Interference detection method based on OFDM communication system

Abstract

The invention relates to the technical field of communication interference detection, and particularly discloses an interference detection method based on an OFDM communication system, and the method comprises the steps: S1, signal preprocessing, S2, time-frequency joint analysis, S3, deep learning model construction, S4, multi-interference source positioning, S5, intelligent interference suppression, and S6, interference information feedback. According to the method, Gaussian noise and non-Gaussian interference can be effectively distinguished through a third-order cumulant and other high-order spectrum analysis method, nonlinear features of complex signals can be captured, and the detection accuracy of complex interference is remarkably improved; the combination of the convolutional neural network and the long short-term memory network can automatically extract complex features in interference signals, provide more accurate interference type classification, utilize a beam forming algorithm and a multi-antenna array technology to accurately position a plurality of interference sources, and through a space-time joint estimation method, under the condition of facing the plurality of interference sources, improve the accuracy of the interference sources. And the positioning precision is obviously improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of communication interference detection, and particularly relates to an interference detection method based on an OFDM communication system. Background Art

[0002] In modern wireless communication systems, especially communication systems based on orthogonal frequency division multiplexing (OFDM) technology, due to their high spectral efficiency and strong anti-multipath interference ability, they are widely used in various high-speed data transmission scenarios. However, with the increasing complexity of the wireless communication environment, various interference signals, including Gaussian noise, non-Gaussian interference, and sudden and time-varying interference, pose a severe challenge to the performance of OFDM systems. These interferences not only affect the transmission quality of signals but may also cause data transmission interruption or system performance degradation.

[0003] In terms of the identification and classification of interference signals, existing detection methods often rely on the statistical characteristics of signals, such as power spectral density, etc. These methods have limited effects in distinguishing Gaussian noise from non-Gaussian interference and are difficult to capture the non-linear characteristics in complex signals, resulting in low detection accuracy for complex interferences. Especially when the characteristics of interference signals are similar to or superimposed on background noise, existing detection methods often have difficulty in effectively distinguishing them, affecting the performance of communication systems.

[0004] In view of this, the inventor proposes an interference detection method based on an OFDM communication system to solve the above problems. Summary of the Invention

[0005] The purpose of the present invention is to provide an interference detection method based on an OFDM communication system to solve the problems raised in the above background art.

[0006] To achieve the above purpose, the present invention provides the following technical solutions:

[0007] An interference detection method based on an OFDM communication system, comprising:

[0008] S1. Signal preprocessing, including synchronization filtering and denoising, synchronizing the received signal, and using Kalman filtering to optimize the quality of the received signal in real time. Combining high-order spectral analysis, Gaussian noise and non-Gaussian interference signals are distinguished, and background noise and other interference signals are removed to obtain a pure signal;

[0009] S2. Time-frequency joint analysis, including frequency domain conversion, time-frequency analysis, and interference pattern extraction. Use the Fast Fourier Transform (FFT) to convert the pure signal obtained in step S1 from the time domain to the frequency domain. Adopt wavelet transform and Short-Time Fourier Transform (STFT) to capture the local changes of the signal, segment the signal, and then perform Fourier transform to provide dual information of time and frequency. Use time-frequency joint analysis to extract the spectrogram of the signal and calculate the Power Spectral Density (PSD).

[0010] S3. Build a deep learning model, including data annotation, feature extraction, and deep learning model training. Use the spectrogram obtained in step S2 and the results of high-order spectrum analysis in step S1 as inputs. Use Convolutional Neural Network (CNN) to automatically extract features, and use Long Short-Term Memory Network (LSTM) to learn and train the extracted time-frequency features to real-time identify whether there is interference in the received signal and classify the interference types.

[0011] S4. Multi-interference source localization. Through Multiple-Input Multiple-Output (MIMO) technology of multi-antenna array, use the spatial characteristics of the received interference signal, combine with beamforming algorithm to locate the interference source. Adopt edge computing to assist in localization, sink part of the computing tasks to edge devices, and share interference information through collaborative computing.

[0012] S5. Intelligent interference suppression. Based on the detected interference information, adopt reinforcement learning algorithm to dynamically adjust the frequency allocation and power allocation of subcarriers, avoid subcarriers or frequency bandwidths severely affected by interference, so as to effectively reduce the impact of interference.

[0013] S6. Interference information feedback. Feed back the detected interference types, locations, and interference source information to the network management center for subsequent interference management and suppression.

[0014] Preferably, the synchronization in step S1 includes symbol synchronization, frequency synchronization, and phase synchronization to ensure the clock alignment between the receiving end and the transmitting end. The Kalman filter is used to optimize the estimation of the received signal in real time, and its expression is:

[0015]

[0016] where The optimal estimation of the signal at time k;

[0017] K k : Kalman gain;

[0018] z k : Measurement value;

[0019] H: Observation matrix, defining the relationship between measurement and state;

[0020] Reduce the received noise interference through a Kalman filter, improve the accuracy of signal synchronization, and adapt to the time-varying channel environment.

[0021] Preferably, in step S1, the high-order spectrum analysis uses the third-order cumulant to analyze the non-Gaussian and non-linear characteristics of the signal. The formula is as follows:

[0022] C3(τ 1 ,τ 2 ) = Ε[x(t)x(t + τ 1 )x(t + τ 2 )] - Ε[x(t)]Ε[x(t + τ 1 )]Ε[x(t + τ 2 )]

[0023] where C3(τ 1 ,τ 2 ) is the third-order cumulant;

[0024] x(t) is the input signal;

[0025] τ 1 ,τ 2 are the delay parameters;

[0026] The third-order cumulant can enhance the ability to distinguish non-linear interference and Gaussian noise, and help better detect non-stationary interference.

[0027] Preferably, in step S2, the FFT converts the signal from the time domain to the frequency domain. The formula is as follows:

[0028]

[0029] x(n) is the time-domain signal;

[0030] X(k) is the frequency-domain signal;

[0031] N is the length of the signal;

[0032] The FFT is used to decompose the OFDM signal into different frequency components to facilitate the analysis of interference on each subcarrier;

[0033] The wavelet transform is used for time-frequency analysis. The formula is as follows:

[0034]

[0035] Wx(a, b) is the wavelet coefficient;

[0036] x(t) is the input signal;

[0037] ψ(t) is the wavelet function;

[0038] a is the scale parameter that controls the frequency resolution;

[0039] b: Translation parameter, controlling the time resolution;

[0040] By adjusting the scale and translation parameter, wavelet transform can analyze the time-domain and frequency-domain characteristics of the signal simultaneously and capture the sudden interference signal.

[0041] Preferably, in step S2, STFT is used for time-frequency joint analysis, and the formula is as follows:

[0042]

[0043] where STFTx(t,f): Time-frequency spectrum;

[0044] x(τ): Input signal;

[0045] w(τ - t): Window function;

[0046] STFT provides the time and frequency information of the signal and is used to analyze the interference signal that changes with time.

[0047] Preferably, in step S3, the forward propagation process of the convolutional neural network CNN is expressed as:

[0048] Z (l) = W (l) * A (l-1) + b (l)

[0049] where, Z (l) : The output of the l-th layer;

[0050] W (l) : Convolution kernel;

[0051] A (l-1) : The activation value of the previous layer;

[0052] b (l) : Bias;

[0053] CNN extracts the features of the spectrogram through convolution operations and automatically identifies the time-frequency features of the interference signal.

[0054] Preferably, in step S3, the expression of the long short-term memory network model LSTM is:

[0055] ft = σ(W f · [h t-1 , x t + b f )

[0056] it = σ(W i · [h t-1 , x t + b i)

[0057]

[0058] ot = σ(W o ·[h t-1 , x t + b o )

[0059] ht = ot Θ tanh(C t )

[0060] Among them, ft, it, ot: forget gate, input gate, and output gate;

[0061] Ct: cell state;

[0062] ht: hidden state;

[0063] xt: input vector;

[0064] LSTM is used to process time - series data, can capture the time - correlation of interference signals, and is suitable for the detection of continuous interference or burst interference.

[0065] Preferably, in step S4, the beam - forming algorithm of multiple antennas is used for multi - interference - source localization, and the formula is as follows:

[0066]

[0067] Among them, P(θ): spatial spectrum;

[0068] a(θ): array response vector;

[0069] En: noise subspace matrix;

[0070] Through the multi - antenna beam - forming method, the positions of multiple interference sources are located with high precision, and the interference suppression ability of the system is improved.

[0071] Preferably, in step S4, for edge computing, its task offloading and cooperation optimization expression is:

[0072]

[0073] Among them, Ctl: total computing cost;

[0074] Cc: computing cost of the i - th task;

[0075] Ct: transmission cost of the i - th task;

[0076] Edge computing can reduce the overall system latency and computing load by optimizing the allocation of computing tasks, and improve the real - time performance and efficiency of interference detection.

[0077] Preferably, in step S5, the reinforcement learning algorithm is used for spectrum hopping and resource allocation, and the formula is as follows:

[0078]

[0079] Among them, Q(s,a): the value of taking action a in state s;

[0080] α: learning rate;

[0081] γ: discount factor;

[0082] r: reward value;

[0083] The Q-Learning algorithm optimizes the spectrum resource allocation of the OFDM system by learning the relationship between the spectrum state and interference, and realizes interference avoidance.

[0084] Compared with the prior art, the beneficial effects of the present invention are:

[0085] (1) Through the third-order cumulant high-order spectrum analysis method, the present invention can effectively distinguish Gaussian noise and non-Gaussian interference, capture the non-linear characteristics of complex signals, and significantly improve the detection accuracy of complex interference; the combination of convolutional neural network and long short-term memory network can automatically extract complex features in interference signals and provide more accurate interference type classification.

[0086] (2) By using the beamforming algorithm and multi-antenna array technology, the present invention can accurately locate the positions of multiple interference sources. Through the space-time joint estimation method, in the face of multiple interference sources, the positioning accuracy is significantly improved. Using intelligent algorithms such as reinforcement learning to dynamically adjust the subcarrier frequency allocation can effectively avoid interference frequency bands and improve the utilization efficiency of spectrum resources.

[0087] (3) By allocating computing tasks to edge nodes, the system response delay is significantly reduced, making interference detection and processing more real-time, improving the system's response speed and processing capacity. After using adaptive filtering and high-order spectrum analysis, the signal quality and transmission stability are improved. Combining wavelet transform and STFT for time-frequency analysis can accurately detect sudden and time-varying interference signals, and the detection of interference signals is more comprehensive and accurate. Sinking the computing tasks to edge nodes reduces the computing pressure and data transmission cost of the central server, reduces the overall cost and complexity of the system. Description of the Drawings

[0088] Figure 1 It is a flowchart of an interference detection method based on an OFDM communication system of the present invention;

[0089] Figure 2 It is a block diagram of the signal preprocessing composition of the present invention;

[0090] Figure 3 It is the block diagram of the time-frequency joint analysis of the present invention;

[0091] Figure 4 It is the block diagram of constructing the deep learning model of the present invention;

[0092] Figure 5 It is the block diagram of the synchronization of the present invention;

[0093] Figure 6 It is the block diagram of the interference information feedback of the present invention. Specific embodiments

[0094] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0095] Embodiment 1:

[0096] Please refer to Figures 1 to 6 As shown, an interference detection method based on an OFDM communication system includes the following steps:

[0097] S1. Signal preprocessing, including synchronization filtering and denoising, synchronize the received signal, and use Kalman filtering to optimize the quality of the received signal in real time. Combining high-order spectrum analysis, distinguish Gaussian noise from non-Gaussian interference signals, remove background noise and other interference signals, and obtain a pure signal;

[0098] S2. Time-frequency joint analysis, including frequency-domain conversion, time-frequency analysis, and interference pattern extraction. Use the fast Fourier transform FFT to convert the pure signal obtained in step S1 from the time domain to the frequency domain. Adopt wavelet transform and short-time Fourier transform STFT to capture the local changes of the signal and segment the signal before performing Fourier transform, providing dual information of time and frequency. Use time-frequency joint analysis to extract the spectrogram of the signal and calculate the power spectral density PSD;

[0099] S3. Construct a deep learning model, including data annotation, feature extraction, and deep learning model training. Use the spectrogram obtained in step S2 and the results of the high-order spectrum analysis in step S1 as inputs, use a convolutional neural network CNN to automatically extract features, and use a long short-term memory network model LSTM to learn and train the extracted time-frequency features to identify in real time whether there is interference in the received signal and classify the interference types;

[0100] S4. Multi-interference source localization. Through the multi-antenna array MIMO technology, utilize the spatial characteristics of the received interference signals, and combine with beamforming algorithms such as MUSIC to localize the interference sources. Adopt edge computing to assist in localization, sink some computing tasks such as interference detection, classification, and localization to the edge devices, and share interference information through collaborative computing.

[0101] S5. Intelligent interference suppression. Based on the detected interference information, adopt reinforcement learning algorithms to dynamically adjust the frequency allocation and power allocation of subcarriers, avoid subcarriers or frequency bandwidths severely affected by interference, thereby effectively reducing the impact of interference.

[0102] S6. Interference information feedback. Feed back the detected interference type, location, and interference source information to the network management center for subsequent interference management and suppression.

[0103] As can be seen from the above, through high-order spectral analysis methods such as third-order cumulants, Gaussian noise and non-Gaussian interference can be effectively distinguished, the non-linear characteristics of complex signals can be captured, and the detection accuracy of complex interference can be significantly improved; the combination of convolutional neural network (CNN) and long short-term memory network (LSTM) can automatically extract complex features in interference signals and provide more accurate interference type classification.

[0104] By using beamforming algorithms and multi-antenna array technologies, the positions of multiple interference sources can be accurately located. Through space-time joint estimation methods, in the face of multiple interference sources, the positioning accuracy has been significantly improved. Using intelligent algorithms such as reinforcement learning to dynamically adjust the subcarrier frequency allocation can effectively avoid interference frequency bands and improve the utilization efficiency of spectrum resources.

[0105] Embodiment 2:

[0106] Reference Figure 1 As shown, the synchronization in step S1 includes symbol synchronization, frequency synchronization, and phase synchronization to ensure the clock alignment between the receiving end and the sending end. The Kalman filter is used to optimize the estimation of the received signal in real time, and its expression is:

[0107]

[0108] Where, The optimal estimation of the signal at time k;

[0109] K k : Kalman gain;

[0110] z k : Measured value;

[0111] H: Observation matrix, defining the relationship between measurement and state;

[0112] Reduce the received noise interference through a Kalman filter, improve the accuracy of signal synchronization, and adapt to the time-varying channel environment.

[0113] Specifically, in step S1, the high-order spectrum analysis uses the third-order cumulant to analyze the non-Gaussian and non-linear characteristics of the signal. The formula is as follows:

[0114] C3(τ 1 ,τ 2 ) = Ε[x(t)x(t + τ 1 )x(t + τ 2 )] - Ε[x(t)]Ε[x(t + τ 1 )]Ε[x(t + τ 2 )]

[0115] Among them, C3(τ 1 ,τ 2 ) is the third-order cumulant;

[0116] x(t) is the input signal;

[0117] τ 1 ,τ 2 are the time-delay parameters;

[0118] The third-order cumulant can enhance the ability to distinguish non-linear interference and Gaussian noise, and help better detect non-stationary interference.

[0119] Specifically, in step S2, the FFT converts the signal from the time domain to the frequency domain. The formula is as follows:

[0120]

[0121] x(n) is the time-domain signal;

[0122] X(k) is the frequency-domain signal;

[0123] N is the length of the signal;

[0124] The FFT is used to decompose the OFDM signal into different frequency components to facilitate the analysis of interference on each subcarrier;

[0125] The wavelet transform is used for time-frequency analysis. The formula is as follows:

[0126]

[0127] Wx(a, b) is the wavelet coefficient;

[0128] x(t) is the input signal;

[0129] ψ(t) is the wavelet function;

[0130] a is the scale parameter, which controls the frequency resolution;

[0131] b: Translation parameter, which controls the time resolution;

[0132] By adjusting the scale and translation parameter, wavelet transform can analyze the time-domain and frequency-domain characteristics of the signal simultaneously and capture the sudden interference signal.

[0133] Specifically, in step S2, STFT is used for time-frequency joint analysis, and the formula is as follows:

[0134]

[0135] where STFTx(t,f): Time-frequency spectrum;

[0136] x(τ): Input signal;

[0137] w(τ - t): Window function;

[0138] STFT provides the time and frequency information of the signal and is used to analyze the interference signal that changes with time.

[0139] Specifically, the forward propagation process of the convolutional neural network CNN in step S3 is expressed as:

[0140] Z (l) = W (l) * A (l-1) + b (l)

[0141] where, Z (l) : The output of the l-th layer;

[0142] W (l) : Convolution kernel;

[0143] A (l-1) : The activation value of the previous layer;

[0144] b (l) : Bias;

[0145] CNN extracts the features of the spectrogram through convolution operations and automatically identifies the time-frequency features of the interference signal.

[0146] Specifically, the expression of the long short-term memory network model LSTM in step S3 is:

[0147] ft = σ(W f · [h t-1 , x t + b f )

[0148] it = σ(W i · [h t-1 , x t + b i)

[0149]

[0150] ot = σ(W o ·[h t-1 , x t + b o )

[0151] ht = ot Θ tanh(C t )

[0152] Among them, ft, it, ot: forgetting gate, input gate, and output gate;

[0153] Ct: cell state;

[0154] ht: hidden state;

[0155] xt: input vector;

[0156] LSTM is used to process time - series data, can capture the time - correlation of interference signals, and is suitable for the detection of continuous interference or burst interference.

[0157] Specifically, the beam - forming algorithm for multiple antennas in step S4 is used for multi - interference - source localization, and the formula is as follows:

[0158]

[0159] Among them, P(θ): spatial spectrum;

[0160] a(θ): array response vector;

[0161] En: noise subspace matrix;

[0162] Through the multi - antenna beam - forming method, the positions of multiple interference sources can be accurately located, and the interference suppression ability of the system can be improved.

[0163] Specifically, for edge computing in step S4, its task offloading and collaborative optimization expression is:

[0164]

[0165] Among them, Ctl: total computing cost;

[0166] Cc: computing cost of the i - th task;

[0167] Ct: transmission cost of the i - th task;

[0168] Edge computing can reduce the overall system latency and computing load by optimizing the allocation of computing tasks, and improve the real - time performance and efficiency of interference detection.

[0169] Specifically, in step S5, the reinforcement learning algorithm is used for spectrum hopping and resource allocation, and the formula is as follows:

[0170]

[0171] Among them, Q(s,a): the value of taking action a in state s;

[0172] α: learning rate;

[0173] γ: discount factor;

[0174] r: reward value;

[0175] The Q-Learning algorithm optimizes the spectrum resource allocation of the OFDM system by learning the relationship between the spectrum state and interference, and realizes interference avoidance.

[0176] As can be seen from the above, by allocating computing tasks to edge nodes, the system response delay is significantly reduced, making interference detection and processing more real-time, improving the system's response speed and processing ability. After using adaptive filtering and high-order spectrum analysis, the signal quality and transmission stability are improved;

[0177] Combining wavelet transform and STFT for time-frequency analysis can accurately detect sudden and time-varying interference signals, making the detection of interference signals more comprehensive and accurate. Sinking computing tasks to edge nodes reduces the computing pressure and data transmission cost of the central server, and reduces the overall cost and complexity of the system.

[0178] Embodiment 3:

[0179] This design is specifically applied to modern mobile communication systems (such as LTE or 5G NR), and OFDM is widely used in broadband wireless communication. However, these systems are vulnerable to various types of interference, such as burst pulse interference, broadband noise, adjacent channel interference, etc. This design demonstrates how to apply the above interference detection methods to improve the accuracy of interference detection, the accuracy of interference localization, and the system's adaptive ability through simulation and experimental analysis.

[0180] Furthermore, a set of 5G base stations is deployed in a certain city, and the OFDM system is used for wireless communication in the 3.5 GHz band. Due to the operation of industrial equipment in this area, the communication system is affected by narrowband interference and burst pulse interference from industrial equipment. The system needs to detect, classify, and locate the interference, and adaptively adjust the frequency resource allocation to reduce the impact of interference on the system performance.

[0181] 1. Signal preprocessing

[0182] The number of OFDM subcarriers is 128, the length of each symbol is 1024 sampling points, the signal-to-noise ratio (SNR) of the channel is 20 dB, and it is subject to narrowband interference during transmission. The center frequency is 3.51 GHz, the bandwidth is 200 kHz, and there is also impulse interference with a duration of 0.5 ms.

[0183] Processing: Use Kalman filtering for synchronization and denoising to filter out part of the broadband noise and correct the symbol synchronization error.

[0184] Effect: After denoising and synchronization correction, the system SNR is increased by about 2 dB, and the symbol error rate (SER) is reduced from 0.02 to 0.015.

[0185] 2. Time-frequency joint analysis

[0186] The narrowband interference is mainly concentrated between subcarriers 32 and 36, and at the same time, the impulse interference appears in the symbol time domain with a short duration.

[0187] Processing: Use the joint wavelet transform and short-time Fourier transform (STFT) for time-frequency analysis. The frequency-domain concentration characteristics of narrowband interference and the time-domain instantaneous changes of impulse interference are captured through the wavelet transform. STFT further analyzes the time-frequency distribution of time-varying interference.

[0188] Effect: Through time-frequency analysis, the frequency band of narrowband interference is accurately located, and the duration and position of impulse interference are determined. At this time, the interference recognition rate reaches more than 95%.

[0189] Higher-order spectrum analysis

[0190] In the denoised signal, there is still some non-Gaussian non-linear interference mixed in.

[0191] Processing: Apply the third-order cumulant for higher-order spectrum analysis, especially focusing on the non-linear characteristics of the signal to distinguish Gaussian noise and non-Gaussian interference signals.

[0192] Effect: The third-order cumulant can further distinguish non-Gaussian impulse interference, successfully detect the non-linear interference in the system, and the detection accuracy is improved by 5%, thus further improving the recognition ability for complex interference.

[0193] 3. Construct a deep learning model

[0194] The base station receives a large amount of data after time-frequency analysis, including spectrograms affected by narrowband interference and impulse interference. The data set contains 2000 labeled training samples, of which 80% are used for training and 20% are used for testing.

[0195] Processing: Use a Convolutional Neural Network (CNN) for spectrogram feature extraction. Subsequently, use a CNN-LSTM combined model to detect time-series related interference signals.

[0196] Effect: After 100 rounds of training, the accuracy of the model on the test set reaches 97%.

[0197] 4. Multi-interference source localization

[0198] Eight antenna arrays are arranged between multiple base stations, and MIMO technology is used. Industrial equipment, as an interference source, is located 500 meters away from the base station, and its position is unknown.

[0199] Processing: Use the multi-antenna beamforming method, i.e., the MUSIC method, for beamforming and interference source localization. Determine the azimuth and distance of the interference source by jointly measuring the Angle of Arrival (AOA) of the signal through multiple antennas.

[0200] Effect: The positioning error of the MUSIC method is less than 5 meters, and the positioning accuracy reaches 99%, which can accurately identify the location of industrial equipment.

[0201] 5. Intelligent interference suppression

[0202] After detecting interference, the base station needs to adjust subcarrier allocation to reduce the impact of interference on communication quality. At this time, the base station uses frequency hopping technology to dynamically adjust subcarriers and avoid subcarrier frequency bands affected by interference.

[0203] Processing: Use the Q-Learning reinforcement learning algorithm. The base station adaptively adjusts subcarrier allocation by continuously learning the changes in the spectrum environment. The initial action space is set as different subcarrier frequency bands, and the state space is set as different channel conditions and interference detection situations.

[0204] Effect: After 500 iterations of training, the reinforcement learning algorithm successfully finds the optimal frequency hopping strategy. The system throughput increases by 15%, the interference suppression effect is significant, and the symbol error rate is reduced to 0.005.

[0205] Edge computing assistance and system optimization

[0206] Multiple base stations cooperate through edge computing nodes to process interference detection tasks, so as to reduce communication latency and the load on the central server. The amount of data processed by each base station is 200MB.

[0207] Processing: Adopt an edge computing architecture. Some interference detection and localization tasks are processed at the edge nodes. Cooperative computing reduces the load of data transmission to the central server and feeds back interference information in real time through the edge nodes.

[0208] Effect: Edge computing significantly reduces communication latency. The system response time is reduced from the original 100 ms to 50 ms, and the interference handling efficiency is increased by 30%.

[0209] 6. Interference information feedback

[0210] Feedback the detected interference type, location, and interference source information to the network management center for subsequent interference management and suppression.

[0211] As can be seen from the above, by combining technologies such as high-order spectrum analysis, time-frequency joint analysis, deep learning models, data augmentation, and edge computing, the interference detection ability of the OFDM communication system can be significantly improved, the spectrum resource allocation can be optimized, the real-time processing ability can be enhanced, and the overall performance and user experience of the system can be improved. These improvements provide an efficient and reliable solution for interference management in complex environments.

[0212] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An interference detection method based on an OFDM communication system, characterized in that: The following steps are involved: S1, signal preprocessing, including synchronous filtering and denoising, synchronizes the received signal, and uses Kalman filtering to optimize the quality of the received signal in real time, combines high-order spectrum analysis to distinguish Gaussian noise from non-Gaussian interference signals, removes background noise and other interference signals, and obtains a pure signal; S2, time-frequency joint analysis, including frequency domain conversion, time-frequency analysis and interference pattern extraction, using fast Fourier transform FFT to convert the pure signal obtained in step S1 from the time domain to the frequency domain, using wavelet transform and short-time Fourier transform STFT to capture the local changes of the signal and perform Fourier transform after segmenting the signal to provide dual information of time and frequency, using time-frequency joint analysis to extract the spectrum of the signal and calculate the power spectral density PSD; S3, building a deep learning model, including data labeling, feature extraction and deep learning model training, taking the spectrum obtained in step S2 and the result of the high-order spectrum analysis in step S1 as input, using the convolutional neural network CNN to automatically extract features, using the long short-term memory network model LSTM to learn and train the extracted time-frequency features, and identifying in real time whether there is interference in the received signal, and classifying the interference type; S4, multiple interference source positioning, through multi-antenna array MIMO technology, using the spatial characteristics of the received interference signal, combined with the beamforming algorithm, the interference source is located, edge computing is used to assist positioning, some computing tasks are transferred to edge devices, and interference information is shared through collaborative computing; S5, intelligent interference suppression, based on the detected interference information, uses reinforcement learning algorithm to dynamically adjust the frequency allocation and power allocation of subcarriers to avoid subcarriers or frequency bandwidths that are severely interfered with, thereby effectively reducing the impact of interference; S6: Interference information feedback: Feedback the detected interference type, location and interference source information to the network management center for subsequent interference management and suppression.

2. The interference detection method based on OFDM communication system according to claim 1, characterized in that: The synchronization in step S1 includes symbol synchronization, frequency synchronization and phase synchronization, ensuring that the clocks of the receiving end and the transmitting end are aligned. The Kalman filter is used to optimize the estimation of the received signal in real time, and its expression is: in, The optimal estimate of the signal at time k; K k : Kalman gain; z k : measurement value; H: Observation matrix, defining the relationship between measurements and states.

3. The interference detection method based on OFDM communication system according to claim 1, characterized in that: In step S1, the high-order spectrum analysis uses the third-order cumulant to analyze the non-Gaussianity and nonlinear characteristics of the signal. The formula is as follows: C3(τ1,τ2)=Ε[x(t)x(t+τ1)x(t+τ2)]-Ε[x(t)]Ε[x(t+τ1)]Ε[x(t+τ2)] Where, C3(τ1,τ2): third-order cumulant; x(t): input signal; τ1,τ2: delay parameters.

4. The interference detection method based on OFDM communication system according to claim 1, characterized in that: In step S2, FFT converts the signal from the time domain to the frequency domain, and the formula is as follows: k=0,1,...,N-1 x(n): time domain signal; X(k): frequency domain signal; N: length of the signal; Wavelet transform is used for time-frequency analysis, and the formula is as follows: Wx(a,b): wavelet coefficients; x(t): input signal; ψ(t): wavelet function; a: scale parameter, controlling frequency resolution; b: Translation parameter, controlling the temporal resolution.

5. The interference detection method based on OFDM communication system according to claim 1, characterized in that: In step S2, STFT is used for time-frequency joint analysis, and the formula is as follows: Where STFTx(t,f): time-frequency spectrum; x(τ): input signal; w(τ-t): window function; STFT provides the time and frequency information of the signal and is used to analyze interference signals that vary over time.

6. The interference detection method based on OFDM communication system according to claim 1, characterized in that: The forward propagation process of the convolutional neural network CNN in step S3 is expressed as: Z (l) =W (l) *A (l-1) +b (l) Among them, Z (l) : The output of the lth layer; W (l) : convolution kernel; A (l-1) : The activation value of the previous layer; b (l) : bias.

7. The interference detection method based on OFDM communication system according to claim 1, characterized in that: The expression of the long short-term memory network model LSTM in step S3 is: ft=σ(W f ·[h t-1 ,x t ]+b f ) it=σ(W i ·[h t-1 ,x t ]+b i ) ot=σ(W o ·[h t-1 ,x t ]+b o ) ht=otΘtanh(C t ) Among them, ft, it, ot: forget gate, input gate and output gate; Ct: cell status; ht: hidden state; xt: input vector.

8. The interference detection method based on OFDM communication system according to claim 1, characterized in that: The beamforming algorithm of multiple antennas in step S4 is used to locate multiple interference sources, and the formula is as follows: Where, P(θ): spatial spectrum; a(θ): array response vector; En: noise subspace matrix.

9. The interference detection method based on OFDM communication system according to claim 1, characterized in that: In step S4, the edge computing task offloading and collaborative optimization expressions are: Where, Ctl: total computation cost; Cc: the computational cost of the i-th task; Ct: The transmission cost of the i-th task.

10. The interference detection method based on OFDM communication system according to claim 1, characterized in that: In step S5, the reinforcement learning algorithm is used for spectrum hopping and resource allocation, and the formula is as follows: Among them, Q(s,a): the value of taking action a in state s; α: learning rate; γ: discount factor; r: reward value.

Images