Wireless charging equipment identification method based on USB signal crosstalk

Abstract

The invention discloses a wireless charging device identification method based on USB signal crosstalk, a USB power supply used for charging a wireless charging device is provided with at least two USB ports, when the wireless charging device is connected with one USB port of the USB power supply through a wireless charger, a voltage signal with the duration of two seconds is collected as a wireless charging device identification signal, and the wireless charging device identification signal is used as a wireless charging device identification signal. After filtering and baseline correction are sequentially carried out on the wireless charging equipment identification signal, a standardized feature vector is extracted and uploaded to a remote server, identification is carried out through a wireless charging equipment identification classifier to obtain a corresponding equipment label, the equipment label is verified, if verification is passed, the equipment label is correct, identification is passed, and if verification is not passed, the equipment label is not correct. If the verification is not passed, the equipment label is incorrect, and the identification is not passed; the wireless charging device has the advantages that the wireless charging device can be effectively charged in the whole wireless charging period, software installation in the wireless charging device is not needed, and hardware modification on the wireless charger is not needed.

Description

Technical Field

[0001] The present invention relates to a method for identifying a wireless charging device, and in particular to a method for identifying a wireless charging device based on USB signal crosstalk. Background Art

[0002] In recent years, wireless charging technology has become an important way to recharge wireless charging devices (such as smartphones, IPADs, laptops, etc.) in public places (such as hotels, offices, cafes, and airports). This technology provides a convenient charging solution, avoiding the use of traditional charging cables and greatly improving user convenience. Wireless charging device identification plays a key role in wireless charging. By accurately identifying the wireless charging device connected to it, the wireless charger can provide the best charging current and voltage to ensure maximum charging efficiency and battery safety of the wireless charging device. In addition, correct wireless charging device identification can also optimize the user experience, for example, providing a customized charging interface and other smart functions.

[0003] Different wireless charging device models bring unique charging parameters or charging protocols. The internal components of different wireless charging device models, such as CPU, GPU, screen driver, screen technology and resolution, will produce unique power consumption patterns during the charging process of wireless charging devices. These patterns are reflected in the frequency and amplitude of the voltage signal, making it possible to identify wireless charging devices.

[0004] The mainstream wireless charging device identification methods on the market are mainly based on direct access to the internal hardware features of wireless charging devices, such as identifying wireless charging devices by analyzing the data of sound sensors and inertial sensors. Although the traditional wireless charging device identification method does not require modification of the wireless charger hardware and will not cause charging interruptions during the charging cycle, allowing the wireless charging device to charge effectively throughout the wireless charging cycle, it requires additional software installation in the wireless charging device.

[0005] Most other wireless charging device identification methods rely on software-level interactions or require physical modifications to the wireless charger. Although this method does not cause charging interruptions during the charging cycle, allowing the wireless charging device to charge effectively throughout the wireless charging cycle, it requires software installation in the wireless charging device or hardware modifications to the wireless charger.

[0006] In recent years, Yang Deliang and others proposed the first system QID that can identify Qi-compliant mobile devices in real time during wireless charging. The accuracy of wireless charging device identification using this system QID is 85.2%. QID extracts the characteristics of wireless charging devices from the clock oscillator and control scheme of the power receiver of the wireless charging device, and uses a lightweight algorithm to classify the wireless charging device. It uses a two-dimensional motion unit to simulate Qi's various multi-coil designs to achieve fine-grained wireless charging device fingerprint recognition. Although this method does not require software installation in the wireless charging device, it will not only cause charging interruptions during the charging cycle, but also fail to ensure that the wireless charging device is effective throughout the wireless charging cycle, and requires hardware modification of the wireless charger. Summary of the invention

[0007] The technical problem to be solved by the present invention is to provide a wireless charging device identification method based on USB signal crosstalk, which can ensure that the wireless charging device is effectively charged throughout the entire wireless charging cycle and does not require software installation in the wireless charging device or hardware modification of the wireless charger.

[0008] The technical solution adopted by the present invention to solve the above technical problems is: a wireless charging device identification method based on USB signal crosstalk, wherein a USB power supply for charging the wireless charging device has at least two USB ports, and when the wireless charging device is connected to a USB port of the USB power supply through a wireless charger, a voltage signal with a collection time of two seconds is used as a wireless charging device identification signal, and the wireless charging device identification signal is filtered and baseline corrected in turn, and a standardized feature vector is extracted and uploaded to a remote server, and the standardized feature vector is identified by a trained wireless charging device identification classifier pre-stored in the remote server to obtain a corresponding device label, and the device label is verified. If the verification passes, the device label is correct and the wireless charging device identification passes; if the verification fails, the device label is incorrect and the wireless charging device identification fails.

[0009] Furthermore, the trained wireless charging device identification classifier pre-stored in the remote server is obtained by the following steps:

[0010] Step 1: Equipment preparation and parameter configuration, specifically:

[0011] S1-1. According to the need for wireless charging device identification, prepare m different types of wireless charging devices, n wireless chargers from different manufacturers, p USB power supplies from different brands and different architectures with at least two USB ports, a digital oscilloscope and a passive probe, where m is an integer greater than or equal to 2, and its value is equal to the total number of wireless charging devices that need to be identified, n is an integer greater than or equal to 2, and the larger the value, the better, and p is an integer greater than or equal to 2, and the larger the value, the better; set a device label for each of the m wireless charging devices, and the device labels of the m wireless charging devices are different;

[0012] S1-2, each of the m types of wireless charging devices, the n types of wireless chargers and the p types of USB power supplies, a wireless charger and a USB power supply are combined as a wireless charging system to form s=m*n*p types of wireless charging system combinations, in which the wireless charger in each wireless charging system combination can charge the wireless charging device after being connected to a USB port of the USB power supply via a USB cable;

[0013] Step 2: Build a standardized data set. The specific process is as follows:

[0014] S2-1, set the sampling frequency of the digital oscilloscope to 2MHz, connect the passive probe to the digital oscilloscope, connect the wireless charger in each wireless charging system combination to a USB port of the USB power supply through a USB cable, and charge the wireless charging device;

[0015] The signal at the USB port where the USB power source in each wireless charging system combination is connected to the wireless charger is called the signal of the wireless charging system combination. The signal of each wireless charging system combination is collected using a passive probe, where the time duration for the passive probe to collect the signal of the i-th wireless charging system combination is t i , t i >2 seconds, the time length collected by the passive probe is t i The signal of the i-th wireless charging system combination is recorded as signal sample P i , signal sample P i It is composed of multiple sampling values, each sampling value corresponds to a sampling point of the digital oscilloscope, i = 1, 2, ..., s, a total of s signal samples are collected, using P1 to P s Construct the original total data set P of signal samples, P = {P1,, P2, ..., P s};

[0016] S2-2, signal sample length standardization, the specific process is:

[0017] Take the i-th signal sample P iThe sampling value of the first two seconds, the number of sampling points in the first two seconds of sampling value is recorded as 4M, M = 1000000, 4M = 4000000, take the i-th signal sample P i The sampling values ​​of the first two seconds constitute the i-th signal sample length standardized data set A i , using s signal sample length standardized data sets to form a standardized data set A, A = {A1, A2, ...A s};

[0018] Step 3: Perform data preprocessing on the standardized data set A to obtain the data set FED. The specific process is as follows:

[0019] S3-1. Use a first-order low-pass filter with a cutoff frequency of 0.1 MHz to perform low-pass filtering on the standardized data set A to remove high-frequency noise components and obtain a low-pass filtered data set, denoted as B, where B = {B1, B2, ..., B s}, where data set B i Normalize the dataset A for the i-th signal sample length i The data set obtained after low-pass filtering is data set B i The number of sampling points contained in data set B is also 4M, that is, i Contains 4M sample values;

[0020] S3-2, use a moving average filter to smooth the data set B, and obtain a smoothed data set C, where C = {C1, C2, ..., C s}, C i For data set B i The data set is obtained by smoothing using a moving mean filter, where B i Use the moving mean filter to smooth the data set C i The specific process is:

[0021] Set a sliding window with a size of 1000 sampling points and place the sliding window on data set B. i From the 1st sampling point to the 1000th sampling point, at this time, data set B i The first sampling point to the 1000th sampling point of is covered by the sliding window as data set B i The first segment of the data set B is then moved backwards to i Segmentation is performed, and the sliding window moves one sampling point each time. Each time, 1000 consecutive sampling points covered by the sliding window are considered as a segment until data set B i The 4Mth sampling point of data set B is covered. iThe 1000 sampling points covered by the sliding window for the kth time are taken as its kth segment, k = 1, 2, ..., 3999001, and the data set B is calculated. i The average value of the 1000 sample values ​​in the kth segment is recorded as C i,k ; Using C i,1 To C i,3999001 Constructing the dataset C i , C i,k For dataset C i The kth sample value in;

[0022] S3-3, using formula (2) to calculate C i，k Perform baseline correction:

[0023] FE i,k =C i,k -min(C i ) (2)

[0024] Among them, FE i,k Indicates C i，k The sampling value obtained by baseline correction, min(C i ) is the data set C i The minimum value of all data in ;

[0025] Using FE i,1 to FE i,3990001 Constructing the dataset FE i , FE i,k For the dataset FE i The kth data in;

[0026] Using FE1 to FE s Construct a data set FE, FE = {FE1, FE2, ..., FE s}, FE i is the i-th data set in data set FE;

[0027] S3-4, to FE i Perform equal length slicing. The specific process is as follows: for the i-th data set FE i , starting from the first sampling point, a division is made every 0.2M sampling points, and it is divided into 20 slice segments, each slice segment contains 0.2M sampling points. The data set FE i Contains 20 slice segments, the dataset FE i The xth slice in the dataset is denoted as FED i，x , x = 1, 2, ..., 20; FED i，x The yth sampling point is recorded as FED i，x [y], y=1, 2,..., n, n=0.2M;

[0028] Adopting FED i，1 To FED i，20 Constructing the data set FED i , FED i，x For the dataset FED i The xth data set in ;

[0029] Using FED1 to FED s Construct the data set FED, FED i is the i-th data set in the data set FED;

[0030] Step 4: FED dataset i The xth dataset FED i，x Extract features and get the data set FED i，x The specific process is:

[0031] S4-1. Calculate FED using formula (3) to formula (13) respectively i，x 11 time-domain features: average value μ i,x , standard deviation σ i,x , root square amplitude RMS i，x , peak-to-peak value PPC i，x , skewness S i，x , Kurtosis K i，x , peak factor CF i，x , Clearance Factor CL i，x , shape factor Sh i，x , Pulse Index I i，x and coefficient of variation CV i，x :

[0032]

[0033]

[0034] PPC i,x =max(FED i,x )-min(FED i,x ) (6)

[0035]

[0036] In the above formula, max is the maximum value symbol, max(FED i,x ) represents the data set FED i,x The maximum value in, min is the minimum value symbol, min(FED i,x ) represents the data set FED i,x The minimum value in ;

[0037] S4-2. The data set FE is transformed by fast Fourier transform (FFT) to obtain a frequency domain data set of the data set FE. The frequency domain data set of the data set FE is recorded as FE′, where FE′={FE′1, FE′2, …, FE′ s}, where FE′ i Represents the dataset FE i The data set obtained by fast Fourier transform (FFT) is converted into FE i The xth slice fragment FED i，x The transformed data set is recorded as FED i，x ′, dataset FED i，x FE′ i The xth data set in FED i，x The yth data in ′ is recorded as FED i，x [y]′,FED i，x [y]′ is FED i，x [y] Data obtained by fast Fourier transform (FFT);

[0038] Calculate FED using formula (14) to formula (25) respectively i，x ′’s 12 frequency domain features: mean μ i,x ′, variance σ i,x ' 2 , skewness S i,x ′, kurtosis K i,x ', root mean square amplitude RMS i,x ′、Peak i,x ′、Form factor WF i,x ', peak factor CF i,x ′、ClF i,x ′、Pulse index IF i,x ′, first-order moment M i,x ′ and frequency center SC i,x ′:

[0039]

[0040] Peak i,x ′=max(FED i,x ′) (19)

[0041]

[0042]

[0043] Where max(FED i,x ′) represents FED i，x The maximum value of ′; f s is the sampling frequency of the digital oscilloscope;

[0044] S4-3. Fed i，x 11 time domain features and FED i，x The 12 frequency domain features of ′ are used as 23 feature elements to form a 23-dimensional vector, which is used as FED i，x The eigenvector of i，x , the feature vector F i，x The rth characteristic element of i，x，r , r = 1, 2, ..., 23;

[0045] Using the eigenvector F i，1 To F i，20 Construct the feature vector set F i , F i ={F i，1 , F i，2 , …, F i，20};

[0046] Using the feature vector set F1 to F s The total set of feature vectors is denoted as F, where F = {F1, F2, ..., F s};

[0047] S4-4. Perform Z-score normalization on each feature vector in the total feature vector set F to obtain the corresponding average value, where formula (26) is used to normalize F i，x Perform Z-score standardization:

[0048]

[0049] in, is the eigenvector F i，x The average value obtained by Z-score standardization;

[0050] Formula (27) is used to calculate F i，x，r The corresponding Z-score value is:

[0051]

[0052] In formula (27), z i，x，r Indicates F i，x，r The Z-score value obtained after Z-score standardization;

[0053] Using Z i，x，1 To Z i，x，23 Construct the feature vector F i，x The normalized eigenvector Z i，x , Z i，x ={z i，x，1, z i，x，2 , …, z i，x，23};

[0054] Use the standardized eigenvector Z i，1 To Z i，20 Construct a standardized feature vector set Z i , Z i = {Z i，1 , Z i，2 , …, Z i，20};

[0055] Use the standardized feature vector set Z1 to Z s The total set of standardized feature vectors is denoted as Z, where Z = {Z1, Z2, ..., Z s};

[0056] Step 5: Use the device labels of the wireless charging devices in the i-th wireless charging system combination as the standardized feature vector Z i，x The device label is the normalized feature vector Z i，x Its device label constitutes a standardized feature vector sample, and the standardized feature vector set Z1 to Z s All the standardized feature vectors in are constructed to obtain 20s standardized feature vector samples, and the 20s standardized feature vector samples are used to construct the training data set;

[0057] Step 6: Integrate the random forest, support vector machine, K nearest neighbor classifier and voting ensemble learning algorithm to form an integrated classification model, wherein the integrated random forest, support vector machine and K nearest neighbor classifier are all used to identify the standardized feature vectors input therein, obtain the corresponding device labels and output them to the voting ensemble learning algorithm, and the voting ensemble learning algorithm is used to process the device labels outputted thereto by the integrated random forest, support vector machine and K nearest neighbor classifier in a voting manner to obtain the final device label output;

[0058] Step 7: Use the training data set to train the integrated classification model to obtain a trained integrated classification model, where the trained integrated classification model is the trained wireless charging device identification classifier pre-stored in the remote server.

[0059] Furthermore, when a wireless charging device is charged in a wireless charging station equipped with a voltage sampling device with a sampling frequency set to 2 MHz, the wireless charging device is connected to a USB port of a USB power supply through the wireless charger. At this time, the same method as steps 2 to 4 is used to obtain a standardized feature vector of the wireless charging device, and the standardized feature vector is remotely transmitted to a trained wireless charging device identification classifier pre-stored in a remote server. The trained wireless charging device identification classifier identifies the standardized feature vector to obtain a device label.

[0060] Furthermore, after the trained wireless charging device identification classifier obtains the device label, the following method is used for verification: obtain the confidence values ​​of the three device labels output by the integrated random forest, support vector machine and K nearest neighbor classifier to the Voting ensemble learning algorithm. If the minimum value of the confidence values ​​of the three device labels is greater than 0.5, the verification passes at this time; if the maximum value of the confidence values ​​of the three device labels is less than 0.5, the verification fails.

[0061] Furthermore, when the verification fails, it means that the device tag of the wireless charging device does not exist in the trained wireless charging device identification classifier. At this time, a device tag is set for the wireless charging device, and the device tag is different from the device tag already existing in the currently trained wireless charging device identification classifier. The standardized feature vector obtained for the wireless charging device and its corresponding device tag constitute a new standardized feature vector sample, and add it to the current training data set, update the current training data set, and then use the updated current training data set to train and update the currently trained wireless charging device identification classifier again, so that the trained wireless charging device identification classifier has the ability to identify the wireless charging device.

[0062] Compared with the prior art, the advantage of the present invention is that when the wireless charging device is connected to a USB port of a USB power source through a wireless charger, a voltage signal with a collection time of two seconds is used as a wireless charging device identification signal, and after the wireless charging device identification signal is filtered and baseline corrected in turn, a feature vector set is extracted and uploaded to a remote server, and the feature vector set is identified by a trained wireless charging device identification classifier pre-stored in the remote server to obtain an identification label of the wireless charging device. Therefore, we can additionally set up a signal collection and processing device connected to the remote server, without the need to install software in the wireless charging device or perform hardware modification on the wireless charger, thereby ensuring that the wireless charging device is effectively charged throughout the entire wireless charging cycle. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1It is a flow chart of the method for identifying a wireless charging device based on USB signal crosstalk of the present invention;

[0064] Figure 2 This is a diagram of an experimental method of a wireless charging device identification method based on USB signal crosstalk of the present invention;

[0065] Figure 3 This is an experimental signal diagram of the wireless charging device identification method based on USB signal crosstalk of the present invention;

[0066] Figure 4 This is a device identification performance diagram of the wireless charging device identification method based on USB signal crosstalk of the present invention. DETAILED DESCRIPTION

[0067] The present invention is further described in detail below with reference to the accompanying drawings.

[0068] Embodiment 1: Figure 1 As shown, a method for identifying a wireless charging device based on USB signal crosstalk is provided, wherein a USB power source for charging the wireless charging device has at least two USB ports, and is characterized in that when the wireless charging device is connected to a USB port of the USB power source through a wireless charger, a voltage signal with a collection time of two seconds is used as a wireless charging device identification signal, the wireless charging device identification signal is filtered and baseline corrected in turn, a standardized feature vector is extracted and uploaded to a remote server, the standardized feature vector is identified by a trained wireless charging device identification classifier pre-stored in the remote server, a corresponding device label is obtained, and the device label is verified. If the verification passes, the device label is correct and the wireless charging device identification passes. If the verification fails, the device label is incorrect and the wireless charging device identification fails.

[0069] In this embodiment, when the wireless charging device is connected to a USB port of a USB power source through a wireless charger, a voltage signal with a collection time of two seconds is used as a wireless charging device identification signal. After the wireless charging device identification signal is filtered and baseline corrected in turn, a feature vector set is extracted and uploaded to a remote server. The feature vector set is identified by a trained wireless charging device identification classifier pre-stored in the remote server to obtain an identification label of the wireless charging device. Therefore, we can additionally set up a signal collection and processing device connected to the remote server. There is no need to install software in the wireless charging device or perform hardware modification on the wireless charger, which can ensure that the wireless charging device is effectively charged throughout the wireless charging cycle.

[0070] Embodiment 2: This embodiment is basically the same as Embodiment 1, except that: in this embodiment, the trained wireless charging device identification classifier pre-stored in the remote server is obtained by the following steps:

[0071] Step 1: Equipment preparation and parameter configuration, specifically:

[0072] S1-1. According to the need for wireless charging device identification, prepare m different types of wireless charging devices, n wireless chargers from different manufacturers, p USB power supplies from different brands and different architectures with at least two USB ports, a digital oscilloscope (as a voltage sampling device) and a passive probe, where m is an integer greater than or equal to 2, and its value is equal to the total number of wireless charging devices that need to be identified, n is an integer greater than or equal to 2, and the larger the value, the better, and p is an integer greater than or equal to 2, and the larger the value, the better; set a device label for each of the m wireless charging devices, and the device labels of the m wireless charging devices are different;

[0073] S1-2, each of the m types of wireless charging devices, the n types of wireless chargers and the p types of USB power supplies, a wireless charger and a USB power supply are combined as a wireless charging system to form s=m*n*p types of wireless charging system combinations, in which the wireless charger in each wireless charging system combination can charge the wireless charging device after being connected to a USB port of the USB power supply via a USB cable;

[0074] Step 2: Build a standardized data set. The specific process is as follows:

[0075] S2-1, set the sampling frequency of the digital oscilloscope to 2MHz, connect the passive probe to the digital oscilloscope, connect the wireless charger in each wireless charging system combination to a USB port of the USB power supply through a USB cable, and charge the wireless charging device;

[0076] The signal at the USB port where the USB power source in each wireless charging system combination is connected to the wireless charger is called the signal of the wireless charging system combination. The signal of each wireless charging system combination is collected using a passive probe, where the time duration for the passive probe to collect the signal of the i-th wireless charging system combination is t i , t i >2 seconds, the time length collected by the passive probe is t i The signal of the i-th wireless charging system combination is recorded as signal sample P i , signal sample P i It is composed of multiple sampling values, each sampling value corresponds to a sampling point of the digital oscilloscope, i = 1, 2, ..., s, a total of s signal samples are collected, using P1 to P sConstruct the original total data set P of signal samples, P = {P1,, P2, ..., P s}; S2-2, signal sample length standardization, the specific process is:

[0077] Take the i-th signal sample P i The sampling value of the first two seconds, the number of sampling points in the first two seconds of sampling value is recorded as 4M, M = 1000000, 4M = 4000000, take the i-th signal sample P i The sampling values ​​of the first two seconds constitute the i-th signal sample length standardized data set A i , using s signal sample length standardized data sets to form a standardized data set A, A = {A1, A2, ...A s};

[0078] Step 3: Perform data preprocessing on the standardized data set A to obtain the data set FED. The specific process is as follows:

[0079] S3-1. Use a first-order low-pass filter with a cutoff frequency of 0.1 MHz to perform low-pass filtering on the standardized data set A to remove high-frequency noise components and obtain a low-pass filtered data set, denoted as B, where B = {B1, B2, ..., B s}, where data set B i Normalize the dataset A for the i-th signal sample length i The data set obtained after low-pass filtering is data set B i The number of sampling points contained in data set B is also 4M, that is, i Contains 4M sample values;

[0080] S3-2, use a moving average filter to smooth the data set B, and obtain a smoothed data set C, where C = {C1, C2, ..., C s}, C i For data set B i The data set is obtained by smoothing using a moving mean filter, where B i Use the moving mean filter to smooth the data set C i The specific process is:

[0081] Set a sliding window with a size of 1000 sampling points and place the sliding window on data set B. i From the 1st sampling point to the 1000th sampling point, at this time, data set B i The first sampling point to the 1000th sampling point of is covered by the sliding window as data set B i The first segment of the data set B is then moved backwards to iSegmentation is performed, and the sliding window moves one sampling point each time. Each time, 1000 consecutive sampling points covered by the sliding window are considered as a segment until data set B i The 4Mth sampling point of data set B is covered. i The 1000 sampling points covered by the sliding window for the kth time are taken as its kth segment, k = 1, 2, ..., 3999001, and the data set B is calculated. i The average value of the 1000 sample values ​​in the kth segment is recorded as C i,k ; Using C i,1 To C i,3999001 Constructing the dataset C i , C i,k For dataset C i The kth sample value in;

[0082] S3-3, using formula (2) to calculate C i，k Perform baseline correction:

[0083] FE i,k =C i,k -min(C i ) (2)

[0084] Among them, FE i,k Indicates C i，k The sampling value obtained by baseline correction, min(C i ) is the data set C i The minimum value of all data in ;

[0085] Using FE i,1 to FE i,3990001 Constructing the dataset FE i , FE i,k For the dataset FE i The kth data in;

[0086] Using FE1 to FE s Construct a data set FE, FE = {FE1, FE2, ..., FE s}, FE i is the i-th data set in data set FE;

[0087] S3-4, to FE i Perform equal length slicing. The specific process is as follows: for the i-th data set FE i , starting from the first sampling point, a division is made every 0.2M sampling points, and it is divided into 20 slice segments, each slice segment contains 0.2M sampling points. The data set FE i Contains 20 slice segments, the dataset FE i The xth slice in the dataset is denoted as FED i，x, x = 1, 2, ..., 20; FED i，x The yth sampling point is recorded as FED i，x [y], y=1, 2,..., n, n=0.2M;

[0088] Adopting FED i，1 To FED i，20 Constructing the data set FED i , FED i，x For the dataset FED i The xth data set in ;

[0089] Using FED1 to FED s Construct the data set FED, FED i is the i-th data set in the data set FED;

[0090] Step 4: FED dataset i The xth dataset FED i，x Extract features and get the data set FED i，x The specific process is:

[0091] S4-1. Calculate FED using formula (3) to formula (13) respectively i，x 11 time-domain features: average value μ i,x , standard deviation σ i,x , root square amplitude RMS i，x , peak-to-peak value PPC i，x , skewness S i，x , Kurtosis K i，x , peak factor CF i，x , Clearance Factor CL i，x , shape factor Sh i，x , Pulse Index I i，x and coefficient of variation CV i，x :

[0092]

[0093] PPC i,x =max(FED i,x )-min(FED i,x ) (6)

[0094]

[0095] In the above formula, max is the maximum value symbol, max(FED i,x ) represents the data set FED i,x The maximum value in, min is the minimum value symbol, min(FED i,x ) represents the data set FED i,xThe minimum value in ;

[0096] S4-2. The data set FE is transformed by fast Fourier transform (FFT) to obtain a frequency domain data set of the data set FE. The frequency domain data set of the data set FE is recorded as FE′, where FE′={FE′1, FE′2, …, FE′ s}, where FE′ i Represents the dataset FE i The data set obtained by fast Fourier transform (FFT) is converted into FE i The xth slice fragment FED i，x The transformed data set is recorded as FED i，x ′, dataset FED i，x FE′ i The xth data set in FED i，x The yth data in ′ is recorded as FED i，x [y]′,FED i，x [y]′ is FED i，x [y] Data obtained by fast Fourier transform (FFT);

[0097] Calculate FED using formula (14) to formula (25) respectively i，x ′’s 12 frequency domain features: mean μ i,x ′, variance σ i,x ' 2 , skewness S i,x ′, kurtosis K i,x ', root mean square amplitude RMS i,x ′、Peak i,x ′、Form factor WF i,x ', peak factor CF i,x ′、ClF i,x ′、Pulse index IF i,x ′, first-order moment M i,x ′ and frequency center SC i,x ′:

[0098]

[0099] Peak i,x ′=max(FED i,x ′) (19)

[0100]

[0101]

[0102] Where max(FED i,x ′) represents FED i，xThe maximum value of ′; f s is the sampling frequency of the voltage sampling device (i.e., digital oscilloscope);

[0103] S4-3. Fed i，x 11 time domain features and FED i，x The 12 frequency domain features of ′ are used as 23 feature elements to form a 23-dimensional vector, which is used as FED i，x The eigenvector of i，x , the feature vector F i，x The rth characteristic element of i，x，r , r = 1, 2, ..., 23;

[0104] Using the eigenvector F i，1 To F i，20 Construct the feature vector set F i , F i ={F i，1 , F i，2 , …, F i，20};

[0105] Using the feature vector set F1 to F s The total set of feature vectors is denoted as F, where F = {F1, F2, ..., F s};

[0106] S4-4. Perform Z-score normalization on each feature vector in the total feature vector set F to obtain the corresponding average value, where formula (26) is used to normalize F i，x Perform Z-score standardization:

[0107]

[0108] in, is the eigenvector F i，x The average value obtained by Z-score standardization;

[0109] Formula (27) is used to calculate F i，x，r The corresponding Z-score value is:

[0110]

[0111] In formula (27), z i，x，r Indicates F i，x，r The Z-score value obtained after Z-score standardization;

[0112] Using Z i，x，1 To Z i，x，23 Construct the feature vector F i，x The normalized eigenvector Zi，x , Z i，x ={z i，x，1 , z i，x，2 , …, z i，x，23};

[0113] Use the standardized eigenvector Z i，1 To Z i，20 Construct a standardized feature vector set Z i , Z i = {Z i，1 , Z i，2 , …, Z i，20};

[0114] Use the standardized feature vector set Z1 to Z s The total set of standardized feature vectors is denoted as Z, where Z = {Z1, Z2, ..., Z s};

[0115] Step 5: Use the device labels of the wireless charging devices in the i-th wireless charging system combination as the standardized feature vector Z i，x The device label is the normalized feature vector Z i，x Its device label constitutes a standardized feature vector sample, and the standardized feature vector set Z1 to Z s All the standardized feature vectors in are constructed to obtain 20s standardized feature vector samples, and the 20s standardized feature vector samples are used to construct the training data set;

[0116] Step 6: Integrate the random forest, support vector machine, K nearest neighbor classifier and voting ensemble learning algorithm to form an integrated classification model, wherein the integrated random forest, support vector machine and K nearest neighbor classifier are all used to identify the standardized feature vectors input therein, obtain the corresponding device labels and output them to the voting ensemble learning algorithm, and the voting ensemble learning algorithm is used to process the device labels outputted thereto by the integrated random forest, support vector machine and K nearest neighbor classifier in a voting manner to obtain the final device label output;

[0117] Step 7: Use the training data set to train the integrated classification model to obtain a trained integrated classification model, where the trained integrated classification model is the trained wireless charging device identification classifier pre-stored in the remote server.

[0118] In this embodiment, when a wireless charging device is charged in a wireless charging station equipped with a voltage sampling device with a sampling frequency set to 2 MHz, the wireless charging device is connected to a USB port of a USB power supply through a wireless charger. At this time, the same method as steps 2 to 4 is used to obtain a standardized feature vector of the wireless charging device, and the standardized feature vector is remotely transmitted to a trained wireless charging device identification classifier pre-stored in a remote server. The trained wireless charging device identification classifier identifies the standardized feature vector to obtain a device label.

[0119] In this embodiment, after the trained wireless charging device identification classifier obtains the device label, the following method is used for verification: obtain the confidence values ​​of the three device labels output by the integrated random forest, support vector machine and K nearest neighbor classifier to the Voting ensemble learning algorithm. If the minimum value of the confidence values ​​of the three device labels is greater than 0.5, the verification is passed at this time. If the maximum value of the confidence values ​​of the three device labels is less than 0.5, the verification fails.

[0120] Embodiment 3: This embodiment is basically the same as Embodiment 2, except that: in this embodiment, when the verification fails, it means that the device tag of the wireless charging device does not exist in the trained wireless charging device identification classifier. At this time, a device tag is set for the wireless charging device, and the device tag is different from the device tags already existing in the currently trained wireless charging device identification classifier. The standardized feature vector obtained by the wireless charging device and its corresponding device tag constitute a new standardized feature vector sample, and are added to the current training data set, the current training data set is updated, and then the current training data set obtained after the update is used to train and update the currently trained wireless charging device identification classifier again, so that the trained wireless charging device identification classifier has the identification ability of the wireless charging device.

[0121] In this embodiment, the training data set can be continuously updated through relevant data of wireless charging devices on the market, and the wireless charging device identification classifier can be updated and trained, so that the trained wireless charging device identification classifier can adapt to the identification requirements of current wireless charging devices.

[0122] In order to verify the performance of the wireless charging device identification method based on USB signal crosstalk of the present invention, Figure 2 As shown, we used an Ugreen USB hub with multiple USB ports (i.e., charging ports). One USB port was connected to a Xiaomi wireless charger to charge a Huawei Mate40 smartphone, and the other USB port was used to record voltage and collect signal samples using a RIGOL MSO5000 digital oscilloscope. Figure 3Voltage signal traces for different wireless charging devices. Our wireless charging devices include different models of mainstream mobile phones: iPhone XR, iPhone12, iPhone14, iPhone15, OnePlus12, OPPOFind X7Pro, Huawei Mate40, Samsung Galaxy S22 Ultra, Xiaomi 13 and Huawei nova6 (due to the lack of built-in wireless charging function, a wireless charging adapter is required). Analysis Figure 3 It can be seen that when the phone is locked or not performing any activity, the voltage traces are very stable, and the patterns of the voltage traces are distinguishable on different phone models.

[0123] The wireless charging device identification method based on USB signal crosstalk of the present invention is used to identify five wireless charging devices (iPhone XR, Huawei Mate40, Huawei nova6, Xiaomi 13, Samsung Galaxy S22 Ultra), and its device identification performance is as follows: Figure 4 Analysis Figure 4 It can be seen that the wireless charging device identification method based on USB signal crosstalk of the present invention has an accuracy rate of 100% for identifying five types of wireless charging devices.

[0124] In summary, the wireless charging device identification method based on USB signal crosstalk of the present invention additionally sets a signal acquisition and processing device connected to a remote server, and can accurately identify the wireless charging device without installing software in the wireless charging device or modifying the hardware of the wireless charger, and can ensure that the wireless charging device is effectively charged throughout the wireless charging cycle. The wireless charging device identification method based on USB signal crosstalk of the present invention has broad application prospects in the current field of wireless charging technology.

Claims

1. A method for identifying a wireless charging device based on USB signal crosstalk, wherein a USB power source for charging the wireless charging device has at least two USB ports, characterized in that When the wireless charging device is connected to a USB port of a USB power source through a wireless charger, a voltage signal with a collection time of two seconds is used as a wireless charging device identification signal. After filtering and baseline correction are performed on the wireless charging device identification signal in turn, a standardized feature vector is extracted and uploaded to a remote server. The standardized feature vector is identified by a trained wireless charging device identification classifier pre-stored in the remote server to obtain a corresponding device label, and the device label is verified. If the verification passes, the device label is correct and the wireless charging device is identified. If the verification fails, the device label is incorrect and the wireless charging device is identified.

2. A method for identifying a wireless charging device based on USB signal crosstalk according to claim 1, characterized in that The trained wireless charging device identification classifier pre-stored in the remote server is obtained through the following steps: Step 1: Equipment preparation and parameter configuration, specifically: S1-1. According to the need for wireless charging device identification, prepare m different types of wireless charging devices, n wireless chargers from different manufacturers, p USB power supplies with at least two USB ports from different brands and different architectures, a digital oscilloscope and a passive probe, where m is an integer greater than or equal to 2, and its value is equal to the total number of wireless charging devices that need to be identified, n is an integer greater than or equal to 2, the larger the value, the better, and p is an integer greater than or equal to 2, the larger the value, the better; A device tag is set for each of the m types of wireless charging devices, and the device tags of the m types of wireless charging devices are different; S1-2, each of the m types of wireless charging devices, the n types of wireless chargers and the p types of USB power supplies, a wireless charger and a USB power supply are combined as a wireless charging system to form s=m*n*p types of wireless charging system combinations, in which the wireless charger in each wireless charging system combination can charge the wireless charging device after being connected to a USB port of the USB power supply via a USB cable; Step 2: Build a standardized data set. The specific process is as follows: S2-1, set the sampling frequency of the digital oscilloscope to 2MHz, connect the passive probe to the digital oscilloscope, connect the wireless charger in each wireless charging system combination to a USB port of the USB power supply through a USB cable, and charge the wireless charging device; The signal at the USB port where the USB power source in each wireless charging system combination is connected to the wireless charger is called the signal of the wireless charging system combination. The signal of each wireless charging system combination is collected using a passive probe, where the time duration for the passive probe to collect the signal of the i-th wireless charging system combination is t i , t i >2 seconds, the time length collected by the passive probe is t i The signal of the i-th wireless charging system combination is recorded as signal sample P i , signal sample P i It is composed of multiple sampling values, each sampling value corresponds to a sampling point of the digital oscilloscope, i = 1, 2, ..., s, a total of s signal samples are collected, using P1 to P s Construct the original total data set P of signal samples, P = {P1,, P2, ..., P s }; S2-2, signal sample length standardization, the specific process is: Take the i-th signal sample P i The sampling value of the first two seconds, the number of sampling points in the first two seconds of sampling value is recorded as 4M, M = 1000000, 4M = 4000000, take the i-th signal sample P i The sampling values ​​of the first two seconds constitute the i-th signal sample length standardized data set A i , using s signal sample length standardized data sets to form a standardized data set A, A = {A1, A2, ...A s }; Step 3: Perform data preprocessing on the standardized data set A to obtain the data set FED. The specific process is as follows: S3-1. Use a first-order low-pass filter with a cutoff frequency of 0.1 MHz to perform low-pass filtering on the standardized data set A to remove high-frequency noise components and obtain a low-pass filtered data set, denoted as B, where B = {B1, B2, ..., B s }, where data set B i Normalize the dataset A for the i-th signal sample length i The data set obtained after low-pass filtering is data set B i The number of sampling points contained in data set B is also 4M, that is, i Contains 4M sample values; S3-2, use a moving average filter to smooth the data set B, and obtain a smoothed data set C, where C = {C1, C2, ..., C s }, C i For data set B i The data set is obtained by smoothing using a moving mean filter, where B i Use the moving mean filter to smooth the data set C i The specific process is: Set a sliding window with a size of 1000 sampling points and place the sliding window on data set B. i From the 1st sampling point to the 1000th sampling point, at this time, data set B i The first sampling point to the 1000th sampling point of is covered by the sliding window as data set B i The first segment of the data set B is then moved backwards to i Segmentation is performed, and the sliding window moves one sampling point each time. Each time, 1000 consecutive sampling points covered by the sliding window are considered as a segment until data set B i The 4Mth sampling point of data set B is covered. i The 1000 sampling points covered by the sliding window for the kth time are taken as its kth segment, k = 1, 2, ..., 3999001, and the data set B is calculated. i The average value of the 1000 sample values ​​in the kth segment is recorded as C i,k ; Using C i,1 To C i,3999001 Constructing the dataset C i , C i,k For dataset C i The kth sample value in; S3-3, using formula (2) to calculate C i，k Perform baseline correction: FAIRY i,k =C i,k -min(C i ) (2) Among them, FE i,k Indicates C i，k The sampling value obtained by baseline correction, min(C i ) is the data set C i The minimum value of all data in ; Using FE i,1 to FE i,3990001 Constructing the FE dataset i , FE i,k For the dataset FE i The kth data in; Using FE1 to FE s Construct a data set FE, FE = {FE1, FE2, ..., FE s }, FE i is the i-th data set in data set FE; S3-4, to FE i Perform equal length slicing. The specific process is as follows: for the i-th data set FE i , starting from the first sampling point, a division is made every 0.2M sampling points, and it is divided into 20 slice segments, each slice segment contains 0.2M sampling points. The data set FE i Contains 20 slice segments, the dataset FE i The xth slice in the dataset is denoted as FED i，x , x = 1, 2, ..., 20; FED i，x The yth sampling point is recorded as FED i，x [y], y=1, 2,..., n, n=0.2M; Adopting FED i，1 To FED i，20 Constructing the data set FED i , FED i，x For the dataset FED i The xth data set in ; Using FED1 to FED s Construct the data set FED, FED i is the i-th data set in the data set FED; Step 4: FED dataset i The xth dataset FED i，x Extract features and get the data set FED i，x The specific process is: S4-1. Calculate FED using formula (3) to formula (13) respectively i，x 11 time-domain features of: average value μ i,x , standard deviation σ i,x , root-mean-square amplitude RMS i，x , peak-to-peak value PPC i，x , skewness S i，x , Kurtosis K i，x , peak factor CF i，x , Clearance Factor CL i，x , shape factor Sh i，x , Pulse Index I i，x and coefficient of variation CV i，x : PPC i,x =max(FED i,x )-min(FED i,x ) (6) In the above formula, max is the maximum value symbol, max(FED i,x ) represents the data set FED i,x The maximum value in, min is the minimum value symbol, min(FED i,x ) represents the data set FED i,x The minimum value in ; S4-2. The data set FE is transformed by fast Fourier transform (FFT) to obtain a frequency domain data set of the data set FE. The frequency domain data set of the data set FE is recorded as FE′, where FE′={FE′1, FE′2, …, FE′ s }, where FE′ i Represents the dataset FE i The data set obtained by fast Fourier transform (FFT) is converted into FE i The xth slice fragment FED i，x The transformed data set is recorded as FED i，x ′, dataset FED i，x FE′ i The xth data set in FED i，x The yth data in ′ is recorded as FED i，x [y]′,FED i，x [y]′ is FED i，x [y] Data obtained by fast Fourier transform (FFT); Calculate FED using formula (14) to formula (25) respectively i，x ′’s 12 frequency domain features: mean μ i,x ′, variance σ i,x ' 2 , skewness S i,x ′, kurtosis K i,x ', root mean square amplitude RMS i,x ′、Peak i,x ′、Form factor WF i,x ′、peak factor CF i,x ′、ClF i,x ′、Pulse index IF i,x ′, first-order moment M i,x ′ and frequency center SC i,x ′: Peak i,x ′=max(FED i,x ′) (19) Where max(FED i,x ′) represents FED i，x The maximum value of ′; f s is the sampling frequency of the digital oscilloscope; S4-3. Fed i，x 11 time domain features and FED i，x The 12 frequency domain features of ′ are used as 23 feature elements to form a 23-dimensional vector, which is used as FED i，x The eigenvector of i，x , the feature vector F i，x The rth characteristic element of i，x，r , r=1,2,…,23; Using the eigenvector F i，1 To F i，20 Construct the feature vector set F i , F i ={F i，1 , F i，2 , …, F i，20 }; Using the feature vector set F1 to F s The total set of feature vectors is denoted as F, where F = {F1, F2, ..., F s }; S4-4. Perform Z-score normalization on each feature vector in the total feature vector set F to obtain the corresponding average value, where formula (26) is used to normalize F i，x Perform Z-score standardization: in, is the eigenvector F i，x The average value obtained by Z-score standardization; Formula (27) is used to calculate F i，x，r The corresponding Z-score value is: In formula (27), z i，x，r Indicates F i，x，r The Z-score value obtained after Z-score standardization; Using Z i，x，1 To Z i，x，23 Construct the feature vector F i，x The normalized eigenvector Z i，x , Z i，x ={z i，x，1 , z i，x，2 , …, z i，x，23 }; Use the standardized eigenvector Z i，1 To Z i，20 Construct a standardized feature vector set Z i , Z i = {Z i，1 , Z i，2 , …, Z i，20 }; Use the standardized feature vector set Z1 to Z s The total set of standardized feature vectors is denoted as Z, where Z = {Z1, Z2, ..., Z s }; Step 5: Use the device labels of the wireless charging devices in the i-th wireless charging system combination as the standardized feature vector Z i，x The device label is the normalized feature vector Z i，x Its device label constitutes a standardized feature vector sample, and the standardized feature vector set Z1 to Z s All the standardized feature vectors in are constructed to obtain 20s standardized feature vector samples, and the 20s standardized feature vector samples are used to construct the training data set; Step 6: Integrate the random forest, support vector machine, K nearest neighbor classifier and voting ensemble learning algorithm to form an integrated classification model, wherein the integrated random forest, support vector machine and K nearest neighbor classifier are all used to identify the standardized feature vectors input therein, obtain the corresponding device labels and output them to the voting ensemble learning algorithm, and the voting ensemble learning algorithm is used to process the device labels outputted thereto by the integrated random forest, support vector machine and K nearest neighbor classifier in a voting manner to obtain the final device label output; Step 7: Use the training data set to train the integrated classification model to obtain a trained integrated classification model, where the trained integrated classification model is the trained wireless charging device identification classifier pre-stored in the remote server.

3. A method for identifying a wireless charging device based on USB signal crosstalk according to claim 2, characterized in that When a wireless charging device is charged in a wireless charging station equipped with a voltage sampling device with a sampling frequency set to 2 MHz, the wireless charging device is connected to a USB port of a USB power supply through a wireless charger. At this time, the same method as steps 2 to 4 is used to obtain a standardized feature vector of the wireless charging device, and the standardized feature vector is remotely transmitted to a pre-stored trained wireless charging device identification classifier in a remote server. The trained wireless charging device identification classifier identifies the standardized feature vector to obtain a device label.

4. A method for identifying a wireless charging device based on USB signal crosstalk according to claim 3, characterized in that After the trained wireless charging device identification classifier obtains the device label, the following method is used for verification: obtain the confidence values ​​of the three device labels output by the integrated random forest, support vector machine and K nearest neighbor classifier to the Voting ensemble learning algorithm. If the minimum value of the confidence values ​​of the three device labels is greater than 0.5, the verification passes at this time. If the maximum value of the confidence values ​​of the three device labels is less than 0.5, the verification fails.

5. A method for identifying a wireless charging device based on USB signal crosstalk according to claim 4, characterized in that When the verification fails, it means that the device label of the wireless charging device does not exist in the trained wireless charging device identification classifier. At this time, a device label is set for the wireless charging device, and the device label is different from the device label that already exists in the currently trained wireless charging device identification classifier. The standardized feature vector obtained by the wireless charging device and its corresponding device label constitute a new standardized feature vector sample, and add it to the current training data set, update the current training data set, and then use the updated current training data set to train and update the currently trained wireless charging device identification classifier again, so that the trained wireless charging device identification classifier has the ability to identify the wireless charging device.

Images