A channel adaptive signal detection method and device

Abstract

The application discloses a kind of channel adaptive signal detection method and device, the method includes: obtaining large bandwidth offline signal data information;The large bandwidth offline signal data information is handled, and training data set is obtained;Using the training data set, the preset signal detection model is trained, and the optimization signal detection model is obtained;Using the optimization signal detection model, the large bandwidth offline signal data information to be detected is handled, and channel adaptive signal detection result is obtained.The method of the application can efficiently screen out signals of different bandwidths by multi-stage channelization, effectively saving detection resources by eliminating channels that cannot detect signals, and speeding up the reasoning speed of signals.Meanwhile, the method of multi-stage channelization is also conducive to screening out signals with smaller bandwidth, improving the accuracy of intelligent detection and identification.

Description

Technical Field

[0001] This invention relates to the field of intelligent signal detection and recognition technology, and in particular to a channel-adaptive signal detection method and apparatus. Background Technology

[0002] In the field of signal detection and recognition, traditional detection algorithms have long been used in related detection equipment. However, these algorithms also have certain limitations, such as low detection accuracy in low signal-to-noise ratio scenarios. The emergence of deep learning has attracted widespread attention in the industry, as it demonstrates excellent performance in image-related detection, especially supervised learning-based target detection algorithms. These algorithms can efficiently learn subtle features of targets in images and achieve intelligent detection and recognition through continuous iterative learning.

[0003] Supervised learning model training for large-bandwidth signal recognition has always faced challenges in data labeling. Different imaging devices generate different features when converting the same raw data into time-frequency maps, leading to labeling difficulties and requiring relabeling during retraining on different devices. In signal recognition, large-bandwidth signals are identified by channelization at different levels and channels. Performing multi-level channelization only when the current channel lacks a signal results in resource waste and slow detection speed. Therefore, efficiently addressing the issues of redundant labeling and the slow, difficult, and low-accuracy recognition of large-bandwidth signals is crucial. Summary of the Invention

[0004] The technical problem to be solved by this invention is to provide a channel-adaptive signal detection method and apparatus, which involves inputting large-bandwidth offline signal data into labeling software; channelizing the signal data into three levels: primary channelization, secondary channelization, and tertiary channelization, with equal division of the channel frequency domain and an overlap rate of 20% for each channel; processing each channel data at each channelization level to convert it into image data; labeling the images containing the signal by mapping the signal's location information onto the original data through time and frequency domain calculations and saving it as label information to construct a dataset; training a signal detection model using the dataset; accessing real-time data and performing channelization processing on the large-bandwidth signal; processing the data in the channel from the primary channelization result, converting it into image data, and then using the trained signal detection model to identify the presence and type of signal in the channel. Then, based on the signal detection results in the channel—whether a signal is detected or not—the system adaptively decides whether to continue to the next level of channelization. Channels with detected signals do not proceed to the next level of channelization, while channels without detected signals undergo a second level of channelization. The data in each channel after the second level of channelization is processed again, converted into image data, and then the trained signal detection model is used to detect the presence and type of signal after the second level of channelization. Channels without detected signals undergo a third level of channelization. The data in each channel is processed, converted into image data, and then the trained signal detection model is used for a final signal detection. This invention, by using data annotation mapping and channelization methods, solves the problem of duplicate labeling caused by different time-frequency map styles generated by different mapping devices for the same data, and also addresses the issues of slow detection, difficulty in detection, and low detection rate of large-bandwidth signals in intelligent detection.

[0005] To address the aforementioned technical problems, a first aspect of the present invention discloses a channel adaptive signal detection method, the method comprising: S1, acquire high-bandwidth offline signal data information; S2, process the high-bandwidth offline signal data information to obtain a training dataset; S3, using the training dataset, train the preset signal detection model to obtain an optimized signal detection model; S4. Using the optimized signal detection model, the large-bandwidth offline signal data information to be detected is processed to obtain the channel adaptive signal detection result.

[0006] As an optional implementation, in the first aspect of the present invention, the processing of the large-bandwidth offline signal data information to obtain a training dataset includes: S21, Channelize the high-bandwidth offline signal data information to obtain channelized data information; The channelization includes primary channelization, secondary channelization, and tertiary channelization; The channelized data information includes primary channelized data information, secondary channelized data information, and tertiary channelized data information; S22, The channelized data information is processed to obtain channel image data information; the channel image data information includes primary channel image data information, secondary channel image data information and tertiary channel image data information. S23, process the channel image data information to obtain the tag information of the channel image data information; The channel image data information and the label information of the channel image data information constitute the training dataset.

[0007] As an optional implementation, in the first aspect of the present invention, the channelization of the large-bandwidth offline signal data information to obtain channelized data information includes: S211, Perform first-level channelization on the large bandwidth offline signal data information to obtain first-level channelized data information; The expression for the first-level channelization data information is: in, For the first level Channel number at frequency The channelized frequency domain data at that location, i.e., the first-level channelized data information. This is the primary channel number. For continuous frequency variables in the frequency domain, This refers to the frequency domain sampling data of the original received signal with ultra-wide bandwidth, i.e., the offline signal data information of large bandwidth, which is the original spectrum of the entire broadband segment without channel segmentation. This is an interval indicator function; frequencies falling within the channel's frequency band retain their original spectrum, while frequencies outside the band are set to zero, thus achieving frequency band truncation. For the first level Channel start frequency, , For the first level Channel termination frequency, This represents the theoretical bandwidth of a single channel at level one. For a fixed overlap rate in the channel, For the total bandwidth of offline signal data information, For frequency domain variables; Frequency domain information for high-bandwidth offline signal data. This is the overall starting frequency of the signal; S212, perform secondary channelization on the high-bandwidth offline signal data information to obtain secondary channelized data information; The expression for the secondary channelized data information is as follows: in, For the second level Channel frequency domain data information, i.e., secondary channelized data information. For secondary single-channel segmentation bandwidth, For the second level Channel start frequency, For the second level Channel termination frequency, This represents the total bandwidth of the remaining idle channels after the first-level detection. This is the frequency domain data of the idle channels after primary filtering. , This is the starting frequency for the secondary idle channel. It is an integer; S213, perform three-level channelization on the large bandwidth offline signal data information to obtain three-level channelized data information; The expression for the three-level channelized data information is as follows: in, For the third level Channel frequency domain data information, i.e., Level 3 channelized data information. It is a three-level single-channel extreme segmentation bandwidth. This represents the total bandwidth of the remaining idle channels after secondary detection. For the third level Channel start frequency, , For the third level Channel termination frequency, This is the starting frequency for the third-level idle channel. This is the frequency domain data of the idle channel after secondary filtering.

[0008] As an optional implementation, in the first aspect of the present invention, processing the channelized data information to obtain channel image data information includes: S221, Process the first-level channelization data information to obtain first-level channel image data information; S222, Process the secondary channelization data information to obtain secondary channel image data information; S223, Process the three-level channelization data information to obtain three-level channel image data information; S224, integrate the first-level channel image data information, the second-level channel image data information and the third-level channel image data information to obtain channel image data information.

[0009] As an optional implementation, in the first aspect of the present invention, processing the first-level channelization data information to obtain first-level channel image data information includes: S2211, Process the first-level channelized data information to obtain first-level time-frequency characteristic information; S2212, The first-level time-frequency feature information is processed using a filtering model to obtain the first-level filtered time-frequency feature information; S2213, The first-level channelization data information is transformed to obtain first-level transformation feature information; S2214, the first-level filter time-frequency feature information and the first-level transform feature information are fused to obtain the first-level channel image data information.

[0010] As an optional implementation, in the first aspect of the present invention, processing the first-level channelization data information to obtain first-level time-frequency feature information includes: The first-level channelized data information is processed using a preset time-frequency feature information calculation model to obtain first-level time-frequency feature information; The preset time-frequency feature information calculation model expression is as follows: in, , This is primary time-frequency characteristic information. For integration variables, This serves as the baseline window length coefficient, ranging from 1.0 to 2.0. It is a bandwidth-adaptive Gaussian-weighted window function; The window length adaptive coefficient varies with the current channel bandwidth. Dynamic changes; This represents the system's maximum channel bandwidth. This is the signal amplitude correction coefficient, with a value ranging from 0.3 to 1.2. For mathematical expectation operations, the first-level channelized data information is frequency domain information. Performing an inverse Fourier transform on it yields the time-domain first-level channelized data. express.

[0011] As an optional implementation, in the first aspect of the present invention, the step of using the optimized signal detection model to process the large-bandwidth offline signal data information to be detected to obtain the channel adaptive signal detection result includes: S41, acquire the large-bandwidth offline signal data information to be detected; S42, perform first-level channelization on the large-bandwidth offline signal data information to be detected to obtain first-level data information; S43, using the optimized signal detection model, the first-level data information is processed to obtain the first-level detection result; S44, when the first-level detection result is that there is a signal, output the first-level detection result; the first-level detection result is the channel adaptive signal detection result; S45, when the first-level detection result is no signal, the first-level detection result is channelized to obtain second-level data information; S46, Process the secondary data information to obtain the secondary detection result; S47, when the secondary detection result indicates the presence of a signal, the secondary detection result is output; the secondary detection result is a channel adaptive signal detection result; S48, when the secondary detection result is no signal, the secondary detection result is channelized to the third level to obtain the third level data information; S49, The three-level data information is processed to obtain the three-level detection results; the three-level detection results are the channel adaptive signal detection results.

[0012] A second aspect of this invention discloses a channel adaptive signal detection device, the device comprising: The data acquisition module is used to acquire high-bandwidth offline signal data information; The training dataset construction module is used to process the high-bandwidth offline signal data information to obtain the training dataset; The model training module is used to train the preset signal detection model using the training dataset to obtain an optimized signal detection model. The signal detection module is used to process the large-bandwidth offline signal data information to be detected using the optimized signal detection model, and obtain the channel adaptive signal detection result.

[0013] As an optional implementation, in the second aspect of the present invention, the processing of the high-bandwidth offline signal data information to obtain a training dataset includes: S21, Channelize the high-bandwidth offline signal data information to obtain channelized data information; The channelization includes primary channelization, secondary channelization, and tertiary channelization; The channelized data information includes primary channelized data information, secondary channelized data information, and tertiary channelized data information; S22, The channelized data information is processed to obtain channel image data information; the channel image data information includes primary channel image data information, secondary channel image data information and tertiary channel image data information. S23, process the channel image data information to obtain the tag information of the channel image data information; The channel image data information and the label information of the channel image data information constitute the training dataset.

[0014] As an optional implementation, in the second aspect of the present invention, the channelization of the large-bandwidth offline signal data information to obtain channelized data information includes: S211, Perform first-level channelization on the large bandwidth offline signal data information to obtain first-level channelized data information; The expression for the first-level channelization data information is: in, For the first level Channel number at frequency The channelized frequency domain data at that location, i.e., the first-level channelized data information. This is the primary channel number. For continuous frequency variables in the frequency domain, This refers to the frequency domain sampling data of the original received signal with ultra-wide bandwidth, i.e., the offline signal data information of large bandwidth, which is the original spectrum of the entire broadband segment without channel segmentation. This is an interval indicator function; frequencies falling within the channel's frequency band retain their original spectrum, while frequencies outside the band are set to zero, thus achieving frequency band truncation. For the first level Channel start frequency, , For the first level Channel termination frequency, This represents the theoretical bandwidth of a single channel at level one. For a fixed overlap rate in the channel, For the total bandwidth of offline signal data information, For frequency domain variables; Frequency domain information for high-bandwidth offline signal data. This is the overall starting frequency of the signal; S212, perform secondary channelization on the high-bandwidth offline signal data information to obtain secondary channelized data information; The expression for the secondary channelized data information is as follows: in, For the second level Channel frequency domain data information, i.e., secondary channelized data information. For secondary single-channel segmentation bandwidth, For the second level Channel start frequency, For the second level Channel termination frequency, This represents the total bandwidth of the remaining idle channels after the first-level detection. This is the frequency domain data of the idle channels after primary filtering. , This is the starting frequency for the secondary idle channel. It is an integer; S213, perform three-level channelization on the large bandwidth offline signal data information to obtain three-level channelized data information; The expression for the three-level channelized data information is as follows: in, For the third level Channel frequency domain data information, i.e., Level 3 channelized data information. It is a three-level single-channel extreme segmentation bandwidth. This represents the total bandwidth of the remaining idle channels after secondary detection. For the third level Channel start frequency, , For the third level Channel termination frequency, This is the starting frequency for the third-level idle channel. This is the frequency domain data of the idle channel after secondary filtering.

[0015] As an optional implementation, in the second aspect of the present invention, processing the channelized data information to obtain channel image data information includes: S221, Process the first-level channelization data information to obtain first-level channel image data information; S222, Process the secondary channelization data information to obtain secondary channel image data information; S223, Process the three-level channelization data information to obtain three-level channel image data information; S224, integrate the first-level channel image data information, the second-level channel image data information and the third-level channel image data information to obtain channel image data information.

[0016] As an optional implementation, in the second aspect of the present invention, processing the first-level channelization data information to obtain first-level channel image data information includes: S2211, Process the first-level channelized data information to obtain first-level time-frequency characteristic information; S2212, The first-level time-frequency feature information is processed using a filtering model to obtain the first-level filtered time-frequency feature information; S2213, The first-level channelization data information is transformed to obtain first-level transformation feature information; S2214, the first-level filter time-frequency feature information and the first-level transform feature information are fused to obtain the first-level channel image data information.

[0017] As an optional implementation, in the second aspect of the present invention, processing the first-level channelization data information to obtain first-level time-frequency feature information includes: The first-level channelized data information is processed using a preset time-frequency feature information calculation model to obtain first-level time-frequency feature information; The preset time-frequency feature information calculation model expression is as follows: in, , This is primary time-frequency characteristic information. For integration variables, This serves as the baseline window length coefficient, ranging from 1.0 to 2.0. It is a bandwidth-adaptive Gaussian-weighted window function; The window length adaptive coefficient varies with the current channel bandwidth. Dynamic changes; This represents the system's maximum channel bandwidth. This is the signal amplitude correction coefficient, with a value ranging from 0.3 to 1.2. For mathematical expectation operations, the first-level channelized data information is frequency domain information. Performing an inverse Fourier transform on it yields the time-domain first-level channelized data. express.

[0018] As an optional implementation, in the second aspect of the present invention, the step of using the optimized signal detection model to process the large-bandwidth offline signal data information to be detected to obtain the channel adaptive signal detection result includes: S41, acquire the large-bandwidth offline signal data information to be detected; S42, perform first-level channelization on the large-bandwidth offline signal data information to be detected to obtain first-level data information; S43, using the optimized signal detection model, the first-level data information is processed to obtain the first-level detection result; S44, when the first-level detection result is that there is a signal, output the first-level detection result; the first-level detection result is the channel adaptive signal detection result; S45, when the first-level detection result is no signal, the first-level detection result is channelized to obtain second-level data information; S46, Process the secondary data information to obtain the secondary detection result; S47, when the secondary detection result indicates the presence of a signal, the secondary detection result is output; the secondary detection result is a channel adaptive signal detection result; S48, when the secondary detection result is no signal, the secondary detection result is channelized to the third level to obtain the third level data information; S49, The three-level data information is processed to obtain the three-level detection results; the three-level detection results are the channel adaptive signal detection results.

[0019] A third aspect of the present invention discloses another channel-adaptive signal detection apparatus, the apparatus comprising: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute some or all of the steps in the channel adaptive signal detection method disclosed in the first aspect of the present invention.

[0020] The fourth aspect of the present invention discloses a computer-storeable medium storing computer instructions, which, when invoked, are used to execute some or all of the steps in the channel adaptive signal detection method disclosed in the first aspect of the present invention.

[0021] Compared with the prior art, the embodiments of the present invention have the following beneficial effects: (1) For data labeling, the characteristics of the signal in the time-frequency graph generated by different hardware for converting the same data into a time-frequency graph will have slight changes. Supervised learning requires a large amount of labeled sample data, so a lot of manpower and resources are needed for data labeling. Once the hardware requirements for generating the time-frequency graph change, it will lead to repeated labeling of the data, which will cause a lot of waste of resources and time costs. This invention proposes a method to effectively solve this problem by mapping the position information on the image to the position information of the channel in the original data. It is only necessary to detect the original data and label data into the pre-trained model at different training times, which can conveniently and quickly complete the training and output of the model.

[0022] (2) Correspondingly, after the model training is completed, the channel adaptation method proposed in this invention can effectively handle the intelligent detection and recognition of large bandwidth signals. First, through multi-level channelization, signals with different bandwidths can be efficiently screened out. Eliminating channels that cannot detect signals can also effectively save detection resources and accelerate the signal inference speed. At the same time, the multi-level channelization method is also conducive to screening out signals with smaller bandwidths, thereby improving the accuracy of intelligent detection and recognition. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a flowchart illustrating a channel adaptive signal detection method disclosed in an embodiment of the present invention; Figure 2 This is a flowchart illustrating another channel adaptive signal detection method disclosed in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of a channel adaptive signal detection device disclosed in an embodiment of the present invention; Figure 4 This is a schematic diagram of another channel adaptive signal detection device disclosed in an embodiment of the present invention. Detailed Implementation

[0025] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0027] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0028] In all embodiments of the present invention, the variables involved in all computational expressions or mathematical functions have been dimensionlessized before computation.

[0029] In all embodiments of the present invention, the values ​​of the independent variables in the input of all computational expressions or mathematical functions meet the reasonable requirements of the input range of the computational expressions or mathematical functions, and can ensure that the computational expressions or mathematical functions can be calculated smoothly without violating physical laws or mathematical rules.

[0030] This invention discloses a channel-adaptive signal detection method and apparatus. The method includes: acquiring large-bandwidth offline signal data; processing the large-bandwidth offline signal data to obtain a training dataset; training a preset signal detection model using the training dataset to obtain an optimized signal detection model; and processing the large-bandwidth offline signal data to be detected using the optimized signal detection model to obtain a channel-adaptive signal detection result. This invention's method, through multi-level channelization, can efficiently screen signals of different bandwidths. Eliminating channels where signals cannot be detected effectively saves detection resources and accelerates signal inference. Simultaneously, the multi-level channelization method is also beneficial for screening signals with smaller bandwidths, improving the accuracy of intelligent detection and recognition. These are described in detail below.

[0031] Example 1 Please see Figure 1 , Figure 1 This is a flowchart illustrating a channel adaptive signal detection method disclosed in an embodiment of the present invention. Wherein, Figure 1 The described channel-adaptive signal detection method is applied to the field of intelligent signal detection and recognition, specifically involving a channel-adaptive intelligent signal detection and recognition method based on supervised learning. This invention does not limit the specific methods described. Figure 1 As shown, the channel adaptive signal detection method may include the following operations: S1, acquire high-bandwidth offline signal data information; Optionally, high-bandwidth offline signal data refers to pre-acquired and stored broadband radio frequency sampling data, which is not received online in real time. The data has a wide spectral span and includes various communication signals, environmental clutter, and noise. S2, process the high-bandwidth offline signal data information to obtain a training dataset; S3, using the training dataset, train the preset signal detection model to obtain an optimized signal detection model; The preset signal detection model sets up three independent and differentiated time-frequency Mamba branches, corresponding to three levels of channel granularity: wideband global Mamba branch, midband detailed Mamba branch, and narrowband fine Mamba branch. Each branch consists of a stack of time-frequency sequence scanning units, selective state-space modeling units, lightweight feature activation units, and downsampling adaptation units.

[0032] The wideband global Mamba branch employs large-scale sequence scanning to model the global spectral distribution and long-distance steady-state characteristics of the first-level wideband channel, compensating for the lack of long-distance dependency in CNNs. The mid-band detailed Mamba branch employs medium-scale sequence modeling to capture the spectral transition details and short-term fluctuation characteristics of the second-level channel. The narrowband fine Mamba branch employs fine-grained dense scanning to mine the weak signals and transient pulse change characteristics of the third-level narrowband channel, achieving adaptive perception of full-scale time-frequency characteristics.

[0033] S4. Using the optimized signal detection model, the large-bandwidth offline signal data information to be detected is processed to obtain the channel adaptive signal detection result.

[0034] Optionally, the processing of the high-bandwidth offline signal data information to obtain the training dataset includes: S21, Channelize the high-bandwidth offline signal data information to obtain channelized data information; The channelization includes primary channelization, secondary channelization, and tertiary channelization; The channelized data information includes primary channelized data information, secondary channelized data information, and tertiary channelized data information; S22, The channelized data information is processed to obtain channel image data information; the channel image data information includes primary channel image data information, secondary channel image data information and tertiary channel image data information. S23, process the channel image data information to obtain the tag information of the channel image data information; Optionally, channel image data containing signals can be manually screened one by one, and the location of the signals can be manually marked. The location information can be mapped to the start and end times of the signals in the original data, as well as the start and end of the frequency, and these tag information can be stored. The channel image data information and the label information of the channel image data information constitute the training dataset.

[0035] Optionally, the high-bandwidth offline signal data information is channelized to obtain channelized data information, including: S211, Perform first-level channelization on the large bandwidth offline signal data information to obtain first-level channelized data information; The expression for the first-level channelization data information is: in, For the first level Channel number at frequency The channelized frequency domain data at that location, i.e., the first-level channelized data information. This is the primary channel number. For continuous frequency variables in the frequency domain, This refers to the frequency domain sampling data of the original received signal with ultra-wide bandwidth, i.e., the offline signal data information of large bandwidth, which is the original spectrum of the entire broadband segment without channel segmentation. This is an interval indicator function; frequencies falling within the channel's frequency band retain their original spectrum, while frequencies outside the band are set to zero, thus achieving frequency band truncation. For the first level Channel start frequency, , For the first level Channel termination frequency, This represents the theoretical bandwidth of a single channel at level one. For a fixed overlap rate in the channel, For the total bandwidth of offline signal data information, For frequency domain variables; Frequency domain information for high-bandwidth offline signal data. This is the overall starting frequency of the signal; S212, perform secondary channelization on the high-bandwidth offline signal data information to obtain secondary channelized data information; The expression for the secondary channelized data information is as follows: in, For the second level Channel frequency domain data information, i.e., secondary channelized data information. For secondary single-channel segmentation bandwidth, For the second level Channel start frequency, For the second level Channel termination frequency, This represents the total bandwidth of the remaining idle channels after the first-level detection. This is the frequency domain data of the idle channels after primary filtering. , This is the starting frequency for the secondary idle channel. It is an integer; S213, perform three-level channelization on the large bandwidth offline signal data information to obtain three-level channelized data information; The expression for the three-level channelized data information is as follows: in, For the third level Channel frequency domain data information, i.e., Level 3 channelized data information. It is a three-level single-channel extreme segmentation bandwidth. This represents the total bandwidth of the remaining idle channels after secondary detection. For the third level Channel start frequency, , For the third level Channel termination frequency, This is the starting frequency for the third-level idle channel. This is the frequency domain data of the idle channel after secondary filtering.

[0036] When dividing into secondary and tertiary levels, the real-time signal-to-noise ratio (SNR) of a single-level channel is calculated and compared with the preset SNR threshold SNR0 ∈ [-8dB, -3dB]. If the channel satisfies SNRc ≤ SNR0, it is determined that the channel is a pure noise idle frequency band with no effective signal, and it enters the next level of refined channel subdivision. If SNRc > SNR0, it is determined that the channel contains an effective signal, the subsequent subdivision operation of this level frequency band is terminated, and the current level channel data is retained for detection.

[0037] The first and second level channels cover a wide spectrum area and use 32 channels to evenly divide the spectrum, achieving efficient coarse screening and mesoscale fine screening across the entire frequency band. The third level channel is only used to target the idle frequency bands with suspected weak signals remaining in a very small range after the second level screening. Although the number of channels is 8, the target bandwidth of the subdivision is extremely small, which can achieve extremely fine decomposition of the local spectrum.

[0038] The number of secondary channels N and the number of tertiary channels M are both adaptively and dynamically determined, and are not fixed values. N represents the total number of secondary subdivision channels, calculated based on the total bandwidth of remaining idle channels after primary detection and filtering, and the subdivision bandwidth of a single secondary channel. M represents the total number of tertiary extreme subdivision channels, adaptively solved based on the remaining idle bandwidth after secondary filtering and the subdivision bandwidth of a single tertiary channel.

[0039] The number of secondary channels N and the number of tertiary channels M are adaptively calculated using the residual idle bandwidth of the previous level and the theoretical bandwidth of a single channel in the current level, rounded to the nearest integer. The calculation formula is as follows: , At the same time, a lower limit threshold for the number of channels is set to ensure that there is at least one channel in each subdivision, so as to avoid invalid subdivision and parameter failure caused by insufficient bandwidth.

[0040] This invention employs a progressive three-level channelization segmentation strategy—first-level coarse segmentation, second-level fine segmentation, and third-level extreme fine segmentation—to perform layered, successive frequency band extraction processing on large-bandwidth offline signal data. This effectively addresses the technical shortcomings of traditional fixed-bandwidth channelization methods, such as limited frequency band division, inability to balance coarse and fine scales, and the susceptibility of weak signals to broadband noise. The first-level channelization uniformly divides the entire bandwidth into 32 overlapping sub-channels, achieving preliminary global spectrum segmentation based on a fixed overlap rate. This enables rapid coarse screening of signals across the entire frequency band, effectively identifying suspected target frequency bands and eliminating most purely noisy idle frequency bands, ensuring overall detection efficiency.

[0041] Second-level channelization further subdivides the remaining idle frequency bands after the first-level screening, refining the frequency bands suspected of containing hidden signals, compressing single-channel bandwidth, improving mid-band spectral resolution, effectively reducing cross-band interference and spectral aliasing, and compensating for the insufficient refinement of the first-level broadband channel, achieving initial separation of noise from effective signals. Third-level channelization performs extreme subdivision of the remaining idle frequency bands from the second level, further reducing single-channel bandwidth, and specifically targeting and extracting detailed features of weak transient signals and low-power hidden signals deeply embedded in the noise floor.

[0042] Optionally, processing the channelized data information to obtain channel image data information includes: S221, Process the first-level channelization data information to obtain first-level channel image data information; S222, Process the secondary channelization data information to obtain secondary channel image data information; The method for obtaining secondary channel image data information is the same as that in S2211~S2214; S223, Process the three-level channelization data information to obtain three-level channel image data information; The method for obtaining Level 3 channel image data information is the same as that in S2211~S2214; S224, integrate the first-level channel image data information, the second-level channel image data information and the third-level channel image data information to obtain channel image data information.

[0043] Optionally, processing the first-level channelization data information to obtain first-level channel image data information includes: S2211, Process the first-level channelized data information to obtain first-level time-frequency characteristic information; S2212, The first-level time-frequency feature information is processed using a filtering model to obtain the first-level filtered time-frequency feature information; In the formula, This provides the time-frequency characteristic information for the first-level filter. For the current time and frequency point ( The local neighborhood window centered on ) has its size adaptively determined based on the signal time-frequency resolution and channel bandwidth, and does not use a fixed size. This embodiment does not impose any restrictions on this size. These are the joint weighting coefficients in the time domain and frequency domain, respectively. These are the time-domain and frequency-domain spatial scale coefficients; The similarity constraint coefficient for time-frequency features. Used to control the spatial smoothing range of the temporal neighborhood, with values ​​ranging from 1.5 to 4. Used to control the smoothing range of the frequency domain neighborhood, with a value ranging from 1.0 to 3.0. The signal-to-noise ratio (SNR) of the current channel is adaptively fine-tuned, with a value ranging from 0.08 to 0.5. Used to constrain the similarity of time-frequency features in the frequency domain, with values ​​ranging from 0.05 to 0.2; S2213, The first-level channelization data information is transformed to obtain first-level transformation feature information; In the formula, This represents the fourth-order frequency domain cumulant transformation characteristics, i.e., first-order transformation characteristic information; for The frequency domain transform result; This is for the statistical expectation operation of frequency domain sequences.

[0044] S2214, the first-level filter time-frequency feature information and the first-level transform feature information are fused to obtain the first-level channel image data information.

[0045] In the formula, This is the final generated image pixel matrix, i.e., the first-level channel image data information; This represents the normalized time-frequency characteristic information of the first-level filter. This refers to the normalized first-level transform feature information; The adaptive feature fusion weights range from 0.1 to 0.9, determined by the current channel signal-to-noise ratio. Dynamically determined, it is the ratio of effective signal energy to noise energy within a single-level channel; The preset signal-to-noise ratio threshold ranges from -5 to -8 dB. This is the weighting adjustment coefficient, with a value ranging from 0.2 to 0.8; This is the time-frequency energy normalization correction factor. , .

[0046] Optionally, the processing of the first-level channelized data information to obtain first-level time-frequency feature information includes: The first-level channelized data information is processed using a preset time-frequency feature information calculation model to obtain first-level time-frequency feature information; The preset time-frequency feature information calculation model expression is as follows: in, , This is primary time-frequency characteristic information. For integration variables, This serves as the baseline window length coefficient, ranging from 1.0 to 2.0. It is a bandwidth-adaptive Gaussian-weighted window function; The window length adaptive coefficient varies with the current channel bandwidth. Dynamic changes; This represents the system's maximum channel bandwidth. This is the signal amplitude correction coefficient, with a value ranging from 0.3 to 1.2. For mathematical expectation operations, the first-level channelized data information is frequency domain information. Performing an inverse Fourier transform on it yields the time-domain first-level channelized data. express.

[0047] This invention completes the generation of first-level channel image data through a multi-level processing method involving time-frequency adaptive filtering, fourth-order frequency domain cumulant feature transformation, and dynamic weight feature fusion. This effectively solves the technical problems of weak anti-interference capability, missing weak features, and severe noise residue in traditional single time-frequency feature imaging.

[0048] Optionally, the step of using the optimized signal detection model to process the large-bandwidth offline signal data to be detected and obtain the channel adaptive signal detection result includes: S41, acquire the large-bandwidth offline signal data information to be detected; S42, perform first-level channelization on the large-bandwidth offline signal data information to be detected to obtain first-level data information; S43, using the optimized signal detection model, the first-level data information is processed to obtain the first-level detection result; S44, when the first-level detection result is that there is a signal, output the first-level detection result; the first-level detection result is the channel adaptive signal detection result; S45, when the first-level detection result is no signal, the first-level detection result is channelized to obtain second-level data information; S46, Process the secondary data information to obtain the secondary detection result; S47, when the secondary detection result indicates the presence of a signal, the secondary detection result is output; the secondary detection result is a channel adaptive signal detection result; S48, when the secondary detection result is no signal, the secondary detection result is channelized to the third level to obtain the third level data information; S49, The three-level data information is processed to obtain the three-level detection results; the three-level detection results are the channel adaptive signal detection results.

[0049] It is evident that supervised learning model training requires a large amount of labeled sample data, and labeling is the most resource-intensive process in the entire process. During training, there is often a need to re-label the data. The method proposed in this invention, which maps image label information to the original data, can effectively avoid this problem.

[0050] In real-world intelligent recognition signals, the bandwidth varies. Therefore, a channelized signal with a large bandwidth can adapt to the detection and recognition of signals of various bandwidths, thereby improving detection accuracy. However, not every channel necessarily contains signal features. Real-time inference processes have high speed requirements. Therefore, a channel adaptive method is used to determine which channels need to undergo further multi-level channelization. This effectively avoids resource waste and improves detection accuracy.

[0051] Example 2 Please see Figure 2 , Figure 2 This is a flowchart illustrating another channel adaptive signal detection method disclosed in an embodiment of the present invention. Figure 2 The described channel-adaptive signal detection method is applied to the field of intelligent signal detection and recognition, specifically involving a channel-adaptive intelligent signal detection and recognition method based on supervised learning. This invention does not limit the specific methods described. Figure 2 As shown, the channel adaptive signal detection method may include the following operations: 1. Input the high-bandwidth offline signal data into the marking software; 2. Channelize the signal data. Channelization is divided into three levels: Level 1 channelization, Level 2 channelization, and Level 3 channelization. 3. Normalize and FFT each channel data of each channelization level to convert it into image data; label the images containing signals, and map the signal location information onto the original data through time and frequency domain calculations and save it as label information to construct a dataset; 4. Train the Mamba-YOLO signal recognition model using the dataset; 5. Access real-time data and perform hierarchical channelization processing on high-bandwidth signals; 6. By using single-level channelization to process the wide bandwidth signal, it is divided into 32 equal channels in the frequency domain, and the channel overlap rate is set to 20%; 7. Normalize and perform FFT processing on the data of each channel to generate the time-frequency diagram of each channel; 8. Input the time-frequency graph of each channel into the trained Mamba-YOLO model to perform signal detection and determine whether a signal exists in the channel and the type of signal. 9. Remove channels that have detected signals, and perform two-level channelization on the remaining channels. Each channel is divided into 32 equal channels in the frequency domain, and the channel overlap rate is set to 20%. 10. Normalize and perform FFT processing on the data of each channel to generate the time-frequency diagram of each channel; 11. Input the time-frequency graph of each channel into the trained Mamba-YOLO model to perform signal detection and determine whether a signal exists in the channel and the type of signal. 12. Remove channels that have detected signals, and perform three-level channelization on the remaining channels. Each channel is divided into 8 equal channels in the frequency domain, and the channel overlap rate is set to 20%. 13. Normalize and perform FFT processing on the data of each channel to generate the time-frequency diagram of each channel; 14. Input the time-frequency graph of each channel into the trained Mamba-YOLO model to perform signal detection and determine whether a signal exists in the channel and the type of the signal. The labeling software can load large-bandwidth offline file data, channelize the offline data, normalize and FFT the data of each channel to generate time-frequency maps, and save all the generated time-frequency maps in a specified folder. The software opens each map one by one and labels the signal positions. The software generates corresponding labels based on the labeling information. The labels contain the position information mapped to the original data.

[0052] By using a hierarchical channelization method, the channel detection signal is classified into different categories, and then the decision to proceed to the next level of channelization is adaptively made based on whether a signal is detected.

[0053] Example 3 This embodiment of a channel-adaptive signal detection method includes the following steps: 1. Model Training: Collect raw signal data files and import them into data labeling software. First, perform channelization processing on the signals. Normalize and perform FFT processing on the channel signal data of each channel level to generate corresponding time-frequency maps. Manually screen out the time-frequency maps containing the signals one by one, and then manually label the signal positions. Based on the position information, map the start and end times of the signals in the raw data, as well as the start and end times of the frequencies. Store this label information to create a dataset. The labeling software can load large-bandwidth offline file data, channelize the offline data, normalize and FFT the data of each channel to generate time-frequency maps, and save all the generated time-frequency maps in a specified folder. The software opens each map one by one and labels the signal positions. The software generates corresponding labels based on the labeling information. The labels contain the position information mapped to the original data.

[0054] Furthermore, the Mamba-YOLO object detection algorithm is used, and the prepared dataset is set into training set, validation set and test set according to the appropriate ratio to train Mamba-YOLO; Furthermore, the trained model is imported into the inference program.

[0055] 2. Real-time inference: Access real-time signal data; detect and identify the bandwidth of the signal and the total bandwidth of the actual input signal according to actual needs, set the number of channels for each channelization level, first perform one-level channelization, and then preprocess the data of each channel, which includes normalization and FFT transformation. The time-frequency map generated by the preprocessing is input into the trained model for inference, which can detect whether there is a signal in each image and the type and location information of the signal. Furthermore, based on the inference results, the channels corresponding to the time-frequency maps where no signal was detected will continue to undergo secondary channelization. After a time-frequency map containing a signal is detected, the type of the signal will be output and its position information on the time-frequency map will be selected. After that, the channels corresponding to the time-frequency map will not continue to undergo the next level of channelization. Furthermore, the signal data obtained after secondary channelization, which has more channels, is processed. After preprocessing, a time-frequency map is generated and input into the trained model for inference. The model adaptively selects whether the current channel should continue to the next level of channelization based on whether the time-frequency map corresponding to the channel contains the signal. At the same time, it outputs the type of the detected signal and its location information, and removes the channel corresponding to the detected signal. Furthermore, the secondary channelized channels that do not detect signals are subjected to a final, more refined tertiary channelization. The results of each channelization are preprocessed, and the data converted into a time-frequency map is input into the inference model. The model outputs the type of the detected signal and its location information. Furthermore, the structured information of all detected signals is summarized and saved.

[0056] Example 4 Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of a channel adaptive signal detection device disclosed in an embodiment of the present invention. Figure 3 The described channel-adaptive signal detection device is applied in the field of intelligent signal detection and recognition, specifically involving a channel-adaptive intelligent signal detection and recognition method based on supervised learning. This invention does not limit the scope of the embodiments. Figure 3 As shown, the channel adaptive signal detection device may include the following operations: S301, data acquisition module, used to acquire high-bandwidth offline signal data information; S302, Training dataset construction module, used to process the high-bandwidth offline signal data information to obtain a training dataset; S303, Model training module, used to train the preset signal detection model using the training dataset to obtain an optimized signal detection model; S304, signal detection module, is used to process the large-bandwidth offline signal data information to be detected using the optimized signal detection model, and obtain the channel adaptive signal detection result.

[0057] Example 5 Please see Figure 4 , Figure 4 This is a schematic diagram of another channel adaptive signal detection device disclosed in an embodiment of the present invention. Figure 4 The described channel-adaptive signal detection device is applied in the field of intelligent signal detection and recognition, specifically involving a channel-adaptive intelligent signal detection and recognition method based on supervised learning. This invention does not limit the scope of the embodiments. Figure 4 As shown, the channel adaptive signal detection device may include the following operations: Memory 401 storing executable program code; Processor 402 coupled to memory 401; The processor 402 calls the executable program code stored in the memory 401 to execute the steps in the channel adaptive signal detection method described in Embodiments 1 to 3.

[0058] Example 6 This invention discloses a computer-readable storage medium storing a computer program for electronic data interchange, wherein the computer program causes a computer to perform the steps of the channel adaptive signal detection method described in Embodiments 1 to 3.

[0059] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0060] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0061] Finally, it should be noted that the channel adaptive signal detection method and apparatus disclosed in the embodiments of the present invention are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A channel-adaptive signal detection method, characterized in that, The method includes: S1, acquire high-bandwidth offline signal data information; S2, process the high-bandwidth offline signal data information to obtain a training dataset; S3, using the training dataset, train the preset signal detection model to obtain an optimized signal detection model; S4. Using the optimized signal detection model, the large-bandwidth offline signal data information to be detected is processed to obtain the channel adaptive signal detection result.

2. The channel adaptive signal detection method according to claim 1, characterized in that, The process of processing the high-bandwidth offline signal data to obtain a training dataset includes: S21, Channelize the high-bandwidth offline signal data information to obtain channelized data information; The channelization includes primary channelization, secondary channelization, and tertiary channelization; The channelized data information includes primary channelized data information, secondary channelized data information, and tertiary channelized data information; S22, The channelized data information is processed to obtain channel image data information; the channel image data information includes primary channel image data information, secondary channel image data information and tertiary channel image data information. S23, process the channel image data information to obtain the tag information of the channel image data information; The channel image data information and the label information of the channel image data information constitute the training dataset.

3. The channel adaptive signal detection method according to claim 2, characterized in that, The process of channelizing the high-bandwidth offline signal data to obtain channelized data information includes: S211, Perform first-level channelization on the large bandwidth offline signal data information to obtain first-level channelized data information; The expression for the first-level channelization data information is: in, For the first level Channel number at frequency The channelized frequency domain data at that location, i.e., the first-level channelized data information. This is the primary channel number. For continuous frequency variables in the frequency domain, This refers to the frequency domain sampling data of the original received signal with ultra-wide bandwidth, i.e., the offline signal data information of large bandwidth, which is the original spectrum of the entire broadband segment without channel segmentation. This is an interval indicator function; frequencies falling within the channel's frequency band retain their original spectrum, while frequencies outside the band are set to zero, thus achieving frequency band truncation. For the first level Channel start frequency, , For the first level Channel termination frequency, This represents the theoretical bandwidth of a single channel at level one. For a fixed overlap rate in the channel, For the total bandwidth of offline signal data information with large bandwidth, For frequency domain variables; Frequency domain information for high-bandwidth offline signal data. This is the overall starting frequency of the signal; S212, perform secondary channelization on the high-bandwidth offline signal data information to obtain secondary channelized data information; The expression for the secondary channelized data information is: in, For the second level Channel frequency domain data information, i.e., secondary channelized data information. For secondary single-channel segmentation bandwidth, For the second level Channel start frequency, For the second level Channel termination frequency, This represents the total bandwidth of the remaining idle channels after the first-level detection. This is the frequency domain data of the idle channels after primary filtering. , This is the starting frequency for the secondary idle channel. It is an integer; S213, perform three-level channelization on the large bandwidth offline signal data information to obtain three-level channelized data information; The expression for the three-level channelized data information is as follows: in, For the third level Channel frequency domain data information, i.e., Level 3 channelized data information. It is a three-level single-channel extreme subdivision bandwidth. This represents the total bandwidth of the remaining idle channels after secondary detection. For the third level Channel start frequency, , For the third level Channel termination frequency, This is the starting frequency for the third-level idle channel. This is the frequency domain data of the idle channel after secondary filtering.

4. The channel adaptive signal detection method according to claim 1, characterized in that, The process of processing the channelized data information to obtain channel image data information includes: S221, Process the first-level channelization data information to obtain first-level channel image data information; S222, Process the secondary channelization data information to obtain secondary channel image data information; S223, Process the three-level channelization data information to obtain three-level channel image data information; S224, integrate the first-level channel image data information, the second-level channel image data information and the third-level channel image data information to obtain channel image data information.

5. The channel adaptive signal detection method according to claim 4, characterized in that, The process of processing the first-level channelization data information to obtain first-level channel image data information includes: S2211, Process the first-level channelized data information to obtain first-level time-frequency characteristic information; S2212, The first-level time-frequency feature information is processed using a filtering model to obtain the first-level filtered time-frequency feature information; S2213, The first-level channelization data information is transformed to obtain first-level transformation feature information; S2214, the first-level filter time-frequency feature information and the first-level transform feature information are fused to obtain the first-level channel image data information.

6. The channel adaptive signal detection method according to claim 5, characterized in that, The process of processing the first-level channelized data information to obtain first-level time-frequency feature information includes: The first-level channelized data information is processed using a preset time-frequency feature information calculation model to obtain first-level time-frequency feature information; The preset time-frequency feature information calculation model expression is as follows: in, , This is primary time-frequency characteristic information. For integration variables, This serves as the baseline window length coefficient, ranging from 1.0 to 2.

0. It is a bandwidth-adaptive Gaussian-weighted window function; The window length adaptive coefficient varies with the current channel bandwidth. Dynamic changes; This represents the system's maximum channel bandwidth. This is the signal amplitude correction coefficient, with a value ranging from 0.3 to 1.

2. For mathematical expectation operations, the first-level channelized data information is frequency domain information. Performing an inverse Fourier transform on it yields the time-domain first-level channelized data. express.

7. The channel adaptive signal detection method according to claim 1, characterized in that, The process of using the optimized signal detection model to process the large-bandwidth offline signal data to be detected, and obtaining the channel-adaptive signal detection result, includes: S41, acquire the large-bandwidth offline signal data information to be detected; S42, perform first-level channelization on the large-bandwidth offline signal data information to be detected to obtain first-level data information; S43, using the optimized signal detection model, the first-level data information is processed to obtain the first-level detection result; S44, when the first-level detection result is that there is a signal, output the first-level detection result; the first-level detection result is the channel adaptive signal detection result; S45, when the first-level detection result is no signal, the first-level detection result is channelized to obtain second-level data information; S46, Process the secondary data information to obtain the secondary detection result; S47, when the secondary detection result indicates the presence of a signal, the secondary detection result is output; the secondary detection result is a channel adaptive signal detection result; S48, when the secondary detection result is no signal, the secondary detection result is channelized to the third level to obtain the third level data information; S49, The three-level data information is processed to obtain the three-level detection results; the three-level detection results are the channel adaptive signal detection results.

8. A channel-adaptive signal detection device, characterized in that, The device includes: The data acquisition module is used to acquire high-bandwidth offline signal data information; The training dataset construction module is used to process the high-bandwidth offline signal data information to obtain the training dataset; The model training module is used to train the preset signal detection model using the training dataset to obtain an optimized signal detection model. The signal detection module is used to process the large-bandwidth offline signal data information to be detected using the optimized signal detection model, and obtain the channel adaptive signal detection result.

9. A channel-adaptive signal detection device, characterized in that, The device includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the channel adaptive signal detection method as described in any one of claims 1-7.

10. A computer-storable medium, characterized in that, The computer storage medium stores computer instructions, which, when invoked, are used to execute the channel adaptive signal detection method as described in any one of claims 1-7.

Images