CN110659656A - Method and device for detecting interference - Google Patents
Method and device for detecting interference Download PDFInfo
- Publication number
- CN110659656A CN110659656A CN201810693247.4A CN201810693247A CN110659656A CN 110659656 A CN110659656 A CN 110659656A CN 201810693247 A CN201810693247 A CN 201810693247A CN 110659656 A CN110659656 A CN 110659656A
- Authority
- CN
- China
- Prior art keywords
- interference
- data
- sample
- interference data
- marking
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000012549 training Methods 0.000 claims abstract description 35
- 238000013507 mapping Methods 0.000 claims abstract description 21
- 238000012545 processing Methods 0.000 claims abstract description 13
- 238000002372 labelling Methods 0.000 claims description 6
- 238000013527 convolutional neural network Methods 0.000 description 13
- 239000003550 marker Substances 0.000 description 7
- 238000001514 detection method Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000013459 approach Methods 0.000 description 3
- 238000010801 machine learning Methods 0.000 description 3
- 238000012935 Averaging Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 238000004140 cleaning Methods 0.000 description 2
- 238000011478 gradient descent method Methods 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000010835 comparative analysis Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 238000007637 random forest analysis Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Image Analysis (AREA)
- Radar Systems Or Details Thereof (AREA)
- Noise Elimination (AREA)
Abstract
The embodiment of the invention discloses a method and a device for detecting interference, wherein the method comprises the following steps: marking the collected sample interference data by using a plurality of classifier voting modes to obtain interference data with marked interference types; performing time-frequency domain mapping processing on the interference data with the marked interference types to obtain a time-frequency domain mapped interference map; training the interference map subjected to time-frequency domain mapping to obtain a trained interference model; and classifying the interference types of the interference data to be verified of the external field by using the trained interference model to obtain the interference types of the interference data to be verified of the external field.
Description
Technical Field
The present invention relates to the technical field of LTE (Long Term Evolution) wireless performance, and in particular, to a method and an apparatus for detecting interference.
Background
In the past, the interference is generally checked on site by a spectrometer or the interference type is judged by drawing the waveform of the interference, but if the interference is more or the stations are more, the method for checking the interference is troublesome and very labor-consuming. Because the interference types of the external field are multiple, the interference form is complex, the manual detection efficiency is low, and the integral concept is difficult to form;
existing automated analysis requires human intervention to identify the characteristics of each disturbance and to identify such disturbance characteristics in a programmed manner. However, although one may sometimes identify a certain interference feature, it is difficult to identify the feature by means of coding.
Disclosure of Invention
The technical problem solved by the scheme provided by the embodiment of the invention is that the interference judgment is very complicated and the interference characteristic needs to be manually extracted.
The method for detecting the interference provided by the embodiment of the invention comprises the following steps:
marking the collected sample interference data by using a plurality of classifier voting modes to obtain interference data with marked interference types;
performing time-frequency domain mapping processing on the interference data with the marked interference types to obtain a time-frequency domain mapped interference map;
training the interference map subjected to time-frequency domain mapping to obtain a trained interference model;
and classifying the interference types of the interference data to be verified of the external field by using the trained interference model to obtain the interference types of the interference data to be verified of the external field.
Preferably, the step of performing interference type labeling processing on the collected sample interference data by using a plurality of classifier voting modes to obtain interference data labeled with interference types includes:
respectively carrying out initial marking on the collected sample interference data of various interference types to obtain sample interference data with the initial marking;
according to the interference energy intensity of the sample interference data with the same initial mark, marking the sample interference data as the sample interference data of a positive interference mark, the sample interference data of other interference marks and the sample interference data of an interference-free mark;
and finally marking the sample interference data of the positive interference mark, the sample interference data of other interference marks and the sample interference data of the non-interference marks in a classifier voting mode to obtain interference data with the marked interference type.
Preferably, the step of finally marking the sample interference data of the positive interference marker, the sample interference data of other interference markers and the sample interference data of the non-interference markers by means of classifier voting to obtain the interference data with the labeled interference types includes:
randomly dividing sample interference data with the same initial mark to obtain a plurality of interference groups containing the sample interference data of the positive interference mark, the sample interference data of other interference marks and the sample interference data of the non-interference mark;
training each interference group to generate a classifier for voting the sample interference data;
and finally marking the sample interference data of the positive interference mark, the sample interference data of other interference marks and the sample interference data of the non-interference marks by using the classifier to obtain the interference data with the marked interference type.
Preferably, the method further comprises the following steps:
and when the external field to-be-verified interference data does not conform to the interference type in the trained interference model, performing recursive training on the interference model again according to the interference type of the external field to-be-verified interference data to obtain and store a new interference model.
According to an embodiment of the present invention, an apparatus for detecting interference includes:
the marking module is used for marking the interference types of the collected sample interference data by utilizing various classifier voting modes to obtain interference data with the marked interference types;
the time-frequency domain mapping processing module is used for performing time-frequency domain mapping processing on the interference data with the marked interference types to obtain a time-frequency domain mapped interference map;
the training module is used for training the interference map subjected to time-frequency domain mapping to obtain a trained interference model;
and the interference classification module is used for classifying the interference types of the external field to-be-verified interference data by using the trained interference model to obtain the interference types of the external field to-be-verified interference data.
Preferably, the marking module comprises:
the initial marking unit is used for respectively carrying out initial marking on the collected sample interference data of various interference types to obtain sample interference data with initial marking, and marking the sample interference data as sample interference data of a positive interference mark, sample interference data of other interference marks and sample interference data of an interference-free mark according to the interference energy intensity of the sample interference data with the same initial marking;
and the final marking unit is used for finally marking the sample interference data of the positive interference marks, the sample interference data of other interference marks and the sample interference data of the non-interference marks in a classifier voting mode to obtain the interference data with the marked interference types.
Preferably, the final marking unit includes:
a dividing subunit, configured to randomly divide the sample interference data with the same initial marker to obtain a plurality of interference groups including the sample interference data with the positive interference marker, the sample interference data with other interference markers, and the sample interference data with no interference markers;
a training subunit, configured to generate a classifier for voting on the sample interference data by training each interference group;
and the final marking subunit is used for finally marking the sample interference data of the positive interference mark, the sample interference data of other interference marks and the sample interference data of the non-interference marks by using the classifier to obtain the interference data with the marked interference type.
Preferably, the training module is further configured to perform recursive training on the interference model again according to the interference type of the external field to-be-verified interference data when the external field to-be-verified interference data does not conform to the interference type in the trained interference model, so as to obtain and store a new interference model.
According to an embodiment of the present invention, an apparatus for detecting interference includes: a processor, and a memory coupled to the processor; the memory stores a program for detecting interference, which is executable on the processor, and when the program for detecting interference is executed by the processor, the method for detecting interference provided by the embodiment of the invention is implemented.
According to an embodiment of the present invention, a computer storage medium is provided, which stores a program for detecting interference, and when the program for detecting interference is executed by a processor, the method for detecting interference provided by the embodiment of the present invention is implemented.
According to the scheme provided by the embodiment of the invention, whether Interference exists or not can be detected through the atlas identification of NI (Noise Interference), which Interference exists, so that the Interference of the network is checked, the network performance is improved, the benefits are brought, the system performances such as LTE (Long term evolution) and the like are better optimized, the user experience of UE (user equipment) is improved, and the user perception degree is increased.
Drawings
Fig. 1 is a flowchart of a method for detecting interference according to an embodiment of the present invention;
FIG. 2 is a diagram illustrating an apparatus for detecting interference according to an embodiment of the present invention;
fig. 3 is a flowchart of a method for detecting interference according to an embodiment of the present invention;
FIG. 4 is a schematic diagram of time-frequency domain provided by an embodiment of the present invention;
fig. 5 is a schematic diagram of radio and television interference provided by an embodiment of the present invention;
fig. 6 is a schematic view of atmospheric waveguide interference provided by an embodiment of the present invention.
Detailed Description
The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, and it should be understood that the preferred embodiments described below are only for the purpose of illustrating and explaining the present invention, and are not to be construed as limiting the present invention.
Fig. 1 is a flowchart of a method for detecting interference according to an embodiment of the present invention, as shown in fig. 1, including:
step S101: marking the collected sample interference data by using a plurality of classifier voting modes to obtain interference data with marked interference types;
step S102: performing time-frequency domain mapping processing on the interference data with the marked interference types to obtain a time-frequency domain mapped interference map;
step S103: training the interference map subjected to time-frequency domain mapping to obtain a trained interference model;
step S104: and classifying the interference types of the interference data to be verified of the external field by using the trained interference model to obtain the interference types of the interference data to be verified of the external field.
Wherein the step S101 includes: respectively carrying out initial marking on the collected sample interference data of various interference types to obtain sample interference data with the initial marking; according to the interference energy intensity of the sample interference data with the same initial mark, marking the sample interference data as the sample interference data of a positive interference mark, the sample interference data of other interference marks and the sample interference data of an interference-free mark; and finally marking the sample interference data of the positive interference mark, the sample interference data of other interference marks and the sample interference data of the non-interference marks in a classifier voting mode to obtain interference data with the marked interference type.
Specifically, the step of finally labeling the sample interference data of the positive interference marker, the sample interference data of other interference markers, and the sample interference data of the non-interference marker by means of classifier voting to obtain the interference data with labeled interference types includes: randomly dividing the sample interference data with the same initial mark to obtain a plurality of interference groups containing the sample interference data of the positive interference mark, the sample interference data of other interference marks and the sample interference data of the non-interference mark; training each interference group to generate a classifier for voting the sample interference data; and finally marking the sample interference data of the positive interference mark, the sample interference data of other interference marks and the sample interference data of the non-interference marks by using the classifier to obtain the interference data with the marked interference type.
The embodiment of the invention also comprises the following steps: and when the external field to-be-verified interference data does not conform to the interference type in the trained interference model, performing recursive training on the interference model again according to the interference type of the external field to-be-verified interference data to obtain and store a new interference model.
Fig. 2 is a schematic diagram of an apparatus for detecting interference according to an embodiment of the present invention, as shown in fig. 2, including: a marking module 201, a time-frequency domain mapping processing module 202, a training module 203 and an interference classification module 204.
The marking module 201 is configured to perform interference type marking processing on the collected sample interference data by using multiple classifier voting modes to obtain interference data with an interference type marked;
the time-frequency domain mapping processing module 202 is configured to perform time-frequency domain mapping processing on the interference data with the labeled interference type to obtain a time-frequency domain mapped interference map;
the training module 203 is configured to train the time-frequency domain mapped interference map to obtain a trained interference model;
the interference classification module 204 is configured to perform interference type classification on the external field to-be-verified interference data by using the trained interference model, so as to obtain an interference type of the external field to-be-verified interference data.
Wherein the marking module 201 comprises: the initial marking unit is used for respectively carrying out initial marking on the collected sample interference data of various interference types to obtain sample interference data with initial marking, and marking the sample interference data as sample interference data of a positive interference mark, sample interference data of other interference marks and sample interference data of an interference-free mark according to the interference energy intensity of the sample interference data with the same initial marking; and the final marking unit is used for finally marking the sample interference data of the positive interference marks, the sample interference data of other interference marks and the sample interference data of the non-interference marks in a classifier voting mode to obtain the interference data with the marked interference types.
Specifically, the final marking unit includes: a dividing subunit, configured to randomly divide the sample interference data with the same initial marker to obtain a plurality of interference groups including the sample interference data with the positive interference marker, the sample interference data with other interference markers, and the sample interference data with no interference markers; a training subunit, configured to generate a classifier for voting on the sample interference data by training each interference group; and the final marking subunit is used for finally marking the sample interference data of the positive interference mark, the sample interference data of other interference marks and the sample interference data of the non-interference marks by using the classifier to obtain the interference data with the marked interference type.
The training module 203 is further configured to perform recursive training on the interference model again according to the interference type of the outfield to-be-verified interference data when the outfield to-be-verified interference data does not conform to the interference type in the trained interference model, so as to obtain and store a new interference model.
The device for detecting interference provided by the embodiment of the invention is characterized by comprising: a processor, and a memory coupled to the processor; the memory stores a program for detecting interference, which is executable on the processor, and when the program for detecting interference is executed by the processor, the method for detecting interference provided by the embodiment of the invention is implemented.
The computer storage medium according to an embodiment of the present invention is characterized in that the computer storage medium stores a program for detecting interference, and the program for detecting interference is executed by a processor to implement the steps of the method for detecting interference according to an embodiment of the present invention.
Fig. 3 is a flowchart of a method for specifically detecting interference according to an embodiment of the present invention, as shown in fig. 3, including:
step 1: various outfield NI uplink interference data are collected.
Step 2: the interference data of different types are labeled, a few samples in certain categories can be labeled in a mode of few samples, and then all samples are labeled in a mode of voting by a multi-classifier.
For multi-classifier voting, it should be noted that:
firstly, due to the large number of unmarked samples of the external field, if the numerous unmarked samples collected for each specific scene are marked by people one by one, the marks needed to be marked are too many to be completed by people, fatigue judgment is caused, and errors are introduced by people.
Therefore, a multi-classifier voting mode is adopted;
firstly, averaging a large number of samples in a cell level by adopting a cell merging mode from a plurality of unmarked samples, and marking the samples in a manual observation mode after averaging so as to divide the samples into a plurality of categories, namely A, B, C and the like;
in any one of the categories, A is assumed. The class A is a sub-sample space, which contains a plurality of unmarked samples, and then the samples with the strongest energy (the threshold needs to be set according to specific situations) are marked as positive marks (interference marks) or non-interference marks in the class according to the conditions of the energy level.
And thirdly, randomly subdividing the class-A subspace into a plurality of small spaces, namely a1, a2.. an (which may or may not be overlapped and have the same size as much as possible), wherein n is determined according to the situation, generally 5 are adopted, each newly divided space contains certain positive samples and unmarked samples, a classifier is trained by adopting the marks, and n classifiers are generated by the n subspaces.
Fourthly, for each large category A, B, C and the like, the obtained multiple classifiers are respectively adopted for voting and scoring,
the total score is n, n-2/3 n are positive sample marks, 2/3 n-1/3 n are other category marks, and 1/3 n-0 are no-mark categories.
Here, some new categories other than the current sample labels can be identified and labeled as others by means of multi-classifier voting. This approach effectively solves the problem of how to find new class labels in the classifier.
This other class, if frequently present in subsequent tests of new samples, requires human intervention to see if it is a new interference class and, if so, to add a new class label.
Through the multi-classifier voting mode, a large number of unmarked samples can be labeled. The method effectively improves the marking accuracy and the marking speed and creates conditions for subsequent identification.
And step 3: according to the labeling result, the interference is identified by adopting a time domain and frequency domain combined mode, specifically, a CNN (Convolutional Neural Network) mode is adopted for machine learning, and the accuracy of machine learning is improved by adjusting proper parameters.
How to form a sample image by using a time domain and frequency domain combined method needs to be explained as follows:
1) the interference itself is not a map, but is received by base stations distributed over 100 RBs as a data input, which is a frequency domain input, and the 100 RBs can be ranked from low to high. This input in the frequency domain reflects the variation of the interference in the frequency domain, which is different for each different interference.
2) The frequency domain characteristics of each dimension also vary from the time domain, so that the NI variation of each RB from the time domain alone can also be used as an input of one dimension. This temporal variation is also a reaction to the interference signature.
3) The two characteristics of the interference in the frequency domain and the time domain are considered comprehensively, the frequency domain is 100RB, the time domain is truncated by 1 hour granularity (which can be changed as appropriate), and the generated atlas pattern of the time-frequency domain is used as the input of the convolutional neural network CNN. The interference is characterized by utilizing the relatively high-efficiency spectrum recognition capability of the CNN, which is shown in figure 4.
Meanwhile, the characteristic of symmetry of the uplink and the downlink can be utilized, and whether the downlink has interference or not can be judged by detecting the uplink interference.
And 4, step 4: and (5) checking the successfully trained interference model.
And (4) carrying out comparative analysis on the outfield data aiming at the network which is trained successfully, checking the model, and recording an incorrect sample. And analyzes whether there is a new type.
And 5: and deploying an external field, analyzing and improving according to actual conditions, and recursing the model.
And performing recursive training on the model according to the collected new samples and the new labeling conditions to obtain a trained model again.
Step 6: and redeploying the new training model to the outfield, and identifying and judging the outfield interference again.
By using the reported NI data, the interference detection is identified by using a machine-learned pattern according to a frequency domain and time domain joint analysis method, and meanwhile, the interference type can be effectively judged by using uplink and downlink symmetry.
The following describes an embodiment of the present invention in detail with reference to fig. 5 and 6
Example 1
For example, the LTE external field has radio and television interference in a certain area 1, and the distribution of the waveform for the radio and television interference over 100 RBs is generally as shown in fig. 5. When such interference occurs, it is necessary to artificially recognize the presence of such interference and to take measures effectively.
The uplink NI data collected by the system is extracted by the background tool or the network management tool, and according to the characteristics of the time-frequency domain, the uplink NI data with the granularity of 15 minutes is extracted (the specific granularity can be set to, for example, 5 minutes or 10 minutes). Performing necessary data cleaning processes, such as removing null values, removing abnormal values and the like; then obtaining sample data needing to be processed; since the extracted data of various interference (including non-interference) samples are not labeled, the collected samples need to be labeled according to a set method flow.
The method comprises the following steps:
301. the method is characterized in that a large number of unmarked samples are averaged in a cell level by adopting a cell merging mode, and are marked in a manual observation mode after being averaged, wherein the broadcast-television interference is marked in the cell level because the broadcast-television interference is aimed at.
302. The samples marked according to the cell level of the radio and television interference are classified into a group, which is marked as A, so that the samples marked by the group A are still a set consisting of a large number of samples, and in the set, part of the samples should be classified into the samples belonging to the radio and television interference theoretically, but part of the samples still belong to the interference-free samples or other interferences.
303. The samples in the class A space are sorted according to energy, and are divided into positive samples and negative samples according to a threshold, a specific threshold value method can be adopted, for example, negative samples below-113 db and positive samples above-70 db are adopted, and the specific threshold can be adjusted according to actual conditions.
304. Dividing the space A into a plurality of subspaces randomly as a1, a2 and aN; for example, it may be divided into 5 subspaces. Taking N as 5, it is necessary to ensure that there should be some positive-labeled samples and some negative-labeled samples in each space
305. The positive and negative samples in each subspace are trained into classifiers, which may be implemented using a random forest or some other classifier (as well as other methods set forth in the examples below).
306. Voting is carried out on the whole space sample by the plurality of classifiers, the total score is n, n-2/3 n are positive sample marks, 2/3 n-1/3 n are other class marks, and 1/3 n-0 are no mark classes.
Through the above steps, a large amount of unlabeled samples are labeled.
Next, time-frequency domain mapping is performed on the labeled data.
Is operated as follows
307. The obtained data is sorted according to the time dimension, the samples can be divided again by sorting the time dimension, and if the collected samples are 15 minutes in granularity, the samples of 1 hour (specific parameters can be set) can be combined into one sample according to the sorted samples.
308. The newly obtained sample comprises information of two dimensions, wherein one dimension is a time dimension, the other dimension is a frequency dimension, the frequency domain dimension is 1-100 RB, the time dimension is 1 hour, and the time span can be set and can be replaced by 2 hours or more.
309. During this merging process, some samples may not be merged, specifically, if merging is performed in one hour, 4 samples need to be merged in 15 minutes granularity, 8 samples need to be merged in 2 hours granularity, but if the total number of samples is not 4, or is a multiple of 8, a phenomenon that some samples cannot be merged occurs. When these unwanted samples are present, the embodiments herein employ either a direct discard approach or other approaches (as will be described in the examples below).
And (5) training the data subjected to the time-frequency domain mapping by a CNN convolutional neural network.
The CNN convolutional neural network trained in the method adopts 3 convolutional sampling layers, two full-connection layers and the last full-connection layer, and the softmax is used for outputting the class probability.
And generating a model for the network by using training data and a gradient descent method to form an interference library, and continuously correcting the model error and continuously improving the identification accuracy rate through external field iteration.
Example 2
For example, the LTE external field has interference of the atmospheric waveguide in a certain area 2, which is generally the case for the interference waveform of the atmospheric waveguide, as shown in fig. 6.
When such interference occurs, it is necessary to artificially recognize the presence of such interference and to take measures effectively.
The uplink NI data collected by the system is extracted by the background tool or the network management tool, and according to the characteristics of the time-frequency domain, the uplink NI data with the granularity of 15 minutes is extracted (the specific granularity can be set to, for example, 5 minutes or 10 minutes). Performing necessary data cleaning processes, such as removing null values, removing abnormal values and the like; then obtaining sample data needing to be processed; since the extracted data of various interference (including non-interference) samples are not labeled, the collected samples need to be labeled according to a set method flow.
The method comprises the following steps:
401. the method is characterized in that a large number of unmarked samples are averaged in a cell level by adopting a cell merging mode, and are marked in a manual observation mode after being averaged, wherein the broadcast-television interference is marked in the cell level because the broadcast-television interference is aimed at.
402. The samples marked according to the cell level of the radio and television interference are classified into a group, which is marked as A, so that the samples marked by the group A are still a set consisting of a large number of samples, and in the set, part of the samples should be classified into the samples belonging to the radio and television interference theoretically, but part of the samples still belong to the interference-free samples or other interferences.
403. The samples in the class-A space are sorted according to energy, and divided into positive samples and negative samples according to a threshold, a relative threshold method can be adopted, for example, less than 20% of the samples are negative samples, more than 80% of the samples are positive samples, and the specific threshold can be adjusted according to actual conditions.
404. Dividing the space A into a plurality of subspaces randomly as a1, a2 and aN; for example, it may be divided into 5 subspaces. Taking N to 5, it is necessary to ensure that there should be some positive and negative labeled samples in each space.
405. Different from the previous embodiment, the positive and negative samples in each subspace may be trained into different classifiers by using different kinds of classifier algorithms, for example, a bayesian classifier, a decision tree classifier, an svm classifier, or other multiple classifiers may be used.
406. Voting is carried out on the whole space sample by the plurality of classifiers, the total score is n, n-2/3 n are positive sample marks, 2/3 n-1/3 n are other class marks, and 1/3 n-0 are no mark classes.
Through the above steps, a large amount of unlabeled samples are labeled.
Next, time-frequency domain mapping is performed on the labeled data.
Is operated as follows
407. The obtained data is sorted according to the time dimension, the samples can be divided again by sorting the time dimension, and if the collected samples are 15 minutes in granularity, the samples of 1 hour (specific parameters can be set) can be combined into one sample according to the sorted samples.
408. The newly obtained sample comprises information of two dimensions, wherein one dimension is a time dimension, the other dimension is a frequency dimension, the frequency domain dimension is 1-100 RB, the time dimension is 1 hour, and the time span can be set and can be replaced by 2 hours or more.
409. During this merging process, some samples may not be merged, specifically, if merging is performed in one hour, 4 samples need to be merged in 15 minutes granularity, 8 samples need to be merged in 2 hours granularity, but if the total number of samples is not 4, or is a multiple of 8, a phenomenon that some samples cannot be merged occurs. When these extra samples occur, unlike the previous embodiment, other samples can be copied to complete the sample and combine the samples into a new sample.
And (5) training the data subjected to the time-frequency domain mapping by a CNN convolutional neural network.
The CNN convolutional neural network trained in the method adopts 3 convolutional sampling layers, two full-connection layers and the last full-connection layer, and the softmax is used for outputting the class probability.
And generating a model for the network by using training data and a gradient descent method to form an interference library, and continuously correcting the model error and continuously improving the identification accuracy rate through external field iteration.
In summary, the embodiment of the present invention has a continuous updating process, and can discover a new interference type and learn to store. The cost of interference detection is reduced, the efficiency of judging interference by an external field can be improved by the interference detection mode, and the network quality is improved. The embodiment of the invention can be applied not only in LTE, but also in FDD and 5G, and has universality
According to the scheme provided by the embodiment of the invention, samples are taken on the basis of an external field or a field, the time-frequency domain data of NI is changed into a recognizable map, and an interference model is formed in a machine learning mode by means of training iteration, so that the interference type can be automatically analyzed, and the method for actively learning the interference characteristics can improve a lot of efficiency and facilitate the interference investigation of the external field.
Although the present invention has been described in detail hereinabove, the present invention is not limited thereto, and various modifications can be made by those skilled in the art in light of the principle of the present invention. Thus, modifications made in accordance with the principles of the present invention should be understood to fall within the scope of the present invention.
Claims (10)
1. A method of detecting interference, comprising:
marking the collected sample interference data by using a plurality of classifier voting modes to obtain interference data with marked interference types;
performing time-frequency domain mapping processing on the interference data with the marked interference types to obtain a time-frequency domain mapped interference map;
training the interference map subjected to time-frequency domain mapping to obtain a trained interference model;
and classifying the interference types of the interference data to be verified of the external field by using the trained interference model to obtain the interference types of the interference data to be verified of the external field.
2. The method of claim 1, wherein the step of labeling the interference types of the collected sample interference data by using a plurality of classifier voting methods to obtain interference data labeled with the interference types comprises:
respectively carrying out initial marking on the collected sample interference data of various interference types to obtain sample interference data with the initial marking;
according to the interference energy intensity of the sample interference data with the same initial mark, marking the sample interference data as the sample interference data of a positive interference mark, the sample interference data of other interference marks and the sample interference data of an interference-free mark;
and finally marking the sample interference data of the positive interference mark, the sample interference data of other interference marks and the sample interference data of the non-interference marks in a classifier voting mode to obtain interference data with the marked interference type.
3. The method of claim 2, wherein the final labeling of the sample interference data of the positive interference markers, the sample interference data of other interference markers, and the sample interference data of non-interference markers by means of classifier voting comprises:
randomly dividing sample interference data with the same initial mark to obtain a plurality of interference groups containing the sample interference data of the positive interference mark, the sample interference data of other interference marks and the sample interference data of the non-interference mark;
training each interference group to generate a classifier for voting the sample interference data;
and finally marking the sample interference data of the positive interference mark, the sample interference data of other interference marks and the sample interference data of the non-interference marks by using the classifier to obtain the interference data with the marked interference type.
4. The method of claim 1, further comprising:
and when the external field to-be-verified interference data does not conform to the interference type in the trained interference model, performing recursive training on the interference model again according to the interference type of the external field to-be-verified interference data to obtain and store a new interference model.
5. An apparatus for detecting interference, comprising:
the marking module is used for marking the interference types of the collected sample interference data by utilizing various classifier voting modes to obtain interference data with the marked interference types;
the time-frequency domain mapping processing module is used for performing time-frequency domain mapping processing on the interference data with the marked interference types to obtain a time-frequency domain mapped interference map;
the training module is used for training the interference map subjected to time-frequency domain mapping to obtain a trained interference model;
and the interference classification module is used for classifying the interference types of the external field to-be-verified interference data by using the trained interference model to obtain the interference types of the external field to-be-verified interference data.
6. The apparatus of claim 5, wherein the tagging module comprises:
the initial marking unit is used for respectively carrying out initial marking on the collected sample interference data of various interference types to obtain sample interference data with initial marking, and marking the sample interference data as sample interference data of a positive interference mark, sample interference data of other interference marks and sample interference data of an interference-free mark according to the interference energy intensity of the sample interference data with the same initial marking;
and the final marking unit is used for finally marking the sample interference data of the positive interference marks, the sample interference data of other interference marks and the sample interference data of the non-interference marks in a classifier voting mode to obtain the interference data with the marked interference types.
7. The apparatus of claim 6, wherein the final marking unit comprises:
the dividing unit is used for randomly dividing the sample interference data with the same initial mark to obtain a plurality of interference groups containing the sample interference data of the positive interference mark, the sample interference data of other interference marks and the sample interference data of the non-interference mark;
a training subunit, configured to generate a classifier for voting on the sample interference data by training each interference group;
and the final marking subunit is used for finally marking the sample interference data of the positive interference mark, the sample interference data of other interference marks and the sample interference data of the non-interference marks by using the classifier to obtain the interference data with the marked interference type.
8. The apparatus of claim 5, wherein the training module is further configured to, when the external field to-be-verified interference data does not conform to the interference type in the trained interference model, perform recursive training on the interference model again according to the interference type of the external field to-be-verified interference data, to obtain and store a new interference model.
9. An apparatus for detecting interference, the apparatus comprising: a processor, and a memory coupled to the processor; the memory has stored thereon a program for detecting disturbances which is executable on the processor, the program for detecting disturbances implementing the steps of the method for detecting disturbances according to any one of claims 1 to 4 when being executed by the processor.
10. A computer storage medium, characterized in that it stores a program for detecting interference, which when executed by a processor implements the steps of the method for detecting interference according to any one of claims 1 to 4.
Priority Applications (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810693247.4A CN110659656B (en) | 2018-06-29 | 2018-06-29 | Method and device for detecting interference |
| CN202311283734.0A CN117216573A (en) | 2018-06-29 | 2018-06-29 | Method and device for detecting interference |
| PCT/CN2019/091548 WO2020001311A1 (en) | 2018-06-29 | 2019-06-17 | Method for detecting interference, apparatus, device, and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810693247.4A CN110659656B (en) | 2018-06-29 | 2018-06-29 | Method and device for detecting interference |
Related Child Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311283734.0A Division CN117216573A (en) | 2018-06-29 | 2018-06-29 | Method and device for detecting interference |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110659656A true CN110659656A (en) | 2020-01-07 |
| CN110659656B CN110659656B (en) | 2024-08-16 |
Family
ID=68986142
Family Applications (2)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311283734.0A Pending CN117216573A (en) | 2018-06-29 | 2018-06-29 | Method and device for detecting interference |
| CN201810693247.4A Active CN110659656B (en) | 2018-06-29 | 2018-06-29 | Method and device for detecting interference |
Family Applications Before (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311283734.0A Pending CN117216573A (en) | 2018-06-29 | 2018-06-29 | Method and device for detecting interference |
Country Status (2)
| Country | Link |
|---|---|
| CN (2) | CN117216573A (en) |
| WO (1) | WO2020001311A1 (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111835443A (en) * | 2020-07-23 | 2020-10-27 | 中国联合网络通信集团有限公司 | Method and device for inhibiting atmospheric waveguide interference |
| CN112083393A (en) * | 2020-10-27 | 2020-12-15 | 西安电子科技大学 | Intermittent sampling forwarding interference identification method based on spectrogram average time characteristic |
| WO2022188467A1 (en) * | 2021-03-09 | 2022-09-15 | 中兴通讯股份有限公司 | Interference detection method, apparatus and device, and storage medium |
| CN115379477A (en) * | 2022-08-10 | 2022-11-22 | 中国联合网络通信集团有限公司 | Method, device, equipment and storage medium for determining cell interference type |
| CN115379476A (en) * | 2022-08-10 | 2022-11-22 | 中国联合网络通信集团有限公司 | Method, device, equipment and storage medium for determining cell interference type |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114375007A (en) * | 2020-10-15 | 2022-04-19 | 中国移动通信集团浙江有限公司 | Uplink interference classification method, device, equipment and readable medium |
| CN114581252B (en) * | 2022-03-03 | 2024-04-05 | 平安科技（深圳）有限公司 | Target case prediction method and device, electronic equipment and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103632168A (en) * | 2013-12-09 | 2014-03-12 | 天津工业大学 | Classifier integration method for machine learning |
| EP3086320A1 (en) * | 2015-04-23 | 2016-10-26 | Adidas AG | Method and device for associating frames in a video of an activity of a person with an event |
| CN106330385A (en) * | 2016-08-29 | 2017-01-11 | 电子科技大学 | Interference type identification method |
| CN107563303A (en) * | 2017-08-09 | 2018-01-09 | 中国科学院大学 | A kind of robustness Ship Target Detection method based on deep learning |
| CN107734507A (en) * | 2016-08-12 | 2018-02-23 | 索尼公司 | Wireless scene identification device and method and Wireless Telecom Equipment and system |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108197545A (en) * | 2017-12-25 | 2018-06-22 | 广东海格怡创科技有限公司 | The recognition methods of interference type and system |
| CN108135003A (en) * | 2017-12-25 | 2018-06-08 | 广东海格怡创科技有限公司 | The construction method and system of interference type identification model |
-
2018
- 2018-06-29 CN CN202311283734.0A patent/CN117216573A/en active Pending
- 2018-06-29 CN CN201810693247.4A patent/CN110659656B/en active Active
-
2019
- 2019-06-17 WO PCT/CN2019/091548 patent/WO2020001311A1/en active Application Filing
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103632168A (en) * | 2013-12-09 | 2014-03-12 | 天津工业大学 | Classifier integration method for machine learning |
| EP3086320A1 (en) * | 2015-04-23 | 2016-10-26 | Adidas AG | Method and device for associating frames in a video of an activity of a person with an event |
| CN107734507A (en) * | 2016-08-12 | 2018-02-23 | 索尼公司 | Wireless scene identification device and method and Wireless Telecom Equipment and system |
| CN106330385A (en) * | 2016-08-29 | 2017-01-11 | 电子科技大学 | Interference type identification method |
| CN107563303A (en) * | 2017-08-09 | 2018-01-09 | 中国科学院大学 | A kind of robustness Ship Target Detection method based on deep learning |
Non-Patent Citations (4)
| Title |
|---|
| QIANG YE AT.AL.: ""Improving a Neural Network Classifier Ensemble with Multi-Task Learning"", 《THE 2006 IEEE INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORK PROCEEDINGS》, 30 October 2006 (2006-10-30) * |
| 司莉;毕贵红;魏永刚;陶然;张寿明;: "基于RQA与SVM的声发射信号检测识别方法", 振动与冲击, no. 02, 28 January 2016 (2016-01-28), pages 97 - 103 * |
| 李言兵;: "基于随机森林算法模型的干扰预测", 山东通信技术, no. 04, 15 December 2017 (2017-12-15), pages 25 - 26 * |
| 苑成成: ""标签噪声的集成学习研究"", 《中国优秀硕士学位论文全文数据库 信息科技辑》, vol. 2017, no. 03, 15 March 2017 (2017-03-15), pages 140 - 152 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111835443A (en) * | 2020-07-23 | 2020-10-27 | 中国联合网络通信集团有限公司 | Method and device for inhibiting atmospheric waveguide interference |
| CN111835443B (en) * | 2020-07-23 | 2023-03-24 | 中国联合网络通信集团有限公司 | Method and device for inhibiting atmospheric waveguide interference |
| CN112083393A (en) * | 2020-10-27 | 2020-12-15 | 西安电子科技大学 | Intermittent sampling forwarding interference identification method based on spectrogram average time characteristic |
| CN112083393B (en) * | 2020-10-27 | 2023-06-30 | 西安电子科技大学 | Intermittent sampling forwarding interference identification method based on spectrogram average time characteristics |
| WO2022188467A1 (en) * | 2021-03-09 | 2022-09-15 | 中兴通讯股份有限公司 | Interference detection method, apparatus and device, and storage medium |
| CN115379477A (en) * | 2022-08-10 | 2022-11-22 | 中国联合网络通信集团有限公司 | Method, device, equipment and storage medium for determining cell interference type |
| CN115379476A (en) * | 2022-08-10 | 2022-11-22 | 中国联合网络通信集团有限公司 | Method, device, equipment and storage medium for determining cell interference type |
| CN115379477B (en) * | 2022-08-10 | 2024-03-15 | 中国联合网络通信集团有限公司 | Method, device, equipment and storage medium for determining cell interference type |
| CN115379476B (en) * | 2022-08-10 | 2024-03-19 | 中国联合网络通信集团有限公司 | Method, device, equipment and storage medium for determining cell interference type |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2020001311A1 (en) | 2020-01-02 |
| CN117216573A (en) | 2023-12-12 |
| CN110659656B (en) | 2024-08-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110659656B (en) | Method and device for detecting interference | |
| CN106357618B (en) | Web anomaly detection method and device | |
| US9923912B2 (en) | Learning detector of malicious network traffic from weak labels | |
| US20230300159A1 (en) | Network traffic anomaly detection method and apparatus, and electronic apparatus and storage medium | |
| US8682813B2 (en) | Sample class prediction method, prediction program, and prediction apparatus | |
| CN107908642B (en) | Industry text entity extraction method based on distributed platform | |
| CN108363701B (en) | Named entity identification method and system | |
| CN111325260B (en) | Data processing method and device, electronic equipment and computer readable medium | |
| CN107483451B (en) | Method and system for processing network security data based on serial-parallel structure and social network | |
| US20230385699A1 (en) | Data boundary deriving system and method | |
| CN116038076B (en) | Welding shielding gas control method based on scene matching | |
| CN113139536A (en) | Text verification code identification method and equipment based on cross-domain meta learning and storage medium | |
| CN111126420B (en) | Method and device for establishing identification model | |
| CN114970694B (en) | Network security situation assessment method and model training method thereof | |
| JP7214822B1 (en) | CAM-based weakly supervised learning object detection apparatus and method | |
| CN113128584B (en) | Mode-level unsupervised sorting method of multifunctional radar pulse sequence | |
| WO2019180868A1 (en) | Image generation device, image generation method, and image generation program | |
| KR20210158740A (en) | Apparatus and method for clustering validation based on machine learning performance | |
| CN113962999B (en) | Noise label segmentation method based on Gaussian mixture model and label correction model | |
| CN111651433B (en) | Sample data cleaning method and system | |
| CN103984960B (en) | A kind of remote sensing images sample intelligent acquisition method | |
| CN111581640A (en) | Malicious software detection method, device and equipment and storage medium | |
| US20160267168A1 (en) | Residual data identification | |
| CN117711608B (en) | Comprehensive analysis system for integrating nerve injury data | |
| JP2020030752A (en) | Information processing device, information processing method and program |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |