Abnormality detection method and device, test equipment, storage medium and program product

Abstract

The invention discloses an anomaly detection method and device, test equipment, a storage medium and a program product, and the method comprises the steps: obtaining a to-be-detected radio frequency signal which is a signal received by a radio frequency chip; preprocessing the radio frequency signal, and determining model input data; inputting the model input data into a preset model, and determining a detection result of the radio frequency signal through the preset model; wherein the detection result is used for indicating whether the radio frequency signal is abnormal due to peripheral factors of the radio frequency chip or not. Therefore, by adopting a machine learning mode based on the preset model, whether abnormity caused by peripheral factors of the radio frequency chip exists or not can be identified, and the accuracy of signal abnormity detection is improved; manual intervention is reduced, and the testing efficiency can be improved; in addition, by introducing a preset model of machine learning, the method can adapt to different abnormal scenes, and robustness is enhanced.

Description

Technical Field

[0001] The present application relates to the field of wireless communication technology, and in particular to an anomaly detection method, device, test equipment, storage medium and program product. Background Art

[0002] RF signal anomaly handling is an important issue in wireless communication systems, which involves detecting, identifying and repairing RF signal anomalies to ensure normal communication and data transmission. In modern wireless communication systems, RF signals may become abnormal due to various reasons, such as noise, interference, multipath effects, etc., resulting in a decrease in communication quality or even interruption.

[0003] In the related art, in order to determine whether there is an abnormality in the RF signal, it is usually achieved by setting a threshold for the test index. However, this threshold setting method still has some defects, such as the inability to accurately distinguish between abnormalities of the chip itself and abnormalities caused by external factors (such as communication interface status problems), and the need to manually adjust the threshold value, which makes the test efficiency low and also reduces the accuracy of abnormality detection. Summary of the invention

[0004] The present application proposes an anomaly detection method, apparatus, test equipment, storage medium and program product.

[0005] The technical solution of this application is implemented as follows:

[0006] In a first aspect, an embodiment of the present application provides an anomaly detection method, the method comprising:

[0007] Acquire a radio frequency signal to be detected, where the radio frequency signal is a signal received by a radio frequency chip;

[0008] Preprocess the RF signal to determine the model input data;

[0009] The model input data is input into a preset model, and the detection result of the RF signal is determined by the preset model; wherein the detection result is used to indicate whether the RF signal has an anomaly caused by peripheral factors of the RF chip.

[0010] In a second aspect, an embodiment of the present application provides an anomaly detection device, the anomaly detection device comprising an acquisition unit, a preprocessing unit and an anomaly detection unit, wherein:

[0011] An acquisition unit, configured to acquire a radio frequency signal to be detected, where the radio frequency signal is a signal received by the radio frequency chip;

[0012] A preprocessing unit configured to preprocess the radio frequency signal and determine model input data;

[0013] The anomaly detection unit is configured to input the model input data into a preset model, and determine the detection result of the radio frequency signal through the preset model; wherein the detection result is used to indicate whether the radio frequency signal has an anomaly caused by peripheral factors of the radio frequency chip.

[0014] In a third aspect, an embodiment of the present application provides a testing device, the testing device comprising a memory and a processor, wherein:

[0015] A memory for storing computer programs that can be run on the processor;

[0016] A processor is used to execute the anomaly detection method as described in the first aspect when running a computer program.

[0017] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the anomaly detection method as described in the first aspect is implemented.

[0018] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program or instructions, which, when executed by a processor, implements the anomaly detection method as described in the first aspect.

[0019] The embodiments of the present application provide an abnormality detection method, device, test equipment, storage medium and program product, which obtain a radio frequency signal to be detected, where the radio frequency signal is a signal received by a radio frequency chip; pre-process the radio frequency signal to determine the model input data; input the model input data into a preset model, and determine the detection result of the radio frequency signal through the preset model; wherein the detection result is used to indicate whether the radio frequency signal has an abnormality caused by factors outside the radio frequency chip. In this way, for the radio frequency signal to be detected, a machine learning method based on a preset model is adopted, which can not only identify whether the radio frequency signal has an abnormality caused by factors outside the radio frequency chip, so as to exclude the abnormality caused by factors outside the radio frequency chip and improve the accuracy of signal abnormality detection; and in the detection process, it is no longer necessary to manually adjust the threshold value, which reduces manual intervention and can also improve test efficiency; in addition, by introducing a preset model of machine learning, it can also adapt to different abnormal scenarios and enhance robustness.

[0020] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and are not intended to limit the technical solutions of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is an illustration of the abnormal spectrum of the 204B protocol synchronization signal. Figure 1 ;

[0022] Figure 2 This is an illustration of the abnormal spectrum of the 204B protocol synchronization signal. Figure 2 ;

[0023] Figure 3 This is an illustration of the abnormal spectrum of the 204B protocol synchronization signal. Figure 3 ;

[0024] Figure 4 This is a schematic diagram of the abnormal spectrum of the chip phase-locked loop losing lock;

[0025] Figure 5 It is a schematic diagram of abnormal spectrum of in-phase and quadrature signals;

[0026] Figure 6 A schematic diagram of a method for detecting anomalies provided in an embodiment of the present application Figure 1 ;

[0027] Figure 7 A schematic diagram of a method for detecting anomalies provided in an embodiment of the present application Figure 2 ;

[0028] Figure 8 A schematic diagram of the structure of a preset model provided in an embodiment of the present application;

[0029] Fig. 9 A schematic diagram of a method for detecting anomalies provided in an embodiment of the present application Figure 3 ;

[0030] Fig.10 A schematic diagram of a process for constructing a preset model provided in an embodiment of the present application;

[0031] Fig.11 A schematic diagram of an application framework of an anomaly detection method provided in an embodiment of the present application;

[0032] Fig.12 A schematic diagram of the structure of an abnormality detection device provided in an embodiment of the present application Figure 1 ;

[0033] Fig.13 A schematic diagram of the structure of an abnormality detection device provided in an embodiment of the present application Figure 2 ;

[0034] Fig.14 A schematic diagram of the hardware structure of a test device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0035] In order to enable a more detailed understanding of the features and technical contents of the embodiments of the present application, the implementation of the embodiments of the present application is described in detail below in conjunction with the accompanying drawings. The attached drawings are for reference only and are not used to limit the embodiments of the present application.

[0036] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0037] It should also be noted that the terms "first\second\third" involved in the embodiments of the present application are only used to distinguish similar objects and do not represent a specific ordering of the objects. It can be understood that "first\second\third" can be interchanged in a specific order or sequence where permitted, so that the embodiments of the present application described here can be implemented in an order other than that illustrated or described here.

[0038] It should also be noted that in the following description, reference is made to “some embodiments”, which describe a subset of all possible embodiments, but it can be understood that “some embodiments” may be the same subset or different subsets of all possible embodiments, and may be combined with each other without conflict.

[0039] To facilitate understanding of the technical solutions of the embodiments of the present application, the relevant technologies of the embodiments of the present application are described below. The following related technologies can be arbitrarily combined with the technical solutions of the embodiments of the present application as optional solutions, and they all belong to the protection scope of the embodiments of the present application.

[0040] For the RF signal received by the RF chip, the abnormal processing of the RF signal is an important issue in the wireless communication system, which involves detecting, identifying and repairing the abnormality of the RF signal to ensure normal communication and data transmission. In modern wireless communication systems, due to various reasons, such as noise, interference, multipath effects, etc., the RF signal may be abnormal, resulting in a decrease in communication quality or even interruption. Therefore, the embodiment of the present application urgently needs to develop an effective RF signal abnormality processing method, which is crucial to improve the reliability and performance of the communication system.

[0041] In the chip performance test, some scenarios need to combine the spectrum calculated by Fast Fourier Transform (FFT) to determine whether the working state of the RF chip is abnormal. In the automated test system, the working state of the RF chip is usually determined by setting thresholds for the test indicators. However, for many test scenarios, this threshold setting method is not suitable for judging abnormal signals. The reason is that some abnormalities may be caused by peripheral factors other than the chip itself. The chip test hopes to intercept the problems of the chip itself and exclude abnormalities caused by other peripheral factors. Therefore, it is necessary to identify the abnormalities caused by peripheral factors through the automated test system.

[0042] For example, Figure 1 to Figure 5It is the data spectrum analysis and processing during the test process. All these spectra are obtained under abnormal signals. Among them, Figure 1 This is an illustration of the abnormal spectrum of the 204B protocol synchronization signal (sync) Figure 1 , Figure 2 This is an illustration of the abnormal spectrum of the 204B protocol synchronization signal (sync) Figure 2 , Figure 3 This is an illustration of the abnormal spectrum of the 204B protocol synchronization signal (sync) Figure 3 , Figure 4 This is a schematic diagram of the abnormal spectrum of the chip phase locked loop (PLL) losing lock. Figure 5 Figure 1 is a schematic diagram of the abnormal spectrum of the in-phase (I) and quadrature (Q) signals. Figure 1 to Figure 5 In the graph, the horizontal axis (abscissa) represents frequency, and the vertical axis (ordinate) represents amplitude.

[0043] Here, some anomalies are caused by abnormal peripheral factors of the RF chip. Among them, the peripheral factors may include the unstable sync state of the peripheral communication interface (Serdes), which will cause the error spectrum to be analyzed in the frequency domain, and the phenomenon of error spectrum is not single, such as Figure 1 to Figure 3 As shown in the figure, some spectra show that both the signal and the background noise are abnormal, some show that the signal is normal but the background noise is abnormal, and some show that the background noise is oscillating. In addition, some abnormalities are caused by the working status of the chip itself, such as Figure 4-5 As shown, this will also cause the noise floor to rise.

[0044] To sum up the above situation, in the automated testing system, if you want to retain the anomalies of the chip itself and intercept the anomalies caused by peripheral factors such as the communication interface, the method of simply setting a threshold to determine whether the chip working status is abnormal is unreasonable. Therefore, it is necessary to provide an optimized and improved anomaly detection method to eliminate the anomalies caused by peripheral factors such as the communication interface.

[0045] In recent years, with the rapid development of neural networks and the rise of deep learning, more and more researchers have begun to apply them to the field of anomaly processing of RF signals. Neural networks have powerful nonlinear modeling and adaptive learning capabilities, and can automatically learn features and patterns from a large amount of training data and use them to detect and identify signal anomalies. Compared with rule-based methods (such as setting thresholds), anomaly detection methods based on neural networks have better adaptability and robustness.

[0046] Based on this, in order to improve the test efficiency and reduce the test cost, a new technical solution is proposed for the current chip performance test. Specifically, a method for detecting anomalies of RF signals based on neural networks is proposed. The core idea of ​​this anomaly detection method is to extract features and detect anomalies of RF signals by constructing an appropriate neural network architecture. First, the RF signal to be detected is obtained, and the RF signal is preprocessed, including denoising, filtering and other operations to reduce the influence of noise and interference. Then, the model input data obtained after processing is input into the preset model, and the model input data is nonlinearly transformed and feature extracted through the combination of multiple hidden layers and activation functions. Finally, the detection result of the RF signal is determined by the judgment of the output layer to realize the anomaly detection of the signal.

[0047] In this way, for the RF signal to be detected, the machine learning method based on the preset model is used, which can not only identify whether there are anomalies caused by peripheral factors of the RF chip, so as to eliminate the anomalies caused by peripheral factors of the RF chip and improve the accuracy of signal anomaly detection; but also during the detection process, there is no need to manually adjust the threshold value, which reduces manual intervention and improves test efficiency; in addition, by introducing the preset model of machine learning, it can also adapt to different abnormal scenarios and enhance robustness.

[0048] The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0049] In one embodiment of the present application, Figure 6 A schematic diagram of a method for detecting anomalies provided in an embodiment of the present application Figure 1 .like Figure 6 As shown, the method includes:

[0050] S601: Acquire a radio frequency signal to be detected.

[0051] In an embodiment of the present application, the anomaly detection method can be applied to an anomaly detection device, or a test device integrated with the device. For example, taking an automated test system as an example, the automated test system may include a test device and a radio frequency chip under test, and the test device can perform a performance test on the radio frequency chip. When the radio frequency chip is working, the signal received by the radio frequency chip is collected and used as the radio frequency signal to be detected. In other words, the radio frequency signal here is the signal received by the radio frequency chip.

[0052] In the embodiments of the present application, radio frequency (RF) refers to radio frequency. Generally speaking, radio frequency refers to high-frequency electromagnetic waves in the frequency range of 300kHz to 300GHz. Electromagnetic waves with a frequency lower than 100kHz will be absorbed by the surface and cannot form effective transmission. Electromagnetic waves with a frequency higher than 100kHz can propagate in the air and be reflected by the ionosphere at the outer edge of the atmosphere, and have long-distance transmission capabilities. Such high-frequency electromagnetic waves with long-distance transmission capabilities can be called radio frequency signals.

[0053] For example, in order to transmit television signals over the air, the video full television signal must be modulated into a high frequency or radio frequency signal, with each signal occupying a channel, so that the television signal can transmit multiple television programs over the air at the same time without causing confusion.

[0054] S602: Preprocess the radio frequency signal to determine model input data.

[0055] In the embodiment of the present application, the radio frequency signal can be represented by x[n]. Considering that the anomaly detection here is usually performed on the spectrum data corresponding to the radio frequency signal, in some embodiments, see Figure 7 , for step S602, the step may include:

[0056] S602-1, perform Fourier transform on the radio frequency signal to obtain spectrum data.

[0057] In the embodiments of the present application, the Fourier transform here mainly refers to the discrete Fourier transform, by which the signal can be transformed from the time domain to the frequency domain, and then the spectral structure and change law of the signal can be studied. In practical applications, the fast Fourier transform (Fast Fourier Transform, FFT) is usually used to efficiently calculate the discrete Fourier transform.

[0058] For example, assuming that the RF signal is x[n], its corresponding spectrum data is X[k], n represents the time index, and k represents the frequency index. The calculation formula of FFT is as follows:

[0059]

[0060] Where N is the number of sampling points, e is the base of the natural logarithm, 0≤k <N。

[0061] S602-2, preprocess the spectrum data to obtain model input data.

[0062] In the embodiment of the present application, after the spectrum data is obtained, data preprocessing may be performed on the spectrum data, including operations such as denoising and filtering, to reduce the influence of noise and interference.

[0063] In some embodiments, performing data preprocessing on the spectrum data to obtain model input data may include: performing filtering on the spectrum data to obtain filtered data; and performing normalization on the filtered data to obtain model input data.

[0064] In the embodiment of the present application, filtering is an operation to filter out a specific frequency band in the radio frequency signal, which is an important measure to suppress and prevent interference. Among them, filtering can refer to low-pass filtering, high-pass filtering, band-pass filtering, band-stop filtering, Wiener filtering, Kalman filtering, Gaussian filtering and other processing.

[0065] Exemplarily, the filtering process here may be Gaussian filtering. Gaussian filtering is a linear smoothing filter suitable for eliminating Gaussian noise. The formula of Gaussian filtering is as follows:

[0066]

[0067] Where G[nk] represents the value of the Gaussian distribution at point nk. Due to the infinite length of the Gaussian distribution, a truncated Gaussian filter is usually used in actual calculations, and only values ​​within a limited range are considered here.

[0068] It is understandable that after obtaining the filtered data, the filtered data may be normalized. In some embodiments, the method may include: performing normalization and / or regularization on the filtered data to obtain model input data.

[0069] It should be noted that in the embodiments of the present application, regularization and standardization are two different data processing methods, each of which has different purposes and application scenarios. Among them, standardization is to scale the data in proportion so that it falls into a small specific interval, usually to facilitate the next step of data processing. Standardization is not for the convenience of processing or comparison with other data, but to make the data more conducive to processing using the properties of the standard normal distribution. For example, Z-Score standardization (zero-mean standardization) is a common standardization method, which makes the data more consistent with the standard normal distribution by subtracting the mean and dividing it by the standard deviation. Regularization mainly solves ill-posed problems, and these constraints can be interpreted as prior knowledge by imposing constraints on the minimized empirical error function. Regularization effectively reduces the phenomenon of overfitting by introducing a regularization factor and increasing the role of guiding constraints. The use of regularization in logistic regression is a typical example, which uses prior knowledge to introduce a regularization factor in the processing process to eliminate the singularity of ill-posed problems and improve the generalization ability of the model.

[0070] That is to say, in the embodiments of the present application, regularization mainly prevents overfitting and improves the generalization ability of the model by introducing constraints; while standardization optimizes the distribution of data by scaling the data to make it more consistent with the assumptions of certain mathematical or statistical models.

[0071] In one possible implementation, the filtered data may be standardized, and the standardization here may refer to Z-Score standardization. Using Z-Score standardization, the difference between each sample and the mean of the overall sample is calculated and divided by the standard deviation of the overall sample. The calculation formula of Z-Score standardization is as follows:

[0072]

[0073] Among them, x represents the sample (i.e., filtered data), μ represents the mean of the entire sample, and σ represents the standard deviation of the entire sample.

[0074] In another possible implementation, the filtered data may be regularized, where the regularization may be mapping the data into a preset value range, for example, scaling the data to between [0, 1]. The calculation formula for the regularization is as follows:

[0075]

[0076] Wherein, x represents the sample (ie, filtered data), min(x) represents the minimum value of the sample, and max(x) represents the maximum value of the sample.

[0077] In another possible implementation, the filtered data may be subjected to Z-Score standardization and regularization at the same time. After Gaussian filtering, that is, after the filtered data is obtained, the filtered data may be subjected to Z-Score standardization first, and then regularization; or, the filtered data may be subjected to regularization first, and then Z-Score standardization. Exemplarily, in the embodiment of the present application, the filtered data may be subjected to Z-Score standardization first, and then regularization, but this is not limited to any limitation.

[0078] In an embodiment of the present application, Z-Score standardization can convert the filtered data into a distribution with zero mean and unit variance, so that different features are comparable and the influence of feature scale is eliminated. This is very important for the effectiveness and stability of many machine learning algorithms such as clustering and regression models. Regularization can map the filtered data to a preset numerical range, usually between [0,1]. It retains the relative relationship of the original data by scaling the data, while limiting the data to a preset numerical range. Regularization is often used in situations where the weights of each feature need to be considered proportionally, and it is more robust to outliers; at the same time, it can avoid model bias problems caused by different numerical ranges of different features.

[0079] In this way, for the spectrum data corresponding to the radio frequency signal, the spectrum data can be Gaussian filtered to obtain filtered data; then the filtered data is normalized by standardization and / or regularization, so as to obtain model input data.

[0080] S603, inputting the model input data into a preset model, and determining the detection result of the RF signal through the preset model; wherein the detection result is used to indicate whether the RF signal has an anomaly caused by peripheral factors of the RF chip.

[0081] In a possible embodiment, the preset model may refer to a binary classification model based on a neural network. A neural network is a computational model that can be composed of a large number of nodes (or neurons) connected to each other. In a neural network, weight (weight, w) and bias (bias, b) are two core concepts that play a vital role in the training and learning process of a neural network.

[0082] For neural networks, weights are used to measure the importance of input signals. Each neuron has several connections with neurons in the previous layer, and each connection has its own weight. The existence of weights allows neural networks to learn complex patterns in data, especially those patterns that have different degrees of importance in different dimensions. The initial value of the weight is usually set randomly and then gradually adjusted during the training process. Bias can be regarded as the "bias" of the neural network. It provides an additional degree of freedom for the activation function of the neuron, allowing the neuron to respond before the input signal reaches a certain threshold. The bias term can be regarded as a kind of "pre-filter" for the input signal, which can eliminate noise to a certain extent and improve the robustness of the neural network.

[0083] In the embodiment of the present application, the preset model may include at least one linear layer and at least one activation layer. The activation layer may include an activation function, and the activation function may be a linear rectification function or a rectified linear unit (ReLU), which generally refers to a nonlinear function represented by a ramp function and its variants.

[0084] For example, Figure 8 A schematic diagram of the composition structure of a preset model provided in an embodiment of the present application. The preset model may include five linear layers and five activation layers, each activation layer is located after the corresponding linear layer; and finally outputs a model with a binary classification result. Figure 8 As shown, the five linear layers include: the first linear layer, the second linear layer, the third linear layer, the fourth linear layer and the fifth linear layer, and the five activation layers include: the first activation layer, the second activation layer, the third activation layer, the fourth activation layer and the fifth activation layer. Among them, the weight of the first linear layer is 4000*8192, and the bias is 4000; the weight of the second linear layer is 2000*4000, and the bias is 2000; the weight of the third linear layer is 500*2000, and the bias is 500; the weight of the fourth linear layer is 150*500, and the bias is 150; the weight of the fifth linear layer is 50*150, and the bias is 50. In addition, the first activation layer, the second activation layer, the third activation layer, the fourth activation layer and the fifth activation layer all use the ReLU activation function.

[0085] It can also be understood that in the embodiment of the present application, this two-classification model can be used to solve the problem of abnormality detection, especially for whether the IQ spectrum data in the RF signal is caused by the external factors of the RF chip, such as the situation caused by the abnormal state of the peripheral communication interface (Serdes). Because in the training process of the preset model, it can learn the correlation and weight between different features, and use this information to classify the newly collected RF signals, which can be classified according to the probability value output by the preset model.

[0086] In some embodiments, a preset model may be used to perform anomaly detection on model input data.

[0087] See also Fig. 9 , for step S603, the step may include:

[0088] S603-1, input the model input data into the preset model, and obtain the probability value output by the preset model.

[0089] S603-2: When the probability value is within the first preset range, determine that the detection result of the radio frequency signal indicates that the radio frequency signal has an abnormality caused by peripheral factors of the radio frequency chip.

[0090] S603-3: When the probability value is within the second preset range, determine that the detection result of the radio frequency signal indicates that there is no abnormality in the radio frequency signal caused by peripheral factors of the radio frequency chip.

[0091] In an embodiment of the present application, the model input data is input into a preset model, and a probability value is output through the preset model, which characterizes the probability that the model input data has an anomaly caused by peripheral factors of the RF chip; then, based on the probability value, the detection result of the RF signal is determined. Among them, the model input data is obtained based on the preprocessing of the RF signal, so the probability value can also be said to characterize the probability that the RF signal has an anomaly caused by peripheral factors of the RF chip. According to the size of the probability value, it can be determined whether the RF signal has an anomaly caused by peripheral factors of the RF chip. Specifically, it can be: when the probability value is in the first preset range, it is determined that the RF signal has an anomaly caused by peripheral factors of the RF chip; when the probability value is in the second preset range, it is determined that the RF signal does not have an anomaly caused by peripheral factors of the RF chip.

[0092] It should also be noted that in the embodiment of the present application, the first preset range is different from the second preset range. The value in the first preset range is close to 0, and the value in the second preset range is close to 1.

[0093] Exemplarily, the first preset range may be a range near 0, such as a range between 0 and 0.2; the second preset range may be a range near 1, such as a range between 0.8 and 1.

[0094] In some embodiments, the method may further include: when the detection result of the RF signal indicates that the RF signal has an abnormality caused by factors peripheral to the RF chip, reacquiring the RF signal to be detected; or, when the detection result of the RF signal indicates that the RF signal does not have an abnormality caused by factors peripheral to the RF chip, transmitting data based on the RF signal.

[0095] Thus, in the embodiment of the present application, for the RF signal to be detected, after the RF signal is processed to obtain the model input data, the model input data can be input into the preset model, and the probability value output by the model can be used to determine whether there is an abnormality caused by peripheral factors (such as the state of the peripheral communication interface). If the output probability value is within the first preset range, that is, the probability value is close to 0, it indicates that the RF signal has an abnormality caused by the peripheral factors of the RF chip, which is considered to be an abnormal situation (or "abnormal situation"), and it can be recommended to re-acquire a new RF signal, that is, resample the data; if the output probability value is within the second preset range, that is, the probability value is close to 1, it indicates that the RF signal does not have an abnormality caused by the peripheral factors of the RF chip, which is considered to be a normal situation, and data processing can continue, such as data transmission based on the RF signal, so as to complete the transmission and reception of relevant data.

[0096] It should also be noted that in the embodiment of the present application, after the RF signal is detected for abnormality, a corresponding repair operation can also be performed. For example, if the output probability value is in the first preset range, that is, the probability value is close to 0, there may be an abnormality caused by peripheral factors. At this time, the chip periphery can be repaired. For example, for a peripheral communication interface with abnormalities such as looseness or poor contact, the connection can be re-fixed so that the subsequent collected RF signal no longer has abnormalities caused by peripheral factors.

[0097] An abnormality detection method provided in an embodiment of the present application obtains a radio frequency signal to be detected; pre-processes the radio frequency signal to determine the model input data; inputs the model input data into a preset model, and determines the detection result of the radio frequency signal through the preset model; wherein the detection result is used to indicate whether the radio frequency signal has an abnormality caused by factors external to the radio frequency chip. In this way, by adopting a machine learning method based on a preset model, it is not only possible to identify whether the radio frequency signal has an abnormality caused by factors external to the radio frequency chip, so as to exclude the abnormality caused by factors external to the radio frequency chip and improve the accuracy of signal abnormality detection; and in the detection process, it is no longer necessary to manually adjust the threshold value, which reduces manual intervention and can also improve test efficiency; in addition, by introducing a preset model of machine learning, it can also adapt to different abnormal scenarios and enhance robustness.

[0098] In another embodiment of the present application, based on the anomaly detection method described in the above embodiment, Fig.10 A schematic diagram of a process for constructing a preset model provided in an embodiment of the present application. Fig.10 As shown, the method may include:

[0099] S1001, obtaining a training sample set, where the training sample set includes multiple groups of training data.

[0100] It should be noted that in the embodiment of the present application, firstly, a lot of sampling data can be collected. These sampling data can be obtained by Fourier transforming multiple sample RF signals respectively, that is, the sampling data is discrete spectrum data, such as IQ spectrum data.

[0101] It should also be noted that in the embodiments of the present application, for model training, it is usually necessary to evaluate the model performance. Therefore, the acquired many sampled data can be divided into training sample sets and test sample sets. Exemplarily, it can be divided into training sample sets and test sample sets in a ratio of 7:3. Among them, the training sample set may include multiple groups of training data, and the test sample set may include multiple groups of test data; the former is used for model training, and the latter is used for model evaluation.

[0102] S1002, performing data preprocessing on the multiple sets of training data respectively, determining multiple sets of sample data, and adding corresponding sample labels to the multiple sets of sample data; wherein the sample labels are used to indicate whether the sample data has anomalies caused by peripheral factors of the RF chip.

[0103] It should be noted that in the embodiments of the present application, data preprocessing is required for each of the multiple sets of training data, including denoising, filtering and other operations, to reduce the impact of noise and interference. In some embodiments, data preprocessing is performed on each of the multiple sets of training data, which may include: filtering each of the multiple sets of training data to obtain filtered data corresponding to each of the multiple sets of training data; and then normalizing the filtered data corresponding to each of the multiple sets of training data to obtain multiple sets of sample data.

[0104] It should also be noted that, in the embodiment of the present application, the filtering process here may be Gaussian filtering, and the normalization process here may include standardization and / or regularization processing. Exemplarily, Gaussian filtering is performed on multiple sets of training data to obtain filtered data corresponding to each of the multiple sets of training data; then, the filtered data corresponding to each of the multiple sets of training data is normalized by Z-Score standardization and / or regularization to obtain multiple sets of sample data.

[0105] It should also be noted that in the embodiment of the present application, data labeling is required for multiple groups of sample data. That is to say, corresponding sample labels are added to multiple groups of sample data, and the sample labels are used to indicate whether the sample data has anomalies caused by peripheral factors of the RF chip. Exemplarily, the sample labels can be used to determine which of these sample data are abnormal samples (for example, there are anomalies caused by peripheral factors of the RF chip) and which are normal samples (for example, there are no anomalies caused by peripheral factors of the RF chip).

[0106] S1003, performing model training on the neural network model according to the multiple groups of sample data and the corresponding sample labels, and when the model training meets the preset convergence condition, determining the trained neural network model as the preset model.

[0107] It should be noted that in the embodiment of the present application, a suitable neural network model is constructed, and then the neural network model is trained according to multiple groups of sample data and corresponding sample labels to obtain a trained preset model.

[0108] It should also be noted that, in an embodiment of the present application, the structure of the neural network model may include at least one linear layer and at least one activation layer, exemplarily including five linear layers and five activation layers. During the training process, the labeled sample data is used to optimize the model parameters so that it can accurately determine whether the sampled data has anomalies caused by peripheral factors of the RF chip, such as those related to the state of the peripheral communication interface. In addition, by selecting an appropriate loss function, such as a binary cross entropy loss function, the prediction error of the model can be minimized. Subsequently, a suitable optimization algorithm, such as a gradient descent method, is used to optimize and adjust the model parameters (weights and biases).

[0109] Specifically, during the model training process, the goal of weight adjustment is to enable the neural network model to produce correct output when faced with training data. This process is achieved by comparing the error between the actual output of the network and the expected output, and then this error is used to update the weights of the neural network model to reduce the error next time it faces the same input. This process is repeated until the output of the neural network model reaches a satisfactory accuracy, or the weights of the neural network model have converged. The adjustment of weights and biases plays a key role in model training. By updating the weights, the neural network model can learn to recognize important features and patterns from the input, while the bias helps the network maintain stability and robustness when processing complex data. For example, with Figure 8 For example, after model training, for the preset model, the weight of the first linear layer is 4000*8192, and the bias is 4000; the weight of the second linear layer is 2000*4000, and the bias is 2000; the weight of the third linear layer is 500*2000, and the bias is 500; the weight of the fourth linear layer is 150*500, and the bias is 150; the weight of the fifth linear layer is 50*150, and the bias is 50; thereby improving the model performance.

[0110] It should also be noted that in the embodiment of the present application, the model training meets the preset convergence condition, which may include: the number of model trainings meets the preset number of times, or the loss function meets the preset condition. If it is determined that the model training meets the preset convergence condition, the model training of the neural network model can be stopped.

[0111] That is to say, in the embodiments of the present application, whether the model training is completed can be measured, for example, by the number of training times. For example, if the number of model training times (or the number of adjustments of model parameters) meets the preset number of times, it indicates that the parameters of the neural network model are adjusted in place and optimal, and the model training of the neural network model is stopped at this time; or it can be measured by the size of the loss function. For example, if the loss function meets the preset conditions, that is, the output value of the loss function is less than the preset threshold, it indicates that the parameters of the neural network model are adjusted in place and optimal, and the model training of the neural network model is also stopped at this time.

[0112] In some embodiments, after obtaining the preset model, the method may further include: obtaining a test sample set, the test sample set including multiple groups of test data; performing model evaluation on the preset model using the multiple groups of test data to determine the model evaluation result; wherein the model evaluation result is used to indicate the anomaly detection capability of the preset model.

[0113] In an embodiment of the present application, a test sample set can also be used to evaluate the performance of a preset model, such as calculating the accuracy, precision, recall rate and other indicators of the preset model on the test sample set to evaluate the anomaly detection capability of the preset model, including its classification capability and robustness.

[0114] Furthermore, after obtaining the model evaluation result, if the model evaluation result is lower than the preset evaluation value, it indicates that the performance of the preset model is not ideal, and the training sample set can continue to be used for model training or model parameter adjustment; if the model evaluation result is greater than or equal to the preset evaluation value, it indicates that the performance of the preset model meets the requirements, and the preset model can be used to perform anomaly detection on the RF signal to be detected.

[0115] It should also be noted that in the embodiments of the present application, the training sample set and the test sample set are independent of each other. The training sample set is used to train the model, while the test sample set is used to evaluate the final detection performance and generalization ability of the model; in addition, the training sample set is used in the model training stage, while the test sample set is used after the model training is completed.

[0116] Specifically, the training sample set is a data set used to train the model. The model learns features and rules by observing and learning sample data in the training set, thereby establishing a neural network model; and during the training process, the model will continuously update parameters based on the data in the training sample set until the model performs at a satisfactory level on the training sample set. The test sample set is a data set used to evaluate the final detection performance and generalization ability of the model. It is a data set used to simulate the performance of the model in real scenarios. After the model training and verification are completed, the test sample set is used to evaluate the overall performance of the model, including indicators such as accuracy, precision, and recall. The evaluation results of the test sample set can provide an important reference for the online application of the model.

[0117] In some embodiments, the method may further include: adjusting some parameters in the preset model according to the detection results of the RF signal and / or the model adjustment requirements; wherein the some parameters include at least one of the following: the network structure of the preset model and the training sample set of the preset model.

[0118] It should be noted that in the embodiments of the present application, the preset model can also be continuously optimized. Among them, some parameters in the preset model can be adjusted according to the detection results of the RF signal and / or the model adjustment requirements, for example, according to the feedback and improvement requirements in the actual application, so as to continuously optimize the performance and accuracy of the preset model. Exemplarily, the adjustment here can include adjusting the network structure of the preset model, updating the training sample set, etc., which is not specifically limited here.

[0119] The embodiment of the present application provides an anomaly detection method, which can be to use a training sample set to train a neural network model to obtain a preset model after training; then use a test sample set to evaluate the preset model to obtain a model evaluation result. Then, the trained and evaluated preset model is used to perform anomaly detection on the RF signal to be detected, which can not only identify whether the RF signal has anomalies caused by peripheral factors of the RF chip, so as to eliminate anomalies caused by peripheral factors of the RF chip and improve the accuracy of signal anomaly detection; and in the detection process, it is no longer necessary to manually adjust the threshold value, reduce manual intervention, and improve test efficiency; at the same time, the use of preset models can also adapt to different abnormal scenarios and enhance robustness.

[0120] In another embodiment of the present application, based on the anomaly detection method described in the above embodiment, Fig.11 Schematic diagram of an application framework of an anomaly detection method provided in an embodiment of the present application. Fig.11As shown, the application framework may include: a data classification module 1101, a data preprocessing module 1102, a data annotation module 1103 and a model design and training module 1104. Among them, the data classification module 1101 is used to divide the sampled data into a training sample set and a test sample set in a ratio of 7:3, and the training sample set includes multiple groups of training data, and the test sample set includes multiple groups of test data. The data preprocessing module 1102 is used to perform data preprocessing on multiple groups of training data, including operations such as denoising and filtering; the data annotation module 1103 is used to add corresponding sample labels to the sample data to be input into the model, for example, to distinguish which data are abnormal samples (for example, there are abnormalities caused by peripheral factors of the RF chip) and which data are normal samples (for example, there are no abnormalities caused by peripheral factors of the RF chip). The model design and training module 1104 is used to perform model training based on these sample data, and finally the detection results can be output according to the preset model after training.

[0121] Here, the definition of data is as shown in the above formula (1). Assume that the RF signal is x[n], and its corresponding spectrum data is X[k], n represents the time index, and k represents the frequency index. In formula (1), N is the number of sampling points, e is the base of the natural logarithm, and 0≤k <N。

[0122] The spectrum data of the RF signal after FFT processing is Gaussian filtered. Here, the calculation of Gaussian filtering is shown in the above formula (2), where G[nk] represents the value of the Gaussian distribution at nk. Due to the infinite length of the Gaussian distribution, a truncated Gaussian filter is usually used in actual calculations, and only values ​​within a limited range are considered here.

[0123] Then the filtered data after Gaussian filtering is normalized by Z-Score standardization and regularization. Among them, Z-Score standardization is used by calculating the difference between each sample and the mean of the entire sample and dividing it by the standard deviation of the entire sample, as shown in the above formula (3). Regularization is used to map the data to a certain range, for example, scaling the data to between [0,1], as shown in the above formula (4).

[0124] In addition, Z-Score standardization can transform data into a distribution with zero mean and unit variance, making different features comparable and eliminating the influence of feature scale. This is very important for the effectiveness and stability of many machine learning algorithms such as clustering and regression models. Regularization maps data to a certain range, usually between [0,1]. It retains the relative relationship of the original data by scaling the data while limiting the data to a specific range. Regularization is often used in situations where the weights of each feature need to be considered proportionally. It is more robust to outliers and can avoid model bias problems caused by different numerical ranges of different features.

[0125] In some embodiments, the model design is as follows Figure 8 As shown, it can include five linear layers and five activation layers, and finally output a model with a binary classification result. This binary classification model can be used to solve the problem of anomaly detection, especially for the case where the sampled IQ spectrum data is caused by the abnormal state of the peripheral communication interface (Serdes). Through model training, it can learn the correlation and weight between different features, and use this information to classify new sampled data.

[0126] During the model training process, the labeled samples are used to optimize the model parameters so that it can accurately determine whether the sampled IQ spectrum data is related to the peripheral communication interface (Serdes) state. By appropriately selecting a loss function, such as a binary cross entropy loss function, the prediction error of the model can be minimized. Subsequently, using a suitable optimization algorithm, such as the gradient descent method, the weights and biases of the model can be updated to improve the performance of the model.

[0127] Thus, in practical applications, the newly sampled IQ spectrum data can be input into the trained preset model, and the output probability value can be calculated to determine whether there is an abnormality caused by the state of the peripheral communication interface (Serdes). If the output probability value is within the first preset range, such as the probability value is close to 0, it indicates that there may be an abnormality, and it is recommended to resample the data; if the output probability value is within the second preset range, such as the probability value is close to 1, it can be determined as a normal situation, and the data can be processed continuously at this time to complete the transmission and reception of the relevant data.

[0128] In simple terms, the above binary classification model is a combination of a linear layer and an activation layer, which can help quickly and accurately determine whether the sampled IQ spectrum data is abnormal due to an abnormal state of the peripheral communication interface (Serdes), thereby improving the efficiency and accuracy of data processing.

[0129] In some embodiments, the present application provides a method for processing radio frequency signal anomalies based on a neural network, and the detailed process may include:

[0130] 1) Data collection: Collect a lot of sampling data and corresponding peripheral communication interface (Serdes) status information.

[0131] 2) Data preprocessing: Preprocess the collected sampling data, including noise removal, missing value processing, filtering and other operations.

[0132] 3) Model training: Use multiple sets of processed sample data to train the neural network model. Use the input data as the training sample set, and provide a sample label (0 or 1) associated with each sample, where the sample label is 0, indicating an abnormal sample, that is, there is an abnormality caused by peripheral factors of the RF chip; the sample label is 1, indicating a normal sample, that is, there is no abnormality caused by peripheral factors of the RF chip. By iteratively optimizing the model parameters, it can subsequently accurately determine whether the newly sampled IQ spectrum data abnormality is caused by peripheral factors of the chip, such as whether it is caused by the state of the peripheral communication interface (Serdes).

[0133] 4) Model evaluation: Use the test sample set to evaluate the model performance, and calculate the model's accuracy, precision, recall and other indicators on the test set to evaluate its classification ability and robustness.

[0134] 5) Anomaly detection: Use the trained and evaluated preset model to input the newly sampled unknown data. By calculating the output probability value of the preset model, determine whether it is related to the peripheral communication interface (Serdes) state. For example, if the probability value is close to 0, it is judged as an abnormal situation; if the probability value is close to 1, it is judged as a normal situation.

[0135] 6) Processing decision: Make corresponding processing decisions based on the judgment results of the model. If it is judged to be an abnormal situation, it is recommended to resample the data; if it is judged to be a normal situation, then you can continue to process the data.

[0136] 7) Continuous optimization: Based on feedback and improvement needs in actual applications, the performance and accuracy of the preset model can be continuously optimized. For example, this may include adjusting the network structure of the model, updating training data, etc.

[0137] That is to say, the embodiment of the present application proposes a RF signal detection method based on a neural network, which aims to solve the problem that the threshold setting method in the related art cannot effectively judge the abnormal working state of the chip. By introducing a machine learning or model-driven abnormality detection method, on the one hand, the accuracy of abnormal signal judgment can be improved: the traditional threshold setting method may not be able to distinguish between the abnormality of the chip itself and the abnormality caused by the communication interface status problem. The use of machine learning or model-driven methods, combined with a large number of known abnormalities and normal spectrum samples for learning, can more accurately judge whether the new sample is abnormal or not, improve the accuracy of abnormal signal judgment, and improve test efficiency. On the other hand, exclude non-essential abnormal factors: Since the goal of chip testing is to retain the abnormal signals of the chip itself and exclude abnormalities caused by other peripheral factors, the technical solution can accurately identify the abnormalities caused by the chip itself and exclude the abnormalities caused by the peripheral communication interface status problem by establishing a model of the chip working behavior and comparing it with the expected mode (real abnormal situation). On the other hand, it enhances the robustness of the automated test system: the traditional threshold setting method often requires setting the threshold value based on experience in advance, which is difficult to adapt to complex and changeable abnormal situations; and this technical solution, by introducing machine learning or model-driven methods, can automatically adapt to different abnormal scenarios according to actual conditions, thereby enhancing the robustness of the automated test system. On the other hand, it improves test efficiency and reduces manual intervention: the traditional threshold setting method requires manual adjustment of the threshold value, involves a lot of manual intervention, and has low test efficiency; and this technical solution adopts machine learning or model-driven methods, which can realize automatic abnormal judgment, reduce manual intervention, and improve test efficiency.

[0138] In summary, the specific implementation of the aforementioned embodiment is elaborated in detail through the above embodiment. It can be seen that through the technical scheme of the aforementioned embodiment, a technical scheme for judging the abnormality of spectrum data in the chip performance test is proposed here. Different from the threshold setting method in the related art, the technical scheme adopts machine learning or model-driven abnormality detection method to accurately judge whether the radio frequency signal is abnormal; by introducing a large number of known abnormal samples and normal samples for learning, the technical scheme can identify and distinguish the abnormalities caused by the chip itself and the abnormalities caused by peripheral factors of the chip (such as communication interface status problems). In addition, by establishing a model of the chip working behavior and comparing it with the expected mode, the technical scheme can exclude non-essential abnormal factors; the technical scheme can improve the accuracy of the abnormal judgment of the radio frequency signal, enhance the robustness of the automated test system, and improve the test efficiency and reduce manual intervention. In short, the abnormal detection method of the embodiment of the present application has the advantages of accuracy, exclusion of non-essential abnormal factors, enhancement of system robustness, improvement of test efficiency and reduction of manual intervention in the abnormal judgment of the radio frequency signal in the chip performance test, and improves the accuracy of signal abnormality detection.

[0139] In yet another embodiment of the present application, Fig.12 A schematic diagram of the structure of an abnormality detection device provided in an embodiment of the present application Figure 1 .like Fig.12 As shown, the anomaly detection device 120 may include an acquisition unit 1201, a preprocessing unit 1202 and an anomaly detection unit 1203, wherein:

[0140] The acquisition unit 1201 is configured to acquire a radio frequency signal to be detected, where the radio frequency signal is a signal received by the radio frequency chip;

[0141] The preprocessing unit 1202 is configured to preprocess the radio frequency signal to determine the model input data;

[0142] The anomaly detection unit 1203 is configured to input the model input data into a preset model, and determine the detection result of the RF signal through the preset model; wherein the detection result is used to indicate whether the RF signal has an anomaly caused by peripheral factors of the RF chip.

[0143] In some embodiments, the preprocessing unit 1202 is configured to perform Fourier transform on the radio frequency signal to obtain spectrum data; and perform data preprocessing on the spectrum data to obtain model input data.

[0144] In some embodiments, the preprocessing unit 1202 is further configured to perform filtering processing on the spectrum data to obtain filtered data; and perform normalization processing on the filtered data to obtain model input data.

[0145] In some embodiments, the preprocessing unit 1202 is further configured to perform standardization and / or regularization on the filtered data to obtain model input data.

[0146] In some embodiments, the anomaly detection unit 1203 is configured to obtain a probability value output by a preset model; and determine a detection result of the radio frequency signal according to the probability value.

[0147] In some embodiments, the anomaly detection unit 1203 is further configured to, when the probability value is within a first preset range, determine that the detection result of the RF signal indicates that the RF signal has an anomaly caused by factors peripheral to the RF chip; and when the probability value is within a second preset range, determine that the detection result of the RF signal indicates that the RF signal does not have an anomaly caused by factors peripheral to the RF chip; wherein the first preset range is different from the second preset range.

[0148] In some embodiments, Fig.13 A schematic diagram of the structure of an abnormality detection device provided in an embodiment of the present application Figure 2 .like Fig.13As shown, the anomaly detection device 120 may also include a model training unit 1204, which is configured to obtain a training sample set, the training sample set including multiple groups of training data; perform data preprocessing on the multiple groups of training data respectively, determine multiple groups of sample data, and add corresponding sample labels for the multiple groups of sample data; wherein the sample label is used to indicate whether the sample data has anomalies caused by peripheral factors of the radio frequency chip; and perform model training on the neural network model according to the multiple groups of sample data and the corresponding sample labels, and when the model training meets the preset convergence conditions, determine the trained neural network model as the preset model.

[0149] In some embodiments, see Fig.13 The anomaly detection device 120 may also include a model testing unit 1205, which is configured to obtain a test sample set after obtaining the preset model, and the test sample set includes multiple groups of test data; and use the multiple groups of test data to perform model evaluation on the preset model to determine the model evaluation result; wherein the model evaluation result is used to indicate the anomaly detection capability of the preset model.

[0150] It is understandable that in the embodiments of the present application, a "unit" may be a part of a circuit, a part of a processor, a part of a program or software, etc., and of course, it may be a module, or it may be non-modular. Moreover, the components in the present embodiment may be integrated into a processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of a software functional module.

[0151] If the integrated unit is implemented in the form of a software function module and is not sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment is essentially or the part that contributes to the prior art or the whole or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) or a processor to perform all or part of the steps of the method described in this embodiment. The aforementioned storage medium includes: U disk, mobile hard disk, read only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0152] Therefore, this embodiment provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by at least one processor, the steps of the anomaly detection method described in any one of the aforementioned embodiments are implemented.

[0153] In some embodiments, Fig.14 A schematic diagram of the hardware structure of a test device provided in an embodiment of the present application. Fig.14 As shown, the test device 140 may include: a communication interface 1401, a memory 1402 and a processor 1403; each component is coupled together via a bus system 1404. It is understood that the bus system 1404 is used to achieve connection and communication between these components. In addition to the data bus, the bus system 1404 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, Fig.14 In the figure, various buses are labeled as bus system 1404. Among them:

[0154] The communication interface 1401 is used to receive and send signals during the process of sending and receiving information with the chip under test;

[0155] A memory 1402, used for storing a computer program that can be run on the processor 1403;

[0156] The processor 1403 is configured to execute, when running the computer program:

[0157] Acquire a radio frequency signal to be detected, where the radio frequency signal is a signal received by a radio frequency chip;

[0158] Preprocessing the radio frequency signal to determine model input data;

[0159] The model input data is input into a preset model, and a detection result of the radio frequency signal is determined by the preset model; wherein the detection result is used to indicate whether the radio frequency signal has an anomaly caused by peripheral factors of the radio frequency chip.

[0160] It can be understood that the memory 1402 in the embodiment of the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct memory bus random access memory (DRRAM). The memory 1402 of the systems and methods described herein is intended to include, but is not limited to, these and any other suitable types of memory.

[0161] The processor 1403 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 1403. The above-mentioned processor 1403 can be a general-purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiments of the present application can be directly embodied as a hardware decoding processor to execute, or the hardware and software modules in the decoding processor are combined and executed. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 1402, and the processor 1403 reads the information in the memory 1402 and completes the steps of the above method in combination with its hardware.

[0162] It is understood that the embodiments described herein may be implemented in hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit may be implemented in one or more application specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), general purpose processors, controllers, microcontrollers, microprocessors, other electronic units for performing the functions described in the present application, or a combination thereof.

[0163] For software implementation, the techniques described herein can be implemented by modules (e.g., procedures, functions, etc.) that perform the functions described herein. The software code can be stored in a memory and executed by a processor. The memory can be implemented in the processor or outside the processor.

[0164] Optionally, as another embodiment, the processor 1403 is further configured to execute the steps of the method described in any one of the aforementioned embodiments when running the computer program.

[0165] In some embodiments, the preset model in the embodiments of the present application is a software model. The preset model can be integrated into the RF chip, or the preset model can also be integrated into the test device 140, without any limitation here.

[0166] The above is only a preferred embodiment of the present application and is not intended to limit the scope of protection of the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the scope of protection of the present application.

[0167] An embodiment of the present application also provides a computer program product including instructions, including a computer program or instructions, which, when executed by a processor, implements the steps in the anomaly detection method provided in the above embodiment.

[0168] The embodiment of the present application also provides a computer program, which can be applied to the test device in the embodiment of the present application. When the computer program is run, the test device can execute the steps in the anomaly detection method provided in the aforementioned embodiment.

[0169] It should be understood that the description of the above storage medium and device embodiments is similar to the description of the above method embodiments, and has similar beneficial effects as the method embodiments. For technical details not disclosed in the storage medium, storage medium and device embodiments of this application, please refer to the description of the method embodiments of this application for understanding.

[0170] It should be understood that "one embodiment" or "an embodiment" or "some embodiments" mentioned throughout the specification means that specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present application. Therefore, "in one embodiment" or "in one embodiment" or "in some embodiments" appearing throughout the specification may not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. It should be understood that in various embodiments of the present application, the size of the sequence number of the above-mentioned processes does not mean the order of execution, and the execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiment of the present application. The above-mentioned sequence numbers of the embodiments of the present application are only for description and do not represent the advantages and disadvantages of the embodiments. The above description of each embodiment tends to emphasize the differences between the various embodiments, and the same or similar aspects can be referenced to each other. For the sake of brevity, this article will not repeat them.

[0171] It should also be noted that in this application, the terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device including the element.

[0172] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0173] The methods disclosed in several method embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments.

[0174] The features disclosed in several product embodiments provided in this application can be arbitrarily combined without conflict to obtain new product embodiments.

[0175] The features disclosed in several method or device embodiments provided in this application can be arbitrarily combined without conflict to obtain new method embodiments or device embodiments.

[0176] The above are only specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0177] The above is only a preferred embodiment of the present application and is not intended to limit the scope of protection of the present application. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. An anomaly detection method, characterized in that: The method comprises: Acquire a radio frequency signal to be detected, where the radio frequency signal is a signal received by a radio frequency chip; Preprocessing the radio frequency signal to determine model input data; The model input data is input into a preset model, and a detection result of the radio frequency signal is determined by the preset model; wherein the detection result is used to indicate whether the radio frequency signal has an anomaly caused by peripheral factors of the radio frequency chip.

2. The method according to claim 1, characterized in that The preprocessing of the radio frequency signal to determine the model input data includes: Performing Fourier transform on the radio frequency signal to obtain spectrum data; The spectrum data is preprocessed to obtain the model input data.

3. The method according to claim 2, characterized in that The preprocessing of the spectrum data to obtain the model input data includes: Performing filtering on the spectrum data to obtain filtered data; The filtered data is normalized to obtain the model input data.

4. The method according to claim 3, characterized in that The normalizing the filtered data to obtain the model input data includes: The filtered data is standardized and / or regularized to obtain the model input data.

5. The method according to claim 1, characterized in that The determining the detection result of the radio frequency signal by using the preset model includes: Obtaining a probability value output by the preset model; wherein the probability value represents the probability that the model input data contains an abnormality caused by peripheral factors of the RF chip; A detection result of the radio frequency signal is determined according to the probability value.

6. The method according to claim 5, characterized in that The determining, according to the probability value, a detection result of the radio frequency signal includes: When the probability value is within a first preset range, determining that the detection result of the radio frequency signal indicates that the radio frequency signal has an abnormality caused by peripheral factors of the radio frequency chip; or, When the probability value is within the second preset range, it is determined that the detection result of the radio frequency signal indicates that there is no abnormality in the radio frequency signal caused by peripheral factors of the radio frequency chip.

7. The method according to any one of claims 1 to 6, characterized in that The method further comprises: Acquire a training sample set, wherein the training sample set includes multiple groups of training data; Performing data preprocessing on the multiple sets of training data respectively, determining multiple sets of sample data, and adding corresponding sample labels to the multiple sets of sample data; wherein the sample labels are used to indicate whether the sample data has anomalies caused by peripheral factors of the radio frequency chip; The neural network model is trained according to the multiple groups of sample data and the corresponding sample labels, and when the model training meets the preset convergence condition, the trained neural network model is determined as the preset model.

8. The method according to claim 7, characterized in that After obtaining the preset model, the method further includes: Acquire a test sample set, wherein the test sample set includes multiple groups of test data; The preset model is evaluated using the multiple groups of test data to determine a model evaluation result; wherein the model evaluation result is used to indicate the anomaly detection capability of the preset model.

9. The method according to any one of claims 1 to 6, characterized in that The method further comprises: When the detection result of the radio frequency signal indicates that the radio frequency signal has an abnormality caused by peripheral factors of the radio frequency chip, reacquire the radio frequency signal to be detected; or, When the detection result of the radio frequency signal indicates that the radio frequency signal does not have an abnormality caused by peripheral factors of the radio frequency chip, data transmission is performed based on the radio frequency signal.

10. The method according to any one of claims 1 to 6, characterized in that The method further comprises: Adjusting some parameters in the preset model according to the detection result of the radio frequency signal and / or the model adjustment requirement; Among them, the partial parameters include at least one of the following: the network structure of the preset model and the training sample set of the preset model.

11. An abnormality detection device, characterized in that: The anomaly detection device comprises an acquisition unit, a preprocessing unit and an anomaly detection unit, wherein: The acquisition unit is configured to acquire a radio frequency signal to be detected, where the radio frequency signal is a signal received by the radio frequency chip; The preprocessing unit is configured to preprocess the radio frequency signal to determine model input data; The abnormality detection unit is configured to input the model input data into a preset model, and determine the detection result of the radio frequency signal through the preset model; wherein the detection result is used to indicate whether the radio frequency signal has an abnormality caused by peripheral factors of the radio frequency chip.

12. A testing device, characterized in that: The test device comprises a memory and a processor, wherein: The memory is used to store a computer program that can be run on the processor; The processor is configured to execute the anomaly detection method according to any one of claims 1 to 10 when running the computer program.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the abnormality detection method according to any one of claims 1 to 10 is implemented.

14. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instruction is executed by a processor, the abnormality detection method according to any one of claims 1 to 10 is implemented.

Images