JP7351082B2 - Interference source prediction model generation method, interference source recognition method, device, and terminal device - Google Patents

Description

The present invention relates to the field of communication technology, and relates to an interference source prediction model generation method, an interference source recognition method, an apparatus, and a terminal device.

IoT (Internet of Things) is widely used in various fields. Entities in IoT typically include sensors, gateways, networks, clouds, and applications.

With advances in technology and increased reliance on short-range wireless networks such as WIFI, Zigbee, and Bluetooth, users need reliable, high-performance, and scalable coverage from their wireless networks. be.

On the other hand, traditional sensor network deployments fail to provide sufficient coverage and predictable performance. Interference is one of the causes of performance degradation. Unlike wired networks, wireless links are susceptible to environmental changes and surrounding wireless activity.

Additionally, protocols associated with wireless networks, such as IEEE 802.11, 802.15.4, and 802.15.1, operate in unlicensed frequency bands. The interference problem can become more serious because multiple wireless networks can interfere with each other. Moreover, the rapid increase in the number of users in the unlicensed spectrum can also seriously exacerbate this.

The inventor discovered the following. That is, the interference is unpredictable and arises from mobile users, modules serving other unlicensed frequency bands, and traffic changes. Therefore, by being able to diagnose different interference sources, IoT service suppliers can propose countermeasures to resolve interference problems or avoid potential problems.

In order to solve at least one of the above-mentioned problems, the present invention provides an interference source prediction model generation method, an interference source recognition method, an apparatus, and a terminal device.

According to a first aspect of an embodiment of the present invention, a method for generating an interference source prediction model is formed, the method comprising:
performing energy detection sampling on the test window at a predetermined sampling rate for each test window of the plurality of test windows under a plurality of preset interference scenarios to obtain a sample data set within each test window;
Each test is based on the test period to which each test window belongs, and/or the timestamp of the samples in the sample data set within each test window, and/or the data probed in each test window by a wireless network probe. labeling the windows; and training a deep neural network to obtain parameters of the interference source prediction model based on the sample data set of each test window and the label corresponding to each test window. include.

According to a second aspect of an embodiment of the present invention, there is provided an interference source recognition method, the method comprising:
forming an interference source prediction model using a sample data set corresponding to each test window of the plurality of test windows and a label corresponding to the test window;
performing energy detection sampling on the received signal strength within a detection window at a predetermined sampling rate to obtain a sample data set corresponding to the detection window; and inputting the sample data set to the interference source prediction model; The method includes obtaining an interferer type corresponding to the window.

According to a third aspect of an embodiment of the present invention, there is provided an interference source prediction model generation device, the device comprising:
Under a plurality of preset interference scenarios, for each test window of the plurality of test windows, energy detection sampling is performed on the test window at a predetermined sampling rate to obtain a sample data set within each test window. one sampling unit;
Each test is based on the test period to which each test window belongs, and/or the timestamp of the samples in the sample data set within each test window, and/or the data probed in each test window by a wireless network probe. a labeling unit for labeling windows; and training a deep neural network based on the sample data set of each test window and the label corresponding to each test window to obtain parameters of the interference source prediction model. including training units.

According to a fourth aspect of an embodiment of the present invention, there is provided an interference source recognition device, wherein the device comprises:
a generation unit that forms an interference source prediction model using a sample data set corresponding to each test window of the plurality of test windows and a label corresponding to the test window;
a second sampling unit that performs energy detection sampling on the received signal strength within a detection window at a predetermined sampling rate to obtain a sample data set corresponding to the detection window; and inputting the sample data set to the interference source prediction model. and includes a determining unit for obtaining an interference source type corresponding to the detection window.

According to a fifth aspect of an embodiment of the present invention, a terminal device is provided, wherein the terminal device includes the above-mentioned device.

The beneficial effects of the present invention are as follows: using a high rate Received Signal Strength Indication (RSSI) sampler to obtain real-time RSSI samples in a wireless network environment; By performing interference analysis using this method, it is possible to determine the source of interference and provide IoT service suppliers with a reference basis for resolving interference problems.

FIG. 3 is a diagram illustrating a deployment scenario in an embodiment of the invention. 3 is a diagram showing an interference source prediction model generation method in Example 1. FIG. FIG. 2 is a diagram illustrating the basic principle of deep learning using an RSSI sampling value sequence as input. FIG. 3 is a diagram illustrating an implementation scheme for marking test windows; FIG. 2 is a diagram showing a network structure of a convolutional neural network. FIG. 7 is a diagram showing an interference source recognition method in Example 2. FIG. 7 is a diagram showing an interference source prediction model generation device in Example 3. FIG. 3 illustrates one implementation of a tagging unit. FIG. 6 shows another implementation of the labeling unit; FIG. 6 shows another implementation of the labeling unit; FIG. 7 is a diagram showing an interference source recognition device in Example 4. FIG. 7 is a diagram showing a terminal device in Example 5.

Hereinafter, preferred embodiments for carrying out the present invention will be described in detail with reference to the accompanying drawings.

In wireless networks, one parameter that reflects the conditions of the wireless channel is the received signal strength indication (RSSI). RSSI is the average received signal power obtained by making energy detection measurements over multiple symbol times; it is a measurement of the radio frequency (RF) power level at the input of a transceiver. be. Since RF energy increases when interference is present, RSSI can be used to detect the occurrence of interference.

In an embodiment of the present invention, it is assumed that the wireless network system is a system based on IEEE 802.15.4 technology, such as Zigbee, 6lowpan, etc. The sampler may use 802.15.4 technology, which means performing RSSI sampling on one of the 802.15.4 channels. Under that scenario, external interference sources may include WIFI interference, Bluetooth interference, microwave interference, etc. Also, for WIFI interference, the interference analysis algorithm may give the traffic taking place, e.g. video streams, file transmissions, WIFI beacons, etc.

In embodiments of the invention, the sampling concept and interference source definition can be extended to other wireless communication systems. For example, in a Bluetooth system, a Bluetooth sampler performs RSSI sampling on the Bluetooth channel. Under that scenario, the interference sources may include 802.15.4 interference, microwave interference, WIFI video streams, WIFI file transmissions, etc.

Embodiments of the present invention do not limit the wireless network system and interference source type. For example, methods of embodiments of the present invention can be applied to IoT, sensor networks, wireless LANs (WLANs), and other wireless networks.

In embodiments of the present invention, interference analysis policy decisions can be completed within a very short time (on the order of tens of milliseconds) and can reflect data packet level interference diagnosis. Specifically, conventional interference detection or source recognition time is quite long, typically taking seconds or minutes. Within such a long period of time, interference may occur only within a certain period of time, or multiple different sources of interference may occur. For example, for interference caused by wireless communication systems, the interference is due to data packet transmission, and the transmission time of data packets is on the order of milliseconds, so within a very long time, e.g. within a time window on the order of seconds. Since the time when the channel is used for data packet transmission has only a very small proportion, it cannot be said that there is interference due to data packet transmission within that time. In fact, in such a case, the channel is unaffected by data packet transmission for most of the time.

In embodiments of the present invention, diagnostic results (interference source types) are reported with finer granularity. A terminal user, for example a network administrator or a network management application, can generate one detailed report for each diagnostic window or can generate statistical results reports for relatively long periodic window sizes.

In embodiments of the present invention, the size of the window of the diagnosis period needs to be selected in order to report diagnostic results with finer granularity. In embodiments of the invention, it is referred to as a detection window. Since the data packet duration of different wireless network systems varies from less than one millisecond to several milliseconds, in order to characterize the different traffic modes, embodiments of the present invention set the size of the detection window to Set it slightly larger than the normal data packet transmission duration. Also, in many countries, 20 ms (milliseconds) is the transmission period of a microwave oven (power supply is 50 Hz), so in the embodiment of the present invention, 20 ms is set as the detection window. This allows the detection window to take into account the wireless network system and the microwave oven at the same time. Note that the embodiments of the present invention are not limited to this, and in other embodiments, the detection window may be set to other values.

FIG. 1 is a diagram illustrating one deployment scenario in an embodiment of the invention. As shown in FIG. 1, a microwave oven (MWO) 101, a wireless LAN AP (WiFi) 102, and an RSSI sampler 103 are deployed in an office environment. Note that the deployment scenario in Figure 1 is only an example; similar deployments may be used in a field area network (FAN), such as a factory, or a home area network (HAN). can. Furthermore, one RSSI sampler 103 may be deployed, a plurality of RSSI samplers 103 may be deployed in a plurality of different areas, or a plurality of RSSI samplers 103 may be deployed in the same area. can be obtained.

Additionally, the RSSI sampler 103 may be one independent device that is responsible only for RSSI sampling and also supports an interference analysis engine, such as a wireless LAN gateway or another device that supports an interference analysis engine. It is connected to a device, such as a computer, notebook, or Raspberry Pi, and the sampling results can be sent to these devices for interference analysis. The RSSI sampler 103 may also be an RSSI sampling module located in a network node, such as a sensor node or a gateway, so that there is no dedicated node for sampling and the sampling function is It can be activated during the Interference Analysis Preparation Stairs, Analysis Stairs, or All-Weather Monitoring Period.

Here, the network node can support an interference analysis engine, which allows it to complete not only the RSSI sampling function but also the interference analysis function. Note that the present embodiment is not limited to this, and the network node may not support the interference analysis engine and may send the sampling result of the RSSI sampling module to a device that supports the interference analysis engine. Note that the above-mentioned interference analysis engine refers to one based on the interference source recognition method in the embodiment of the present invention.

In embodiments of the invention, the sampling rate is selected to obtain sufficient RSSI samples within a diagnostic period. In embodiments of the present invention, a sampling rate of once every 100 μs (microseconds) may be used, such as a sampling frequency of 10 kHz, and a sampling rate of once every 95 μs (sampling frequency of 10.5 μs). kHz). The RSSI sampler captures the energy on the channel due to interfering transmissions, e.g. WIFI beacons or Bluetooth data packets, and it obtains RSSI samples by sequentially accessing the RSSI register of the wireless chip (chip) of the sensor node. be able to.

In an embodiment of the invention, first, interference detection is performed under a preset interference scenario to generate an interference source prediction model, and then RSSI sampling is performed in an actual scenario (for example, the scenario shown in FIG. 1). The type of interference source can be obtained by performing the following steps, acquiring RSSI samples, and inputting the RSSI samples acquired in the actual scenario into the interference source prediction model.

Embodiments of the present invention provide a method for generating an interference source prediction model. FIG. 2 is a diagram illustrating one implementation of the method. As shown in FIG. 2, the method includes the following steps.

Step 201: Under a plurality of preset interference scenarios, for each test window of the plurality of test windows, perform energy detection sampling on the received signal strength within the test window at a predetermined sampling rate; Obtain a sample data set of;
Step 202: The test period to which each test window belongs and/or the timestamp of the sample in the sample data set within each test window and/or the time stamp of the sample detected in each test window by a wireless network sniffer. label each test window based on the data provided;
Step 203: Based on the sample data set of each test window and the indicators corresponding to each test window, train a deep neural network to obtain the parameters of the interference source prediction model.

In this embodiment, an interference source prediction model can be generated using a deep learning method. First, in step 201, training data can be obtained.

In step 201, multiple interference scenarios are preset to obtain training data, and RSSI sampling is performed under the multiple interference scenarios (i.e., energy detection sampling for received signal strength within a test window). ), it is possible to obtain the necessary training data (sample data set within each detection window).

In step 201, the size of the test window is the same as the size of the detection window described above, for example, both are set to 20ms. Note that this embodiment is not limited to this, and the test window may be set to other values depending on other purposes. Further, in step 201, the sampling rate may be set in advance so that sufficient RSSI samples can be acquired within the diagnostic cycle (test window), and may be set, for example, once every approximately 100 μs. In this embodiment, explanation will be given using an example of once every 95 μs. However, as described above, the present embodiment is not limited to this, and depending on needs, the sampling rate may be set to other values in advance, such as a sampling rate of once every 200 μs. It may be higher, ie the sampling period is less than 200 μs. Further, this embodiment is not limited to a specific sampling method and sampling process, and can specifically refer to conventional techniques.

Next, in step 202, a label is attached to each test window based on the detection result of step 201, and a label (label) corresponding to each test window is obtained. The signs or labels here represent the interference source type that corresponds to each test window.

In one implementation, a test window may be labeled based on the test period to which the test window belongs. For example, it is determined whether the test period to which the test window belongs is the working period of a certain interference source (for convenience of explanation, it will be referred to as the first interference source), and if it is determined "yes", the test window A label indicating that interference has been received from the first interference source is attached to the first interference source. If it is determined "no", the first interference source is excluded in the test window, ie, there is no interference from the first interference source in the test window.

In this implementation method, the first interference source is, for example, a microwave oven, and in this implementation method, a mark is attached to a test window due to microwave oven interference. The microwave interference is preset, e.g. artificially generated, and the microwave transmission period (i.e. the alternating current period) is the same as the period of the test window, i.e. Since both are 20 ms, all sample data sets of the working period of the microwave oven can be labeled as "MWO", meaning that they have been interfered with by the microwave oven. However, this embodiment is not limited to this, and the first interference source may be another interference source.

Another implementation is to determine, for a given test window, the test period to which the test window belongs, the timestamps of the samples in the sample data set within the test window, and the data detected by the wireless network probe. You may also add a label to the test window. For example, it is determined whether the test period to which the test window belongs is the working period of a certain interference source (referred to as a second interference source for convenience of explanation), and When it is determined whether the timestamp of at least one sample is within the time from the start to the end of air propagation of the frame belonging to the specified address pair detected by the wireless network detector, and it is determined to be "yes"; , a mark is placed on the test window indicating that it has received interference from the second interference source. If it is determined "no", the second interference source is excluded in the test window, ie, there is no interference from the second interference source in the test window.

In this implementation method, the designated address pair refers to the MAC address pair of two nodes where data transmission has occurred. Specifically, it consists of one transmitting STA address (TA) and one receiving address (RA). Since different types of MAC frames may occur in the data transmission process, TA and RA may be exchanged, or there may be only one. Further, the frames belonging to the specified address pair are, for example, data frames, control frames, etc., and this embodiment is not limited to this.

In this embodiment, the second interference source is, for example, data transmission, and in this embodiment, a test window due to data transmission interference is marked. Since the data transmission interference is preset, for example, artificially generated, and the data of the transmission belongs to the specified address pair, in this implementation method, the test period to which the test window belongs and , In addition to comparing the data transmission period, wireless network probes are also used. The test period to which the test window belongs is a data transmission period, and the timestamp of at least one sample in the sample data set within the test window is of a frame belonging to the specified address pair detected by the wireless network probe. Within the time period between start and end, the test window is marked with an indication that there is interference due to data transmission.

In this implementation method, the wireless network detector is, for example, a WIFI detector, a Bluetooth detector, a ZigBee detector, etc. If the wireless network is different, the type of detector is different, so in this embodiment, any one One or more probes can be used to probe the data of the transmission.

In this embodiment, the data is, for example, a large amount of data such as video data and text data, but the present embodiment is not limited to this. Further, in this embodiment, the type of the second interference source is not limited, and may be WIFI data transmission, Bluetooth data transmission, or other types of data transmission. In addition, the second interference source may be further divided into multiple types based on traffic types, such as different high-volume traffic under WIFI scenario, such as video streams, file downloads, etc.

Other implementations include marking a test window based on time stamps of samples in a sample data set within the test window and data detected by a wireless network probe. It's okay. For example, determine whether the timestamp of at least one sample in the sample data set within the test window is within the time from the start to the end of the beacon frame detected by the wireless network detector, and select "Yes". If it is determined, the test window is marked with an indication that it has received interference from the above-mentioned beacon frame. If the answer is "no", the test window excludes beacon frame interference.

In the present implementation method, the beacon frame is, for example, a Bluetooth beacon frame, a WIFI beacon frame, etc., and the present implementation method also uses a wireless network detector, and the time of the sample in the sample data set within the test window. Compare the stamp with the time from the start to the end of a certain beacon frame detected by the wireless network detector, and if the timestamp of at least one sample is within the time from the start to the end of the beacon frame. Next, it is assumed that the interference received during the test period to which the test window belongs is due to the beacon frame, and in this case, a mark is attached to the test window to indicate that the test window has received interference from the beacon frame.

Similarly, in this implementation method, the wireless network detector is, for example, a WIFI detector, a Bluetooth detector, a ZigBee detector, etc. Also, depending on the wireless network, the type of detector also differs, so this implementation method In examples, any one or more probes may be used to probe the data of the transmission.

In this embodiment, the above-mentioned implementation methods can be used in combination, and the priority level of the label is set based on the difference in interference type, for example, the first interference source is labeled first. , then tag the second interferer, then tag the beacon frame (which could be a third interferer), and if the current test window is set to any one of the above. If none of the cases apply, the current test window can be marked with a meaning such as normal.

Furthermore, this embodiment is not limited to the several implementation methods described above; depending on the type of interference source, one or more of the implementation methods described above may be omitted, or different types of interference sources may be supported. Alternatively, each type of interference source may be subdivided and then labeled. Each test is based on the test period to which each test window belongs, and/or the timestamp of the samples in the sample data set within each test window, and/or the data probed in each test window by a wireless network probe. All methods of marking windows fall within the scope of the present invention.

Next, in step 203, a deep By training the neural network, parameters of the deep neural network, such as weights and deviations of the neural network, can be obtained as parameters of the interference source prediction model. That is, a classifier can be generated. Then, by using the classifier, the interference source can be recognized.

In one implementation, all samples in the sample data set in all test windows are input to a convolutional neural network, labels corresponding to all test windows are outputs of the convolutional neural network, and the convolutional neural network is By performing training based on the neural network architecture, parameters of each term of the convolutional neural network can be obtained as parameters of the interference source prediction model. In this embodiment, the specific processing method is not limited, and prior art can be referred to for details.

In order to make the method of this example more clear, the method of this example will be explained below using one specific example.

FIG. 3 is a diagram showing the basic principle of deep learning using an RSSI sampling value sequence as input. As shown in Figure 3, suppose the detection window is 20ms and all RSSI samples within the detection window form an N-element input vector. If we use a sampling rate of once every 95μs, N=20ms/95μs=210. The N-element input vector is passed through one deep learning classification model to generate an M×1 output vector. M represents the number of defined classifications, such as normal, WIFI beacon, WIFI video stream, WIFI file download, Bluetooth interference, etc. Unlike the conventional method of performing interference analysis using RSSI sampling values, in the method according to the embodiment of the present invention, the input is the original RSSI sampling value and there is no need to perform feature extraction.

In this embodiment, in order to obtain training data, first, an interference scenario is manually generated, for example, WIFI traffic is intentionally introduced into the environment, and RSSI sampled by a sampler is collected. For the microwave interference scenario, since the microwave transmission period is the same as the test window period, all sample data sets of the microwave working period can be labeled as "MWO". In order to give an accurate label to each sample data set in the wireless data transmission process, the present embodiment further uses a wireless network detector, such as a WIFI sniffer or a Bluetooth sniffer.

For example, connect or incorporate the sampler and probe into the same device (eg, Raspberry Pi) to avoid synchronization problems. When the sampler samples one RSSI, it also saves the timestamp of that RSSI. When the detector detects one data packet (frame), it also records the start time and duration of the data packet (frame). Using the detector log and sampler timestamp, a label for the sample data set can be determined.

Also, detailed labeling rules can reflect a certain priority level. One example is shown below.

If the test window belongs to the working period of the microwave oven, the indicator is “MWO”;
If not, the test window belongs to a transmission period of a video stream or file, and the timestamp of at least one sample in the test window is from the beginning to the end of a frame belonging to the specified address pair, which is detected by the detector. if it is within the time period, mark the test window with a label such as “Video Stream or File Download”;
Otherwise, if the timestamp of at least one sample in the test window is within the time from the start to the end of the detected WIFI beacon frame, mark the test window with a label such as "WIFI Beacon". Attach;
Otherwise, if the timestamp of at least one sample in the test window is within the time from the start to the end of the probed Bluetooth frame, label the window as "Bluetooth";
Otherwise, the indicator is "normal".

Although microwave oven interference, video stream or file transmission interference, WIFI interference, and Bluetooth interference have been explained above as examples, this embodiment is not limited to these, and based on the above-mentioned contents, several It is also possible to increase or decrease the labeling rules for the interferer types.

FIG. 4 describes a WIFI detector and RSSI sample as an example. As shown in Figure 4, the beacon duration and data packet duration are determined based on the WIFI detector log, and for each RSSI sample based on the RSSI sample log and the WIFI detector log. Perform labeling, obtain labeled sample logs, and label each test window based on the labeled sample logs in the test window. As shown in Figure 4, one test window (20ms) includes a data packet duration, that is, the timestamp of at least one RSSI sample within the test window is detected by the WIFI detector. The test window is marked as data transmission within the time period between the start and end of the detected frame belonging to the specified address pair.

The process of finding the parameters of a deep convolutional neural network (CNN or ConvNet) using the acquired training data (RSSI samples and corresponding labels in each test window) is a process of deep learning.

FIG. 5 is a diagram showing the network structure of CNN in this example. This example uses CNN because CNN is suitable for matrix data that is correlated with neighboring data, which can reduce the number of parameters for the deep network. Furthermore, although conventional CNN for image recognition uses two-dimensional input, in this embodiment, only one-dimensional vectors are used as input. In this example, a CNN with two convolutional layers is used, the input is a vector of 210 × 1 elements, and the output is 5 × 1, that is, if the interference sources are divided into 5 types, then The network structure shown in Figure 5 allows existing training methods, such as stochastic gradient descent, to be used to obtain the final convolutional neural network parameters. The final trained neural network is an interference source prediction model.

Note that although only each step or process related to the present invention has been described above, the present invention is not limited thereto. The method may further include other steps or processes or a similar CNN network structure, and prior art can be referred to for the specific contents of these steps or processes.

As can be seen from the above embodiment, in this embodiment, a deep learning method is used to form an interference source prediction model, and the forming process can be completed offline. Therefore, by using a high-rate received signal strength indication (RSSI) sampler to obtain real-time RSSI samples in the wireless network environment, and by performing interference analysis using the formed interference source prediction model. , it can determine the interference source and provide IoT service suppliers with a reference basis to solve the interference problem.

This embodiment provides an interference source recognition method. FIG. 6 is a diagram illustrating the method. As shown in FIG. 6, the method includes the following steps.

Step 601: forming an interference source prediction model using a sample data set corresponding to each test window of the plurality of test windows and a label corresponding to the test window;
Step 602: Perform energy detection sampling on the received signal strength within the detection window at a predetermined sampling rate to obtain a sample data set corresponding to the detection window;
Step 603: Input the sample data set into the interference source prediction model and obtain an interference source type corresponding to the detection window.

In this embodiment, step 601 can be realized by the method of embodiment 1, and since the method has already been described in detail in embodiment 1, the contents thereof are merged here, and here, The detailed explanation will be omitted. Furthermore, in this embodiment, step 601 is not limited, and even if the method of the first embodiment is appropriately modified, an interference source prediction model can be formed.

In this embodiment, step 601 can be completed offline, that is, after generating the interference source prediction model, use the model to perform interference detection for various actual scenarios, and perform interference detection for various actual scenarios. It is possible to determine the interference source type below. For example, by collecting sample data (sample data set) in the actual scenario in step 602, and inputting the sample data set acquired in the actual scenario into the interference source prediction model, the corresponding interference source can be detected. A typology can be obtained.

In this embodiment, the detection method in step 602 is not limited, and the implementation process may be the same as or different from step 201 in FIG. 2, so a detailed explanation thereof will be omitted here. Further, in step 601, the detection window is set to, for example, 20 ms, similar to the test window of Example 1, and the sampling rate is also the same as that of Example 1, for example, once every approximately 100 μs. be.

This embodiment is not limited to the implementation process of step 603, and specifically, prior art can be referred to.

Note that although only each step or process related to the present invention has been described above, the present invention is not limited thereto. The method may further include other steps or processes, and prior art can be referred to for the specific contents of such steps or processes.

As can be seen from the above embodiment, in this embodiment, a deep learning method is used to form an interference source prediction model, and the forming process can be completed offline. This allows us to acquire real-time RSSI samples in a wireless network environment using a high-rate received signal strength indication (RSSI) sampler and perform interference analysis using the formed interference source prediction model. , it can determine the interference source and provide IoT service suppliers with a reference basis to solve the interference problem.

This embodiment provides an interference source prediction model generation device. The device may be, for example, a node in the network, or a separate device from a node in the network, and may include one or more parts or assemblies located at said node or device. It's okay. In addition, the principle by which the device solves the problem is similar to the method of Example 1, so its specific implementation can refer to the implementation method of the method of Example 1, and the content is the same. Description is omitted.

FIG. 7 is a diagram showing the device 700. As shown in FIG. 7, the apparatus 700 includes a first sampling unit 701, a marking unit 702, and a training unit 703.

The first sampling unit 701 performs energy detection sampling on each test window of the plurality of test windows at a predetermined sampling rate under a plurality of preset interference scenarios, and performs energy detection sampling on the test window at a predetermined sampling rate. Obtain sample data sets; labeling unit 702 may include the test period to which each test window belongs and/or the timestamp of the samples in the sample data set within each test window; Label each test window based on the data probed in the test window; training unit 703 trains the deep neural network based on the sample data set of each test window and the label corresponding to each test window. and obtain parameters of the interference source prediction model.

In this embodiment, the wireless network probe mentioned above may be any one or more of the following: WIFI probe, Bluetooth probe, ZigBee probe, etc.

In one implementation, labeling unit 702 labels the test window based on the test period to which the test window belongs.

FIG. 8 is a diagram illustrating the labeling unit 702 of this implementation. As shown in FIG. 8, in the implementation scheme, the labeling unit 702 includes a first determining unit 801 and a first labeling unit 802, and the first determining unit 801 determines that the test period to which the test window belongs is the first. Determine if it is the working period of the interference source; if the first determination unit 801 determines "yes", the first labeling unit 802 marks the test window with interference from the first interference source; Attach a sign indicating that In this embodiment, the first interference source is, for example, a microwave oven.

In one implementation, the labeling unit 702 configures the labeling unit 702 based on the test period to which the test window belongs, a timestamp of a sample in a sample data set within the test window, and data probed by a wireless network probe. Test windows can be labeled.

FIG. 9 is a diagram illustrating the labeling unit 702 of this implementation. As shown in FIG. 9, in the implementation scheme, the labeling unit 702 includes a second determining unit 901 and a second labeling unit 902, and the second determining unit 901 determines that the test period to which the test window belongs is the second one. is the working period of the interferer and the timestamp of at least one sample in the sample data set within said test window is from the start to the end of the frame belonging to the specified address pair detected by the wireless network probe. determine whether the time is within; the second labeling unit 902 determines that if the second determining unit 901 determines "yes", the test window indicates that there has been interference from the second interference source; Attach a sign. In this embodiment, the second interference source is, for example, data transmission. The data here is, for example, a large amount of data such as video data and/or file data.

In another implementation, the labeling unit 702 labels the test window based on timestamps of samples in a sample data set within the test window and data probed by a wireless network probe. .

FIG. 10 is a diagram illustrating the labeling unit 702 of this implementation method. As shown in FIG. 10, in the implementation scheme, the labeling unit 702 includes a third determining unit 1001 and a third labeling unit 1002, and the third determining unit 1001 determines the number of samples in the sample data set within the test window. The third labeling unit 1002 determines whether the time stamp of at least one sample is within the time from the start to the end of the beacon frame detected by the wireless network detector, and the third marking unit 1002 determines whether the time stamp of the at least one sample is within the time from the start to the end of the beacon frame detected by the wireless network detector, ”, an indicator is added to the test window indicating that interference from the beacon frame has occurred. In this implementation method, the beacon frame is a WIFI beacon frame, a Bluetooth beacon frame, or the like.

In this embodiment, the training unit 703 takes all samples in the sample data set in all test windows as inputs of a convolutional neural network, labels corresponding to all test windows as outputs of said convolutional neural network, and By performing training based on the architecture of the convolutional neural network, each term parameter of the convolutional neural network can be obtained as a parameter of the interference source prediction model.

Note that although only each component or module related to the present invention has been described above, the present invention is not limited thereto. Apparatus 700 may further include other components or modules, and reference may be made to the related art for specific details of such components or modules.

As can be seen from the above embodiment, in this embodiment, a deep learning method is used to form an interference source prediction model, and the forming process can be completed offline. This allows us to acquire real-time RSSI samples in a wireless network environment using a high-rate received signal strength indication (RSSI) sampler and perform interference analysis using the formed interference source prediction model. , it can determine the interference source and provide IoT service suppliers with a reference basis to solve the interference problem.

This embodiment provides an interference source recognition device. The device may be, for example, a node in the network, a separate device from a node in the network, or one or more components or devices located at the node or device described above. It may be an assembly. In addition, the principle by which the device solves the problem is similar to the method of Example 2, so its specific implementation can refer to the implementation method of the method of Example 2, and the content is the same. Description is omitted.

FIG. 11 is a diagram showing the device 1100. As shown in FIG. 11, the apparatus 1100 includes a generation unit 1101, a second sampling unit 1102, and a determination unit 1103.

The generation unit 1101 forms an interference source prediction model using the sample data set corresponding to each test window of the plurality of test windows and the label corresponding to the test window; the second sampling unit 1102 forms an interference source prediction model at a predetermined sampling rate. Perform energy detection sampling on the received signal strength within a detection window to obtain a sample data set corresponding to the detection window; Obtain the interferer type corresponding to the window.

In this example, the generation unit 1101 is realized by the apparatus described in Example 3, and can execute the method described in Example 1, and in Example 1 and Example 3, the generation unit 1101 is realized by the apparatus described in Example 3, and the generation unit 1101 is implemented by the apparatus described in Example 3. Since it has already been explained in detail, its contents will be merged here, and detailed explanation thereof will be omitted here.

As can be seen from the above embodiment, in this embodiment, a deep learning method is used to form an interference source prediction model, and the forming process can be completed offline. This allows us to acquire real-time RSSI samples in a wireless network environment using a high-rate received signal strength indication (RSSI) sampler and perform interference analysis using the formed interference source prediction model. , it can determine the interference source and provide IoT service suppliers with a reference basis to solve the interference problem.

Embodiments of the present invention provide a terminal device. The terminal device includes the interference source prediction model generation device described in the third embodiment, or the interference source recognition device described in the fourth embodiment. The terminal device may be a certain node in the network, or may be a device different from the node in the network, and specifically is the same as described above.

FIG. 12 is a system configuration diagram of the terminal device of this embodiment. As shown in FIG. 12, the terminal device 1200 may include a central processing unit 1201 and a memory 1202, and the memory 1202 is connected to the central processing unit 1201. It should be noted that the figure is only an example, and further, telecommunication functions or other functions can be realized by supplementing or replacing this structure with other types of structures.

In one implementation method, the functions of the interference source prediction model generation device described in Example 3 or the interference source recognition device described in Example 4 are integrated into the central processing unit 1201, and the functions of the interference source prediction model generation device described in Example 3 are integrated into the central processing unit 1201, and the functions of the interference source prediction model generation device described in Example 3 are integrated into the central processing unit 1201, The model generation method or the interference source recognition method described in Example 2 can be implemented.

For example, the central processing unit 1201 may be configured as follows: Under a plurality of preset interference scenarios, for each test window of a plurality of test windows, the central processing unit 1201 may be configured to energy detection sampling to obtain a sample data set within each test window; the test period to which each test window belongs; and/or the timestamp of the sample in the sample data set within each test window; or label each test window based on the data detected in each test window by the wireless network detector; apply a label to a deep neural network based on the sample data set of each test window and the label corresponding to each test window; Parameters of the interference source prediction model are obtained by performing training on the interference source prediction model.

Further, for example, the central processing unit 1201 may be configured as follows, that is, using a sample data set corresponding to each test window of a plurality of test windows and a label corresponding to the test window, forming a source prediction model; performing energy detection sampling on the received signal strength within a detection window at a predetermined sampling rate to obtain a sample data set corresponding to the detection window; forming the sample data set into the interference source prediction model; and obtain the interference source type corresponding to the detection window.

In another implementation method, the interference source prediction model generation device according to the third embodiment or the interference source recognition device according to the fourth embodiment may be arranged separately from the central processing unit 1201, and for example, The interference source prediction model generation device or the interference source recognition device is configured as a chip connected to the central processing unit 1201, and under the control of the central processing unit 1201, the interference source prediction model generation device or the interference source recognition device described in the third embodiment or the collective embodiment The function of the interference source recognition device described in 4 may be realized.

As shown in FIG. 12, the terminal device 1200 may further include a communication module 1203, an input unit 1204, an audio processor 1205, a display 1206, a power source 1207, and the like. Note that the terminal device 1200 does not necessarily need to include all the parts shown in FIG. Furthermore, the terminal device 1200 may further include components not shown in FIG. 12, and related art may be referred to in this regard.

As shown in FIG. 12, central processing unit 1201 may be referred to as a controller or operating controller and may include a microprocessor or other processing and/or logic device; , can receive input and control the operation of each component of the terminal device 1200.

The storage 1202 may be, for example, one or more of a buffer, fresh memory, HDD, movable media, volatile storage, non-volatile storage, or other suitable device. Various types of information can be stored, and programs for information processing can also be stored. The central processing unit 1201 can execute the program stored in the storage device 1202 so as to realize storage or processing of information. Note that the functions of other parts are similar to those of the prior art, so detailed explanation thereof will be omitted here. Furthermore, each component of the terminal device 1200 may be realized by dedicated hardware, firmware, software, or a combination thereof, all of which fall within the technical scope of the present invention.

The terminal device of this embodiment forms an interference source prediction model using a deep learning method, and the forming process can be completed offline. It uses a high-rate received signal strength indication (RSSI) sampler to acquire real-time RSSI samples in the wireless network environment and performs interference analysis using the formed interference source prediction model. It can determine the source and provide IoT service suppliers with a reference basis for resolving interference issues.

The embodiment of the present invention further provides a computer readable program, wherein when the program is executed in a terminal device, the program causes the computer to execute the method described in the first embodiment or the second embodiment. let

Embodiments of the present invention further provide a storage medium storing a computer readable program, wherein the computer readable program causes a computer to perform the method described in the first embodiment or the second embodiment.

Further, the apparatuses, methods, etc. according to the embodiments of the present invention may be implemented by software, hardware, or a combination of hardware and software. The invention also relates to such a computer readable program, i.e. said program, when executed by a logic component, is capable of causing said logic component to implement the above-mentioned device or component; A component may be capable of implementing the method or steps thereof described above. Furthermore, the present invention also relates to a storage medium for storing the above-mentioned program, such as a hard disk, magnetic disk, optical disk, DVD, fresh memory, etc.

Although the preferred embodiments of the present invention have been described above, the present invention is not limited to these embodiments, and any changes to the present invention fall within the technical scope of the present invention unless they depart from the spirit of the present invention.

Claims (9)

An apparatus for generating an interference source prediction model, the apparatus comprising:
Under multiple interference scenarios set in advance, RSSI (Received Signal Strength Indicator) sampling is performed on each test window at a predetermined sampling rate for each test window of multiple test windows, and the sample data in each test window is a first sampling unit to obtain a set;
Based on the test period to which each test window belongs, and/or the timestamp of the sample in the sample data set within each test window, and/or the data probed in each test window by a wireless network probe, the test window a labeling unit for labeling the interference source prediction model; and a training unit for training a deep neural network based on the sample data set of each test window and the label corresponding to each test window to obtain parameters of the interference source prediction model. including;
The apparatus , wherein the indicator represents an interference source type corresponding to each test window . The device according to claim 1,
The length of the test window is the same as the transmission period of the microwave oven or an integral multiple of the transmission period of the microwave oven, and the predetermined sampling rate is higher than once every 200 μs. The device according to claim 1,
The apparatus, wherein the wireless network probe is one or more of a WiFi probe, a Bluetooth probe, and a ZigBee probe. The device according to claim 1,
the labeling unit labels the test window based on a test period to which the test window belongs;
The labeling unit includes:
a first determining unit that determines whether a test period to which the test window belongs is a working period of a first interference source; An apparatus comprising a first marking unit for marking interference by a source. The device according to claim 1 or 4,
The marking unit marks the test window based on a test period to which the test window belongs, a timestamp of a sample in a sample data set within the test window, and data probed by the wireless network probe. Attached,
The labeling unit includes:
The test period to which the test window belongs is a working period of a second interference source, and the timestamp of at least one sample in the sample data set within the test window is a specified address detected by the wireless network detector. a second determination unit that determines whether the time from the start to the end of the frame belonging to the pair is within the time; a second labeling unit for labeling the apparatus. The device according to claim 1,
the marking unit tags the test window based on timestamps of samples in the sample data set within the test window and data probed by the wireless network probe;
The labeling unit includes:
a third determining unit determining whether a timestamp of at least one sample in the sample data set within the test window is within the time from the start to the end of a beacon frame detected by the wireless network detector; The apparatus includes a third marking unit that marks the test window as interference by a beacon when the third judgment unit judges yes. The device according to claim 1,
The training unit takes all samples in the sample data set in all test windows as inputs of a convolutional neural network, takes the indicators of all test windows as outputs of the convolutional neural network, and is based on the architecture of the convolutional neural network. an apparatus that performs training using a convolutional neural network, and obtains parameters of each term of the convolutional neural network as parameters of the interference source prediction model. A device for recognizing an interference source, the device comprising:
a second sampling unit that performs RSSI (Received Signal Strength Indicator) sampling on a detection window at a predetermined sampling rate to obtain a sample data set corresponding to the detection window; 8. An apparatus comprising a determination unit that inputs the interference source prediction model generated by the apparatus according to any one of items 1 to 7 and obtains an interference source type corresponding to the detection window. A terminal device comprising the device according to any one of claims 1 to 8.

Images