WO2022188467A1 - Interference detection method, apparatus and device, and storage medium - Google Patents

Abstract

An interference detection method, apparatus and device, and a storage medium. The method comprises: when an interference type of data to be subjected to detection is not detected by using a pre-created first time granularity detection model, filtering a combination of said data and a predetermined first interference detection probability, so as to obtain a first filtering result (S110); determining an interference detection probability of the first filtering result in a pre-created current time granularity detection model, and taking same as a second interference detection probability (S120), wherein the current time granularity detection model is a detection model, the time granularity of which is coarser than that of the first time granularity detection model; and determining the interference type of said data according to a comparison result of threshold values of the second interference detection probability and the first preset detection probability (S130).

Description

Interference detection method, apparatus, device and storage medium

CROSS-REFERENCE TO RELATED APPLICATIONS

This application is based on the Chinese patent application with the application number of 202110257673.5 and the filing date of March 09, 2021, and claims the priority of the Chinese patent application. The entire content of the Chinese patent application is incorporated herein by reference.

technical field

The present application relates to communications, and in particular, to an interference detection method, apparatus, device, and storage medium.

Background technique

Interference is one of the important problems faced by wireless networks. With the continuous construction of various wireless networks, various potential sources of interference are constantly being generated at an alarming rate, and wireless networks are faced with a complex interference environment. Communication products occupy the existing frequency resources of other networks, improper network configuration of operators, problems of the transmitter itself, overlapping of spectrum resources, and deliberate interference are all causes of wireless network interference. Network operators hope to optimize network performance and improve communication quality through interference identification. The existing interference identification methods have low accuracy and cannot adapt to the current complex network environment. Therefore, how to improve the interference detection rate to adapt to the complex network environment is a problem that needs to be solved at present.

SUMMARY OF THE INVENTION

In view of this, embodiments of the present application provide an interference detection method, apparatus, device, and storage medium.

An embodiment of the present application provides an interference detection method, which includes: in the case that the interference type of the data to be detected is not detected by using a pre-created first time granularity detection model, detecting the data to be detected and the predetermined first interference The combination of detection probabilities is filtered to obtain a first filtering result; the interference detection probability of the first filtering result in the pre-created current time granularity detection model is determined as the second interference detection probability, wherein the current time granularity detection The model is a detection model with a coarser time granularity than the first time granularity detection model; the interference of the data to be detected is determined according to the comparison result of the second interference detection probability and the first preset detection probability threshold value type.

An embodiment of the present application provides an interference detection apparatus, including: a first filter configured to detect the interference type of the data to be detected by using a pre-created first time granularity detection model to detect the interference type of the data to be detected. Perform filtering with a combination of a predetermined first interference detection probability to obtain a first filtering result; a first determining module is configured to determine the interference detection probability of the first filtering result in the pre-created current time granularity detection model, as The second interference detection probability, wherein the current time granularity detection model is a detection model with a coarser time granularity than the first time granularity detection model; the second determination module is configured to be based on the second interference detection probability and The comparison result of the first preset detection probability threshold value determines the interference type of the data to be detected.

An embodiment of the present application provides an interference detection device, including: a communication module, a memory, and one or more processors; the communication module is configured to perform communication interaction between communication nodes; the memory is configured to store One or more programs; when the one or more programs are executed by the one or more processors, the one or more processors enable the one or more processors to implement the method described in any of the above embodiments.

An embodiment of the present application provides a storage medium, where a computer program is stored in the storage medium, and when the computer program is executed by a processor, the method described in any of the foregoing embodiments is implemented.

Description of drawings

1 is a flowchart of a method for detecting interference provided by an embodiment of the present application;

2 is a structural block diagram of an interference detection apparatus provided by an embodiment of the present application;

3 is a structural block diagram of a training unit provided by an embodiment of the present application;

4 is a structural block diagram of a detection unit provided by an embodiment of the present application;

5 is a schematic diagram of the creation of a time granularity detection model provided by an embodiment of the present application;

6 is a schematic diagram of interference detection provided by an embodiment of the present application;

7 is a schematic diagram of the creation of another time granularity detection model provided by an embodiment of the present application;

FIG. 8 is another schematic diagram of interference detection provided by an embodiment of the present application;

FIG. 9 is another schematic diagram of interference detection provided by an embodiment of the present application;

10 is a structural block diagram of another interference detection apparatus provided by an embodiment of the present application;

FIG. 11 is a schematic structural diagram of an interference detection device provided by an embodiment of the present application.

Detailed ways

Hereinafter, the embodiments of the present application will be described with reference to the accompanying drawings. The present application will be described below with reference to the accompanying drawings. The examples are only used to explain the present application and are not intended to limit the scope of the present application.

In an embodiment, FIG. 1 is a flowchart of an interference detection method provided by an embodiment of the present application. This embodiment may be performed by an interference detection device. The interference detection device may be the terminal side (eg, user equipment). As shown in FIG. 1 , this embodiment includes: S110-S130.

S110. If the interference type of the data to be detected is not detected by using the pre-created first time granularity detection model, filter the combination of the data to be detected and the predetermined first interference detection probability to obtain a first filtering result.

The first time granularity detection model refers to a pre-created first time granularity level training model. In the actual operation process, the pre-created AI model is trained by using the first time granularity feature value until a satisfactory training effect is obtained, and the first time granularity level training model is output and used as the first time granularity detection model. Exemplarily, the first time granularity detection model may include one of the following: a slot-level detection model; a minute-level detection model; a day-level detection model; and a week-level detection model.

In the embodiment, if the interference type of the data to be detected is not detected by the first time granularity detection model, it indicates that the interference feature of the data to be detected is not in the first time granularity, in this case, the first time granularity detection can be used The time granularity detection model corresponding to the time granularity of the previous level corresponding to the time granularity of the model performs further interference detection on the data to be detected. First, the combination of the data to be detected and the predetermined first interference detection probability is filtered to obtain a first filtering result. The first interference detection probability refers to the probability that the data to be detected is output in the first time granularity detection model.

In an embodiment, since different time granularity detection models correspond to different time lengths, time domain filtering is performed on the combination of the data to be detected and the predetermined first interference detection probability to obtain a filtering result.

S120. Determine the interference detection probability of the first filtering result in the pre-created current time granularity detection model as the second interference detection probability.

The current time granularity detection model is a detection model with a coarser time granularity than the first time granularity detection model. In an embodiment, the time granularity corresponding to the current time granularity detection model is coarser than the time granularity corresponding to the first time granularity detection model. Exemplarily, when the first time granularity detection model is a slot-level detection model, the current time granularity detection model is a minute-level detection model; when the first time granularity detection model is a minute-level detection model, the current time The granularity detection model is a sky-level detection model; when the first-time granularity detection model is a sky-level detection model, the current time granularity detection model is a week-level detection model.

In the embodiment, the interference detection probability can be understood as the frequency at which the first filtering result appears at the current time granularity, that is, the interference detection probability can be directly calculated by the frequency at which the first filtering result appears at the current time granularity. Exemplarily, it is assumed that the current time granularity detection model is a slot-level detection model, and the total detection duration is 10 slots, wherein, within the total detection duration, there are 2 slots with interference 1 and 4 slots with interference 2. , then the interference detection probability of interference 1 is 0.2, and the interference detection probability of interference 2 is 0.4.

S130. Determine the interference type of the data to be detected according to the comparison result between the second interference detection probability and the first preset detection probability threshold.

The first preset detection probability threshold is a preconfigured detection probability threshold. In the embodiment, when the second interference detection probability is greater than the first preset detection probability threshold value, it indicates that interference within the current time granularity in the data to be detected has been detected; when the second interference detection probability is smaller than the first In the case of preset detection probability threshold value, indicating that no interference is detected within the current time granularity, the filtering result is not marked as any interference. This solution can adapt to complex network scenarios by intelligently detecting the type of interference in the data to be detected by combining AI.

In an embodiment, when the second interference detection probability is smaller than the corresponding first preset detection probability threshold value, the method further includes: switching the current time granularity detection model to a time granularity detection model with a coarser time granularity, and returning to The step of filtering the combination of the data to be detected and the predetermined first interference detection probability to obtain a first filtering result, until the interference type of the data to be detected is detected. In an embodiment, in the case where the interference type of the data to be detected is not detected by the first time granularity detection model, the current time granularity detection model is switched to a time granularity detection model with a coarser time granularity, and the next time granularity is used The coarser time granularity detection model performs interference detection on the data to be detected until the type of interference in the data to be detected is detected, so that the multi-level time granularity detection model is used to perform interference detection on the data to be detected at different time granularities. The detection results at all levels of time granularity are obtained, thereby improving the interference detection accuracy in complex network scenarios.

In one embodiment, filtering the combination of the data to be detected and the predetermined first interference detection probability to obtain the first filtering result includes: combining the data to be detected and the predetermined first interference detection probability to obtain combined data; The combined data is filtered by the time length of the current time granularity to obtain a first filtering result. In the embodiment, the process of combining the data to be detected and the first interference detection probability refers to combining the first interference detection probability and the data to be detected as combined data. Exemplarily, it is assumed that the character'_slot of the data to be detected includes: [RSSI ^' , NI ^' , Cha ^' _space , Cha ^' _time , Cha ^' _freq1 , Cha ^' _freq2 ], and the first interference detection probability includes P1'_slot, P2'_slot ,...,P6′_slot, then the combined data is [RSSI ^' ,NI ^' ,Cha ^' _space ,Cha ^' _time ,Cha ^' _freq1 ,Cha ^' _freq2 ,P1'_slot,P2'_slot,...,P6' _slot].

In one embodiment, the combined data is filtered by the time length of the current time granularity to obtain a first filtering result, including: according to the instantaneous combined data of the current time granularity, the combined data filtered by the previous current time granularity, and the first prediction result. A weight coefficient and a second preset weight coefficient are set to filter the combined data to obtain a first filtering result. In an embodiment, the first preset weight coefficient and the second preset weight coefficient are empirical values obtained through a large number of experiments in advance. It can be understood that different preset weight coefficients are set according to different time granularities. It should be noted that, the first preset weight coefficient and the second preset weight coefficient may be equal or unequal, which is not limited. Of course, in the actual operation process, the first preset weight coefficient and the second preset weight coefficient may also be adjusted according to the actual situation, which is not limited. Exemplarily, assuming that the current time granularity is minutes, the instantaneous combined data of the current time granularity refers to the instantaneous combined data of the current 1 minute; the combined data filtered by the previous current time granularity refers to the filtered data of the previous one minute. Combine data. For example, suppose that traindata_minute' _t represents the combined data obtained after filtering for the current 1 minute, traindata_minute' _t-1 represents the combined data filtered for the previous 1 minute, traindata_minute _t represents the instantaneous combined data of the current 1 minute, and the first preset weight coefficient is 0.1, and the second preset weight coefficient is 0.9, then traindata_minute′ _t =0.9*traindata_minute′ _t−1 +0.1*traindata_minute _t .

In an embodiment, before the interference type of the data to be detected is not detected by using the pre-created first time granularity detection model, the method further includes:

inputting the predetermined first time granularity feature value into the pre-created first AI training model, until the detection rate reaches the first preset detection rate threshold, and outputting the first AI training model and the first time granularity interference probability distribution value, and use the first AI training model as the first time granularity detection model;

combining the first time granularity characteristic value and the first time granularity interference probability distribution value as the second time granularity characteristic value;

Perform time domain filtering on the second time granularity feature value to obtain the second time granularity level training data; input the second time granularity level training data into the pre-created second AI training model until the detection rate reaches the second preset detection rate rate threshold, output the second AI training model and the second time granularity interference probability distribution value, and use the second AI training model as the second time granularity detection model; wherein, the difference between the first time granularity detection model and the second time granularity detection model The corresponding time granularity becomes thicker in turn.

In an embodiment, after the second AI training model is used as the second time granularity detection model, the method further includes:

combining the second time granularity level training data and the second time granularity interference probability distribution value as a third time granularity feature value;

Perform time domain filtering on the third time granularity feature value to obtain the third time granularity level training data;

Input the training data at the third time granularity level into the pre-created third AI training model until the detection rate reaches the third preset detection rate threshold, output the third AI training model and the third time granularity interference probability distribution value, and use The third AI training model is used as the third time granularity detection model; wherein, the time granularities corresponding to the first time granularity detection model, the second time granularity detection model, and the third time granularity detection model become thicker in sequence.

Exemplarily, when the first time granularity detection model is a slot-level detection model, the time granularity corresponding to the second time granularity detection model is coarser than the time granularity corresponding to the first time granularity detection model, for example, the second time granularity detection model The model may be a minute-level detection model or a day-level detection model; correspondingly, the time granularity corresponding to the third time granularity detection model is coarser than the time granularity corresponding to the second time granularity detection model. For example, the third time granularity detection model may be Weekly detection model.

It should be noted here that the process of training the time granularity detection model refers to training the time granularity detection models corresponding to at least two time granularities, that is, the time granularity detection models corresponding to the two time granularities can be trained. , it is also possible to train time granularity detection models corresponding to three time granularities, or even four, five...N time granularity detection models corresponding to N time granularities. Of course, if more time granularity detection models corresponding to the time granularity are trained, the higher the computing power of the device is required.

In an embodiment, the first time granularity feature value includes one of the following: time domain RSSI; frequency domain NI; spatial domain matched filtering; time domain correlation value; frequency domain correlation value; intermediate RB energy distribution.

In one embodiment, the time granularity detection model includes one of the following: a slot-level detection model; a minute-level detection model; a day-level detection model; and a week-level detection model.

In an embodiment, the interference detection process is described by taking an example that the interference detection apparatus includes two processing units. FIG. 2 is a structural block diagram of an interference detection apparatus provided by an embodiment of the present application. As shown in FIG. 2 , the interference detection apparatus in this embodiment includes: a training unit 210 and a detection unit 220 .

FIG. 3 is a structural block diagram of a training unit provided by an embodiment of the present application. As shown in FIG. 3 , the training unit 210 is composed of several sub-units, which are a data preprocessing unit 2101 and a neural network training unit 2102 respectively. Among them, the data preprocessing unit 2101 is used to perform feature value extraction, data filtering and other operations on the data; the neural network training unit 2102 is used to send the obtained training sample data into the neural network for training, adjust the training parameters, and finally get the trained Detection model.

FIG. 4 is a structural block diagram of a detection unit provided by an embodiment of the present application. As shown in FIG. 4 , the detection unit 220 is composed of several sub-units, which are respectively a data preprocessing unit 2201 , a model selection unit 2202 , a decision unit 2203 and a result collection unit 2204 . Among them, the data preprocessing unit 2201 is used to perform feature value extraction, data filtering and other operations on the data to generate data of corresponding time granularity; the model selection unit 2202 is used to select the corresponding trained model; the judgment unit 2203 is used to use the model The model selected by the selection unit judges the data type; the result collection unit 2204 is used to collect and organize the output results of each level of model.

In an embodiment, taking the first time granularity feature value including: time domain RSSI, frequency domain NI, spatial domain matched filtering, time domain correlation value, frequency domain correlation value and intermediate RB energy distribution as an example, for the 5G uplink system The interference detection process is explained. Table 1 is a schematic diagram of an initial training data structure provided by an embodiment of the present application. As shown in Table 1, in this embodiment, the combination of measurement quantities in the 5G uplink system is: time domain RSSI, frequency domain NI, spatial domain matched filtering, time domain correlation value, frequency domain correlation value and intermediate RB energy distribution. Table 2 is a schematic diagram of interference types in a 5G NR uplink system provided by an embodiment of the present application. As shown in Table 2, the interference types of the 5G NR uplink system can include: atmospheric waveguide interference, frame out-of-sync interference, D1D2 interference, D4D5 interference, sideband interference and narrowband interference.

Table 1 A schematic diagram of an initial training data structure

Table 2 Schematic diagram of interference types in a 5G NR uplink system

Interference number	Kind of interference
1	Atmospheric duct interference
2	frame out-of-sync interference
3	D1D2 interference
4	D4D5 Interference
5	sideband interference
6	narrowband interference

Before using the detection unit to perform interference detection on the data to be detected, the training unit is used to create a time granularity detection model. FIG. 5 is a schematic diagram of creating a time granularity detection model provided by an embodiment of the present application. Exemplarily, in this embodiment, the training process of the time granularity detection models corresponding to the three time granularities is described. For example, the first time granularity detection model, the second time granularity detection model and the third time granularity detection model are in sequence: time slot level detection model, minute level detection model and day level detection model. process is explained. As shown in Figure 5, the creation process of the time granularity detection model includes the following steps:

Step 1: Obtain RSSI values of 14 symbols in one slot (slot), RSSI _symbol (also referred to as time domain RSSI), where symbol=0, 1, . . . 13.

Step 2: Calculate the correlation coefficient between adjacent symbols Coef _{symbol1, symbol2} (also called time domain correlation value), where symbol1=0,1,...12, symbol2=symbol2+1, so the final result will be 13 time domain features Cha _time .

Step 3: Obtain the NI value NI _rb (also referred to as frequency domain NI) of 273 RBs in a slot, where rb=0, 1, . . . , 272.

Step 4: Calculate the correlation coefficient Coef _rb1,rb2 (also called frequency domain correlation value) between adjacent RBs, where rb1=0,1,...271, rb2=rb1+1, and finally 272 are obtained frequency domain features.

Step 5: Obtain the middle RB, that is, the received power value Power _sc of all subcarriers on the 137th RB (also called the middle RB energy distribution), sc=0, 1, . . . , 11, and obtain 12 frequency domains Feature Cha1 _freq .

Step 6: Obtain matched filtering results of different beams (also called spatial matching filtering), assuming that the number of beams is n, and obtain n spatial domain features Cha _space .

Step 7: The final extracted feature is [RSSI,NI,Cha _space ,Cha _time ,Cha _freq1 ,Cha _freq2 ], denoted as character_slot.

Step 8: Send the character_slot into the pre-created AI model for training, and obtain a satisfactory training effect (you can judge whether the interference detection accuracy rate reaches the first preset detection rate threshold, and when the first preset detection rate threshold is reached, it is considered that After obtaining satisfactory training effect), output two results, one is the slot-level training model (AI_slot, i.e. the first AI training model in the above-mentioned embodiment), and the slot-level training model is saved; the other is the preset interference probability of the slot Distribution value (Pn_slot, namely the first time granularity interference probability distribution value in the above-mentioned embodiment), wherein Pn_slot=[P1_slot, P2_slot, P3_slot, P4_slot, P5_slot, P6_slot].

Step 9: The data preprocessing unit of the training unit combines the character_slot output in step 7 and the Pn_slot output in step 8 into traindata_slot, where traindata_slot=[RSSI, NI, Chaspace, Chatime, Chafreq1, Chafreq2, P1_slot, P2_slot, P3_slot, P4_slot, P5_slot, P6_slot], filter the time domain of traindata_slot for 1 minute to obtain minute-level training data, which is recorded as character_minute.

Among them, the filtering method of traindata_slot is:

traindata_slot' _t =0.9*traindata_slot' _t-1 +0.1*traindata_slot _t , where traindata_slot' _t represents the training data obtained after the current slot filtering, traindata_slot' _t-1 represents the training data after the previous slot filtering, and traindata_slot _t represents Instantaneous training data for the current slot.

Step 10: Send character_minute into the pre-created AI model for training, and obtain a satisfactory training effect (you can judge whether the interference detection accuracy rate reaches the second preset detection rate threshold, and when the second preset detection rate threshold is reached, it is considered that After obtaining satisfactory training effect), output two results, one is the minute-level training model (AI_minute, namely the second AI training model in the above-described embodiment), saves the minute-level training model; The second is the preset interference probability of this minute Distribution value (Pn_minute, namely the second time granularity interference probability distribution value in the above-mentioned embodiment), wherein Pn_minute=[P1_mint, P2_mint, P3_mint, P4_mint, P5_mint, P6_mint].

Step 11: The data preprocessing unit of the training unit combines the character_minute output in step 9 and the Pn_minute output in step 10 into traindata_minute, where traindata_minute=[RSSI, NI, Chaspace, Chatime, Chafreq1, Chafreq2, P1_minute, P2_minute, P3_minute, P4_minute ,P5_minute,P6_minute], filter the traindata_minute time domain for 1 day to obtain the training data of day level, which is recorded as character_day.

Among them, the filtering method of traindata_minute is:

traindata_minute′ _t = 0.9*traindata_minute′ _t-1 +0.1*traindata_minute _t , where traindata_minute′ _t represents the training data obtained after the current 1-minute filtering, traindata_minute′ _t-1 represents the training data after the previous 1-minute filtering, traindata_minute′ t _t represents the instantaneous training data of the current 1 minute.

Step 12: Send character_day into the pre-created AI model for training, and get a satisfactory training effect (you can judge whether the interference detection accuracy reaches the third preset detection rate threshold, and when it reaches the third preset detection rate threshold, After it is considered that a satisfactory training effect is obtained), two results are output, one is the day-level training model (AI_day, that is, the third AI training model in the above embodiment), and the day-level training model is saved; the second is the preset interference of the day The probability distribution value (Pn_day, that is, the third time granularity interference probability distribution value in the above embodiment), wherein Pn_day=[P1_day, P2_day, P3_day, P4_day, P5_day, P6_day].

So far, in the training phase, a total of three detection models with time granularity have been generated, namely the slot-level detection model, the minute-level detection model, and the day-level detection model.

It should be noted here that in the process of creating the time granularity detection model, steps 1 to 7, step 9 and step 11 are performed by the data preprocessing unit in the training unit, and step 8, step 10 and step 12 are performed by The neural network training unit in the training unit executes.

FIG. 6 is a schematic diagram of interference detection provided by an embodiment of the present application. Exemplarily, in the embodiment, the time granularity detection model created as shown in FIG. 5 is used to detect the interference type of the data to be detected, that is, three time granularity detection models are used to detect the interference type of the data to be detected. The process of interference detection may be performed by the detection unit in the above embodiment. As shown in Figure 6, the interference detection process includes the following steps:

Step 1: The data preprocessing unit extracts feature values consistent with the feature values extracted from the training set on the original data to obtain character'_slot (that is, the data to be detected in the above embodiment).

Step 2: The model selection unit calls the slot-level detection model (AI_slot), the character'_slot of the data to be detected enters the slot-level detection model, and outputs the interference detection probability P1'_slot, P2'_slot,...,P6'_slot (ie the first interference detection probability in the above embodiment).

Step 3: The preset time-slot-level threshold configured in the judgment unit is Thr _slot , and it is judged whether the detection probability in step 2 exceeds the preset time-slot-level threshold Thr _slot , if so, it is considered that the corresponding interference occurs, and the current time slot is determined. The data is marked as its corresponding interference, and the judgment result is set to 1 and output to the result collection unit; if not, it is considered that no interference is detected in the current time slot, and the current time slot data is not marked as any interference, and is divided into an unknown branch, and the judgment result Set to 0 and output to the result collection unit. Exemplarily, taking the frame out-of-synchronization interference in this embodiment as an example, this kind of interference will show obvious characteristics in a time slot, so it can be successfully detected under the time slot level model.

Step 4: the data preprocessing unit merges the data of the unknown branch that the step 3 produces and the Pn '_slot that the step 2 outputs, carries out filtering, after 1 minute of filtering time, obtains the filtering result character'_minute (that is, the first in the above-mentioned embodiment). a filter result).

Step 5: The model selection unit calls the minute-level detection model AI_minu (as the current time granularity detection model), the character′_minute of the data to be detected enters the minute-level model, and outputs the interference detection probability P1′_minu, P2′_minu,...,P6′ _minu (ie the second interference detection probability in the above embodiment).

Step 6: The preset minute-level threshold configured in the judgment unit is Thr _minu , and it is judged whether the detection probability in step 5 exceeds the preset minute-level threshold Thr _minu , if so, it is considered that the corresponding interference occurs, and the current minute data is marked as For the corresponding interference, the judgment result is set to 1 and output to the result collection unit; if not, it is considered that no interference has been detected in the current minute, and the current minute data is not marked as any interference, divided into the unknown branch, and the judgment result is set to 0 and output to the Results collection unit. Exemplarily, taking the narrowband interference in this embodiment as an example, this narrowband interference will show obvious characteristics within one minute. Although it cannot be detected at the time slot level, it can be successfully detected in the minute level model. .

Step 7: The data preprocessing unit combines the data of the unknown branch generated in step 6 with the Pn'_minu output in step 5, and performs filtering respectively. After the filtering time is 1 day, the filtering result character'_day is obtained.

Step 8: The model selection unit calls the sky-level detection model AI_day, the data character'_day to be detected enters the sky-level model, and outputs the interference detection probability P1′_day, P2′_day,...,P6′_day.

Step 9: The preset sky-level threshold configured in the judgment unit is Thr _day , and it is judged whether the detection probability in step 8 exceeds the preset sky-level threshold Thr _day . If so, it is considered that the corresponding interference occurs, and the current sky-level data is marked. For its corresponding interference, the judgment result is set to 1 and output to the result collection unit; if not, it is considered that the current sky level has not detected interference, and the current sky level data is not marked as any interference, divided into the unknown branch, and the judgment result is set as 0 is output to the result collection unit. Taking the atmospheric waveguide interference in this implementation as an example, this kind of interference will show obvious characteristics within a day. Although it cannot be detected at the time slot level and the minute level, it can be successfully detected under the sky-level model.

Step 10: The result collection unit provides the final interference type of the data to be detected according to the judgment result of the three-level model.

It should be noted that this embodiment adopts the current time granularity detection model (that is, the minute-level detection model) to continue when the first time granularity detection model (ie, the slot-level detection model) does not detect the interference type of the data to be detected. Detect the interference type of the data to be detected. If the interference type is still not detected, switch the current time granularity detection model to a time granularity detection model with a coarser time granularity (ie, the sky-level detection model), and return the data to be detected and The combination of the first interference detection probabilities is filtered to obtain a new first filtering result, and the interference type of the data to be detected continues to be detected until the interference type of the data to be detected is detected. Of course, in the actual operation process, when the interference type of the data to be detected is not detected by the sky-level detection model, a time granularity detection model with coarser time granularity than that of the sky-level detection model can be used to continue to perform interference detection on the data to be detected.

In the technical solution of this embodiment, a multi-level time granularity detection model is generated by adapting data of different time granularities in the detection model training stage. Then, the detection model of multi-level time granularity is used to adapt to the interference of different time granularities, which improves the ability of interference type identification, and further improves the accuracy of interference type identification.

In an embodiment, the interference detection process in the 5G NR uplink system is described by taking the first time granularity feature value including: 15-minute granularity frequency domain NI as an example. Table 3 is a schematic diagram of another initial training data structure provided by the embodiment of the present application. Table 4 is a schematic diagram of interference types in another 5G NR uplink system provided by the embodiment of this application. As shown in Table 3, a combination of measurement quantities that can be obtained by the 5G NR uplink system is: [15-minute granularity frequency domain NI], and as shown in Table 4, the interference types of the 5G NR uplink system can include: narrowband interference, D4D5 jamming, full band jamming and zero band jamming.

Table 3 Another schematic representation of the initial training data structure

Table 4 Schematic diagram of interference types in another 5G NR uplink system

Interference number	Kind of interference
1	narrowband interference
2	D4D5 Interference
3	full-band interference
4	Zero frequency interference

Before using the detection unit to perform interference detection on the data to be detected, the training unit is used to create a time granularity detection model. FIG. 7 is a schematic diagram of creating another time granularity detection model provided by an embodiment of the present application. In this embodiment, the creation of three time granularity detection models is taken as an example, and the first time granularity detection model, the second time granularity detection model and the third time granularity detection model are in order: a minute-level detection model, a day-level detection model Taking the weekly detection model as an example, the creation process of the time granularity detection model is described. As shown in Figure 7, the creation process of the time granularity detection model includes the following steps:

Step 1: Obtain an NI value NI _rb of 273 RBs within 15 minutes (also referred to as 15-minute granularity frequency domain NI), where rb=0,1,...,272.

Step 2: Calculate the correlation coefficient Coef _{rb1, rb2} between adjacent RBs (that is, the frequency domain correlation value in the above embodiment), where rb1=0,1,...271, rb2=rb1+1, and finally get 272 frequency domain features Cha _freq .

Step 3: The final extracted feature is [15-minute granularity frequency domain NI, Cha _freq ], denoted as character_15m.

Step 4: Send character_15m into the pre-created AI model for training, and get a satisfactory training effect (you can judge whether the interference detection accuracy reaches the first preset detection rate threshold, and when it reaches the first preset detection rate threshold, it is considered that After obtaining a satisfactory training effect), after obtaining a satisfactory training effect, output two results, one is the 15-minute level training model (AI_15m, namely the first AI training model in the above embodiment), and save the 15-minute level training model AI_15m The second is the 15-minute preset interference probability distribution value (Pn_15m, namely the first time granularity interference probability distribution value in the above-mentioned embodiment), wherein Pn_15m=[P1_15m, P2_15m, P3_15m, P4_15m].

Step 5: The preprocessing unit of the training unit merges the character_15m output in step 3 and the Pn_15m output in step 4 into traindata_15m, where traindata_15m=[NI, P1_15m, P2_15m, P3_15m, P4_15m], and filters traindata_15m to obtain sky-level training data , denoted as character_day.

Among them, the filtering method of traindata_15m is:

traindata_15m′ _t =0.9*traindata_15m′ _t-1 +0.1*traindata_15m _t , where traindata_15m′ _t represents the training data obtained after the current 15-minute filtering, traindata_15m′ _t-1 represents the training data after the previous 15-minute filtering, traindata_15m _t represents the instantaneous training data of the current 15 minutes.

Step 6: Send character_day into the pre-created AI model for training, and get a satisfactory training effect (you can judge whether the interference detection accuracy rate reaches the second preset detection rate threshold, and when the second preset detection rate threshold is reached, it is considered that After obtaining a satisfactory training effect), output two results, one is the day-level training model (AI_day, that is, the second AI training model in the above embodiment), and the day-level training model is saved; The second is the preset interference probability of the day Distribution value (Pn_day, the second time granularity interference probability distribution value in the above embodiment), where Pn_day=[P1_day, P2_day, P3_day, P4_day].

Step 7: The preprocessing unit of the training unit merges the character_day output in step 5 and the Pn_day output in step 6 into traindata_day, where traindata_day=[NI, P1_day, P2_day, P3_day, P4_day], filter the time domain of traindata_day for 1 week, and get the week Class training data, denoted as character_week.

Among them, the filtering method of traindata_day is:

traindata_day' _t =0.9*traindata_day' _t-1 +0.1*traindata_day _t , where traindata_day' _t represents the training data obtained after the current 1-day filtering, traindata_day' _t-1 represents the training data after the previous day's filtering, traindata_day _t Represents the instantaneous training data for the current day.

Step 8: Send character_week into the AI model for training, and obtain a satisfactory training effect (you can judge whether the interference detection accuracy rate reaches the third preset detection rate threshold, and when the third preset detection rate threshold is reached, it is considered to be satisfactory. After training effect), output two results, one is the week-level training model (AI_week, namely the third AI training model in the above-mentioned embodiment), and the weekly-level training model is saved; the second is the preset interference probability distribution value of the week ( Pn_week, that is, the third time granularity interference probability distribution value in the above embodiment), wherein Pn_week=[P1_week, P2_week, P3_week, P4_week].

So far, in the training phase, a total of three detection models with time granularity have been generated, namely, a 15-minute-level detection model, a day-level detection model, and a weekly-level detection model.

It should be noted here that steps 1 to 3, step 5 and step 7 in the creation process of the time granularity detection model shown in FIG. 7 are performed by the data preprocessing unit in the training unit; step 4, step 6 and step 8 Executed by the neural network training unit in the training unit.

FIG. 8 is a schematic diagram of another interference detection provided by an embodiment of the present application. In the embodiment, the time granularity detection model created as shown in FIG. 7 is used to detect the interference type of the data to be detected. The process of interference detection may be performed by the detection unit in the above embodiment. Exemplarily, the interference detection process is described by taking the first time granularity detection model as a 15-minute-level detection model, the current time granularity detection model as a sky-level detection model, and a week-level detection model as examples. As shown in Figure 8, the interference detection process includes the following steps:

Step 1: The data preprocessing unit extracts feature values consistent with the feature values extracted from the training set on the original data to obtain character'_15m (ie, the data to be detected in the above embodiment).

Step 2: The model selection unit calls the pre-created 15-minute detection model, the data character'_15m to be detected enters the 15-minute detection model, and outputs the interference detection probabilities P1', P2',...,P4' (that is, the above embodiment). The first jammer detection probability in ).

Step 3: The preset 15-minute level threshold configured in the judgment unit is Thr _15m , and it is determined whether the detection probability in step 2 exceeds the preset 15-minute level threshold Thr _15m , if so, it is considered that the corresponding interference occurs, and the current 15 minutes The data is marked as its corresponding interference, and the judgment result is set to 1 and output to the result collection unit; if not, it is considered that no interference has been detected in the current 15 minutes, and the current 15-minute data is not marked as any interference, and is divided into the unknown branch, and the judgment result Set to 0 and output to the result collection unit. Taking the narrowband interference in this implementation as an example, this narrowband interference will show obvious characteristics within 15 minutes, so it can be successfully detected under the 15-minute detection model.

Step 4: The data preprocessing unit performs merging and filtering for the data of the unknown branch generated in step 3 and the Pn' output in step 2. After the filtering time is 1 day, the filtering result character'_day (that is, the first filtering result in the above-mentioned embodiment) is obtained. ).

Step 5: The model selection unit calls the pre-created sky-level detection model (as the current time granularity detection model), the data character'_day to be detected enters the sky-level detection model, and outputs the interference detection probability P1', P2',...,P4 ' (that is, the second interference detection probability in the above embodiment).

Step 6: The preset sky-level threshold configured in the judgment unit is Thr _day , and it is judged whether the detection probability in step 5 exceeds the sky-level threshold Thr _day , if so, it is considered that the corresponding interference occurs, and the current day data is marked as its corresponding If there is no interference, it is considered that no interference has been detected in the current day, and the data of the current day is not marked as any interference, and is divided into the unknown branch, and the judgment result is set to 0 and output to the result collection. unit. Taking the full-bandwidth interference in this implementation as an example, this kind of interference will show obvious characteristics within a day, so it can be successfully detected under the sky-level detection model.

Step 7: The data preprocessing unit performs merging and filtering on the data of the unknown branch generated in Step 6 and the Pn' output in Step 5. After the filtering time is 1 week, the filtering result character'_week is obtained.

Step 8: The model selection unit calls the week-level detection model, the data character'_week to be detected enters the week-level detection model, and outputs the interference detection probabilities P1', P2',...,P4'.

Step 9: The preset weekly threshold configured in the judgment unit is Thr _week , and it is judged whether the detection probability in step 8 exceeds the weekly threshold Thr _week , if so, it is considered that the corresponding interference occurs, and the current weekly data is marked as Corresponding interference, the judgment result is set to 1 and output to the result collection unit; if not, it is considered that the current cycle-level detection model has not detected interference, and the current cycle-level data is not marked as any interference, divided into unknown branches, and the judgment result is set as 0 is output to the result collection unit. Taking the D4D5 interference in this implementation as an example, this kind of interference will show obvious characteristics within a week, so it can be successfully detected under the weekly detection model.

Step 10: The result collection unit provides the final interference type of the data to be detected according to the judgment result of the three-level model.

It should be noted that this embodiment adopts the current time granularity detection model (ie, the sky-level detection model) when the first time granularity detection model (ie, the 15-minute-level detection model) does not detect the interference type of the data to be detected. Continue to detect the interference type of the data to be detected. If the interference type is still not detected, switch the current time granularity detection model to a time granularity detection model with a coarser time granularity (ie, the weekly detection model), and return the data to be detected. Perform filtering with the combination of the first interference detection probability to obtain a new first filtering result, and continue to detect the interference type of the data to be detected until the interference type of the data to be detected is detected. Of course, in the actual operation process, when the interference type of the data to be detected is not detected by the weekly detection model, a time granularity detection model with coarser time granularity than that of the weekly detection model can be used to continue the interference detection of the data to be detected.

In the technical solution of this embodiment, a multi-level time granularity detection model is generated by adapting data of different time granularities in the detection model training stage. Then, the detection model of multi-level time granularity is used to adapt to the interference of different time granularities, which improves the ability of interference type identification, and further improves the accuracy of interference type identification.

In one embodiment, if the trained detection model and the method for extracting feature values from the training set have been obtained, the detection unit can be used independently. Table 5 is a schematic table of obtained interference features provided by the embodiment of the present application. As shown in Table 5, it is assumed that the feature extraction method of the training set is [time domain RSSI, frequency domain NI]. Table 6 is a schematic table of an interference type provided by this embodiment of the present application. As shown in Table 6, the types of interference include frame out-of-sync interference, narrowband interference, and full-band interference. Among them, the trained detection models are the time-slot-level detection model AI_slot, the 15-minute-level detection model AI_15m, and the hour-level detection model AI__hour.

Table 5 A schematic diagram of the interference characteristics that have been obtained

time-domain raw data	Frequency domain raw data
RSSI0,RSSI1,...RSSI13	NI0,NI1,...,NI272

Table 6 A schematic diagram of a type of interference

Interference number	Kind of interference
1	frame out-of-sync interference
2	narrowband interference
3	full-band interference

FIG. 9 is another schematic diagram of interference detection provided by an embodiment of the present application. The process of interference detection may be performed by the detection unit in the above embodiment. Exemplarily, the interference detection process is described by taking the first time granularity detection model as a time slot-level detection model, the current time granularity detection model as a 15-minute-level detection model, and an hour-level detection model as examples. As shown in Figure 9, the interference detection process includes the following steps:

Step 1: The data preprocessing unit extracts feature values consistent with the feature values extracted from the training set on the original data to obtain character'_slot (that is, the data to be detected in the above embodiment).

Step 2: The model selection unit calls the time-slot-level detection model AI_slot, the character'_slot of the data to be detected enters the time-slot-level model, and outputs the interference detection probabilities P1'_slot, P2'_slot, P3'_slot (that is, the first in the above-mentioned embodiment). Interference detection probability).

Step 3: The preset time-slot-level threshold in the judgment unit is Thr _slot , and it is judged whether the detection probability in step 2 exceeds the time-slot-level threshold Thr _slot , if so, it is considered that the corresponding interference occurs, and the current time slot data is marked as The corresponding interference, the judgment result is set to 1 and output to the result collection unit; if not, it is considered that no interference is detected in the current time slot, and the current time slot data is not marked as any interference, divided into the unknown branch, and the judgment result is set to 0 Output to the result collection unit. Taking the frame out-of-synchronization interference in this implementation as an example, this kind of interference will show obvious characteristics in one time slot, so it can be successfully detected under the time slot level model.

Step 4: The data preprocessing unit merges the data of the unknown branch generated in step 3 and the Pn'_slot output in step 2, and performs filtering. a filter result).

Step 5: The model selection unit calls the 15-minute level detection model AI_15m (as the current time granularity detection model), the data character'_15m to be detected enters the 15-minute level detection model, and outputs the interference detection probability P1'_15m, P2'_15m, P3'_15m (ie the second interference detection probability in the above embodiment).

Step 6: The preset 15-minute level threshold in the judgment unit is Thr _15m , and it is judged whether the detection probability in step 5 exceeds the 15-minute level threshold Thr _15m , if so, it is considered that the corresponding interference occurs, and the current 15-minute data is marked as For the corresponding interference, the judgment result is set to 1 and output to the result collection unit; if not, it is considered that no interference has been detected in the current 15 minutes, and the current 15-minute data is not marked as any interference, divided into the unknown branch, and the judgment result is set to 0 Output to the result collection unit. Taking the narrowband interference in this implementation as an example, this narrowband interference will show obvious characteristics within 15 minutes. Although it cannot be detected at the time slot level, it can be successfully detected under the 15 minute detection model.

Step 7: The data preprocessing unit combines the data of the unknown branch generated in step 6 with the Pn'_15m output in step 5, and performs filtering respectively. After the filtering time is 1 hour, the filtering result character'_hour is obtained.

Step 8: The model selection unit calls the hour-level detection model AI_hour, the character'_hour of the data to be detected enters the hour-level detection model, and outputs the interference detection probability P1′_hour, P2′_hour, P3′_hour.

Step 9: The preset hour-level threshold in the judgment unit is Thr _hour , determine whether the detection probability in step 8 exceeds the hour-level threshold Thr _hour , if so, it is considered that the corresponding interference occurs, and the current hour-level data is marked as its corresponding If there is no interference, it is considered that no interference has been detected in the current hour, and the current hour-level data is not marked as any interference, divided into the unknown branch, and the judgment result is set to 0 and output to the result collection unit. Taking the full-band interference in this implementation as an example, this kind of interference will show obvious characteristics within an hour. Although it cannot be detected at the time slot level and 15 minutes level, it can be successfully detected under the hour-level detection model. .

Step 10: The result collection unit provides the final interference type of the data to be detected according to the judgment result of the three-level model.

It should be noted that this embodiment adopts the current time granularity detection model (ie, the 15-minute level detection model) when the first time granularity detection model (ie, the slot-level detection model) does not detect the interference type of the data to be detected. Continue to detect the interference type of the data to be detected. If the interference type is still not detected, switch the current time granularity detection model to a time granularity detection model with a coarser time granularity (ie, an hour-level detection model), and return the data to be detected. Perform filtering with the combination of the first interference detection probability to obtain a new first filtering result, and continue to detect the interference type of the data to be detected until the interference type of the data to be detected is detected. Of course, in the actual operation process, when the hour-level detection model fails to detect the interference type of the data to be detected, a time granularity detection model with coarser time granularity than the hour-level detection model can be used to continue to perform interference detection on the to-be-detected data.

The technical solution of this embodiment utilizes a multi-level time granularity detection model to adapt to interference of different time granularities, thereby improving the capability of identifying the type of interference, and further improving the accuracy of identifying the type of interference.

In the above embodiment, the multi-level time granularity detection model is used to adapt to the interference of different time granularities, and the ability to identify the type of interference is greatly improved. When the interference exhibits short-term time granularity features, such as slot-level features, symbol-level features, etc., the detection model can successfully detect the interference at the early stage of detection and give the detection result; when the interference exhibits medium- and long-term time granularity features, such as Minute-level features, hour-level features, sky-level features, etc., the detection model can also successfully detect interference in the middle and late stages of detection, and give the detection results. When the data to be detected only contains features of one dimension, the multi-level time granularity detection model can also perform interference detection, so that the interference can be effectively avoided and eliminated.

In an embodiment, FIG. 10 is a structural block diagram of another interference detection apparatus provided by an embodiment of the present application. This embodiment is applied to an interference detection device. As shown in FIG. 10 , this embodiment includes: a first filter 310 , a first determination module 320 and a second determination module 330 .

The first filter 310 is configured to filter the combination of the data to be detected and the predetermined first interference detection probability when the interference type of the data to be detected is not detected by using the pre-created first time granularity detection model , obtain the first filtering result;

The first determination module 320 is configured to determine the interference detection probability of the first filtering result in the pre-created current time granularity detection model, as the second interference detection probability, wherein the current time granularity detection model is larger than the first time granularity detection model. detection model with coarser time granularity;

The second determination module 330 is configured to determine the interference type of the data to be detected according to the comparison result between the second interference detection probability and the first preset detection probability threshold.

In an embodiment, when the second interference detection probability is less than the first preset detection probability threshold, the method further includes:

Switching the current time granularity detection model to a time granularity detection model with a coarser time granularity, and returning to the combination of the data to be detected and the predetermined first interference detection probability to filter to obtain the filtering result, until the detection of the data to be detected. Type of interference.

In an embodiment, the combination of the data to be detected and the predetermined first interference detection probability is filtered to obtain a first filtering result, including:

combining the data to be detected and the predetermined first interference detection probability to obtain combined data;

The combined data is filtered by the time length of the current time granularity to obtain a first filtering result.

In one embodiment, the combined data is filtered by the time length of the current time granularity to obtain a first filtering result, including:

The combined data is filtered according to the instantaneous combined data of the current time granularity, the filtered combined data of the previous current time granularity, the first preset weight coefficient and the second preset weight coefficient to obtain a first filtering result.

In one embodiment, the interference detection device further includes:

The first creation module is configured to input the pre-determined first time granularity feature value into the pre-created first AI training model before the interference type of the data to be detected is not detected by using the pre-created first time granularity detection model , until the detection rate reaches the first preset detection rate threshold, output the first AI training model and the first time granularity interference probability distribution value, and use the first AI training model as the first time granularity detection model;

a first combiner, configured to combine the first time granularity feature value and the first time granularity interference probability distribution value as the second time granularity feature value;

a second filter, configured to perform time-domain filtering on the second time granularity feature value to obtain the second time granularity level training data;

The second creation module is configured to input the training data at the second time granularity level into the pre-created second AI training model until the detection rate reaches the second preset detection rate threshold, and output the second AI training model and the second time granularity Interfering with the probability distribution value, and using the second AI training model as the second time granularity detection model; wherein, the time granularity corresponding to the first time granularity detection model and the second time granularity detection model becomes coarser in turn.

In one embodiment, the interference detection device further includes:

The second combiner is configured to combine the second time granularity level training data and the second time granularity interference probability distribution value as the third time granularity feature value after using the second AI training model as the second time granularity detection model ;

a third filter, configured to perform time domain filtering on the third time granularity feature value to obtain training data at the third time granularity level;

The third creation module is configured to input the training data at the third time granularity level into the pre-created third AI training model until the detection rate reaches the third preset detection rate threshold, and output the third AI training model and the third time granularity Interfering with the probability distribution value, and using the third AI training model as the third time granularity detection model; wherein, the time granularity corresponding to the first time granularity detection model, the second time granularity detection model and the third time granularity detection model becomes thicker in turn .

In an embodiment, the first time granularity feature value includes one of the following: time domain received signal strength indicator RSSI; frequency domain noise indicator NI; spatial domain matched filtering; time domain correlation value; frequency domain correlation value; intermediate resource block RB energy distribution.

In one embodiment, both the first time granularity detection model and the current time granularity detection model include one of the following: a time slot-level detection model; a minute-level detection model; a day-level detection model; and a week-level detection model.

The interference detection apparatus provided in this embodiment is set to implement the interference detection method of the embodiment shown in FIG. 1 . The implementation principle and technical effect of the interference detection apparatus provided in this embodiment are similar, and details are not described herein again.

FIG. 11 is a schematic structural diagram of an interference detection device provided by an embodiment of the present application. As shown in FIG. 10 , the device provided by this application includes: a processor 410 , a memory 420 and a communication module 430 . The number of processors 410 in the device may be one or more, and one processor 410 is taken as an example in FIG. 10 . The number of memories 420 in the device may be one or more, and one memory 420 is taken as an example in FIG. 10 . The processor 410 , the memory 420 and the communication module 430 of the device may be connected by a bus or in other ways, and the connection by a bus is taken as an example in FIG. 10 . In this embodiment, the device may be a terminal side (eg, user equipment).

As a computer-readable storage medium, the memory 420 may be configured to store software programs, computer-executable programs, and modules, such as program instructions/modules corresponding to the device in any embodiment of the present application (for example, the first filter in the interference detection apparatus). 310, the first determination module 320 and the second determination module 330). The memory 420 may include a storage program area and a storage data area, wherein the storage program area may store an operating system, an application program required for at least one function; the storage data area may store data created according to the use of the device, and the like. Additionally, memory 420 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid-state storage device. In some examples, memory 420 may include memory located remotely from processor 410, which may be connected to the device through a network. Examples of such networks include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

The communication module 430 is configured to perform communication interaction among various communication nodes.

The interference detection device provided above may be configured to execute the interference detection method provided by any of the above embodiments, and has corresponding functions and effects.

Embodiments of the present application further provide a storage medium containing computer-executable instructions, where the computer-executable instructions are used to execute an interference detection method when executed by a computer processor, the method comprising: using a pre-created first time granularity In the case where the detection model does not detect the interference type of the data to be detected, the combination of the data to be detected and the predetermined first interference detection probability is filtered to obtain a first filtering result; it is determined that the first filtering result is within the predetermined range. The interference detection probability in the created current time granularity detection model is used as the second interference detection probability, wherein the current time granularity detection model is a detection model with a coarser time granularity than the first time granularity detection model; The comparison result between the detection probability and the first preset detection probability threshold value determines the interference type of the data to be detected.

In the technical solution of the embodiment of the present application, when the interference type of the data to be detected is not detected by using the pre-created first time granularity detection model, the interference detection is performed on the data to be detected in the pre-created current time granularity detection model, Until the interference type of the data to be detected is detected, the multi-level time granularity detection model is used to detect the interference type of the data to be detected, so as to adapt to the interference of different time granularity features, improve the ability to identify the interference type, and improve the performance of the complex network. Interference recognition accuracy in the scene.

As will be understood by those skilled in the art, the term user equipment encompasses any suitable type of wireless user equipment such as a mobile telephone, portable data processing device, portable web browser or vehicle mounted mobile station.

In general, the various embodiments of the present application may be implemented in hardware or special purpose circuits, software, logic, or any combination thereof. For example, some aspects may be implemented in hardware, while other aspects may be implemented in firmware or software that may be executed by a controller, microprocessor or other computing device, although the application is not limited thereto.

Embodiments of the present application may be implemented by the execution of computer program instructions by a data processor of a mobile device, eg in a processor entity, or by hardware, or by a combination of software and hardware. Computer program instructions may be assembly instructions, Instruction Set Architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or written in any combination of one or more programming languages source or object code.

The block diagrams of any logic flow in the figures of this application may represent program steps, or may represent interconnected logic circuits, modules and functions, or may represent a combination of program steps and logic circuits, modules and functions. Computer programs can be stored on memory. The memory may be of any type suitable for the local technical environment and may be implemented using any suitable data storage technology, such as, but not limited to, Read-Only Memory (ROM), Random Access Memory (RAM), optical Memory devices and systems (Digital Video Disc (DVD) or Compact Disk (CD)), etc. Computer-readable media may include non-transitory storage media. The data processor may be of any type suitable for the local technical environment, such as, but not limited to, a general purpose computer, a special purpose computer, a microprocessor, a Digital Signal Processing (DSP), an Application Specific Integrated Circuit (ASIC) ), programmable logic devices (Field-Programmable Gate Array, FGPA) and processors based on multi-core processor architecture.

The above descriptions are only several embodiments of the present application, and are not intended to limit the present application. For those skilled in the art, the present application may have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of this application shall be included within the protection scope of this application.

Claims (11)

1. A method of interference detection, comprising:

   In the case that the interference type of the data to be detected is not detected by using the pre-created first time granularity detection model, filtering the combination of the data to be detected and the predetermined first interference detection probability to obtain a first filtering result;

   Determine the interference detection probability of the first filtering result in the pre-created current time granularity detection model as the second interference detection probability, wherein the current time granularity detection model is a time longer than the first time granularity detection model A detection model with coarser granularity;

   The interference type of the data to be detected is determined according to the comparison result between the second interference detection probability and the first preset detection probability threshold.
2. The method according to claim 1, wherein, in the case that the second interference detection probability is less than the first preset detection probability threshold value, further comprising:

   Switching the current time granularity detection model to a time granularity detection model with a coarser time granularity, and returning to the step of filtering the combination of the data to be detected and the predetermined first interference detection probability to obtain a first filtering result, Until the interference type of the data to be detected is detected.
3. The method according to claim 1, wherein the filtering the combination of the data to be detected and the predetermined first interference detection probability to obtain the first filtering result comprises:

   combining the data to be detected and the predetermined first interference detection probability to obtain combined data;

   The combined data is filtered by the time length of the current time granularity to obtain a first filtering result.
4. The method according to claim 3, wherein the filtering of the time length of the current time granularity on the combined data to obtain the first filtering result comprises:

   The combined data is filtered according to the instantaneous combined data of the current time granularity, the filtered combined data of the previous current time granularity, the first preset weight coefficient and the second preset weight coefficient to obtain a first filtering result.
5. The method according to claim 1, wherein before the interference type of the data to be detected is not detected by using the pre-created first time granularity detection model, the method further comprises:

   inputting the predetermined first time granularity feature value into the pre-created first AI training model, until the detection rate reaches the first preset detection rate threshold, and outputting the first AI training model and the first time granularity interference probability distribution value, and using the first AI training model as the first time granularity detection model;

   combining the first time granularity feature value and the first time granularity interference probability distribution value as a second time granularity feature value;

   performing time domain filtering on the second time granularity feature value to obtain second time granularity level training data;

   inputting the second time granularity level training data into the pre-created second AI training model, until the detection rate reaches the second preset detection rate threshold, and outputting the second AI training model and the second time granularity interference probability distribution value, The second AI training model is used as a second time granularity detection model; wherein, the time granularities corresponding to the first time granularity detection model and the second time granularity detection model become thicker in sequence.
6. The method of claim 5, further comprising:

   combining the second time granularity level training data and the second time granularity interference probability distribution value as a third time granularity feature value;

   performing time domain filtering on the third time granularity feature value to obtain third time granularity level training data;

   inputting the third time granularity level training data into the pre-created third AI training model, until the detection rate reaches the third preset detection rate threshold, and outputting the third AI training model and the third time granularity interference probability distribution value, and use the third AI training model as a third time granularity detection model; wherein, the time granularity corresponding to the first time granularity detection model, the second time granularity detection model and the third time granularity detection model thicken sequentially.
7. The method according to claim 5 or 6, wherein the first time granularity feature value comprises one of the following: time domain received signal strength indication RSSI; frequency domain noise indication NI; spatial domain matched filtering; time domain correlation value; Frequency domain correlation value; middle resource block RB energy distribution.
8. The method according to any one of claims 1-4, wherein the first time granularity detection model and the current time granularity detection model both include one of the following: a time slot-level detection model; a minute-level detection model; level detection model; week level detection model.
9. An interference detection device, comprising:

   a first filter, configured to filter the combination of the data to be detected and the predetermined first interference detection probability when the interference type of the data to be detected is not detected by using the pre-created first time granularity detection model , obtain the first filtering result;

   The first determination module is configured to determine the interference detection probability of the first filtering result in the pre-created current time granularity detection model as the second interference detection probability, wherein the current time granularity detection model is larger than the first time granularity detection model. A detection model with coarser time granularity of a time granularity detection model;

   The second determination module is configured to determine the interference type of the data to be detected according to the comparison result between the second interference detection probability and the first preset detection probability threshold value.
10. An interference detection device, comprising:

    a communication module, configured to perform communication interaction between each communication node;

    memory, configured to store one or more programs; and

    one or more processors;

    Wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of the above claims 1-8.
11. A storage medium, wherein the storage medium stores a computer program, and the computer program implements the method according to any one of the above claims 1-8 when the computer program is executed by a processor.

Images