterized in that: The adaptive channel selection optimization module includes an interference prediction and analysis unit and a dynamic adjustment and multiple switching determination unit; The interference prediction and analysis unit predicts the channel feature sets of all current channels by using the trained multi-stage prediction model, obtains the channel interference prediction value of each channel, sorts all channels according to the channel interference prediction value, and selects the interference threshold channel C*; The interference valley channel C* is screened by the following sorting method; In the formula, fhistory(c i ) represents channel c i The history of the function; α represents the channel interference prediction value ICP (c i );β represents the channel history representation function fhistory(c i ); γ represents the channel path loss factor PLF(c i ) weight value; c n represents n channels; The dynamic adjustment and multiple switching determination unit is used to collect the channel feature set of the current communication environment in real time after selecting the current interference valley channel C*, and input the multi-stage prediction model, output the channel interference prediction value ICP of the current interference valley channel C*, and preset the channel threshold ICPthresh based on the signal interference standard, compare and evaluate the channel threshold ICPthresh with the current channel interference prediction value ICP, and trigger the multiple channel switching mechanism Switch; The triggering multi-channel switching mechanism Switch is triggered by the following evaluation content; When Switch is True, the channel switching is triggered and switched to the interference valley channel C*; True means switching, False means not switching, c d Indicates the current channel.

9. The communication signal interference optimization system based on machine learning according to claim 1, characterized in that: The iterative optimization module collects interference data in real time through a signal interference detection device group after the initial channel selection, performs comprehensive calculations to obtain comprehensive communication quality indicators, presets signal quality thresholds, performs comparative evaluation and analysis on the signal quality of the current channel, and executes an optimization mechanism based on the evaluation results, further optimizes the channel selection strategy, dynamically adjusts the channel threshold, and generates an iterative mechanism.

10. The communication signal interference optimization system based on machine learning according to claim 9, characterized in that: The iterative optimization module includes a channel quality analysis unit, a dynamic threshold adjustment unit and an iterative optimization unit; The channel quality analysis unit includes a channel quality calculation unit and a channel quality assessment unit; The channel quality calculation unit is used to perform comprehensive calculation to obtain a comprehensive communication quality indicator CQV based on the extracted quality data set and the spatial channel correlation coefficient SCC in the channel feature set AUCSD; The channel quality evaluation unit is used to preset a signal quality threshold X based on the average value of historical normal communication quality, and then compare and evaluate it with the obtained comprehensive communication quality indicator CQV to analyze the communication quality of the current channel; The dynamic threshold adjustment unit is used to combine the channel feature set AUCSD of the current channel with the channel interference prediction value ICP of the current channel, update the multi-stage prediction model ICPmodel through an adaptive weighting mechanism, further optimize the channel selection strategy, and dynamically adjust the channel threshold ICPthresh when identifying that the communication quality of the current channel is abnormal; The iterative optimization unit is used to execute the iterative strategy after updating the multi-stage prediction model and dynamically adjusting the channel threshold.

11. The communication signal interference optimization system based on machine learning according to claim 10, characterized in that: The comprehensive communication quality index CQV is calculated and obtained by the following algorithm formula; Wherein, h represents the demand function, and its specific form is set according to the design requirements of the system; log represents the logarithmic function; BER is the signal quality data including the bit error rate; SINR is the signal-to-interference-plus-noise ratio; PLR is the packet loss rate; The specific evaluation contents of the comparative evaluation are as follows: When the comprehensive communication quality index CQV ≥ the signal quality threshold X, it indicates that the current channel communication quality is abnormal; When the comprehensive communication quality index CQV is less than the signal quality threshold X, it means that the current channel communication quality is normal; The channel threshold ICPthresh is adjusted by the following dynamic adjustment method; ΔICPthresh=μ×(CQV-CQVtarget); In the formula, μ represents the adjustment coefficient, CQVtarget represents the target communication quality value, and △ICPthresh represents the dynamically adjusted channel threshold; The iteration strategy is as follows; When the comprehensive communication quality index CQV ≥ the signal quality threshold X, continue to iterate; When the comprehensive communication quality indicator CQV is less than the signal quality threshold X, the iteration is stopped.

12. An optimization method for a communication signal interference optimization system based on machine learning applied to any one of claims 1 to 11, characterized in that: The following steps are involved: S1. Collecting channel data in the communication environment through a channel detection device group installed in the communication environment, constructing a channel data group, and transmitting the channel data group wirelessly; S2, receiving a channel data group in real time, and performing preprocessing and multiple feature engineering on the channel data group to obtain a channel feature set and a signal quality data set; S3. Construct a multi-level model by setting a model architecture; use historical communication data to train the multi-level model, further generate a multi-stage prediction model, and input the feature data group into the multi-stage prediction model to output a channel interference prediction value; S4. Sort the channels according to the output channel interference prediction values, obtain the channel with the least interference, set a channel threshold, compare and analyze the set channel threshold with the channel interference prediction value of the current channel, and trigger a multi-channel switching mechanism; S5. After the initial channel selection, interference data is collected in real time through the signal interference detection equipment group, and comprehensive calculation is performed to obtain comprehensive communication quality indicators; A preset signal quality threshold X is compared and evaluated with the comprehensive communication quality index to analyze the signal quality of the current channel; based on the evaluation result, an optimization mechanism is executed to further optimize the channel selection strategy, dynamically adjust the channel threshold, and generate an iterative mechanism.

13. The optimization method according to claim 12, characterized in that: The channel data in step S1 includes signal data and signal quality data; The signal data includes transmitter phase infection TPI, multipath reflection angle MRA, dynamic frequency shift DFS, electromagnetic reflection coefficient ERC, path loss factor PLF, spatial channel correlation coefficient SCC, delay spread time DLT and frequency phase modulation FPM; The signal quality data includes the number of error bits Cw, the total number of transmitted bits Zc, the signal power Xg, the interference power Pi, the noise power Pn, the number of lost packets Db and the total number of sent packets Zf.

14. The optimization method according to claim 13, characterized in that: In step S1, a channel data group UCSDF is generated by introducing signal data into nonlinear transformation and interaction terms; The channel data set UCSDF is as follows:

15. The optimization method according to claim 14, characterized in that: The preprocessing in step S2 is to preprocess the channel data group and signal quality data, and the preprocessing includes filling in missing values ​​and removing outliers, and then normalizing; the multiple feature engineering processing includes polynomial features, logarithmic transformation, sine and cosine transformation and feature interaction terms.

16. The optimization method according to claim 15, characterized in that: The step S3 is to construct a multi-level model by selecting a multi-level model architecture combining a deep neural network DNN, a spatiotemporal convolutional network STCN and integrated learning, predict channel interference, extract channel nonlinear characteristics through the deep neural network DNN, and the spatiotemporal correlation of the spatiotemporal convolutional network STCN channel, and then integrate multiple models through integrated learning to optimize the stability of the multi-level model; Use the historical channel data group for preprocessing, obtain the historical channel feature set AUCSDl, train the multi-level model, obtain the multi-stage prediction model, and output the channel interference prediction value ICP; The channel interference prediction value ICP is output by the following multi-stage prediction model ICPmodel; ICP(c i )=f DNN+STCN+Ensemble (AUCSD l ); In the formula, f DNN+STCN+Ensemble represents the combined function of the deep neural network DNN, the spatiotemporal convolutional network STCN and the integrated learning model, c i represents the i-th channel.

17. The optimization method according to claim 16, characterized in that: The step S4 predicts the channel feature sets of all current channels by using the trained multi-stage prediction model to obtain the channel interference prediction value of each channel, and sorts all channels according to the channel interference prediction value to select the interference threshold channel C*; The interference valley channel C* is screened by the following sorting method; Where, ICP is the channel interference prediction value; fhistory(c i ) represents channel c i The history of the function; α represents the channel interference prediction value ICP (c i );β represents the channel history representation function fhistory(c i ); γ represents the channel path loss factor PLF(c i ) weight value; c n represents n channels; After selecting the current interference valley channel C*, the channel feature set of the current communication environment is collected in real time and input into the multi-stage prediction model to output the channel interference prediction value ICP of the current interference valley channel C*, and the channel threshold ICPthresh is preset based on the signal interference standard, and the channel threshold ICPthresh is compared and evaluated with the current channel interference prediction value ICP, and the multiple channel switching mechanism Switch is triggered; The triggering multi-channel switching mechanism Switch is triggered by the following evaluation content; When Switch is True, the channel switching is triggered and switched to the interference valley channel C*; True means switching, False means not switching, c d Indicates the current channel.

Images