A method for suppressing noise in power line carrier communication based on confidence model

Abstract

The present invention relates to the field of noise suppression, and in particular to a method for suppressing noise in power line carrier communication based on a confidence model. The method comprises: performing denoising processing on each signal copy in each clustering cluster to generate an initial denoised signal corresponding to each signal copy. All initial denoised signals are fused to generate a target denoised signal. In the denoising process, firstly, a time-frequency diagram corresponding to the signal copy is generated by short-time Fourier change, and then, the signal noise features in the signal copy and the time-frequency diagram are extracted respectively through two inputs, and then the two features are fused to generate the confidence of different types of noise in the signal copy. Finally, the noise category confidence vector is spliced ​​with the signal copy and input into the noise removal model, so as to use the noise category confidence vector to better guide the model, remove the noise information in the signal copy, and obtain a higher quality denoised signal.

Description

Background Art

[0002] Power Line Communication (PLC) refers to a communication method that uses existing power lines as a transmission medium to send and receive data. It can superimpose high-frequency signals on power lines to transmit information without interfering with normal power supply. PLC has the advantages of low cost and no need for additional wiring, so it is widely used in smart homes, automatic meter reading, smart grids and other fields.

[0003] Due to the complex topology and electrical characteristics of the power line network, the signal may reach the receiving end through multiple paths, which may include direct paths and indirect paths through reflection, refraction or scattering. The length of each path is different, resulting in different signal arrival times, causing the signal received by the receiving end to be the superposition of multiple copies of the signal with different time delays.

[0004] In power line carrier communication (PLC), since power lines are not designed for data transmission, their electrical characteristics are complex and changeable, which will introduce various types of noise into the carrier signal. Such as transient noise generated by electrical equipment, noise introduced by external environmental factors, noise introduced by electromagnetic interference, and noise introduced by multiple branches and connection points in the power line network. As a result, the receiving end will receive signal copies of the same carrier signal with different time delays, and each signal copy will have different degrees of noise interference, making it impossible for the receiving end to accurately demodulate a more accurate real signal. Summary of the invention

[0005] In view of the above technical problems, the technical solution adopted by the present invention is:

[0006] According to one aspect of the present invention, a method for suppressing noise in power line carrier communication based on a confidence model is provided, the method comprising the following steps:

[0007] All power carrier signals acquired in the sampling period are clustered to generate a cluster corresponding to each power carrier signal. Each cluster includes multiple signal copies of the same power carrier signal.

[0008] De-noising is performed on each signal copy in each cluster to generate an initial de-noised signal corresponding to each signal copy.

[0009] All initial denoised signals corresponding to each cluster are fused to generate a target denoised signal corresponding to each power carrier signal.

[0010] Denoising, including:

[0011] Perform short-time Fourier transform on the signal copy to generate the time-frequency diagram corresponding to the signal copy.

[0012] Using one-dimensional convolution, the signal replica is subjected to signal noise feature extraction to generate a one-dimensional feature vector.

[0013] Use two-dimensional convolution to extract signal noise features from the time-frequency graph and generate a two-dimensional feature vector.

[0014] The two-dimensional feature vector is expanded and concatenated with the one-dimensional feature vector to generate a one-dimensional fusion noise feature.

[0015] The one-dimensional fusion noise feature is input into the multi-layer perceptron to generate the noise category confidence vector (A1, A2, ..., Ai, ..., Az), where Ai is the confidence corresponding to the i-th noise category, z is the total number of noise categories, i=1, 2, ..., z.

[0016] After the noise category confidence vector is concatenated with the signal copy, it is input into the noise removal model to generate the initial denoised signal corresponding to the signal copy. The noise removal model is a neural network model.

[0017] Furthermore, all power carrier signals acquired in the sampling period are clustered to generate a cluster corresponding to each power carrier signal, including:

[0018] Acquire all power carrier signals received during the sampling period.

[0019] According to the number of categories of power carrier signals received in a sampling period, all power carrier signals are clustered to generate at least one initial cluster.

[0020] If the number of members in the initial cluster is within a preset number range, the initial cluster is used as the cluster corresponding to the power carrier signal.

[0021] Furthermore, after generating at least one initial cluster, the method further includes:

[0022] If the number of members in the initial clustering cluster is outside the preset number range, all members in the initial clustering cluster are added to the next adjacent sampling period for re-clustering.

[0023] Furthermore, according to the number of categories of power carrier signals received in the sampling period, all power carrier signals are clustered, including:

[0024] According to the historical signal reception frequency, the number of categories of power carrier signals received in a sampling period is determined.

[0025] Use K-Means clustering to cluster all power carrier signals.

[0026] Furthermore, all the initial denoised signals corresponding to each cluster are fused to generate a target denoised signal corresponding to each power carrier signal, including:

[0027] After all the initial denoised signals corresponding to each cluster are concatenated, they are input into the target fusion model to generate the target denoised signal corresponding to each power carrier signal. The target fusion model includes a recurrent neural network model or a long short-term memory network model.

[0028] Furthermore, before expanding the two-dimensional feature vector and concatenating it with the one-dimensional feature vector to generate a one-dimensional fusion noise feature, the method further includes:

[0029] The expanded two-dimensional feature vector is normalized so that the feature value of the expanded two-dimensional feature vector is located in a preset normalization interval.

[0030] Furthermore, before expanding the two-dimensional feature vector and concatenating it with the one-dimensional feature vector to generate a one-dimensional fusion noise feature, the method further includes:

[0031] The one-dimensional feature vector is normalized so that the eigenvalue of the one-dimensional feature vector is located in a preset normalization interval.

[0032] Furthermore, the noise removal model is a generative adversarial network model.

[0033] Furthermore, before fusing all the initial denoised signals corresponding to each cluster, the method further includes:

[0034] K-Means clustering is performed on all initial denoised signals corresponding to each cluster to generate a fused cluster corresponding to each cluster.

[0035] All members of the fused cluster corresponding to each cluster are used as all initial denoised signals corresponding to each cluster.

[0036] Furthermore, according to the number of categories of power carrier signals received in the sampling period, all power carrier signals are clustered, including:

[0037] Use DBSCAN clustering to cluster all power carrier signals.

[0038] According to a second aspect of the present invention, a non-transitory computer-readable storage medium is provided, wherein the non-transitory computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned power line carrier communication noise suppression method based on a confidence model is implemented.

[0039] According to a third aspect of the present invention, an electronic device is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned power line carrier communication noise suppression method based on a confidence model when executing the computer program.

[0040] The present invention has at least the following beneficial effects:

[0041] Since the multiple signal copies received by the receiving end are all signals of the same power carrier signal doped with different noise data, there is still a certain degree of similarity between the multiple signal copies. Based on this, the power carrier signal obtained in the sampling period can be clustered to cluster the signal copies of the same power carrier signal with different delays into the same cluster cluster for subsequent denoising processing.

[0042] At the same time, in the denoising process, short-time Fourier transform is first used to generate the time-frequency diagram corresponding to the signal replica. The time-frequency diagram can not only show the time-frequency distribution of the signal, but also show the characteristics of the noise and interference in the signal. Then, the signal noise characteristics in the signal replica and the time-frequency diagram are extracted respectively through two inputs, and the two features are fused to generate the confidence of different types of noise in the signal replica. Finally, the noise category confidence vector is spliced ​​with the signal replica and input into the noise removal model to use the noise category confidence vector to better guide the model, remove the noise information in the signal replica, and obtain a higher quality denoised signal.

[0043] Finally, since the denoising process will inevitably cause the obtained initial denoised signal to have different degrees of distortion, in order to minimize the distortion of the signal, the present invention also fuses multiple denoised initial denoised signals to generate a higher precision target denoised signal corresponding to the power carrier signal. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0045] Figure 1 A flowchart of a power line carrier communication noise suppression method based on a confidence model provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0047] As a possible embodiment of the present invention, Figure 1 As shown, a method for suppressing noise in power line carrier communication based on a confidence model comprises the following steps:

[0048] S100: Clustering all power carrier signals acquired in a sampling period to generate a cluster corresponding to each power carrier signal. Each cluster includes multiple signal copies of the same power carrier signal.

[0049] Power line carrier communication can be used in a wide range of fields, such as smart home: realizing the interconnection and remote control of home appliances. Industrial automation: monitoring and controlling equipment in factories and other places. And scenarios such as smart transportation, security monitoring, and agricultural irrigation.

[0050] Regardless of the specific scenario in which it is used, the power topology network in the scenario will usually be in a long-term stable state, so when the signal is transmitted through power line carrier communication, the electrical characteristics and various noise characteristics will gradually become stable in later use. For example, when the transmitter sends information to the receiver, how long does it generally take for multiple copies of information with different delays to all reach the transmitter, and what are the characteristics of the noise from a certain transmitter to a certain receiver. Based on this, the length of the sampling period can be smoothly determined. Usually, the sampling period can be selected as the maximum value of the delay length of the information copy in the corresponding scenario.

[0051] In addition, the sender usually carries some of its own attribute information when sending information, so that the receiver can correctly decode and process the received data. For example, the frame header information includes synchronization words, sequence numbers, frame lengths, destination addresses, and source addresses. Therefore, the sampling period from a certain sender to a certain receiver can also be determined based on the above frame header information.

[0052] Specifically, S100 includes:

[0053] S101: Acquire all power carrier signals received in a sampling period.

[0054] S102: Clustering all power carrier signals according to the number of categories of power carrier signals received in a sampling period to generate at least one initial cluster.

[0055] Since clustering is used to filter out all signal copies belonging to the same power carrier signal in the sampling period for subsequent denoising, it is necessary to more accurately determine the number of categories of power carrier signals received in the sampling period. This feature value can usually be determined by identifying the type of serial number included in all frame header information, or by counting the number of different power carrier signals (i.e., signal reception frequency) usually received in a sampling period during historical communication.

[0056] Specifically, S102 includes:

[0057] S112: Determine the number of types of power carrier signals received in a sampling period according to the historical signal reception frequency.

[0058] S122: Use K-Means clustering to cluster all power carrier signals.

[0059] When the number of categories of power carrier signals can be determined through existing information, K-Means can be used directly for clustering.

[0060] In addition, S102 also includes:

[0061] S132: Use DBSCAN clustering to cluster all power carrier signals.

[0062] When the number of categories of power carrier signals cannot be determined by existing information, DBSCAN can be used for clustering. DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is a density-based clustering algorithm that identifies clusters by defining "core points", "boundary points" and "noise points". DBSCAN does not require the number of clusters to be specified in advance, but automatically determines the number of clusters based on the density distribution between data points.

[0063] Of course, clustering can also be performed by combining DBSCAN with K-Means, that is, DBSCAN first determines the number of clusters, and then K-Means performs clustering according to the determined number of clusters.

[0064] S103: If the number of members in the initial cluster is within a preset number range, the initial cluster is used as the cluster corresponding to the power carrier signal.

[0065] S104: If the number of members in the initial clustering cluster is outside the preset number range, all members in the initial clustering cluster are added to the next adjacent sampling period for re-clustering.

[0066] Usually, when the power network topology is determined, the number of signal copies generated by a power carrier signal between a certain transmitting end and a certain receiving end also basically has a constant range (that is, a preset number range). At the same time, since in actual use, a sampling cycle may contain all information copies of a certain power carrier signal and a small number of information copies of the next power carrier signal, it is possible to ensure that all information copies of the power carrier signal can be obtained as much as possible by screening within the preset number range.

[0067] S200: performing denoising processing on each signal copy in each cluster to generate an initial denoised signal corresponding to each signal copy.

[0068] A model group is used in the denoising process for specific processing, one of which is a noise recognition model, used to perform steps S202 to S205, and the other is a noise removal model, used to perform step S206. The specific denoising process includes:

[0069] S201: Performing a short-time Fourier transform on the signal replica to generate a time-frequency diagram corresponding to the signal replica.

[0070] Short-Time Fourier Transform (STFT) is a method for analyzing non-stationary signals. In power carrier communication (PLC), the short-time Fourier transform captures the time and frequency information of the signal by dividing the signal into multiple short time periods and performing Fourier transform on each time period. Specifically, the horizontal axis of the obtained two-dimensional time-frequency diagram represents time, the vertical axis represents frequency, and the color or grayscale represents the intensity of the frequency component (i.e., the square of the amplitude). For power carrier signals, STFT can reflect the time-frequency characteristics of the signal, identify the noise source and its impact, and provide a reference for denoising.

[0071] For the main power carrier signal in the signal replica, it appears as a horizontal bright line or strip on the time-frequency diagram. By observing the time-frequency diagram, the main frequency components of the signal and their changes over time can be intuitively identified.

[0072] For transient noise in the signal replica (such as the switching action of electrical equipment), it usually appears as a short pulse or spike on the time-frequency diagram. These noises may appear at a specific time point and have a wide frequency range.

[0073] For periodic interference signals generated by certain noise sources in the signal replica (such as motors, fluorescent lights, etc.), these signals usually appear as periodic frequency strips or pulses in the time-frequency diagram.

[0074] Broadband noise in the signal copy (such as white noise, electromagnetic interference, or external environmental influences such as weather) usually appears as background noise on the time-frequency diagram, covering a wide frequency range. The intensity of broadband noise may vary over time, but it usually appears as a uniformly distributed noise background on the time-frequency diagram.

[0075] S202: Use one-dimensional convolution to extract signal noise features from the signal replica to generate a one-dimensional feature vector.

[0076] S203: Use two-dimensional convolution to extract signal noise features from the time-frequency graph and generate a two-dimensional feature vector.

[0077] S204: Expand the two-dimensional feature vector and concatenate it with the one-dimensional feature vector to generate a one-dimensional fusion noise feature.

[0078] In this embodiment, two inputs are set to extract noise features through the one-dimensional data of the signal itself and the two-dimensional data of its corresponding time-frequency graph. In addition, when splicing two feature vectors, in order for the model to better consider the influence of learning each feature, it is necessary to normalize the two-dimensional feature vector and the one-dimensional feature vector to the same value scale. Specifically, the normalization method is as follows:

[0079] Furthermore, before S204, the method further includes:

[0080] S214: normalizing the expanded two-dimensional feature vector so that the feature value of the expanded two-dimensional feature vector is within a preset normalization interval.

[0081] S224: normalize the one-dimensional feature vector so that the eigenvalue of the one-dimensional feature vector is within a preset normalization range. The normalization ranges of the two feature vectors are the same, such as [0, 1].

[0082] S205: Input the one-dimensional fusion noise feature into the multi-layer perceptron to generate a noise category confidence vector (A1, A2, ..., Ai, ..., Az), where Ai is the confidence corresponding to the i-th noise category, z is the total number of noise categories, i=1, 2, ..., z.

[0083] Multilayer Perceptron (MLP) is a feedforward artificial neural network, which consists of multiple levels of nodes (or neurons), each node is connected to all nodes in the previous layer. Specifically, the multilayer perceptron in this embodiment includes an input layer: receiving input data. Each input node corresponds to a dimension of the input feature. Hidden layer: a layer located between the input layer and the output layer. There can be multiple hidden layers, each containing a number of neurons. Each neuron is connected to all neurons in the previous layer through weights, and an activation function is applied to introduce nonlinearity. Output layer: produces the final prediction result. For the classification task in this embodiment, the output is a noise category confidence vector.

[0084] S206: After concatenating the noise category confidence vector with the signal copy, the noise removal model is input to generate an initial denoised signal corresponding to the signal copy. The noise removal model is a neural network model. Specifically, the noise removal model can be a generative adversarial network model or a convolutional neural network (CNN).

[0085] The Generative Adversarial Network (GAN) is composed of two neural networks: a generator and a discriminator. The two networks compete with each other through adversarial training, and ultimately the generator can generate realistic data samples, while the discriminator can distinguish between real data and generated data. This network model can be used in this embodiment to remove noise in the guidance of the noise category confidence vector to reduce the signal copy.

[0086] In addition, convolutional neural networks (CNNs) are also particularly suitable for processing two-dimensional or one-dimensional time series data, such as the power carrier signal in this embodiment. Through the convolution layer, CNN can extract local features in the signal and reduce the impact of noise through the pooling layer. For the denoising task in this embodiment, an architecture such as U-Net can be used, which combines an encoder and a decoder to effectively denoise while maintaining spatial information.

[0087] S300: performing fusion processing on all initial denoised signals corresponding to each cluster to generate a target denoised signal corresponding to each power carrier signal.

[0088] Specifically, S300 includes:

[0089] S301: After concatenating all initial denoised signals corresponding to each cluster, the signals are input into a target fusion model to generate a target denoised signal corresponding to each power carrier signal. The target fusion model includes a recurrent neural network model or a long short-term memory network model.

[0090] After concatenating all the initial denoised signals corresponding to each cluster, if the overall vector length is less than the preset input length, it is padded with 0.

[0091] In this embodiment, the target fusion model can also be a model for obtaining the average values ​​of multiple initial denoised signals at different points, that is, the fusion processing of multiple initial denoised signals is to take the average value of multiple initial denoised signals to generate the target denoised signal corresponding to the power carrier signal.

[0092] Since the denoising process will inevitably result in different degrees of distortion in the obtained initial denoised signal, in order to minimize the distortion of the signal, the present invention also fuses multiple denoised initial denoised signals to generate a higher precision target denoised signal corresponding to the power carrier signal.

[0093] In order to improve the processing speed, multiple model groups are also set to synchronously denoise different signal copies in the cluster. In this case, the fusion model using RNN or LSTM structure can take into account the time series characteristics to integrate the initial denoised signals from different model groups, which can not only utilize the spatial diversity of different model groups, but also maintain the continuity in time.

[0094] Furthermore, before S300, the method further includes:

[0095] S310: Perform K-Means clustering processing on all initial denoised signals corresponding to each cluster to generate a fused cluster corresponding to each cluster.

[0096] S311: All members in the fused cluster corresponding to each cluster are used as all initial denoised signals corresponding to each cluster.

[0097] In this embodiment, clustering is performed again to further remove noise signals that do not belong to the same cluster, so as to improve the accuracy of the final signal fusion.

[0098] The deep learning neural network model used in the present invention can generate corresponding training samples by collecting historical information in the corresponding power network scene to train the neural network model accordingly, so that it has the corresponding noise recognition, noise removal and signal fusion capabilities. Specifically, the training method and the architecture of each model are existing technologies and will not be repeated here.

[0099] In addition, although the steps of the method in the present disclosure are described in a specific order in the drawings, this does not require or imply that the steps must be performed in this specific order, or that all the steps shown must be performed to achieve the desired results. Additionally or alternatively, some steps may be omitted, multiple steps may be combined into one step, and / or one step may be decomposed into multiple steps, etc.

[0100] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by combining software with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a mobile terminal, or a network device, etc.) to execute the method according to the implementation of the present disclosure.

[0101] In an exemplary embodiment of the present disclosure, an electronic device capable of implementing the above method is also provided.

[0102] It will be appreciated by those skilled in the art that various aspects of the present invention may be implemented as a system, method or program product. Therefore, various aspects of the present invention may be specifically implemented in the following forms, namely: a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software, which may be collectively referred to herein as a "circuit", "module" or "system".

[0103] The electronic device according to this embodiment of the present invention is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0104] The electronic device is presented in the form of a general-purpose computing device. The components of the electronic device may include, but are not limited to: the at least one processor mentioned above, the at least one storage device mentioned above, and a bus connecting different system components (including storage devices and processors).

[0105] The storage stores program codes, which can be executed by the processor, so that the processor executes the steps according to various exemplary embodiments of the present invention described in the above “Exemplary Method” section of this specification.

[0106] The memory may include readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory, and may further include read only memory (ROM).

[0107] The storage may also include a program / utility having a set (at least one) of program modules, such program modules including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0108] The bus may represent one or more of several types of bus structures including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures.

[0109] The electronic device may also communicate with one or more external devices (e.g., keyboards, pointing devices, Bluetooth devices, etc.), may communicate with one or more devices that enable a user to interact with the electronic device, and / or may communicate with any device that enables the electronic device to communicate with one or more other computing devices (e.g., routers, modems, etc.). This communication may be performed through an input / output (I / O) interface. In addition, the electronic device may also communicate with one or more networks (e.g., local area networks (LANs), wide area networks (WANs), and / or public networks, such as the Internet) through a network adapter. The network adapter communicates with other modules of the electronic device through a bus. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with the electronic device, including but not limited to: microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0110] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes a number of instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the implementation of the present disclosure.

[0111] In an exemplary embodiment of the present disclosure, a computer-readable storage medium is also provided, on which a program product capable of implementing the above method of the present specification is stored. In some possible implementations, various aspects of the present invention may also be implemented in the form of a program product, which includes a program code, and when the program product is run on a terminal device, the program code is used to enable the terminal device to execute the steps according to various exemplary embodiments of the present invention described in the above "Exemplary Method" section of the present specification.

[0112] The program product may adopt any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to, a system, device or component of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0113] Computer readable signal media may include data signals propagated in baseband or as part of a carrier wave, in which readable program code is carried. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Readable signal media may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0114] The program code embodied on the readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the foregoing.

[0115] Program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0116] In addition, the above-mentioned figures are only schematic illustrations of the processes included in the method according to an exemplary embodiment of the present invention, and are not intended to be limiting. It is easy to understand that the processes shown in the above-mentioned figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0117] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into multiple modules or units to be embodied.

[0118] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed by the present invention should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the technical solution.

Claims

1. A method for suppressing noise in power line carrier communication based on a confidence model, characterized in that: The method comprises the following steps: Clustering all power carrier signals acquired in a sampling period to generate a cluster corresponding to each power carrier signal; each of the clusters includes multiple signal copies of the same power carrier signal; Performing denoising processing on each signal copy in each cluster to generate an initial denoised signal corresponding to each signal copy; All initial denoised signals corresponding to each cluster are fused to generate a target denoised signal corresponding to each power carrier signal; The denoising process comprises: Performing a short-time Fourier transform on the signal copy to generate a time-frequency diagram corresponding to the signal copy; Using one-dimensional convolution, extracting signal noise features from the signal replica to generate a one-dimensional feature vector; Using two-dimensional convolution, extracting signal noise features from the time-frequency graph to generate a two-dimensional feature vector; Expanding the two-dimensional feature vector and concatenating it with the one-dimensional feature vector to generate a one-dimensional fusion noise feature; Input the one-dimensional fusion noise feature into a multilayer perceptron to generate a noise category confidence vector (A1, A2, ..., Ai, ..., Az); wherein Ai is the confidence corresponding to the i-th noise category, z is the total number of noise categories, i=1, 2, ..., z; After the noise category confidence vector is concatenated with the signal copy, the noise removal model is input to generate an initial denoised signal corresponding to the signal copy; the noise removal model is a neural network model.

2. The method according to claim 1, characterized in that All power carrier signals acquired in the sampling period are clustered to generate a cluster corresponding to each power carrier signal, including: Acquire all power carrier signals received during a sampling period; Clustering all power carrier signals according to the number of categories of power carrier signals received in a sampling period to generate at least one initial clustering cluster; If the number of members in the initial cluster is within a preset number range, the initial cluster is used as the cluster corresponding to the power carrier signal.

3. The method according to claim 2, characterized in that After generating at least one initial cluster, the method further includes: If the number of members in the initial clustering cluster is outside the preset number range, all members in the initial clustering cluster are added to the next adjacent sampling period for re-clustering.

4. The method according to claim 2, characterized in that: According to the number of categories of power carrier signals received in the sampling period, all power carrier signals are clustered, including: According to the historical signal reception frequency, the number of categories of power carrier signals received in a sampling period is determined; Use K-Means clustering to cluster all power carrier signals.

5. The method according to claim 1, characterized in that All the initial denoised signals corresponding to each cluster are fused to generate the target denoised signal corresponding to each power carrier signal, including: All initial denoised signals corresponding to each cluster are concatenated and input into a target fusion model to generate a target denoised signal corresponding to each power carrier signal; the target fusion model includes a recurrent neural network model or a long short-term memory network model.

6. The method according to claim 1, characterized in that Before expanding the two-dimensional feature vector and concatenating it with the one-dimensional feature vector to generate a one-dimensional fusion noise feature, the method further includes: The expanded two-dimensional feature vector is normalized so that the feature value of the expanded two-dimensional feature vector is located in a preset normalization interval.

7. The method according to claim 6, characterized in that Before expanding the two-dimensional feature vector and concatenating it with the one-dimensional feature vector to generate a one-dimensional fusion noise feature, the method further includes: The one-dimensional feature vector is normalized so that the feature value of the one-dimensional feature vector is located in the preset normalization interval.

8. The method according to claim 1, characterized in that The noise removal model is a generative adversarial network model.

9. The method according to claim 1, characterized in that: Before fusing all the initial denoised signals corresponding to each cluster, the method further includes: Perform K-Means clustering processing on all initial denoised signals corresponding to each cluster to generate a fused cluster corresponding to each cluster; All members of the fused cluster corresponding to each cluster are used as all initial denoised signals corresponding to each cluster.

10. The method according to claim 2, characterized in that According to the number of categories of power carrier signals received in the sampling period, all power carrier signals are clustered, including: Use DBSCAN clustering to cluster all power carrier signals.

Images